From 23393312b7183fa61d4a6ba9e97af21f2337a8af Mon Sep 17 00:00:00 2001 From: jacqueline Date: Tue, 1 Aug 2023 10:14:23 +1000 Subject: [PATCH] checkin in opus reference decoder --- lib/opus/.appveyor.yml | 37 + lib/opus/.gitlab-ci.yml | 114 + lib/opus/.travis.yml | 21 + lib/opus/AUTHORS | 6 + lib/opus/CMakeLists.txt | 646 ++ lib/opus/COPYING | 44 + lib/opus/ChangeLog | 0 lib/opus/LICENSE_PLEASE_READ.txt | 22 + lib/opus/Makefile.am | 382 + lib/opus/Makefile.mips | 161 + lib/opus/Makefile.unix | 159 + lib/opus/NEWS | 0 lib/opus/README | 161 + lib/opus/README.draft | 54 + lib/opus/autogen.sh | 14 + lib/opus/celt/_kiss_fft_guts.h | 182 + lib/opus/celt/arch.h | 291 + lib/opus/celt/arm/arm2gnu.pl | 353 + lib/opus/celt/arm/arm_celt_map.c | 160 + lib/opus/celt/arm/armcpu.c | 194 + lib/opus/celt/arm/armcpu.h | 77 + lib/opus/celt/arm/armopts.s.in | 37 + lib/opus/celt/arm/celt_fft_ne10.c | 173 + lib/opus/celt/arm/celt_mdct_ne10.c | 258 + lib/opus/celt/arm/celt_neon_intr.c | 211 + lib/opus/celt/arm/celt_pitch_xcorr_arm.s | 551 ++ lib/opus/celt/arm/fft_arm.h | 71 + lib/opus/celt/arm/fixed_arm64.h | 35 + lib/opus/celt/arm/fixed_armv4.h | 80 + lib/opus/celt/arm/fixed_armv5e.h | 151 + lib/opus/celt/arm/kiss_fft_armv4.h | 121 + lib/opus/celt/arm/kiss_fft_armv5e.h | 118 + lib/opus/celt/arm/mdct_arm.h | 59 + lib/opus/celt/arm/pitch_arm.h | 160 + lib/opus/celt/arm/pitch_neon_intr.c | 281 + lib/opus/celt/bands.c | 1673 ++++ lib/opus/celt/bands.h | 123 + lib/opus/celt/celt.c | 316 + lib/opus/celt/celt.h | 251 + lib/opus/celt/celt_decoder.c | 1375 +++ lib/opus/celt/celt_encoder.c | 2613 ++++++ lib/opus/celt/celt_lpc.c | 344 + lib/opus/celt/celt_lpc.h | 66 + lib/opus/celt/cpu_support.h | 71 + lib/opus/celt/cwrs.c | 715 ++ lib/opus/celt/cwrs.h | 48 + lib/opus/celt/dump_modes/Makefile | 32 + lib/opus/celt/dump_modes/dump_modes.c | 353 + lib/opus/celt/dump_modes/dump_modes_arch.h | 45 + .../celt/dump_modes/dump_modes_arm_ne10.c | 152 + lib/opus/celt/ecintrin.h | 91 + lib/opus/celt/entcode.c | 153 + lib/opus/celt/entcode.h | 152 + lib/opus/celt/entdec.c | 245 + lib/opus/celt/entdec.h | 100 + lib/opus/celt/entenc.c | 294 + lib/opus/celt/entenc.h | 110 + lib/opus/celt/fixed_c5x.h | 79 + lib/opus/celt/fixed_c6x.h | 70 + lib/opus/celt/fixed_debug.h | 836 ++ lib/opus/celt/fixed_generic.h | 188 + lib/opus/celt/float_cast.h | 152 + lib/opus/celt/kiss_fft.c | 604 ++ lib/opus/celt/kiss_fft.h | 210 + lib/opus/celt/laplace.c | 134 + lib/opus/celt/laplace.h | 48 + lib/opus/celt/mathops.c | 209 + lib/opus/celt/mathops.h | 290 + lib/opus/celt/mdct.c | 343 + lib/opus/celt/mdct.h | 112 + lib/opus/celt/meson.build | 69 + lib/opus/celt/mfrngcod.h | 48 + lib/opus/celt/mips/celt_mipsr1.h | 152 + lib/opus/celt/mips/fixed_generic_mipsr1.h | 126 + lib/opus/celt/mips/kiss_fft_mipsr1.h | 167 + lib/opus/celt/mips/mdct_mipsr1.h | 288 + lib/opus/celt/mips/pitch_mipsr1.h | 161 + lib/opus/celt/mips/vq_mipsr1.h | 116 + lib/opus/celt/modes.c | 445 + lib/opus/celt/modes.h | 75 + lib/opus/celt/opus_custom_demo.c | 210 + lib/opus/celt/os_support.h | 91 + lib/opus/celt/pitch.c | 546 ++ lib/opus/celt/pitch.h | 192 + lib/opus/celt/quant_bands.c | 563 ++ lib/opus/celt/quant_bands.h | 66 + lib/opus/celt/rate.c | 646 ++ lib/opus/celt/rate.h | 101 + lib/opus/celt/stack_alloc.h | 184 + lib/opus/celt/static_modes_fixed.h | 892 ++ lib/opus/celt/static_modes_fixed_arm_ne10.h | 388 + lib/opus/celt/static_modes_float.h | 888 ++ lib/opus/celt/static_modes_float_arm_ne10.h | 404 + lib/opus/celt/tests/meson.build | 19 + lib/opus/celt/tests/test_unit_cwrs32.c | 161 + lib/opus/celt/tests/test_unit_dft.c | 180 + lib/opus/celt/tests/test_unit_entropy.c | 383 + lib/opus/celt/tests/test_unit_laplace.c | 93 + lib/opus/celt/tests/test_unit_mathops.c | 266 + lib/opus/celt/tests/test_unit_mdct.c | 228 + lib/opus/celt/tests/test_unit_rotation.c | 86 + lib/opus/celt/tests/test_unit_types.c | 50 + lib/opus/celt/vq.c | 442 + lib/opus/celt/vq.h | 79 + lib/opus/celt/x86/celt_lpc_sse.h | 67 + lib/opus/celt/x86/celt_lpc_sse4_1.c | 89 + lib/opus/celt/x86/pitch_sse.c | 185 + lib/opus/celt/x86/pitch_sse.h | 192 + lib/opus/celt/x86/pitch_sse2.c | 95 + lib/opus/celt/x86/pitch_sse4_1.c | 240 + lib/opus/celt/x86/vq_sse.h | 50 + lib/opus/celt/x86/vq_sse2.c | 217 + lib/opus/celt/x86/x86_celt_map.c | 167 + lib/opus/celt/x86/x86cpu.c | 178 + lib/opus/celt/x86/x86cpu.h | 73 + lib/opus/celt_headers.mk | 53 + lib/opus/celt_sources.mk | 52 + lib/opus/cmake/CFeatureCheck.cmake | 39 + lib/opus/cmake/OpusBuildtype.cmake | 27 + lib/opus/cmake/OpusConfig.cmake | 117 + lib/opus/cmake/OpusConfig.cmake.in | 20 + lib/opus/cmake/OpusFunctions.cmake | 229 + lib/opus/cmake/OpusPackageVersion.cmake | 70 + lib/opus/cmake/OpusSources.cmake | 49 + lib/opus/cmake/README.md | 132 + lib/opus/cmake/RunTest.cmake | 61 + lib/opus/cmake/config.h.cmake.in | 1 + lib/opus/cmake/cpu_info_by_asm.c | 31 + lib/opus/cmake/cpu_info_by_c.c | 9 + lib/opus/cmake/vla.c | 7 + lib/opus/configure.ac | 951 ++ lib/opus/doc/Doxyfile.in | 340 + lib/opus/doc/Makefile.am | 45 + lib/opus/doc/build_draft.sh | 113 + lib/opus/doc/build_isobmff.sh | 50 + lib/opus/doc/build_oggdraft.sh | 52 + lib/opus/doc/customdoxygen.css | 1011 ++ lib/opus/doc/draft-ietf-codec-oggopus.xml | 1873 ++++ lib/opus/doc/draft-ietf-codec-opus-update.xml | 513 + lib/opus/doc/draft-ietf-codec-opus.xml | 8276 +++++++++++++++++ lib/opus/doc/draft-ietf-payload-rtp-opus.xml | 960 ++ lib/opus/doc/footer.html | 32 + lib/opus/doc/header.html | 61 + lib/opus/doc/meson.build | 33 + lib/opus/doc/opus_in_isobmff.css | 60 + lib/opus/doc/opus_in_isobmff.html | 692 ++ lib/opus/doc/opus_logo.svg | 157 + lib/opus/doc/opus_update.patch | 244 + lib/opus/doc/release.txt | 43 + lib/opus/doc/trivial_example.c | 171 + lib/opus/include/meson.build | 13 + lib/opus/include/opus.h | 981 ++ lib/opus/include/opus_custom.h | 343 + lib/opus/include/opus_defines.h | 801 ++ lib/opus/include/opus_multistream.h | 660 ++ lib/opus/include/opus_projection.h | 568 ++ lib/opus/include/opus_types.h | 166 + lib/opus/m4/as-gcc-inline-assembly.m4 | 98 + lib/opus/m4/ax_add_fortify_source.m4 | 53 + lib/opus/m4/opus-intrinsics.m4 | 29 + lib/opus/meson.build | 675 ++ lib/opus/meson/get-version.py | 86 + lib/opus/meson/read-sources-list.py | 28 + lib/opus/meson_options.txt | 22 + lib/opus/opus-uninstalled.pc.in | 12 + lib/opus/opus.m4 | 117 + lib/opus/opus.pc.in | 16 + lib/opus/opus_headers.mk | 9 + lib/opus/opus_sources.mk | 16 + lib/opus/releases.sha2 | 79 + lib/opus/scripts/dump_rnn.py | 57 + lib/opus/scripts/rnn_train.py | 67 + lib/opus/silk/A2NLSF.c | 267 + lib/opus/silk/API.h | 135 + lib/opus/silk/CNG.c | 188 + lib/opus/silk/HP_variable_cutoff.c | 77 + lib/opus/silk/Inlines.h | 188 + lib/opus/silk/LPC_analysis_filter.c | 111 + lib/opus/silk/LPC_fit.c | 82 + lib/opus/silk/LPC_inv_pred_gain.c | 141 + lib/opus/silk/LP_variable_cutoff.c | 135 + lib/opus/silk/MacroCount.h | 710 ++ lib/opus/silk/MacroDebug.h | 945 ++ lib/opus/silk/NLSF2A.c | 141 + lib/opus/silk/NLSF_VQ.c | 76 + lib/opus/silk/NLSF_VQ_weights_laroia.c | 80 + lib/opus/silk/NLSF_decode.c | 93 + lib/opus/silk/NLSF_del_dec_quant.c | 215 + lib/opus/silk/NLSF_encode.c | 124 + lib/opus/silk/NLSF_stabilize.c | 142 + lib/opus/silk/NLSF_unpack.c | 55 + lib/opus/silk/NSQ.c | 437 + lib/opus/silk/NSQ.h | 101 + lib/opus/silk/NSQ_del_dec.c | 733 ++ lib/opus/silk/PLC.c | 446 + lib/opus/silk/PLC.h | 62 + lib/opus/silk/SigProc_FIX.h | 643 ++ lib/opus/silk/VAD.c | 360 + lib/opus/silk/VQ_WMat_EC.c | 131 + lib/opus/silk/ana_filt_bank_1.c | 74 + lib/opus/silk/arm/LPC_inv_pred_gain_arm.h | 57 + .../silk/arm/LPC_inv_pred_gain_neon_intr.c | 288 + lib/opus/silk/arm/NSQ_del_dec_arm.h | 100 + lib/opus/silk/arm/NSQ_del_dec_neon_intr.c | 1124 +++ lib/opus/silk/arm/NSQ_neon.c | 112 + lib/opus/silk/arm/NSQ_neon.h | 114 + lib/opus/silk/arm/SigProc_FIX_armv4.h | 47 + lib/opus/silk/arm/SigProc_FIX_armv5e.h | 61 + lib/opus/silk/arm/arm_silk_map.c | 123 + lib/opus/silk/arm/biquad_alt_arm.h | 68 + lib/opus/silk/arm/biquad_alt_neon_intr.c | 156 + lib/opus/silk/arm/macros_arm64.h | 39 + lib/opus/silk/arm/macros_armv4.h | 110 + lib/opus/silk/arm/macros_armv5e.h | 220 + lib/opus/silk/biquad_alt.c | 121 + lib/opus/silk/bwexpander.c | 51 + lib/opus/silk/bwexpander_32.c | 51 + lib/opus/silk/check_control_input.c | 106 + lib/opus/silk/code_signs.c | 115 + lib/opus/silk/control.h | 150 + lib/opus/silk/control_SNR.c | 113 + lib/opus/silk/control_audio_bandwidth.c | 132 + lib/opus/silk/control_codec.c | 423 + lib/opus/silk/debug.c | 174 + lib/opus/silk/debug.h | 267 + lib/opus/silk/dec_API.c | 419 + lib/opus/silk/decode_core.c | 237 + lib/opus/silk/decode_frame.c | 129 + lib/opus/silk/decode_indices.c | 151 + lib/opus/silk/decode_parameters.c | 115 + lib/opus/silk/decode_pitch.c | 77 + lib/opus/silk/decode_pulses.c | 115 + lib/opus/silk/decoder_set_fs.c | 108 + lib/opus/silk/define.h | 235 + lib/opus/silk/enc_API.c | 587 ++ lib/opus/silk/encode_indices.c | 181 + lib/opus/silk/encode_pulses.c | 206 + lib/opus/silk/errors.h | 98 + lib/opus/silk/fixed/LTP_analysis_filter_FIX.c | 90 + lib/opus/silk/fixed/LTP_scale_ctrl_FIX.c | 58 + lib/opus/silk/fixed/apply_sine_window_FIX.c | 101 + .../arm/warped_autocorrelation_FIX_arm.h | 68 + .../warped_autocorrelation_FIX_neon_intr.c | 265 + lib/opus/silk/fixed/autocorr_FIX.c | 48 + lib/opus/silk/fixed/burg_modified_FIX.c | 280 + lib/opus/silk/fixed/corrMatrix_FIX.c | 150 + lib/opus/silk/fixed/encode_frame_FIX.c | 448 + lib/opus/silk/fixed/find_LPC_FIX.c | 151 + lib/opus/silk/fixed/find_LTP_FIX.c | 99 + lib/opus/silk/fixed/find_pitch_lags_FIX.c | 143 + lib/opus/silk/fixed/find_pred_coefs_FIX.c | 146 + lib/opus/silk/fixed/k2a_FIX.c | 54 + lib/opus/silk/fixed/k2a_Q16_FIX.c | 54 + lib/opus/silk/fixed/main_FIX.h | 244 + .../mips/noise_shape_analysis_FIX_mipsr1.h | 336 + .../silk/fixed/mips/prefilter_FIX_mipsr1.h | 184 + .../mips/warped_autocorrelation_FIX_mipsr1.h | 165 + .../silk/fixed/noise_shape_analysis_FIX.c | 407 + lib/opus/silk/fixed/pitch_analysis_core_FIX.c | 721 ++ lib/opus/silk/fixed/process_gains_FIX.c | 117 + .../silk/fixed/regularize_correlations_FIX.c | 47 + lib/opus/silk/fixed/residual_energy16_FIX.c | 103 + lib/opus/silk/fixed/residual_energy_FIX.c | 98 + lib/opus/silk/fixed/schur64_FIX.c | 93 + lib/opus/silk/fixed/schur_FIX.c | 107 + lib/opus/silk/fixed/structs_FIX.h | 116 + lib/opus/silk/fixed/vector_ops_FIX.c | 102 + .../silk/fixed/warped_autocorrelation_FIX.c | 92 + .../silk/fixed/x86/burg_modified_FIX_sse4_1.c | 400 + .../silk/fixed/x86/vector_ops_FIX_sse4_1.c | 93 + lib/opus/silk/float/LPC_analysis_filter_FLP.c | 249 + lib/opus/silk/float/LPC_inv_pred_gain_FLP.c | 73 + lib/opus/silk/float/LTP_analysis_filter_FLP.c | 75 + lib/opus/silk/float/LTP_scale_ctrl_FLP.c | 58 + lib/opus/silk/float/SigProc_FLP.h | 197 + lib/opus/silk/float/apply_sine_window_FLP.c | 81 + lib/opus/silk/float/autocorrelation_FLP.c | 52 + lib/opus/silk/float/burg_modified_FLP.c | 186 + lib/opus/silk/float/bwexpander_FLP.c | 49 + lib/opus/silk/float/corrMatrix_FLP.c | 93 + lib/opus/silk/float/encode_frame_FLP.c | 435 + lib/opus/silk/float/energy_FLP.c | 59 + lib/opus/silk/float/find_LPC_FLP.c | 104 + lib/opus/silk/float/find_LTP_FLP.c | 64 + lib/opus/silk/float/find_pitch_lags_FLP.c | 132 + lib/opus/silk/float/find_pred_coefs_FLP.c | 117 + lib/opus/silk/float/inner_product_FLP.c | 59 + lib/opus/silk/float/k2a_FLP.c | 54 + lib/opus/silk/float/main_FLP.h | 286 + .../silk/float/noise_shape_analysis_FLP.c | 350 + lib/opus/silk/float/pitch_analysis_core_FLP.c | 630 ++ lib/opus/silk/float/process_gains_FLP.c | 103 + .../silk/float/regularize_correlations_FLP.c | 48 + lib/opus/silk/float/residual_energy_FLP.c | 117 + lib/opus/silk/float/scale_copy_vector_FLP.c | 57 + lib/opus/silk/float/scale_vector_FLP.c | 56 + lib/opus/silk/float/schur_FLP.c | 70 + lib/opus/silk/float/sort_FLP.c | 83 + lib/opus/silk/float/structs_FLP.h | 112 + .../silk/float/warped_autocorrelation_FLP.c | 73 + lib/opus/silk/float/wrappers_FLP.c | 209 + lib/opus/silk/gain_quant.c | 142 + lib/opus/silk/init_decoder.c | 57 + lib/opus/silk/init_encoder.c | 64 + lib/opus/silk/inner_prod_aligned.c | 47 + lib/opus/silk/interpolate.c | 51 + lib/opus/silk/lin2log.c | 46 + lib/opus/silk/log2lin.c | 58 + lib/opus/silk/macros.h | 151 + lib/opus/silk/main.h | 476 + lib/opus/silk/meson.build | 63 + lib/opus/silk/mips/NSQ_del_dec_mipsr1.h | 410 + lib/opus/silk/mips/macros_mipsr1.h | 92 + lib/opus/silk/mips/sigproc_fix_mipsr1.h | 60 + lib/opus/silk/pitch_est_defines.h | 88 + lib/opus/silk/pitch_est_tables.c | 99 + lib/opus/silk/process_NLSFs.c | 107 + lib/opus/silk/quant_LTP_gains.c | 132 + lib/opus/silk/resampler.c | 215 + lib/opus/silk/resampler_down2.c | 74 + lib/opus/silk/resampler_down2_3.c | 103 + lib/opus/silk/resampler_private.h | 88 + lib/opus/silk/resampler_private_AR2.c | 55 + lib/opus/silk/resampler_private_IIR_FIR.c | 107 + lib/opus/silk/resampler_private_down_FIR.c | 194 + lib/opus/silk/resampler_private_up2_HQ.c | 113 + lib/opus/silk/resampler_rom.c | 96 + lib/opus/silk/resampler_rom.h | 68 + lib/opus/silk/resampler_structs.h | 60 + lib/opus/silk/shell_coder.c | 151 + lib/opus/silk/sigm_Q15.c | 76 + lib/opus/silk/sort.c | 154 + lib/opus/silk/stereo_LR_to_MS.c | 229 + lib/opus/silk/stereo_MS_to_LR.c | 85 + lib/opus/silk/stereo_decode_pred.c | 73 + lib/opus/silk/stereo_encode_pred.c | 62 + lib/opus/silk/stereo_find_predictor.c | 79 + lib/opus/silk/stereo_quant_pred.c | 73 + lib/opus/silk/structs.h | 329 + lib/opus/silk/sum_sqr_shift.c | 83 + lib/opus/silk/table_LSF_cos.c | 70 + lib/opus/silk/tables.h | 114 + lib/opus/silk/tables_LTP.c | 294 + lib/opus/silk/tables_NLSF_CB_NB_MB.c | 195 + lib/opus/silk/tables_NLSF_CB_WB.c | 234 + lib/opus/silk/tables_gain.c | 63 + lib/opus/silk/tables_other.c | 124 + lib/opus/silk/tables_pitch_lag.c | 69 + lib/opus/silk/tables_pulses_per_block.c | 264 + lib/opus/silk/tests/meson.build | 8 + .../silk/tests/test_unit_LPC_inv_pred_gain.c | 129 + lib/opus/silk/tuning_parameters.h | 155 + lib/opus/silk/typedef.h | 81 + lib/opus/silk/x86/NSQ_del_dec_sse4_1.c | 915 ++ lib/opus/silk/x86/NSQ_sse4_1.c | 772 ++ lib/opus/silk/x86/SigProc_FIX_sse.h | 99 + lib/opus/silk/x86/VAD_sse4_1.c | 289 + lib/opus/silk/x86/VQ_WMat_EC_sse4_1.c | 173 + lib/opus/silk/x86/main_sse.h | 247 + lib/opus/silk/x86/x86_silk_map.c | 159 + lib/opus/silk_headers.mk | 44 + lib/opus/silk_sources.mk | 158 + lib/opus/src/analysis.c | 983 ++ lib/opus/src/analysis.h | 103 + lib/opus/src/mapping_matrix.c | 378 + lib/opus/src/mapping_matrix.h | 133 + lib/opus/src/meson.build | 45 + lib/opus/src/mlp.c | 144 + lib/opus/src/mlp.h | 60 + lib/opus/src/mlp_data.c | 672 ++ lib/opus/src/opus.c | 356 + lib/opus/src/opus_compare.c | 382 + lib/opus/src/opus_decoder.c | 1041 +++ lib/opus/src/opus_demo.c | 892 ++ lib/opus/src/opus_encoder.c | 2780 ++++++ lib/opus/src/opus_multistream.c | 92 + lib/opus/src/opus_multistream_decoder.c | 552 ++ lib/opus/src/opus_multistream_encoder.c | 1329 +++ lib/opus/src/opus_private.h | 201 + lib/opus/src/opus_projection_decoder.c | 258 + lib/opus/src/opus_projection_encoder.c | 468 + lib/opus/src/repacketizer.c | 349 + lib/opus/src/repacketizer_demo.c | 247 + lib/opus/src/tansig_table.h | 45 + lib/opus/tests/meson.build | 34 + lib/opus/tests/opus_build_test.sh | 29 + lib/opus/tests/opus_decode_fuzzer.c | 124 + lib/opus/tests/opus_decode_fuzzer.options | 2 + lib/opus/tests/opus_encode_regressions.c | 1034 ++ lib/opus/tests/random_config.sh | 126 + lib/opus/tests/run_vectors.sh | 143 + lib/opus/tests/test_opus_api.c | 1904 ++++ lib/opus/tests/test_opus_common.h | 85 + lib/opus/tests/test_opus_decode.c | 463 + lib/opus/tests/test_opus_encode.c | 707 ++ lib/opus/tests/test_opus_padding.c | 90 + lib/opus/tests/test_opus_projection.c | 393 + lib/opus/training/rnn_dump.py | 66 + lib/opus/training/rnn_train.py | 177 + lib/opus/training/txt2hdf5.py | 12 + lib/opus/win32/VS2015/common.props | 82 + lib/opus/win32/VS2015/opus.sln | 168 + lib/opus/win32/VS2015/opus.vcxproj | 399 + lib/opus/win32/VS2015/opus.vcxproj.filters | 585 ++ lib/opus/win32/VS2015/opus_demo.vcxproj | 171 + .../win32/VS2015/opus_demo.vcxproj.filters | 22 + lib/opus/win32/VS2015/test_opus_api.vcxproj | 171 + .../VS2015/test_opus_api.vcxproj.filters | 14 + .../win32/VS2015/test_opus_decode.vcxproj | 171 + .../VS2015/test_opus_decode.vcxproj.filters | 14 + .../win32/VS2015/test_opus_encode.vcxproj | 172 + .../VS2015/test_opus_encode.vcxproj.filters | 17 + lib/opus/win32/config.h | 64 + lib/opus/win32/genversion.bat | 37 + 414 files changed, 101944 insertions(+) create mode 100644 lib/opus/.appveyor.yml create mode 100644 lib/opus/.gitlab-ci.yml create mode 100644 lib/opus/.travis.yml create mode 100644 lib/opus/AUTHORS create mode 100644 lib/opus/CMakeLists.txt create mode 100644 lib/opus/COPYING create mode 100644 lib/opus/ChangeLog create mode 100644 lib/opus/LICENSE_PLEASE_READ.txt create mode 100644 lib/opus/Makefile.am create mode 100644 lib/opus/Makefile.mips create mode 100644 lib/opus/Makefile.unix create mode 100644 lib/opus/NEWS create mode 100644 lib/opus/README create mode 100644 lib/opus/README.draft create mode 100755 lib/opus/autogen.sh create mode 100644 lib/opus/celt/_kiss_fft_guts.h create mode 100644 lib/opus/celt/arch.h create mode 100755 lib/opus/celt/arm/arm2gnu.pl create mode 100644 lib/opus/celt/arm/arm_celt_map.c create mode 100644 lib/opus/celt/arm/armcpu.c create mode 100644 lib/opus/celt/arm/armcpu.h create mode 100644 lib/opus/celt/arm/armopts.s.in create mode 100644 lib/opus/celt/arm/celt_fft_ne10.c create mode 100644 lib/opus/celt/arm/celt_mdct_ne10.c create mode 100644 lib/opus/celt/arm/celt_neon_intr.c create mode 100644 lib/opus/celt/arm/celt_pitch_xcorr_arm.s create mode 100644 lib/opus/celt/arm/fft_arm.h create mode 100644 lib/opus/celt/arm/fixed_arm64.h create mode 100644 lib/opus/celt/arm/fixed_armv4.h create mode 100644 lib/opus/celt/arm/fixed_armv5e.h create mode 100644 lib/opus/celt/arm/kiss_fft_armv4.h create mode 100644 lib/opus/celt/arm/kiss_fft_armv5e.h create mode 100644 lib/opus/celt/arm/mdct_arm.h create mode 100644 lib/opus/celt/arm/pitch_arm.h create mode 100644 lib/opus/celt/arm/pitch_neon_intr.c create mode 100644 lib/opus/celt/bands.c create mode 100644 lib/opus/celt/bands.h create mode 100644 lib/opus/celt/celt.c create mode 100644 lib/opus/celt/celt.h create mode 100644 lib/opus/celt/celt_decoder.c create mode 100644 lib/opus/celt/celt_encoder.c create mode 100644 lib/opus/celt/celt_lpc.c create mode 100644 lib/opus/celt/celt_lpc.h create mode 100644 lib/opus/celt/cpu_support.h create mode 100644 lib/opus/celt/cwrs.c create mode 100644 lib/opus/celt/cwrs.h create mode 100644 lib/opus/celt/dump_modes/Makefile create mode 100644 lib/opus/celt/dump_modes/dump_modes.c create mode 100644 lib/opus/celt/dump_modes/dump_modes_arch.h create mode 100644 lib/opus/celt/dump_modes/dump_modes_arm_ne10.c create mode 100644 lib/opus/celt/ecintrin.h create mode 100644 lib/opus/celt/entcode.c create mode 100644 lib/opus/celt/entcode.h create mode 100644 lib/opus/celt/entdec.c create mode 100644 lib/opus/celt/entdec.h create mode 100644 lib/opus/celt/entenc.c create mode 100644 lib/opus/celt/entenc.h create mode 100644 lib/opus/celt/fixed_c5x.h create mode 100644 lib/opus/celt/fixed_c6x.h create mode 100644 lib/opus/celt/fixed_debug.h create mode 100644 lib/opus/celt/fixed_generic.h create mode 100644 lib/opus/celt/float_cast.h create mode 100644 lib/opus/celt/kiss_fft.c create mode 100644 lib/opus/celt/kiss_fft.h create mode 100644 lib/opus/celt/laplace.c create mode 100644 lib/opus/celt/laplace.h create mode 100644 lib/opus/celt/mathops.c create mode 100644 lib/opus/celt/mathops.h create mode 100644 lib/opus/celt/mdct.c create mode 100644 lib/opus/celt/mdct.h create mode 100644 lib/opus/celt/meson.build create mode 100644 lib/opus/celt/mfrngcod.h create mode 100644 lib/opus/celt/mips/celt_mipsr1.h create mode 100644 lib/opus/celt/mips/fixed_generic_mipsr1.h create mode 100644 lib/opus/celt/mips/kiss_fft_mipsr1.h create mode 100644 lib/opus/celt/mips/mdct_mipsr1.h create mode 100644 lib/opus/celt/mips/pitch_mipsr1.h create mode 100644 lib/opus/celt/mips/vq_mipsr1.h create mode 100644 lib/opus/celt/modes.c create mode 100644 lib/opus/celt/modes.h create mode 100644 lib/opus/celt/opus_custom_demo.c create mode 100644 lib/opus/celt/os_support.h create mode 100644 lib/opus/celt/pitch.c create mode 100644 lib/opus/celt/pitch.h create mode 100644 lib/opus/celt/quant_bands.c create mode 100644 lib/opus/celt/quant_bands.h create mode 100644 lib/opus/celt/rate.c create mode 100644 lib/opus/celt/rate.h create mode 100644 lib/opus/celt/stack_alloc.h create mode 100644 lib/opus/celt/static_modes_fixed.h create mode 100644 lib/opus/celt/static_modes_fixed_arm_ne10.h create mode 100644 lib/opus/celt/static_modes_float.h create mode 100644 lib/opus/celt/static_modes_float_arm_ne10.h create mode 100644 lib/opus/celt/tests/meson.build create mode 100644 lib/opus/celt/tests/test_unit_cwrs32.c create mode 100644 lib/opus/celt/tests/test_unit_dft.c create mode 100644 lib/opus/celt/tests/test_unit_entropy.c create mode 100644 lib/opus/celt/tests/test_unit_laplace.c create mode 100644 lib/opus/celt/tests/test_unit_mathops.c create mode 100644 lib/opus/celt/tests/test_unit_mdct.c create mode 100644 lib/opus/celt/tests/test_unit_rotation.c create mode 100644 lib/opus/celt/tests/test_unit_types.c create mode 100644 lib/opus/celt/vq.c create mode 100644 lib/opus/celt/vq.h create mode 100644 lib/opus/celt/x86/celt_lpc_sse.h create mode 100644 lib/opus/celt/x86/celt_lpc_sse4_1.c create mode 100644 lib/opus/celt/x86/pitch_sse.c create mode 100644 lib/opus/celt/x86/pitch_sse.h create mode 100644 lib/opus/celt/x86/pitch_sse2.c create mode 100644 lib/opus/celt/x86/pitch_sse4_1.c create mode 100644 lib/opus/celt/x86/vq_sse.h create mode 100644 lib/opus/celt/x86/vq_sse2.c create mode 100644 lib/opus/celt/x86/x86_celt_map.c create mode 100644 lib/opus/celt/x86/x86cpu.c create mode 100644 lib/opus/celt/x86/x86cpu.h create mode 100644 lib/opus/celt_headers.mk create mode 100644 lib/opus/celt_sources.mk create mode 100644 lib/opus/cmake/CFeatureCheck.cmake create mode 100644 lib/opus/cmake/OpusBuildtype.cmake create mode 100644 lib/opus/cmake/OpusConfig.cmake create mode 100644 lib/opus/cmake/OpusConfig.cmake.in create mode 100644 lib/opus/cmake/OpusFunctions.cmake create mode 100644 lib/opus/cmake/OpusPackageVersion.cmake create mode 100644 lib/opus/cmake/OpusSources.cmake create mode 100644 lib/opus/cmake/README.md create mode 100644 lib/opus/cmake/RunTest.cmake create mode 100644 lib/opus/cmake/config.h.cmake.in create mode 100644 lib/opus/cmake/cpu_info_by_asm.c create mode 100644 lib/opus/cmake/cpu_info_by_c.c create mode 100644 lib/opus/cmake/vla.c create mode 100644 lib/opus/configure.ac create mode 100644 lib/opus/doc/Doxyfile.in create mode 100644 lib/opus/doc/Makefile.am create mode 100755 lib/opus/doc/build_draft.sh create mode 100755 lib/opus/doc/build_isobmff.sh create mode 100755 lib/opus/doc/build_oggdraft.sh create mode 100644 lib/opus/doc/customdoxygen.css create mode 100644 lib/opus/doc/draft-ietf-codec-oggopus.xml create mode 100644 lib/opus/doc/draft-ietf-codec-opus-update.xml create mode 100644 lib/opus/doc/draft-ietf-codec-opus.xml create mode 100644 lib/opus/doc/draft-ietf-payload-rtp-opus.xml create mode 100644 lib/opus/doc/footer.html create mode 100644 lib/opus/doc/header.html create mode 100644 lib/opus/doc/meson.build create mode 100644 lib/opus/doc/opus_in_isobmff.css create mode 100644 lib/opus/doc/opus_in_isobmff.html create mode 100644 lib/opus/doc/opus_logo.svg create mode 100644 lib/opus/doc/opus_update.patch create mode 100644 lib/opus/doc/release.txt create mode 100644 lib/opus/doc/trivial_example.c create mode 100644 lib/opus/include/meson.build create mode 100644 lib/opus/include/opus.h create mode 100644 lib/opus/include/opus_custom.h create mode 100644 lib/opus/include/opus_defines.h create mode 100644 lib/opus/include/opus_multistream.h create mode 100644 lib/opus/include/opus_projection.h create mode 100644 lib/opus/include/opus_types.h create mode 100644 lib/opus/m4/as-gcc-inline-assembly.m4 create mode 100644 lib/opus/m4/ax_add_fortify_source.m4 create mode 100644 lib/opus/m4/opus-intrinsics.m4 create mode 100644 lib/opus/meson.build create mode 100755 lib/opus/meson/get-version.py create mode 100755 lib/opus/meson/read-sources-list.py create mode 100644 lib/opus/meson_options.txt create mode 100644 lib/opus/opus-uninstalled.pc.in create mode 100644 lib/opus/opus.m4 create mode 100644 lib/opus/opus.pc.in create mode 100644 lib/opus/opus_headers.mk create mode 100644 lib/opus/opus_sources.mk create mode 100644 lib/opus/releases.sha2 create mode 100755 lib/opus/scripts/dump_rnn.py create mode 100755 lib/opus/scripts/rnn_train.py create mode 100644 lib/opus/silk/A2NLSF.c create mode 100644 lib/opus/silk/API.h create mode 100644 lib/opus/silk/CNG.c create mode 100644 lib/opus/silk/HP_variable_cutoff.c create mode 100644 lib/opus/silk/Inlines.h create mode 100644 lib/opus/silk/LPC_analysis_filter.c create mode 100644 lib/opus/silk/LPC_fit.c create mode 100644 lib/opus/silk/LPC_inv_pred_gain.c create mode 100644 lib/opus/silk/LP_variable_cutoff.c create mode 100644 lib/opus/silk/MacroCount.h create mode 100644 lib/opus/silk/MacroDebug.h create mode 100644 lib/opus/silk/NLSF2A.c create mode 100644 lib/opus/silk/NLSF_VQ.c create mode 100644 lib/opus/silk/NLSF_VQ_weights_laroia.c create mode 100644 lib/opus/silk/NLSF_decode.c create mode 100644 lib/opus/silk/NLSF_del_dec_quant.c create mode 100644 lib/opus/silk/NLSF_encode.c create mode 100644 lib/opus/silk/NLSF_stabilize.c create mode 100644 lib/opus/silk/NLSF_unpack.c create mode 100644 lib/opus/silk/NSQ.c create mode 100644 lib/opus/silk/NSQ.h create mode 100644 lib/opus/silk/NSQ_del_dec.c create mode 100644 lib/opus/silk/PLC.c create mode 100644 lib/opus/silk/PLC.h create mode 100644 lib/opus/silk/SigProc_FIX.h create mode 100644 lib/opus/silk/VAD.c create mode 100644 lib/opus/silk/VQ_WMat_EC.c create mode 100644 lib/opus/silk/ana_filt_bank_1.c create mode 100644 lib/opus/silk/arm/LPC_inv_pred_gain_arm.h create mode 100644 lib/opus/silk/arm/LPC_inv_pred_gain_neon_intr.c create mode 100644 lib/opus/silk/arm/NSQ_del_dec_arm.h create mode 100644 lib/opus/silk/arm/NSQ_del_dec_neon_intr.c create mode 100644 lib/opus/silk/arm/NSQ_neon.c create mode 100644 lib/opus/silk/arm/NSQ_neon.h create mode 100644 lib/opus/silk/arm/SigProc_FIX_armv4.h create mode 100644 lib/opus/silk/arm/SigProc_FIX_armv5e.h create mode 100644 lib/opus/silk/arm/arm_silk_map.c create mode 100644 lib/opus/silk/arm/biquad_alt_arm.h create mode 100644 lib/opus/silk/arm/biquad_alt_neon_intr.c create mode 100644 lib/opus/silk/arm/macros_arm64.h create mode 100644 lib/opus/silk/arm/macros_armv4.h create mode 100644 lib/opus/silk/arm/macros_armv5e.h create mode 100644 lib/opus/silk/biquad_alt.c create mode 100644 lib/opus/silk/bwexpander.c create mode 100644 lib/opus/silk/bwexpander_32.c create mode 100644 lib/opus/silk/check_control_input.c create mode 100644 lib/opus/silk/code_signs.c create mode 100644 lib/opus/silk/control.h create mode 100644 lib/opus/silk/control_SNR.c create mode 100644 lib/opus/silk/control_audio_bandwidth.c create mode 100644 lib/opus/silk/control_codec.c create mode 100644 lib/opus/silk/debug.c create mode 100644 lib/opus/silk/debug.h create mode 100644 lib/opus/silk/dec_API.c create mode 100644 lib/opus/silk/decode_core.c create mode 100644 lib/opus/silk/decode_frame.c create mode 100644 lib/opus/silk/decode_indices.c create mode 100644 lib/opus/silk/decode_parameters.c create mode 100644 lib/opus/silk/decode_pitch.c create mode 100644 lib/opus/silk/decode_pulses.c create mode 100644 lib/opus/silk/decoder_set_fs.c create mode 100644 lib/opus/silk/define.h create mode 100644 lib/opus/silk/enc_API.c create mode 100644 lib/opus/silk/encode_indices.c create mode 100644 lib/opus/silk/encode_pulses.c create mode 100644 lib/opus/silk/errors.h create mode 100644 lib/opus/silk/fixed/LTP_analysis_filter_FIX.c create mode 100644 lib/opus/silk/fixed/LTP_scale_ctrl_FIX.c create mode 100644 lib/opus/silk/fixed/apply_sine_window_FIX.c create mode 100644 lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_arm.h create mode 100644 lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_neon_intr.c create mode 100644 lib/opus/silk/fixed/autocorr_FIX.c create mode 100644 lib/opus/silk/fixed/burg_modified_FIX.c create mode 100644 lib/opus/silk/fixed/corrMatrix_FIX.c create mode 100644 lib/opus/silk/fixed/encode_frame_FIX.c create mode 100644 lib/opus/silk/fixed/find_LPC_FIX.c create mode 100644 lib/opus/silk/fixed/find_LTP_FIX.c create mode 100644 lib/opus/silk/fixed/find_pitch_lags_FIX.c create mode 100644 lib/opus/silk/fixed/find_pred_coefs_FIX.c create mode 100644 lib/opus/silk/fixed/k2a_FIX.c create mode 100644 lib/opus/silk/fixed/k2a_Q16_FIX.c create mode 100644 lib/opus/silk/fixed/main_FIX.h create mode 100644 lib/opus/silk/fixed/mips/noise_shape_analysis_FIX_mipsr1.h create mode 100644 lib/opus/silk/fixed/mips/prefilter_FIX_mipsr1.h create mode 100644 lib/opus/silk/fixed/mips/warped_autocorrelation_FIX_mipsr1.h create mode 100644 lib/opus/silk/fixed/noise_shape_analysis_FIX.c create mode 100644 lib/opus/silk/fixed/pitch_analysis_core_FIX.c create mode 100644 lib/opus/silk/fixed/process_gains_FIX.c create mode 100644 lib/opus/silk/fixed/regularize_correlations_FIX.c create mode 100644 lib/opus/silk/fixed/residual_energy16_FIX.c create mode 100644 lib/opus/silk/fixed/residual_energy_FIX.c create mode 100644 lib/opus/silk/fixed/schur64_FIX.c create mode 100644 lib/opus/silk/fixed/schur_FIX.c create mode 100644 lib/opus/silk/fixed/structs_FIX.h create mode 100644 lib/opus/silk/fixed/vector_ops_FIX.c create mode 100644 lib/opus/silk/fixed/warped_autocorrelation_FIX.c create mode 100644 lib/opus/silk/fixed/x86/burg_modified_FIX_sse4_1.c create mode 100644 lib/opus/silk/fixed/x86/vector_ops_FIX_sse4_1.c create mode 100644 lib/opus/silk/float/LPC_analysis_filter_FLP.c create mode 100644 lib/opus/silk/float/LPC_inv_pred_gain_FLP.c create mode 100644 lib/opus/silk/float/LTP_analysis_filter_FLP.c create mode 100644 lib/opus/silk/float/LTP_scale_ctrl_FLP.c create mode 100644 lib/opus/silk/float/SigProc_FLP.h create mode 100644 lib/opus/silk/float/apply_sine_window_FLP.c create mode 100644 lib/opus/silk/float/autocorrelation_FLP.c create mode 100644 lib/opus/silk/float/burg_modified_FLP.c create mode 100644 lib/opus/silk/float/bwexpander_FLP.c create mode 100644 lib/opus/silk/float/corrMatrix_FLP.c create mode 100644 lib/opus/silk/float/encode_frame_FLP.c create mode 100644 lib/opus/silk/float/energy_FLP.c create mode 100644 lib/opus/silk/float/find_LPC_FLP.c create mode 100644 lib/opus/silk/float/find_LTP_FLP.c create mode 100644 lib/opus/silk/float/find_pitch_lags_FLP.c create mode 100644 lib/opus/silk/float/find_pred_coefs_FLP.c create mode 100644 lib/opus/silk/float/inner_product_FLP.c create mode 100644 lib/opus/silk/float/k2a_FLP.c create mode 100644 lib/opus/silk/float/main_FLP.h create mode 100644 lib/opus/silk/float/noise_shape_analysis_FLP.c create mode 100644 lib/opus/silk/float/pitch_analysis_core_FLP.c create mode 100644 lib/opus/silk/float/process_gains_FLP.c create mode 100644 lib/opus/silk/float/regularize_correlations_FLP.c create mode 100644 lib/opus/silk/float/residual_energy_FLP.c create mode 100644 lib/opus/silk/float/scale_copy_vector_FLP.c create mode 100644 lib/opus/silk/float/scale_vector_FLP.c create mode 100644 lib/opus/silk/float/schur_FLP.c create mode 100644 lib/opus/silk/float/sort_FLP.c create mode 100644 lib/opus/silk/float/structs_FLP.h create mode 100644 lib/opus/silk/float/warped_autocorrelation_FLP.c create mode 100644 lib/opus/silk/float/wrappers_FLP.c create mode 100644 lib/opus/silk/gain_quant.c create mode 100644 lib/opus/silk/init_decoder.c create mode 100644 lib/opus/silk/init_encoder.c create mode 100644 lib/opus/silk/inner_prod_aligned.c create mode 100644 lib/opus/silk/interpolate.c create mode 100644 lib/opus/silk/lin2log.c create mode 100644 lib/opus/silk/log2lin.c create mode 100644 lib/opus/silk/macros.h create mode 100644 lib/opus/silk/main.h create mode 100644 lib/opus/silk/meson.build create mode 100644 lib/opus/silk/mips/NSQ_del_dec_mipsr1.h create mode 100644 lib/opus/silk/mips/macros_mipsr1.h create mode 100644 lib/opus/silk/mips/sigproc_fix_mipsr1.h create mode 100644 lib/opus/silk/pitch_est_defines.h create mode 100644 lib/opus/silk/pitch_est_tables.c create mode 100644 lib/opus/silk/process_NLSFs.c create mode 100644 lib/opus/silk/quant_LTP_gains.c create mode 100644 lib/opus/silk/resampler.c create mode 100644 lib/opus/silk/resampler_down2.c create mode 100644 lib/opus/silk/resampler_down2_3.c create mode 100644 lib/opus/silk/resampler_private.h create mode 100644 lib/opus/silk/resampler_private_AR2.c create mode 100644 lib/opus/silk/resampler_private_IIR_FIR.c create mode 100644 lib/opus/silk/resampler_private_down_FIR.c create mode 100644 lib/opus/silk/resampler_private_up2_HQ.c create mode 100644 lib/opus/silk/resampler_rom.c create mode 100644 lib/opus/silk/resampler_rom.h create mode 100644 lib/opus/silk/resampler_structs.h create mode 100644 lib/opus/silk/shell_coder.c create mode 100644 lib/opus/silk/sigm_Q15.c create mode 100644 lib/opus/silk/sort.c create mode 100644 lib/opus/silk/stereo_LR_to_MS.c create mode 100644 lib/opus/silk/stereo_MS_to_LR.c create mode 100644 lib/opus/silk/stereo_decode_pred.c create mode 100644 lib/opus/silk/stereo_encode_pred.c create mode 100644 lib/opus/silk/stereo_find_predictor.c create mode 100644 lib/opus/silk/stereo_quant_pred.c create mode 100644 lib/opus/silk/structs.h create mode 100644 lib/opus/silk/sum_sqr_shift.c create mode 100644 lib/opus/silk/table_LSF_cos.c create mode 100644 lib/opus/silk/tables.h create mode 100644 lib/opus/silk/tables_LTP.c create mode 100644 lib/opus/silk/tables_NLSF_CB_NB_MB.c create mode 100644 lib/opus/silk/tables_NLSF_CB_WB.c create mode 100644 lib/opus/silk/tables_gain.c create mode 100644 lib/opus/silk/tables_other.c create mode 100644 lib/opus/silk/tables_pitch_lag.c create mode 100644 lib/opus/silk/tables_pulses_per_block.c create mode 100644 lib/opus/silk/tests/meson.build create mode 100644 lib/opus/silk/tests/test_unit_LPC_inv_pred_gain.c create mode 100644 lib/opus/silk/tuning_parameters.h create mode 100644 lib/opus/silk/typedef.h create mode 100644 lib/opus/silk/x86/NSQ_del_dec_sse4_1.c create mode 100644 lib/opus/silk/x86/NSQ_sse4_1.c create mode 100644 lib/opus/silk/x86/SigProc_FIX_sse.h create mode 100644 lib/opus/silk/x86/VAD_sse4_1.c create mode 100644 lib/opus/silk/x86/VQ_WMat_EC_sse4_1.c create mode 100644 lib/opus/silk/x86/main_sse.h create mode 100644 lib/opus/silk/x86/x86_silk_map.c create mode 100644 lib/opus/silk_headers.mk create mode 100644 lib/opus/silk_sources.mk create mode 100644 lib/opus/src/analysis.c create mode 100644 lib/opus/src/analysis.h create mode 100644 lib/opus/src/mapping_matrix.c create mode 100644 lib/opus/src/mapping_matrix.h create mode 100644 lib/opus/src/meson.build create mode 100644 lib/opus/src/mlp.c create mode 100644 lib/opus/src/mlp.h create mode 100644 lib/opus/src/mlp_data.c create mode 100644 lib/opus/src/opus.c create mode 100644 lib/opus/src/opus_compare.c create mode 100644 lib/opus/src/opus_decoder.c create mode 100644 lib/opus/src/opus_demo.c create mode 100644 lib/opus/src/opus_encoder.c create mode 100644 lib/opus/src/opus_multistream.c create mode 100644 lib/opus/src/opus_multistream_decoder.c create mode 100644 lib/opus/src/opus_multistream_encoder.c create mode 100644 lib/opus/src/opus_private.h create mode 100644 lib/opus/src/opus_projection_decoder.c create mode 100644 lib/opus/src/opus_projection_encoder.c create mode 100644 lib/opus/src/repacketizer.c create mode 100644 lib/opus/src/repacketizer_demo.c create mode 100644 lib/opus/src/tansig_table.h create mode 100644 lib/opus/tests/meson.build create mode 100755 lib/opus/tests/opus_build_test.sh create mode 100644 lib/opus/tests/opus_decode_fuzzer.c create mode 100644 lib/opus/tests/opus_decode_fuzzer.options create mode 100644 lib/opus/tests/opus_encode_regressions.c create mode 100755 lib/opus/tests/random_config.sh create mode 100755 lib/opus/tests/run_vectors.sh create mode 100644 lib/opus/tests/test_opus_api.c create mode 100644 lib/opus/tests/test_opus_common.h create mode 100644 lib/opus/tests/test_opus_decode.c create mode 100644 lib/opus/tests/test_opus_encode.c create mode 100644 lib/opus/tests/test_opus_padding.c create mode 100644 lib/opus/tests/test_opus_projection.c create mode 100755 lib/opus/training/rnn_dump.py create mode 100755 lib/opus/training/rnn_train.py create mode 100755 lib/opus/training/txt2hdf5.py create mode 100644 lib/opus/win32/VS2015/common.props create mode 100644 lib/opus/win32/VS2015/opus.sln create mode 100644 lib/opus/win32/VS2015/opus.vcxproj create mode 100644 lib/opus/win32/VS2015/opus.vcxproj.filters create mode 100644 lib/opus/win32/VS2015/opus_demo.vcxproj create mode 100644 lib/opus/win32/VS2015/opus_demo.vcxproj.filters create mode 100644 lib/opus/win32/VS2015/test_opus_api.vcxproj create mode 100644 lib/opus/win32/VS2015/test_opus_api.vcxproj.filters create mode 100644 lib/opus/win32/VS2015/test_opus_decode.vcxproj create mode 100644 lib/opus/win32/VS2015/test_opus_decode.vcxproj.filters create mode 100644 lib/opus/win32/VS2015/test_opus_encode.vcxproj create mode 100644 lib/opus/win32/VS2015/test_opus_encode.vcxproj.filters create mode 100644 lib/opus/win32/config.h create mode 100644 lib/opus/win32/genversion.bat diff --git a/lib/opus/.appveyor.yml b/lib/opus/.appveyor.yml new file mode 100644 index 00000000..a0f4a776 --- /dev/null +++ b/lib/opus/.appveyor.yml @@ -0,0 +1,37 @@ +image: Visual Studio 2015 +configuration: +- Debug +- DebugDLL +- DebugDLL_fixed +- Release +- ReleaseDLL +- ReleaseDLL_fixed + +platform: +- Win32 +- x64 + +environment: + api_key: + secure: kR3Ac0NjGwFnTmXdFrR8d6VXjdk5F7L4F/BilC4nvaM= + +build: + project: win32\VS2015\opus.sln + parallel: true + verbosity: minimal + +after_build: +- cd %APPVEYOR_BUILD_FOLDER% +- 7z a opus.zip win32\VS2015\%PLATFORM%\%CONFIGURATION%\opus.??? include\*.h + +test_script: +- cd %APPVEYOR_BUILD_FOLDER%\win32\VS2015\%PLATFORM%\%CONFIGURATION% +- test_opus_api.exe +- test_opus_decode.exe +- test_opus_encode.exe + +artifacts: +- path: opus.zip + +on_success: +- ps: if ($env:api_key -and "$env:configuration/$env:platform" -eq "ReleaseDLL_fixed/x64") { Start-AppveyorBuild -ApiKey $env:api_key -ProjectSlug 'opus-tools' } diff --git a/lib/opus/.gitlab-ci.yml b/lib/opus/.gitlab-ci.yml new file mode 100644 index 00000000..d3303350 --- /dev/null +++ b/lib/opus/.gitlab-ci.yml @@ -0,0 +1,114 @@ +include: + - template: 'Workflows/Branch-Pipelines.gitlab-ci.yml' + +# https://docs.gitlab.com/ee/ci/yaml/workflow.html#switch-between-branch-pipelines-and-merge-request-pipelines +workflow: + rules: + - if: $CI_PIPELINE_SOURCE == "merge_request_event" + - if: $CI_COMMIT_BRANCH && $CI_OPEN_MERGE_REQUESTS && $CI_PIPELINE_SOURCE == "push" + when: never + - if: $CI_COMMIT_BRANCH + - if: $CI_COMMIT_TAG + +default: + tags: + - docker + image: 'debian:bookworm-slim' + +# https://docs.gitlab.com/ee/ci/yaml/yaml_optimization.html#reference-tags +.snippets: + git_prep: + # Make sure we have a recent annotated tag, otherwise meson/get-version.py + # might fail later (e.g. shallow clone without enough history) or return + # a bogus version based on a much older tag. This can happen in merge request + # pipelines from a personal fork, as the fork might not have the latest + # upstream tags if it has been forked a long time ago. Also affects the + # git version picked up by autotools and cmake, not just meson. + - git fetch https://gitlab.xiph.org/xiph/opus.git refs/tags/v1.4:refs/tags/v1.4 + - git describe + +whitespace: + stage: test + before_script: + - apt-get update && + apt-get install -y git + script: + - git diff-tree --check origin/master HEAD + +# Make sure commits are GPG signed +ci-fairy: + stage: test + script: + - apt update + - apt install -y python3-pip git + - pip3 install --break-system-packages git+https://gitlab.freedesktop.org/freedesktop/ci-templates@7811ba9814a3bad379377241c6c6b62d78b20eac + - echo Checking commits $CI_FAIRY_BASE_COMMIT..HEAD + - ci-fairy check-commits --gpg-signed-commit $CI_FAIRY_BASE_COMMIT..HEAD + tags: + - 'docker' + rules: + - if: $CI_PIPELINE_SOURCE == "merge_request_event" + variables: + CI_FAIRY_BASE_COMMIT: $CI_MERGE_REQUEST_DIFF_BASE_SHA + - if: $CI_PIPELINE_SOURCE != "merge_request_event" + variables: + CI_FAIRY_BASE_COMMIT: 'HEAD^1' + +autoconf: + stage: build + before_script: + - apt-get update && + apt-get install -y zip doxygen git automake libtool make + - !reference [.snippets, git_prep] + script: + - ./autogen.sh + - ./configure + - make -j4 + - make distcheck + cache: + paths: + - "src/*.o" + - "src/.libs/*.o" + - "silk/*.o" + - "silk/.libs/*.o" + - "celt/*.o" + - "celt/.libs/*.o" + +cmake: + stage: build + before_script: + - apt-get update && + apt-get install -y cmake ninja-build git + - !reference [.snippets, git_prep] + script: + - mkdir build + - cmake -S . -B build -G "Ninja" -DCMAKE_BUILD_TYPE=Release -DOPUS_BUILD_TESTING=ON -DOPUS_BUILD_PROGRAMS=ON + - cmake --build build + - cd build && ctest --output-on-failure + +.meson: + image: 'debian:bookworm-slim' + stage: build + before_script: + - apt-get update && + apt-get install -y ninja-build doxygen meson git + - !reference [.snippets, git_prep] + script: + - mkdir builddir + - meson setup -Dtests=enabled -Ddocs=enabled -Dbuildtype=release builddir ${MESON_EXTRA_ARGS} + - meson compile -C builddir + - meson test -C builddir + #- meson dist --no-tests -C builddir + +meson x86_64: + extends: '.meson' + variables: + MESON_EXTRA_ARGS: '--werror' + +meson arm64: + extends: '.meson' + tags: + - 'gstreamer-arm64-linux-docker' + variables: + # arm64 build has a compiler warning still, so let's not use --werror for now + MESON_EXTRA_ARGS: '-Dwerror=false' diff --git a/lib/opus/.travis.yml b/lib/opus/.travis.yml new file mode 100644 index 00000000..821c813e --- /dev/null +++ b/lib/opus/.travis.yml @@ -0,0 +1,21 @@ +language: c + +compiler: + - gcc + - clang + +os: + - linux + - osx + +env: + - CONFIG="" + - CONFIG="--enable-assertions" + - CONFIG="--enable-fixed-point" + - CONFIG="--enable-fixed-point --disable-float-api" + - CONFIG="--enable-fixed-point --enable-assertions" + +script: + - ./autogen.sh + - ./configure $CONFIG + - make distcheck diff --git a/lib/opus/AUTHORS b/lib/opus/AUTHORS new file mode 100644 index 00000000..b3d22a20 --- /dev/null +++ b/lib/opus/AUTHORS @@ -0,0 +1,6 @@ +Jean-Marc Valin (jmvalin@jmvalin.ca) +Koen Vos (koenvos74@gmail.com) +Timothy Terriberry (tterribe@xiph.org) +Karsten Vandborg Sorensen (karsten.vandborg.sorensen@skype.net) +Soren Skak Jensen (ssjensen@gn.com) +Gregory Maxwell (greg@xiph.org) diff --git a/lib/opus/CMakeLists.txt b/lib/opus/CMakeLists.txt new file mode 100644 index 00000000..9d824cdc --- /dev/null +++ b/lib/opus/CMakeLists.txt @@ -0,0 +1,646 @@ +cmake_minimum_required(VERSION 3.1) +list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake") + +include(OpusPackageVersion) +get_package_version(PACKAGE_VERSION PROJECT_VERSION) + +project(Opus LANGUAGES C VERSION ${PROJECT_VERSION}) + +include(OpusFunctions) +include(OpusBuildtype) +include(OpusConfig) +include(OpusSources) +include(GNUInstallDirs) +include(CMakeDependentOption) +include(FeatureSummary) + +set(OPUS_BUILD_SHARED_LIBRARY_HELP_STR "build shared library.") +option(OPUS_BUILD_SHARED_LIBRARY ${OPUS_BUILD_SHARED_LIBRARY_HELP_STR} OFF) +if(OPUS_BUILD_SHARED_LIBRARY OR BUILD_SHARED_LIBS OR OPUS_BUILD_FRAMEWORK) + # Global flag to cause add_library() to create shared libraries if on. + set(BUILD_SHARED_LIBS ON) + set(OPUS_BUILD_SHARED_LIBRARY ON) +endif() +add_feature_info(OPUS_BUILD_SHARED_LIBRARY OPUS_BUILD_SHARED_LIBRARY ${OPUS_BUILD_SHARED_LIBRARY_HELP_STR}) + +set(OPUS_BUILD_TESTING_HELP_STR "build tests.") +option(OPUS_BUILD_TESTING ${OPUS_BUILD_TESTING_HELP_STR} OFF) +if(OPUS_BUILD_TESTING OR BUILD_TESTING) + set(OPUS_BUILD_TESTING ON) + set(BUILD_TESTING ON) +endif() +add_feature_info(OPUS_BUILD_TESTING OPUS_BUILD_TESTING ${OPUS_BUILD_TESTING_HELP_STR}) + +set(OPUS_CUSTOM_MODES_HELP_STR "enable non-Opus modes, e.g. 44.1 kHz & 2^n frames.") +option(OPUS_CUSTOM_MODES ${OPUS_CUSTOM_MODES_HELP_STR} OFF) +add_feature_info(OPUS_CUSTOM_MODES OPUS_CUSTOM_MODES ${OPUS_CUSTOM_MODES_HELP_STR}) + +set(OPUS_BUILD_PROGRAMS_HELP_STR "build programs.") +option(OPUS_BUILD_PROGRAMS ${OPUS_BUILD_PROGRAMS_HELP_STR} OFF) +add_feature_info(OPUS_BUILD_PROGRAMS OPUS_BUILD_PROGRAMS ${OPUS_BUILD_PROGRAMS_HELP_STR}) + +set(OPUS_DISABLE_INTRINSICS_HELP_STR "disable all intrinsics optimizations.") +option(OPUS_DISABLE_INTRINSICS ${OPUS_DISABLE_INTRINSICS_HELP_STR} OFF) +add_feature_info(OPUS_DISABLE_INTRINSICS OPUS_DISABLE_INTRINSICS ${OPUS_DISABLE_INTRINSICS_HELP_STR}) + +set(OPUS_FIXED_POINT_HELP_STR "compile as fixed-point (for machines without a fast enough FPU).") +option(OPUS_FIXED_POINT ${OPUS_FIXED_POINT_HELP_STR} OFF) +add_feature_info(OPUS_FIXED_POINT OPUS_FIXED_POINT ${OPUS_FIXED_POINT_HELP_STR}) + +set(OPUS_ENABLE_FLOAT_API_HELP_STR "compile with the floating point API (for machines with float library).") +option(OPUS_ENABLE_FLOAT_API ${OPUS_ENABLE_FLOAT_API_HELP_STR} ON) +add_feature_info(OPUS_ENABLE_FLOAT_API OPUS_ENABLE_FLOAT_API ${OPUS_ENABLE_FLOAT_API_HELP_STR}) + +set(OPUS_FLOAT_APPROX_HELP_STR "enable floating point approximations (Ensure your platform supports IEEE 754 before enabling).") +option(OPUS_FLOAT_APPROX ${OPUS_FLOAT_APPROX_HELP_STR} OFF) +add_feature_info(OPUS_FLOAT_APPROX OPUS_FLOAT_APPROX ${OPUS_FLOAT_APPROX_HELP_STR}) + +set(OPUS_ASSERTIONS_HELP_STR "additional software error checking.") +option(OPUS_ASSERTIONS ${OPUS_ASSERTIONS_HELP_STR} OFF) +add_feature_info(OPUS_ASSERTIONS OPUS_ASSERTIONS ${OPUS_ASSERTIONS_HELP_STR}) + +set(OPUS_HARDENING_HELP_STR "run-time checks that are cheap and safe for use in production.") +option(OPUS_HARDENING ${OPUS_HARDENING_HELP_STR} ON) +add_feature_info(OPUS_HARDENING OPUS_HARDENING ${OPUS_HARDENING_HELP_STR}) + +set(OPUS_FUZZING_HELP_STR "causes the encoder to make random decisions (do not use in production).") +option(OPUS_FUZZING ${OPUS_FUZZING_HELP_STR} OFF) +add_feature_info(OPUS_FUZZING OPUS_FUZZING ${OPUS_FUZZING_HELP_STR}) + +set(OPUS_CHECK_ASM_HELP_STR "enable bit-exactness checks between optimized and c implementations.") +option(OPUS_CHECK_ASM ${OPUS_CHECK_ASM_HELP_STR} OFF) +add_feature_info(OPUS_CHECK_ASM OPUS_CHECK_ASM ${OPUS_CHECK_ASM_HELP_STR}) + +set(OPUS_INSTALL_PKG_CONFIG_MODULE_HELP_STR "install pkg-config module.") +option(OPUS_INSTALL_PKG_CONFIG_MODULE ${OPUS_INSTALL_PKG_CONFIG_MODULE_HELP_STR} ON) +add_feature_info(OPUS_INSTALL_PKG_CONFIG_MODULE OPUS_INSTALL_PKG_CONFIG_MODULE ${OPUS_INSTALL_PKG_CONFIG_MODULE_HELP_STR}) + +set(OPUS_INSTALL_CMAKE_CONFIG_MODULE_HELP_STR "install CMake package config module.") +option(OPUS_INSTALL_CMAKE_CONFIG_MODULE ${OPUS_INSTALL_CMAKE_CONFIG_MODULE_HELP_STR} ON) +add_feature_info(OPUS_INSTALL_CMAKE_CONFIG_MODULE OPUS_INSTALL_CMAKE_CONFIG_MODULE ${OPUS_INSTALL_CMAKE_CONFIG_MODULE_HELP_STR}) + +if(APPLE) + set(OPUS_BUILD_FRAMEWORK_HELP_STR "build Framework bundle for Apple systems.") + option(OPUS_BUILD_FRAMEWORK ${OPUS_BUILD_FRAMEWORK_HELP_STR} OFF) + add_feature_info(OPUS_BUILD_FRAMEWORK OPUS_BUILD_FRAMEWORK ${OPUS_BUILD_FRAMEWORK_HELP_STR}) +endif() + +set(OPUS_FIXED_POINT_DEBUG_HELP_STR "debug fixed-point implementation.") +cmake_dependent_option(OPUS_FIXED_POINT_DEBUG + ${OPUS_FIXED_POINT_DEBUG_HELP_STR} + ON + "OPUS_FIXED_POINT; OPUS_FIXED_POINT_DEBUG" + OFF) +add_feature_info(OPUS_FIXED_POINT_DEBUG OPUS_FIXED_POINT_DEBUG ${OPUS_FIXED_POINT_DEBUG_HELP_STR}) + +set(OPUS_VAR_ARRAYS_HELP_STR "use variable length arrays for stack arrays.") +cmake_dependent_option(OPUS_VAR_ARRAYS + ${OPUS_VAR_ARRAYS_HELP_STR} + ON + "VLA_SUPPORTED; NOT OPUS_USE_ALLOCA; NOT OPUS_NONTHREADSAFE_PSEUDOSTACK" + OFF) +add_feature_info(OPUS_VAR_ARRAYS OPUS_VAR_ARRAYS ${OPUS_VAR_ARRAYS_HELP_STR}) + +set(OPUS_USE_ALLOCA_HELP_STR "use alloca for stack arrays (on non-C99 compilers).") +cmake_dependent_option(OPUS_USE_ALLOCA + ${OPUS_USE_ALLOCA_HELP_STR} + ON + "USE_ALLOCA_SUPPORTED; NOT OPUS_VAR_ARRAYS; NOT OPUS_NONTHREADSAFE_PSEUDOSTACK" + OFF) +add_feature_info(OPUS_USE_ALLOCA OPUS_USE_ALLOCA ${OPUS_USE_ALLOCA_HELP_STR}) + +set(OPUS_NONTHREADSAFE_PSEUDOSTACK_HELP_STR "use a non threadsafe pseudostack when neither variable length arrays or alloca is supported.") +cmake_dependent_option(OPUS_NONTHREADSAFE_PSEUDOSTACK + ${OPUS_NONTHREADSAFE_PSEUDOSTACK_HELP_STR} + ON + "NOT OPUS_VAR_ARRAYS; NOT OPUS_USE_ALLOCA" + OFF) +add_feature_info(OPUS_NONTHREADSAFE_PSEUDOSTACK OPUS_NONTHREADSAFE_PSEUDOSTACK ${OPUS_NONTHREADSAFE_PSEUDOSTACK_HELP_STR}) + +set(OPUS_FAST_MATH_HELP_STR "enable fast math (unsupported and discouraged use, as code is not well tested with this build option).") +cmake_dependent_option(OPUS_FAST_MATH + ${OPUS_FAST_MATH_HELP_STR} + ON + "OPUS_FLOAT_APPROX; OPUS_FAST_MATH; FAST_MATH_SUPPORTED" + OFF) +add_feature_info(OPUS_FAST_MATH OPUS_FAST_MATH ${OPUS_FAST_MATH_HELP_STR}) + +set(OPUS_STACK_PROTECTOR_HELP_STR "use stack protection.") +cmake_dependent_option(OPUS_STACK_PROTECTOR + ${OPUS_STACK_PROTECTOR_HELP_STR} + ON + "STACK_PROTECTOR_SUPPORTED" + OFF) +add_feature_info(OPUS_STACK_PROTECTOR OPUS_STACK_PROTECTOR ${OPUS_STACK_PROTECTOR_HELP_STR}) + +if(NOT MSVC) + set(OPUS_FORTIFY_SOURCE_HELP_STR "add protection against buffer overflows.") + cmake_dependent_option(OPUS_FORTIFY_SOURCE + ${OPUS_FORTIFY_SOURCE_HELP_STR} + ON + "FORTIFY_SOURCE_SUPPORTED" + OFF) + add_feature_info(OPUS_FORTIFY_SOURCE OPUS_FORTIFY_SOURCE ${OPUS_FORTIFY_SOURCE_HELP_STR}) +endif() + +if(MINGW AND (OPUS_FORTIFY_SOURCE OR OPUS_STACK_PROTECTOR)) + # ssp lib is needed for security features for MINGW + list(APPEND OPUS_REQUIRED_LIBRARIES ssp) +endif() + +if(OPUS_CPU_X86 OR OPUS_CPU_X64) + set(OPUS_X86_MAY_HAVE_SSE_HELP_STR "does runtime check for SSE1 support.") + cmake_dependent_option(OPUS_X86_MAY_HAVE_SSE + ${OPUS_X86_MAY_HAVE_SSE_HELP_STR} + ON + "SSE1_SUPPORTED; NOT OPUS_DISABLE_INTRINSICS" + OFF) + add_feature_info(OPUS_X86_MAY_HAVE_SSE OPUS_X86_MAY_HAVE_SSE ${OPUS_X86_MAY_HAVE_SSE_HELP_STR}) + + set(OPUS_X86_MAY_HAVE_SSE2_HELP_STR "does runtime check for SSE2 support.") + cmake_dependent_option(OPUS_X86_MAY_HAVE_SSE2 + ${OPUS_X86_MAY_HAVE_SSE2_HELP_STR} + ON + "SSE2_SUPPORTED; NOT OPUS_DISABLE_INTRINSICS" + OFF) + add_feature_info(OPUS_X86_MAY_HAVE_SSE2 OPUS_X86_MAY_HAVE_SSE2 ${OPUS_X86_MAY_HAVE_SSE2_HELP_STR}) + + set(OPUS_X86_MAY_HAVE_SSE4_1_HELP_STR "does runtime check for SSE4.1 support.") + cmake_dependent_option(OPUS_X86_MAY_HAVE_SSE4_1 + ${OPUS_X86_MAY_HAVE_SSE4_1_HELP_STR} + ON + "SSE4_1_SUPPORTED; NOT OPUS_DISABLE_INTRINSICS" + OFF) + add_feature_info(OPUS_X86_MAY_HAVE_SSE4_1 OPUS_X86_MAY_HAVE_SSE4_1 ${OPUS_X86_MAY_HAVE_SSE4_1_HELP_STR}) + + set(OPUS_X86_MAY_HAVE_AVX_HELP_STR "does runtime check for AVX support.") + cmake_dependent_option(OPUS_X86_MAY_HAVE_AVX + ${OPUS_X86_MAY_HAVE_AVX_HELP_STR} + ON + "AVX_SUPPORTED; NOT OPUS_DISABLE_INTRINSICS" + OFF) + add_feature_info(OPUS_X86_MAY_HAVE_AVX OPUS_X86_MAY_HAVE_AVX ${OPUS_X86_MAY_HAVE_AVX_HELP_STR}) + + # PRESUME depends on MAY HAVE, but PRESUME will override runtime detection + set(OPUS_X86_PRESUME_SSE_HELP_STR "assume target CPU has SSE1 support (override runtime check).") + set(OPUS_X86_PRESUME_SSE2_HELP_STR "assume target CPU has SSE2 support (override runtime check).") + if(OPUS_CPU_X64) # Assume x86_64 has up to SSE2 support + cmake_dependent_option(OPUS_X86_PRESUME_SSE + ${OPUS_X86_PRESUME_SSE_HELP_STR} + ON + "OPUS_X86_MAY_HAVE_SSE; NOT OPUS_DISABLE_INTRINSICS" + OFF) + + cmake_dependent_option(OPUS_X86_PRESUME_SSE2 + ${OPUS_X86_PRESUME_SSE2_HELP_STR} + ON + "OPUS_X86_MAY_HAVE_SSE2; NOT OPUS_DISABLE_INTRINSICS" + OFF) + else() + cmake_dependent_option(OPUS_X86_PRESUME_SSE + ${OPUS_X86_PRESUME_SSE_HELP_STR} + OFF + "OPUS_X86_MAY_HAVE_SSE; NOT OPUS_DISABLE_INTRINSICS" + OFF) + + cmake_dependent_option(OPUS_X86_PRESUME_SSE2 + ${OPUS_X86_PRESUME_SSE2_HELP_STR} + OFF + "OPUS_X86_MAY_HAVE_SSE2; NOT OPUS_DISABLE_INTRINSICS" + OFF) + endif() + add_feature_info(OPUS_X86_PRESUME_SSE OPUS_X86_PRESUME_SSE ${OPUS_X86_PRESUME_SSE_HELP_STR}) + add_feature_info(OPUS_X86_PRESUME_SSE2 OPUS_X86_PRESUME_SSE2 ${OPUS_X86_PRESUME_SSE2_HELP_STR}) + + set(OPUS_X86_PRESUME_SSE4_1_HELP_STR "assume target CPU has SSE4.1 support (override runtime check).") + cmake_dependent_option(OPUS_X86_PRESUME_SSE4_1 + ${OPUS_X86_PRESUME_SSE4_1_HELP_STR} + OFF + "OPUS_X86_MAY_HAVE_SSE4_1; NOT OPUS_DISABLE_INTRINSICS" + OFF) + add_feature_info(OPUS_X86_PRESUME_SSE4_1 OPUS_X86_PRESUME_SSE4_1 ${OPUS_X86_PRESUME_SSE4_1_HELP_STR}) + + set(OPUS_X86_PRESUME_AVX_HELP_STR "assume target CPU has AVX support (override runtime check).") + cmake_dependent_option(OPUS_X86_PRESUME_AVX + ${OPUS_X86_PRESUME_AVX_HELP_STR} + OFF + "OPUS_X86_MAY_HAVE_AVX; NOT OPUS_DISABLE_INTRINSICS" + OFF) + add_feature_info(OPUS_X86_PRESUME_AVX OPUS_X86_PRESUME_AVX ${OPUS_X86_PRESUME_AVX_HELP_STR}) +endif() + +feature_summary(WHAT ALL) + +set_package_properties(Git + PROPERTIES + TYPE + REQUIRED + DESCRIPTION + "fast, scalable, distributed revision control system" + URL + "https://git-scm.com/" + PURPOSE + "required to set up package version") + +set(Opus_PUBLIC_HEADER + ${CMAKE_CURRENT_SOURCE_DIR}/include/opus.h + ${CMAKE_CURRENT_SOURCE_DIR}/include/opus_defines.h + ${CMAKE_CURRENT_SOURCE_DIR}/include/opus_multistream.h + ${CMAKE_CURRENT_SOURCE_DIR}/include/opus_projection.h + ${CMAKE_CURRENT_SOURCE_DIR}/include/opus_types.h) + +if(OPUS_CUSTOM_MODES) + list(APPEND Opus_PUBLIC_HEADER ${CMAKE_CURRENT_SOURCE_DIR}/include/opus_custom.h) +endif() + +add_library(opus ${opus_headers} ${opus_sources} ${opus_sources_float} ${Opus_PUBLIC_HEADER}) +add_library(Opus::opus ALIAS opus) + +get_library_version(OPUS_LIBRARY_VERSION OPUS_LIBRARY_VERSION_MAJOR) +message(DEBUG "Opus library version: ${OPUS_LIBRARY_VERSION}") + +set_target_properties(opus + PROPERTIES SOVERSION + ${OPUS_LIBRARY_VERSION_MAJOR} + VERSION + ${OPUS_LIBRARY_VERSION} + PUBLIC_HEADER + "${Opus_PUBLIC_HEADER}") + +target_include_directories( + opus + PUBLIC $ + $ + $ + PRIVATE ${CMAKE_CURRENT_BINARY_DIR} + ${CMAKE_CURRENT_SOURCE_DIR} + celt + silk) + +target_link_libraries(opus PRIVATE ${OPUS_REQUIRED_LIBRARIES}) +target_compile_definitions(opus PRIVATE OPUS_BUILD) + +if(OPUS_FIXED_POINT_DEBUG) + target_compile_definitions(opus PRIVATE FIXED_DEBUG) +endif() + +if(OPUS_FORTIFY_SOURCE AND NOT MSVC) + target_compile_definitions(opus PRIVATE + $<$>:_FORTIFY_SOURCE=2>) +endif() + +if(OPUS_FLOAT_APPROX) + target_compile_definitions(opus PRIVATE FLOAT_APPROX) +endif() + +if(OPUS_ASSERTIONS) + target_compile_definitions(opus PRIVATE ENABLE_ASSERTIONS) +endif() + +if(OPUS_HARDENING) + target_compile_definitions(opus PRIVATE ENABLE_HARDENING) +endif() + +if(OPUS_FUZZING) + target_compile_definitions(opus PRIVATE FUZZING) +endif() + +if(OPUS_CHECK_ASM) + target_compile_definitions(opus PRIVATE OPUS_CHECK_ASM) +endif() + +if(OPUS_VAR_ARRAYS) + target_compile_definitions(opus PRIVATE VAR_ARRAYS) +elseif(OPUS_USE_ALLOCA) + target_compile_definitions(opus PRIVATE USE_ALLOCA) +elseif(OPUS_NONTHREADSAFE_PSEUDOSTACK) + target_compile_definitions(opus PRIVATE NONTHREADSAFE_PSEUDOSTACK) +else() + message(ERROR "Need to set a define for stack allocation") +endif() + +if(OPUS_CUSTOM_MODES) + target_compile_definitions(opus PRIVATE CUSTOM_MODES) +endif() + +if(OPUS_FAST_MATH) + if(MSVC) + target_compile_options(opus PRIVATE /fp:fast) + else() + target_compile_options(opus PRIVATE -ffast-math) + endif() +endif() + +if(OPUS_STACK_PROTECTOR) + if(MSVC) + target_compile_options(opus PRIVATE /GS) + else() + target_compile_options(opus PRIVATE -fstack-protector-strong) + endif() +elseif(STACK_PROTECTOR_DISABLED_SUPPORTED) + target_compile_options(opus PRIVATE /GS-) +endif() + +if(BUILD_SHARED_LIBS) + if(WIN32) + target_compile_definitions(opus PRIVATE DLL_EXPORT) + elseif(HIDDEN_VISIBILITY_SUPPORTED) + set_target_properties(opus PROPERTIES C_VISIBILITY_PRESET hidden) + endif() +endif() + +add_sources_group(opus silk ${silk_headers} ${silk_sources}) +add_sources_group(opus celt ${celt_headers} ${celt_sources}) + +if(OPUS_FIXED_POINT) + add_sources_group(opus silk ${silk_sources_fixed}) + target_include_directories(opus PRIVATE silk/fixed) + target_compile_definitions(opus PRIVATE FIXED_POINT=1) +else() + add_sources_group(opus silk ${silk_sources_float}) + target_include_directories(opus PRIVATE silk/float) +endif() + +if(NOT OPUS_ENABLE_FLOAT_API) + target_compile_definitions(opus PRIVATE DISABLE_FLOAT_API) +endif() + +if(NOT OPUS_DISABLE_INTRINSICS) + if(((OPUS_X86_MAY_HAVE_SSE AND NOT OPUS_X86_PRESUME_SSE) OR + (OPUS_X86_MAY_HAVE_SSE2 AND NOT OPUS_X86_PRESUME_SSE2) OR + (OPUS_X86_MAY_HAVE_SSE4_1 AND NOT OPUS_X86_PRESUME_SSE4_1) OR + (OPUS_X86_MAY_HAVE_AVX AND NOT OPUS_X86_PRESUME_AVX)) AND + RUNTIME_CPU_CAPABILITY_DETECTION) + target_compile_definitions(opus PRIVATE OPUS_HAVE_RTCD) + if(NOT MSVC) + if(CPU_INFO_BY_ASM_SUPPORTED) + target_compile_definitions(opus PRIVATE CPU_INFO_BY_ASM) + elseif(CPU_INFO_BY_C_SUPPORTED) + target_compile_definitions(opus PRIVATE CPU_INFO_BY_C) + else() + message(ERROR "Runtime cpu capability detection is enabled while CPU_INFO is not supported") + endif() + endif() + add_sources_group(opus celt ${celt_sources_x86_rtcd}) + add_sources_group(opus silk ${silk_sources_x86_rtcd}) + endif() + + if(SSE1_SUPPORTED) + if(OPUS_X86_MAY_HAVE_SSE) + add_sources_group(opus celt ${celt_sources_sse}) + target_compile_definitions(opus PRIVATE OPUS_X86_MAY_HAVE_SSE) + if(NOT MSVC) + set_source_files_properties(${celt_sources_sse} PROPERTIES COMPILE_FLAGS -msse) + endif() + endif() + if(OPUS_X86_PRESUME_SSE) + target_compile_definitions(opus PRIVATE OPUS_X86_PRESUME_SSE) + if(NOT MSVC) + target_compile_options(opus PRIVATE -msse) + endif() + endif() + endif() + + if(SSE2_SUPPORTED) + if(OPUS_X86_MAY_HAVE_SSE2) + add_sources_group(opus celt ${celt_sources_sse2}) + target_compile_definitions(opus PRIVATE OPUS_X86_MAY_HAVE_SSE2) + if(NOT MSVC) + set_source_files_properties(${celt_sources_sse2} PROPERTIES COMPILE_FLAGS -msse2) + endif() + endif() + if(OPUS_X86_PRESUME_SSE2) + target_compile_definitions(opus PRIVATE OPUS_X86_PRESUME_SSE2) + if(NOT MSVC) + target_compile_options(opus PRIVATE -msse2) + endif() + endif() + endif() + + if(SSE4_1_SUPPORTED) + if(OPUS_X86_MAY_HAVE_SSE4_1) + add_sources_group(opus celt ${celt_sources_sse4_1}) + add_sources_group(opus silk ${silk_sources_sse4_1}) + target_compile_definitions(opus PRIVATE OPUS_X86_MAY_HAVE_SSE4_1) + if(NOT MSVC) + set_source_files_properties(${celt_sources_sse4_1} ${silk_sources_sse4_1} PROPERTIES COMPILE_FLAGS -msse4.1) + endif() + + if(OPUS_FIXED_POINT) + add_sources_group(opus silk ${silk_sources_fixed_sse4_1}) + if(NOT MSVC) + set_source_files_properties(${silk_sources_fixed_sse4_1} PROPERTIES COMPILE_FLAGS -msse4.1) + endif() + endif() + endif() + if(OPUS_X86_PRESUME_SSE4_1) + target_compile_definitions(opus PRIVATE OPUS_X86_PRESUME_SSE4_1) + if(NOT MSVC) + target_compile_options(opus PRIVATE -msse4.1) + endif() + endif() + endif() + + if(AVX_SUPPORTED) + # mostly placeholder in case of avx intrinsics is added + if(OPUS_X86_MAY_HAVE_AVX) + target_compile_definitions(opus PRIVATE OPUS_X86_MAY_HAVE_AVX) + endif() + if(OPUS_X86_PRESUME_AVX) + target_compile_definitions(opus PRIVATE OPUS_X86_PRESUME_AVX) + if(NOT MSVC) + target_compile_options(opus PRIVATE -mavx) + endif() + endif() + endif() + + if(MSVC) + if(AVX_SUPPORTED AND OPUS_X86_PRESUME_AVX) # on 64 bit and 32 bits + add_definitions(/arch:AVX) + elseif(OPUS_CPU_X86) # if AVX not supported then set SSE flag + if((SSE4_1_SUPPORTED AND OPUS_X86_PRESUME_SSE4_1) + OR (SSE2_SUPPORTED AND OPUS_X86_PRESUME_SSE2)) + target_compile_definitions(opus PRIVATE /arch:SSE2) + elseif(SSE1_SUPPORTED AND OPUS_X86_PRESUME_SSE) + target_compile_definitions(opus PRIVATE /arch:SSE) + endif() + endif() + endif() + + if(COMPILER_SUPPORT_NEON) + if(OPUS_MAY_HAVE_NEON) + if(RUNTIME_CPU_CAPABILITY_DETECTION) + message(STATUS "OPUS_MAY_HAVE_NEON enabling runtime detection") + target_compile_definitions(opus PRIVATE OPUS_HAVE_RTCD) + add_sources_group(opus celt ${celt_sources_arm_rtcd}) + add_sources_group(opus silk ${silk_sources_arm_rtcd}) + else() + message(ERROR "Runtime cpu capability detection needed for MAY_HAVE_NEON") + endif() + # Do runtime check for NEON + target_compile_definitions(opus + PRIVATE + OPUS_ARM_MAY_HAVE_NEON + OPUS_ARM_MAY_HAVE_NEON_INTR) + endif() + + add_sources_group(opus celt ${celt_sources_arm_neon_intr}) + add_sources_group(opus silk ${silk_sources_arm_neon_intr}) + + # silk arm neon depends on main_Fix.h + target_include_directories(opus PRIVATE silk/fixed) + + if(OPUS_FIXED_POINT) + add_sources_group(opus silk ${silk_sources_fixed_arm_neon_intr}) + endif() + + if(OPUS_PRESUME_NEON) + target_compile_definitions(opus + PRIVATE + OPUS_ARM_PRESUME_NEON + OPUS_ARM_PRESUME_NEON_INTR) + endif() + endif() +endif() + +target_compile_definitions(opus + PRIVATE + $<$:HAVE_LRINT> + $<$:HAVE_LRINTF>) + +if(OPUS_BUILD_FRAMEWORK) + set_target_properties(opus PROPERTIES + FRAMEWORK TRUE + FRAMEWORK_VERSION ${PROJECT_VERSION} + MACOSX_FRAMEWORK_IDENTIFIER org.xiph.opus + MACOSX_FRAMEWORK_SHORT_VERSION_STRING ${PROJECT_VERSION} + MACOSX_FRAMEWORK_BUNDLE_VERSION ${PROJECT_VERSION} + XCODE_ATTRIBUTE_INSTALL_PATH "@rpath" + OUTPUT_NAME Opus) +endif() + +install(TARGETS opus + EXPORT OpusTargets + ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR} + LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR} + RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR} + FRAMEWORK DESTINATION ${CMAKE_INSTALL_PREFIX} + PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/opus) + +if(OPUS_INSTALL_PKG_CONFIG_MODULE) + set(prefix ${CMAKE_INSTALL_PREFIX}) + set(exec_prefix ${CMAKE_INSTALL_PREFIX}) + set(libdir ${CMAKE_INSTALL_FULL_LIBDIR}) + set(includedir ${CMAKE_INSTALL_FULL_INCLUDEDIR}) + set(VERSION ${PACKAGE_VERSION}) + if(HAVE_LIBM) + set(LIBM "-lm") + endif() + configure_file(opus.pc.in opus.pc) + install(FILES ${CMAKE_CURRENT_BINARY_DIR}/opus.pc + DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig) +endif() + +if(OPUS_INSTALL_CMAKE_CONFIG_MODULE) + set(CPACK_GENERATOR TGZ) + include(CPack) + set(CMAKE_INSTALL_PACKAGEDIR ${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}) + install(EXPORT OpusTargets + NAMESPACE Opus:: + DESTINATION ${CMAKE_INSTALL_PACKAGEDIR}) + + include(CMakePackageConfigHelpers) + + set(INCLUDE_INSTALL_DIR ${CMAKE_INSTALL_INCLUDEDIR}) + configure_package_config_file(${PROJECT_SOURCE_DIR}/cmake/OpusConfig.cmake.in + OpusConfig.cmake + INSTALL_DESTINATION + ${CMAKE_INSTALL_PACKAGEDIR} + PATH_VARS + INCLUDE_INSTALL_DIR + INSTALL_PREFIX + ${CMAKE_INSTALL_PREFIX}) + write_basic_package_version_file(OpusConfigVersion.cmake + VERSION ${PROJECT_VERSION} + COMPATIBILITY SameMajorVersion) + install(FILES ${CMAKE_CURRENT_BINARY_DIR}/OpusConfig.cmake + ${CMAKE_CURRENT_BINARY_DIR}/OpusConfigVersion.cmake + DESTINATION ${CMAKE_INSTALL_PACKAGEDIR}) +endif() + +if(OPUS_BUILD_PROGRAMS) + # demo + if(OPUS_CUSTOM_MODES) + add_executable(opus_custom_demo ${opus_custom_demo_sources}) + target_include_directories(opus_custom_demo + PRIVATE ${CMAKE_CURRENT_BINARY_DIR}) + target_link_libraries(opus_custom_demo PRIVATE opus) + target_compile_definitions(opus_custom_demo PRIVATE OPUS_BUILD) + endif() + + add_executable(opus_demo ${opus_demo_sources}) + target_include_directories(opus_demo PRIVATE ${CMAKE_CURRENT_BINARY_DIR}) + target_include_directories(opus_demo PRIVATE silk) # debug.h + target_include_directories(opus_demo PRIVATE celt) # arch.h + target_link_libraries(opus_demo PRIVATE opus ${OPUS_REQUIRED_LIBRARIES}) + target_compile_definitions(opus_demo PRIVATE OPUS_BUILD) + + # compare + add_executable(opus_compare ${opus_compare_sources}) + target_include_directories(opus_compare PRIVATE ${CMAKE_CURRENT_BINARY_DIR}) + target_link_libraries(opus_compare PRIVATE opus ${OPUS_REQUIRED_LIBRARIES}) + if(MSVC) + # move cosmetic warning to level 4 for opus_compare + target_compile_options(opus_compare PRIVATE /w44244) + endif() +endif() + +if(BUILD_TESTING AND NOT BUILD_SHARED_LIBS) + enable_testing() + + # tests + add_executable(test_opus_decode ${test_opus_decode_sources}) + target_include_directories(test_opus_decode + PRIVATE ${CMAKE_CURRENT_BINARY_DIR}) + target_link_libraries(test_opus_decode PRIVATE opus) + target_compile_definitions(test_opus_decode PRIVATE OPUS_BUILD) + if(OPUS_FIXED_POINT) + target_compile_definitions(test_opus_decode PRIVATE DISABLE_FLOAT_API) + endif() + add_test(NAME test_opus_decode COMMAND ${CMAKE_COMMAND} + -DTEST_EXECUTABLE=$ + -DCMAKE_SYSTEM_NAME=${CMAKE_SYSTEM_NAME} + -P "${PROJECT_SOURCE_DIR}/cmake/RunTest.cmake") + + add_executable(test_opus_padding ${test_opus_padding_sources}) + target_include_directories(test_opus_padding + PRIVATE ${CMAKE_CURRENT_BINARY_DIR}) + target_link_libraries(test_opus_padding PRIVATE opus) + add_test(NAME test_opus_padding COMMAND ${CMAKE_COMMAND} + -DTEST_EXECUTABLE=$ + -DCMAKE_SYSTEM_NAME=${CMAKE_SYSTEM_NAME} + -P "${PROJECT_SOURCE_DIR}/cmake/RunTest.cmake") + + add_executable(test_opus_api ${test_opus_api_sources}) + target_include_directories(test_opus_api + PRIVATE ${CMAKE_CURRENT_BINARY_DIR} celt) + target_link_libraries(test_opus_api PRIVATE opus) + target_compile_definitions(test_opus_api PRIVATE OPUS_BUILD) + if(OPUS_FIXED_POINT) + target_compile_definitions(test_opus_api PRIVATE DISABLE_FLOAT_API) + endif() + add_test(NAME test_opus_api COMMAND ${CMAKE_COMMAND} + -DTEST_EXECUTABLE=$ + -DCMAKE_SYSTEM_NAME=${CMAKE_SYSTEM_NAME} + -P "${PROJECT_SOURCE_DIR}/cmake/RunTest.cmake") + + add_executable(test_opus_encode ${test_opus_encode_sources}) + target_include_directories(test_opus_encode + PRIVATE ${CMAKE_CURRENT_BINARY_DIR} celt) + target_link_libraries(test_opus_encode PRIVATE opus) + target_compile_definitions(test_opus_encode PRIVATE OPUS_BUILD) + add_test(NAME test_opus_encode COMMAND ${CMAKE_COMMAND} + -DTEST_EXECUTABLE=$ + -DCMAKE_SYSTEM_NAME=${CMAKE_SYSTEM_NAME} + -P "${PROJECT_SOURCE_DIR}/cmake/RunTest.cmake") +endif() diff --git a/lib/opus/COPYING b/lib/opus/COPYING new file mode 100644 index 00000000..9c739c34 --- /dev/null +++ b/lib/opus/COPYING @@ -0,0 +1,44 @@ +Copyright 2001-2011 Xiph.Org, Skype Limited, Octasic, + Jean-Marc Valin, Timothy B. Terriberry, + CSIRO, Gregory Maxwell, Mark Borgerding, + Erik de Castro Lopo + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: + +- Redistributions of source code must retain the above copyright +notice, this list of conditions and the following disclaimer. + +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. + +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +Opus is subject to the royalty-free patent licenses which are +specified at: + +Xiph.Org Foundation: +https://datatracker.ietf.org/ipr/1524/ + +Microsoft Corporation: +https://datatracker.ietf.org/ipr/1914/ + +Broadcom Corporation: +https://datatracker.ietf.org/ipr/1526/ diff --git a/lib/opus/ChangeLog b/lib/opus/ChangeLog new file mode 100644 index 00000000..e69de29b diff --git a/lib/opus/LICENSE_PLEASE_READ.txt b/lib/opus/LICENSE_PLEASE_READ.txt new file mode 100644 index 00000000..bc88efa6 --- /dev/null +++ b/lib/opus/LICENSE_PLEASE_READ.txt @@ -0,0 +1,22 @@ +Contributions to the collaboration shall not be considered confidential. + +Each contributor represents and warrants that it has the right and +authority to license copyright in its contributions to the collaboration. + +Each contributor agrees to license the copyright in the contributions +under the Modified (2-clause or 3-clause) BSD License or the Clear BSD License. + +Please see the IPR statements submitted to the IETF for the complete +patent licensing details: + +Xiph.Org Foundation: +https://datatracker.ietf.org/ipr/1524/ + +Microsoft Corporation: +https://datatracker.ietf.org/ipr/1914/ + +Skype Limited: +https://datatracker.ietf.org/ipr/1602/ + +Broadcom Corporation: +https://datatracker.ietf.org/ipr/1526/ diff --git a/lib/opus/Makefile.am b/lib/opus/Makefile.am new file mode 100644 index 00000000..492fc09d --- /dev/null +++ b/lib/opus/Makefile.am @@ -0,0 +1,382 @@ +# Provide the full test output for failed tests when using the parallel +# test suite (which is enabled by default with automake 1.13+). +export VERBOSE = yes + +AUTOMAKE_OPTIONS = subdir-objects +ACLOCAL_AMFLAGS = -I m4 + +lib_LTLIBRARIES = libopus.la + +DIST_SUBDIRS = doc + +AM_CPPFLAGS = -I$(top_srcdir)/include -I$(top_srcdir)/celt -I$(top_srcdir)/silk \ + -I$(top_srcdir)/silk/float -I$(top_srcdir)/silk/fixed $(NE10_CFLAGS) + +include celt_sources.mk +include silk_sources.mk +include opus_sources.mk + +if FIXED_POINT +SILK_SOURCES += $(SILK_SOURCES_FIXED) +if HAVE_SSE4_1 +SILK_SOURCES += $(SILK_SOURCES_SSE4_1) $(SILK_SOURCES_FIXED_SSE4_1) +endif +if HAVE_ARM_NEON_INTR +SILK_SOURCES += $(SILK_SOURCES_FIXED_ARM_NEON_INTR) +endif +else +SILK_SOURCES += $(SILK_SOURCES_FLOAT) +if HAVE_SSE4_1 +SILK_SOURCES += $(SILK_SOURCES_SSE4_1) +endif +endif + +if DISABLE_FLOAT_API +else +OPUS_SOURCES += $(OPUS_SOURCES_FLOAT) +endif + +if CPU_X86 +if HAVE_RTCD +CELT_SOURCES += $(CELT_SOURCES_X86_RTCD) +SILK_SOURCES += $(SILK_SOURCES_X86_RTCD) +endif +if HAVE_SSE +CELT_SOURCES += $(CELT_SOURCES_SSE) +endif +if HAVE_SSE2 +CELT_SOURCES += $(CELT_SOURCES_SSE2) +endif +if HAVE_SSE4_1 +CELT_SOURCES += $(CELT_SOURCES_SSE4_1) +endif +endif + +if CPU_ARM +if HAVE_RTCD +CELT_SOURCES += $(CELT_SOURCES_ARM_RTCD) +SILK_SOURCES += $(SILK_SOURCES_ARM_RTCD) +endif + +if HAVE_ARM_NEON_INTR +CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) +SILK_SOURCES += $(SILK_SOURCES_ARM_NEON_INTR) +endif + +if HAVE_ARM_NE10 +CELT_SOURCES += $(CELT_SOURCES_ARM_NE10) +endif + +if OPUS_ARM_EXTERNAL_ASM +noinst_LTLIBRARIES = libarmasm.la +libarmasm_la_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) +BUILT_SOURCES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) \ + $(CELT_AM_SOURCES_ARM_ASM:.s.in=.s) \ + $(CELT_AM_SOURCES_ARM_ASM:.s.in=-gnu.S) +endif +endif + +CLEANFILES = $(CELT_SOURCES_ARM_ASM:.s=-gnu.S) \ + $(CELT_AM_SOURCES_ARM_ASM:.s.in=-gnu.S) + +include celt_headers.mk +include silk_headers.mk +include opus_headers.mk + +libopus_la_SOURCES = $(CELT_SOURCES) $(SILK_SOURCES) $(OPUS_SOURCES) +libopus_la_LDFLAGS = -no-undefined -version-info @OPUS_LT_CURRENT@:@OPUS_LT_REVISION@:@OPUS_LT_AGE@ +libopus_la_LIBADD = $(NE10_LIBS) $(LIBM) +if OPUS_ARM_EXTERNAL_ASM +libopus_la_LIBADD += libarmasm.la +endif + +pkginclude_HEADERS = include/opus.h include/opus_multistream.h include/opus_types.h include/opus_defines.h include/opus_projection.h + +noinst_HEADERS = $(OPUS_HEAD) $(SILK_HEAD) $(CELT_HEAD) + +if EXTRA_PROGRAMS +noinst_PROGRAMS = celt/tests/test_unit_cwrs32 \ + celt/tests/test_unit_dft \ + celt/tests/test_unit_entropy \ + celt/tests/test_unit_laplace \ + celt/tests/test_unit_mathops \ + celt/tests/test_unit_mdct \ + celt/tests/test_unit_rotation \ + celt/tests/test_unit_types \ + opus_compare \ + opus_demo \ + repacketizer_demo \ + silk/tests/test_unit_LPC_inv_pred_gain \ + tests/test_opus_api \ + tests/test_opus_decode \ + tests/test_opus_encode \ + tests/test_opus_padding \ + tests/test_opus_projection \ + trivial_example + +TESTS = celt/tests/test_unit_cwrs32 \ + celt/tests/test_unit_dft \ + celt/tests/test_unit_entropy \ + celt/tests/test_unit_laplace \ + celt/tests/test_unit_mathops \ + celt/tests/test_unit_mdct \ + celt/tests/test_unit_rotation \ + celt/tests/test_unit_types \ + silk/tests/test_unit_LPC_inv_pred_gain \ + tests/test_opus_api \ + tests/test_opus_decode \ + tests/test_opus_encode \ + tests/test_opus_padding \ + tests/test_opus_projection + +opus_demo_SOURCES = src/opus_demo.c + +opus_demo_LDADD = libopus.la $(NE10_LIBS) $(LIBM) + +repacketizer_demo_SOURCES = src/repacketizer_demo.c + +repacketizer_demo_LDADD = libopus.la $(NE10_LIBS) $(LIBM) + +opus_compare_SOURCES = src/opus_compare.c +opus_compare_LDADD = $(LIBM) + +trivial_example_SOURCES = doc/trivial_example.c +trivial_example_LDADD = libopus.la $(LIBM) + +tests_test_opus_api_SOURCES = tests/test_opus_api.c tests/test_opus_common.h +tests_test_opus_api_LDADD = libopus.la $(NE10_LIBS) $(LIBM) + +tests_test_opus_encode_SOURCES = tests/test_opus_encode.c tests/opus_encode_regressions.c tests/test_opus_common.h +tests_test_opus_encode_LDADD = libopus.la $(NE10_LIBS) $(LIBM) + +tests_test_opus_decode_SOURCES = tests/test_opus_decode.c tests/test_opus_common.h +tests_test_opus_decode_LDADD = libopus.la $(NE10_LIBS) $(LIBM) + +tests_test_opus_padding_SOURCES = tests/test_opus_padding.c tests/test_opus_common.h +tests_test_opus_padding_LDADD = libopus.la $(NE10_LIBS) $(LIBM) + +CELT_OBJ = $(CELT_SOURCES:.c=.lo) +SILK_OBJ = $(SILK_SOURCES:.c=.lo) +OPUS_OBJ = $(OPUS_SOURCES:.c=.lo) + +tests_test_opus_projection_SOURCES = tests/test_opus_projection.c tests/test_opus_common.h +tests_test_opus_projection_LDADD = $(OPUS_OBJ) $(SILK_OBJ) $(CELT_OBJ) $(NE10_LIBS) $(LIBM) +if OPUS_ARM_EXTERNAL_ASM +tests_test_opus_projection_LDADD += libarmasm.la +endif + +silk_tests_test_unit_LPC_inv_pred_gain_SOURCES = silk/tests/test_unit_LPC_inv_pred_gain.c +silk_tests_test_unit_LPC_inv_pred_gain_LDADD = $(SILK_OBJ) $(CELT_OBJ) $(NE10_LIBS) $(LIBM) +if OPUS_ARM_EXTERNAL_ASM +silk_tests_test_unit_LPC_inv_pred_gain_LDADD += libarmasm.la +endif + +celt_tests_test_unit_cwrs32_SOURCES = celt/tests/test_unit_cwrs32.c +celt_tests_test_unit_cwrs32_LDADD = $(LIBM) + +celt_tests_test_unit_dft_SOURCES = celt/tests/test_unit_dft.c +celt_tests_test_unit_dft_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM) +if OPUS_ARM_EXTERNAL_ASM +celt_tests_test_unit_dft_LDADD += libarmasm.la +endif + +celt_tests_test_unit_entropy_SOURCES = celt/tests/test_unit_entropy.c +celt_tests_test_unit_entropy_LDADD = $(LIBM) + +celt_tests_test_unit_laplace_SOURCES = celt/tests/test_unit_laplace.c +celt_tests_test_unit_laplace_LDADD = $(LIBM) + +celt_tests_test_unit_mathops_SOURCES = celt/tests/test_unit_mathops.c +celt_tests_test_unit_mathops_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM) +if OPUS_ARM_EXTERNAL_ASM +celt_tests_test_unit_mathops_LDADD += libarmasm.la +endif + +celt_tests_test_unit_mdct_SOURCES = celt/tests/test_unit_mdct.c +celt_tests_test_unit_mdct_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM) +if OPUS_ARM_EXTERNAL_ASM +celt_tests_test_unit_mdct_LDADD += libarmasm.la +endif + +celt_tests_test_unit_rotation_SOURCES = celt/tests/test_unit_rotation.c +celt_tests_test_unit_rotation_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM) +if OPUS_ARM_EXTERNAL_ASM +celt_tests_test_unit_rotation_LDADD += libarmasm.la +endif + +celt_tests_test_unit_types_SOURCES = celt/tests/test_unit_types.c +celt_tests_test_unit_types_LDADD = $(LIBM) +endif + +if CUSTOM_MODES +pkginclude_HEADERS += include/opus_custom.h +if EXTRA_PROGRAMS +noinst_PROGRAMS += opus_custom_demo +opus_custom_demo_SOURCES = celt/opus_custom_demo.c +opus_custom_demo_LDADD = libopus.la $(LIBM) +endif +endif + +EXTRA_DIST = opus.pc.in \ + opus-uninstalled.pc.in \ + opus.m4 \ + Makefile.mips \ + Makefile.unix \ + CMakeLists.txt \ + cmake/CFeatureCheck.cmake \ + cmake/OpusBuildtype.cmake \ + cmake/OpusConfig.cmake \ + cmake/OpusConfig.cmake.in \ + cmake/OpusFunctions.cmake \ + cmake/OpusPackageVersion.cmake \ + cmake/OpusSources.cmake \ + cmake/RunTest.cmake \ + cmake/config.h.cmake.in \ + cmake/vla.c \ + cmake/cpu_info_by_asm.c \ + cmake/cpu_info_by_c.c \ + meson/get-version.py \ + meson/read-sources-list.py \ + meson.build \ + meson_options.txt \ + include/meson.build \ + celt/meson.build \ + celt/tests/meson.build \ + silk/meson.build \ + silk/tests/meson.build \ + src/meson.build \ + tests/meson.build \ + doc/meson.build \ + tests/run_vectors.sh \ + celt/arm/arm2gnu.pl \ + celt/arm/celt_pitch_xcorr_arm.s \ + win32/VS2015/opus.vcxproj \ + win32/VS2015/test_opus_encode.vcxproj.filters \ + win32/VS2015/test_opus_encode.vcxproj \ + win32/VS2015/opus_demo.vcxproj \ + win32/VS2015/test_opus_api.vcxproj.filters \ + win32/VS2015/test_opus_api.vcxproj \ + win32/VS2015/test_opus_decode.vcxproj.filters \ + win32/VS2015/opus_demo.vcxproj.filters \ + win32/VS2015/opus.vcxproj.filters \ + win32/VS2015/test_opus_decode.vcxproj \ + win32/VS2015/opus.sln \ + win32/VS2015/common.props \ + win32/genversion.bat \ + win32/config.h + +pkgconfigdir = $(libdir)/pkgconfig +pkgconfig_DATA = opus.pc + +m4datadir = $(datadir)/aclocal +m4data_DATA = opus.m4 + +# Targets to build and install just the library without the docs +opus check-opus install-opus: export NO_DOXYGEN = 1 + +opus: all +check-opus: check +install-opus: install + + +# Or just the docs +docs: + ( cd doc && $(MAKE) $(AM_MAKEFLAGS) ) + +install-docs: + ( cd doc && $(MAKE) $(AM_MAKEFLAGS) install ) + + +# Or everything (by default) +all-local: + @[ -n "$(NO_DOXYGEN)" ] || ( cd doc && $(MAKE) $(AM_MAKEFLAGS) ) + +install-data-local: + @[ -n "$(NO_DOXYGEN)" ] || ( cd doc && $(MAKE) $(AM_MAKEFLAGS) install ) + +clean-local: + -( cd doc && $(MAKE) $(AM_MAKEFLAGS) clean ) + +uninstall-local: + ( cd doc && $(MAKE) $(AM_MAKEFLAGS) uninstall ) + + +# We check this every time make is run, with configure.ac being touched to +# trigger an update of the build system files if update_version changes the +# current PACKAGE_VERSION (or if package_version was modified manually by a +# user with either AUTO_UPDATE=no or no update_version script present - the +# latter being the normal case for tarball releases). +# +# We can't just add the package_version file to CONFIGURE_DEPENDENCIES since +# simply running autoconf will not actually regenerate configure for us when +# the content of that file changes (due to autoconf dependency checking not +# knowing about that without us creating yet another file for it to include). +# +# The MAKECMDGOALS check is a gnu-make'ism, but will degrade 'gracefully' for +# makes that don't support it. The only loss of functionality is not forcing +# an update of package_version for `make dist` if AUTO_UPDATE=no, but that is +# unlikely to be a real problem for any real user. +$(top_srcdir)/configure.ac: force + @case "$(MAKECMDGOALS)" in \ + dist-hook) exit 0 ;; \ + dist-* | dist | distcheck | distclean) _arg=release ;; \ + esac; \ + if ! $(top_srcdir)/update_version $$_arg 2> /dev/null; then \ + if [ ! -e $(top_srcdir)/package_version ]; then \ + echo 'PACKAGE_VERSION="unknown"' > $(top_srcdir)/package_version; \ + fi; \ + . $(top_srcdir)/package_version || exit 1; \ + [ "$(PACKAGE_VERSION)" != "$$PACKAGE_VERSION" ] || exit 0; \ + fi; \ + touch $@ + +force: + +# Create a minimal package_version file when make dist is run. +dist-hook: + echo 'PACKAGE_VERSION="$(PACKAGE_VERSION)"' > $(top_distdir)/package_version + + +.PHONY: opus check-opus install-opus docs install-docs + +# automake doesn't do dependency tracking for asm files, that I can tell +$(CELT_SOURCES_ARM_ASM:%.s=%-gnu.S): celt/arm/armopts-gnu.S +$(CELT_SOURCES_ARM_ASM:%.s=%-gnu.S): $(top_srcdir)/celt/arm/arm2gnu.pl + +# convert ARM asm to GNU as format +%-gnu.S: $(top_srcdir)/%.s + $(top_srcdir)/celt/arm/arm2gnu.pl @ARM2GNU_PARAMS@ < $< > $@ +# For autoconf-modified sources (e.g., armopts.s) +%-gnu.S: %.s + $(top_srcdir)/celt/arm/arm2gnu.pl @ARM2GNU_PARAMS@ < $< > $@ + +OPT_UNIT_TEST_OBJ = $(celt_tests_test_unit_mathops_SOURCES:.c=.o) \ + $(celt_tests_test_unit_rotation_SOURCES:.c=.o) \ + $(celt_tests_test_unit_mdct_SOURCES:.c=.o) \ + $(celt_tests_test_unit_dft_SOURCES:.c=.o) \ + $(silk_tests_test_unit_LPC_inv_pred_gain_SOURCES:.c=.o) + +if HAVE_SSE +SSE_OBJ = $(CELT_SOURCES_SSE:.c=.lo) +$(SSE_OBJ): CFLAGS += $(OPUS_X86_SSE_CFLAGS) +endif + +if HAVE_SSE2 +SSE2_OBJ = $(CELT_SOURCES_SSE2:.c=.lo) +$(SSE2_OBJ): CFLAGS += $(OPUS_X86_SSE2_CFLAGS) +endif + +if HAVE_SSE4_1 +SSE4_1_OBJ = $(CELT_SOURCES_SSE4_1:.c=.lo) \ + $(SILK_SOURCES_SSE4_1:.c=.lo) \ + $(SILK_SOURCES_FIXED_SSE4_1:.c=.lo) +$(SSE4_1_OBJ): CFLAGS += $(OPUS_X86_SSE4_1_CFLAGS) +endif + +if HAVE_ARM_NEON_INTR +ARM_NEON_INTR_OBJ = $(CELT_SOURCES_ARM_NEON_INTR:.c=.lo) \ + $(SILK_SOURCES_ARM_NEON_INTR:.c=.lo) \ + $(SILK_SOURCES_FIXED_ARM_NEON_INTR:.c=.lo) +$(ARM_NEON_INTR_OBJ): CFLAGS += \ + $(OPUS_ARM_NEON_INTR_CFLAGS) $(NE10_CFLAGS) +endif diff --git a/lib/opus/Makefile.mips b/lib/opus/Makefile.mips new file mode 100644 index 00000000..e9bfc22e --- /dev/null +++ b/lib/opus/Makefile.mips @@ -0,0 +1,161 @@ +#################### COMPILE OPTIONS ####################### + +# Uncomment this for fixed-point build +FIXED_POINT=1 + +# It is strongly recommended to uncomment one of these +# VAR_ARRAYS: Use C99 variable-length arrays for stack allocation +# USE_ALLOCA: Use alloca() for stack allocation +# If none is defined, then the fallback is a non-threadsafe global array +CFLAGS := -DUSE_ALLOCA $(CFLAGS) +#CFLAGS := -DVAR_ARRAYS $(CFLAGS) + +# These options affect performance +# HAVE_LRINTF: Use C99 intrinsics to speed up float-to-int conversion +CFLAGS := -DHAVE_LRINTF $(CFLAGS) + +###################### END OF OPTIONS ###################### + +-include package_version + +include silk_sources.mk +include celt_sources.mk +include opus_sources.mk + +ifdef FIXED_POINT +SILK_SOURCES += $(SILK_SOURCES_FIXED) +else +SILK_SOURCES += $(SILK_SOURCES_FLOAT) +OPUS_SOURCES += $(OPUS_SOURCES_FLOAT) +endif + +EXESUFFIX = +LIBPREFIX = lib +LIBSUFFIX = .a +OBJSUFFIX = .o + +CC = $(TOOLCHAIN_PREFIX)cc$(TOOLCHAIN_SUFFIX) +AR = $(TOOLCHAIN_PREFIX)ar +RANLIB = $(TOOLCHAIN_PREFIX)ranlib +CP = $(TOOLCHAIN_PREFIX)cp + +cppflags-from-defines = $(addprefix -D,$(1)) +cppflags-from-includes = $(addprefix -I,$(1)) +ldflags-from-ldlibdirs = $(addprefix -L,$(1)) +ldlibs-from-libs = $(addprefix -l,$(1)) + +WARNINGS = -Wall -W -Wstrict-prototypes -Wextra -Wcast-align -Wnested-externs -Wshadow + +CFLAGS += -mips32r2 -mno-mips16 -std=gnu99 -O2 -g $(WARNINGS) -DENABLE_ASSERTIONS -DMIPSr1_ASM -DOPUS_BUILD -mdspr2 -march=74kc -mtune=74kc -mmt -mgp32 + +CINCLUDES = include silk celt + +ifdef FIXED_POINT +CFLAGS += -DFIXED_POINT=1 -DDISABLE_FLOAT_API +CINCLUDES += silk/fixed +else +CINCLUDES += silk/float +endif + + +LIBS = m + +LDLIBDIRS = ./ + +CFLAGS += $(call cppflags-from-defines,$(CDEFINES)) +CFLAGS += $(call cppflags-from-includes,$(CINCLUDES)) +LDFLAGS += $(call ldflags-from-ldlibdirs,$(LDLIBDIRS)) +LDLIBS += $(call ldlibs-from-libs,$(LIBS)) + +COMPILE.c.cmdline = $(CC) -c $(CFLAGS) -o $@ $< +LINK.o = $(CC) $(LDPREFLAGS) $(LDFLAGS) +LINK.o.cmdline = $(LINK.o) $^ $(LDLIBS) -o $@$(EXESUFFIX) + +ARCHIVE.cmdline = $(AR) $(ARFLAGS) $@ $^ && $(RANLIB) $@ + +%$(OBJSUFFIX):%.c + $(COMPILE.c.cmdline) + +%$(OBJSUFFIX):%.cpp + $(COMPILE.cpp.cmdline) + +# Directives + + +# Variable definitions +LIB_NAME = opus +TARGET = $(LIBPREFIX)$(LIB_NAME)$(LIBSUFFIX) + +SRCS_C = $(SILK_SOURCES) $(CELT_SOURCES) $(OPUS_SOURCES) + +OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(SRCS_C)) + +OPUSDEMO_SRCS_C = src/opus_demo.c +OPUSDEMO_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(OPUSDEMO_SRCS_C)) + +TESTOPUSAPI_SRCS_C = tests/test_opus_api.c +TESTOPUSAPI_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSAPI_SRCS_C)) + +TESTOPUSDECODE_SRCS_C = tests/test_opus_decode.c +TESTOPUSDECODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSDECODE_SRCS_C)) + +TESTOPUSENCODE_SRCS_C = tests/test_opus_encode.c tests/opus_encode_regressions.c +TESTOPUSENCODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSENCODE_SRCS_C)) + +TESTOPUSPADDING_SRCS_C = tests/test_opus_padding.c +TESTOPUSPADDING_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSPADDING_SRCS_C)) + +OPUSCOMPARE_SRCS_C = src/opus_compare.c +OPUSCOMPARE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(OPUSCOMPARE_SRCS_C)) + +TESTS := test_opus_api test_opus_decode test_opus_encode test_opus_padding + +# Rules +all: lib opus_demo opus_compare $(TESTS) + +lib: $(TARGET) + +check: all + for test in $(TESTS); do ./$$test; done + +$(TARGET): $(OBJS) + $(ARCHIVE.cmdline) + +opus_demo$(EXESUFFIX): $(OPUSDEMO_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_api$(EXESUFFIX): $(TESTOPUSAPI_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_decode$(EXESUFFIX): $(TESTOPUSDECODE_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_encode$(EXESUFFIX): $(TESTOPUSENCODE_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_padding$(EXESUFFIX): $(TESTOPUSPADDING_OBJS) $(TARGET) + $(LINK.o.cmdline) + +opus_compare$(EXESUFFIX): $(OPUSCOMPARE_OBJS) + $(LINK.o.cmdline) + +celt/celt.o: CFLAGS += -DPACKAGE_VERSION='$(PACKAGE_VERSION)' +celt/celt.o: package_version + +package_version: force + @if [ -x ./update_version ]; then \ + ./update_version || true; \ + elif [ ! -e ./package_version ]; then \ + echo 'PACKAGE_VERSION="unknown"' > ./package_version; \ + fi + +force: + +clean: + rm -f opus_demo$(EXESUFFIX) opus_compare$(EXESUFFIX) $(TARGET) \ + test_opus_api$(EXESUFFIX) test_opus_decode$(EXESUFFIX) \ + test_opus_encode$(EXESUFFIX) test_opus_padding$(EXESUFFIX) \ + $(OBJS) $(OPUSDEMO_OBJS) $(OPUSCOMPARE_OBJS) $(TESTOPUSAPI_OBJS) \ + $(TESTOPUSDECODE_OBJS) $(TESTOPUSENCODE_OBJS) $(TESTOPUSPADDING_OBJS) + +.PHONY: all lib clean force check diff --git a/lib/opus/Makefile.unix b/lib/opus/Makefile.unix new file mode 100644 index 00000000..90a48f0c --- /dev/null +++ b/lib/opus/Makefile.unix @@ -0,0 +1,159 @@ +#################### COMPILE OPTIONS ####################### + +# Uncomment this for fixed-point build +#FIXED_POINT=1 + +# It is strongly recommended to uncomment one of these +# VAR_ARRAYS: Use C99 variable-length arrays for stack allocation +# USE_ALLOCA: Use alloca() for stack allocation +# If none is defined, then the fallback is a non-threadsafe global array +CFLAGS := -DUSE_ALLOCA $(CFLAGS) +#CFLAGS := -DVAR_ARRAYS $(CFLAGS) + +# These options affect performance +# HAVE_LRINTF: Use C99 intrinsics to speed up float-to-int conversion +#CFLAGS := -DHAVE_LRINTF $(CFLAGS) + +###################### END OF OPTIONS ###################### + +-include package_version + +include silk_sources.mk +include celt_sources.mk +include opus_sources.mk + +ifdef FIXED_POINT +SILK_SOURCES += $(SILK_SOURCES_FIXED) +else +SILK_SOURCES += $(SILK_SOURCES_FLOAT) +OPUS_SOURCES += $(OPUS_SOURCES_FLOAT) +endif + +EXESUFFIX = +LIBPREFIX = lib +LIBSUFFIX = .a +OBJSUFFIX = .o + +CC = $(TOOLCHAIN_PREFIX)cc$(TOOLCHAIN_SUFFIX) +AR = $(TOOLCHAIN_PREFIX)ar +RANLIB = $(TOOLCHAIN_PREFIX)ranlib +CP = $(TOOLCHAIN_PREFIX)cp + +cppflags-from-defines = $(addprefix -D,$(1)) +cppflags-from-includes = $(addprefix -I,$(1)) +ldflags-from-ldlibdirs = $(addprefix -L,$(1)) +ldlibs-from-libs = $(addprefix -l,$(1)) + +WARNINGS = -Wall -W -Wstrict-prototypes -Wextra -Wcast-align -Wnested-externs -Wshadow +CFLAGS += -O2 -g $(WARNINGS) -DOPUS_BUILD +CINCLUDES = include silk celt + +ifdef FIXED_POINT +CFLAGS += -DFIXED_POINT=1 -DDISABLE_FLOAT_API +CINCLUDES += silk/fixed +else +CINCLUDES += silk/float +endif + + +LIBS = m + +LDLIBDIRS = ./ + +CFLAGS += $(call cppflags-from-defines,$(CDEFINES)) +CFLAGS += $(call cppflags-from-includes,$(CINCLUDES)) +LDFLAGS += $(call ldflags-from-ldlibdirs,$(LDLIBDIRS)) +LDLIBS += $(call ldlibs-from-libs,$(LIBS)) + +COMPILE.c.cmdline = $(CC) -c $(CFLAGS) -o $@ $< +LINK.o = $(CC) $(LDPREFLAGS) $(LDFLAGS) +LINK.o.cmdline = $(LINK.o) $^ $(LDLIBS) -o $@$(EXESUFFIX) + +ARCHIVE.cmdline = $(AR) $(ARFLAGS) $@ $^ && $(RANLIB) $@ + +%$(OBJSUFFIX):%.c + $(COMPILE.c.cmdline) + +%$(OBJSUFFIX):%.cpp + $(COMPILE.cpp.cmdline) + +# Directives + + +# Variable definitions +LIB_NAME = opus +TARGET = $(LIBPREFIX)$(LIB_NAME)$(LIBSUFFIX) + +SRCS_C = $(SILK_SOURCES) $(CELT_SOURCES) $(OPUS_SOURCES) + +OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(SRCS_C)) + +OPUSDEMO_SRCS_C = src/opus_demo.c +OPUSDEMO_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(OPUSDEMO_SRCS_C)) + +TESTOPUSAPI_SRCS_C = tests/test_opus_api.c +TESTOPUSAPI_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSAPI_SRCS_C)) + +TESTOPUSDECODE_SRCS_C = tests/test_opus_decode.c +TESTOPUSDECODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSDECODE_SRCS_C)) + +TESTOPUSENCODE_SRCS_C = tests/test_opus_encode.c tests/opus_encode_regressions.c +TESTOPUSENCODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSENCODE_SRCS_C)) + +TESTOPUSPADDING_SRCS_C = tests/test_opus_padding.c +TESTOPUSPADDING_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSPADDING_SRCS_C)) + +OPUSCOMPARE_SRCS_C = src/opus_compare.c +OPUSCOMPARE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(OPUSCOMPARE_SRCS_C)) + +TESTS := test_opus_api test_opus_decode test_opus_encode test_opus_padding + +# Rules +all: lib opus_demo opus_compare $(TESTS) + +lib: $(TARGET) + +check: all + for test in $(TESTS); do ./$$test; done + +$(TARGET): $(OBJS) + $(ARCHIVE.cmdline) + +opus_demo$(EXESUFFIX): $(OPUSDEMO_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_api$(EXESUFFIX): $(TESTOPUSAPI_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_decode$(EXESUFFIX): $(TESTOPUSDECODE_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_encode$(EXESUFFIX): $(TESTOPUSENCODE_OBJS) $(TARGET) + $(LINK.o.cmdline) + +test_opus_padding$(EXESUFFIX): $(TESTOPUSPADDING_OBJS) $(TARGET) + $(LINK.o.cmdline) + +opus_compare$(EXESUFFIX): $(OPUSCOMPARE_OBJS) + $(LINK.o.cmdline) + +celt/celt.o: CFLAGS += -DPACKAGE_VERSION='$(PACKAGE_VERSION)' +celt/celt.o: package_version + +package_version: force + @if [ -x ./update_version ]; then \ + ./update_version || true; \ + elif [ ! -e ./package_version ]; then \ + echo 'PACKAGE_VERSION="unknown"' > ./package_version; \ + fi + +force: + +clean: + rm -f opus_demo$(EXESUFFIX) opus_compare$(EXESUFFIX) $(TARGET) \ + test_opus_api$(EXESUFFIX) test_opus_decode$(EXESUFFIX) \ + test_opus_encode$(EXESUFFIX) test_opus_padding$(EXESUFFIX) \ + $(OBJS) $(OPUSDEMO_OBJS) $(OPUSCOMPARE_OBJS) $(TESTOPUSAPI_OBJS) \ + $(TESTOPUSDECODE_OBJS) $(TESTOPUSENCODE_OBJS) $(TESTOPUSPADDING_OBJS) + +.PHONY: all lib clean force check diff --git a/lib/opus/NEWS b/lib/opus/NEWS new file mode 100644 index 00000000..e69de29b diff --git a/lib/opus/README b/lib/opus/README new file mode 100644 index 00000000..4b13076f --- /dev/null +++ b/lib/opus/README @@ -0,0 +1,161 @@ +== Opus audio codec == + +Opus is a codec for interactive speech and audio transmission over the Internet. + + Opus can handle a wide range of interactive audio applications, including +Voice over IP, videoconferencing, in-game chat, and even remote live music +performances. It can scale from low bit-rate narrowband speech to very high +quality stereo music. + + Opus, when coupled with an appropriate container format, is also suitable +for non-realtime stored-file applications such as music distribution, game +soundtracks, portable music players, jukeboxes, and other applications that +have historically used high latency formats such as MP3, AAC, or Vorbis. + + Opus is specified by IETF RFC 6716: + https://tools.ietf.org/html/rfc6716 + + The Opus format and this implementation of it are subject to the royalty- +free patent and copyright licenses specified in the file COPYING. + +This package implements a shared library for encoding and decoding raw Opus +bitstreams. Raw Opus bitstreams should be used over RTP according to + https://tools.ietf.org/html/rfc7587 + +The package also includes a number of test tools used for testing the +correct operation of the library. The bitstreams read/written by these +tools should not be used for Opus file distribution: They include +additional debugging data and cannot support seeking. + +Opus stored in files should use the Ogg encapsulation for Opus which is +described at: + https://tools.ietf.org/html/rfc7845 + +An opus-tools package is available which provides encoding and decoding of +Ogg encapsulated Opus files and includes a number of useful features. + +Opus-tools can be found at: + https://gitlab.xiph.org/xiph/opus-tools.git +or on the main Opus website: + https://opus-codec.org/ + +== Compiling libopus == + +To build from a distribution tarball, you only need to do the following: + + % ./configure + % make + +To build from the git repository, the following steps are necessary: + +0) Set up a development environment: + +On an Ubuntu or Debian family Linux distribution: + + % sudo apt-get install git autoconf automake libtool gcc make + +On a Fedora/Redhat based Linux: + + % sudo dnf install git autoconf automake libtool gcc make + +Or for older Redhat/Centos Linux releases: + + % sudo yum install git autoconf automake libtool gcc make + +On Apple macOS, install Xcode and brew.sh, then in the Terminal enter: + + % brew install autoconf automake libtool + +1) Clone the repository: + + % git clone https://gitlab.xiph.org/xiph/opus.git + % cd opus + +2) Compiling the source + + % ./autogen.sh + % ./configure + % make + +3) Install the codec libraries (optional) + + % sudo make install + +Once you have compiled the codec, there will be a opus_demo executable +in the top directory. + +Usage: opus_demo [-e] + [options] + opus_demo -d [options] + + +mode: voip | audio | restricted-lowdelay +options: + -e : only runs the encoder (output the bit-stream) + -d : only runs the decoder (reads the bit-stream as input) + -cbr : enable constant bitrate; default: variable bitrate + -cvbr : enable constrained variable bitrate; default: + unconstrained + -bandwidth + : audio bandwidth (from narrowband to fullband); + default: sampling rate + -framesize <2.5|5|10|20|40|60> + : frame size in ms; default: 20 + -max_payload + : maximum payload size in bytes, default: 1024 + -complexity + : complexity, 0 (lowest) ... 10 (highest); default: 10 + -inbandfec : enable SILK inband FEC + -forcemono : force mono encoding, even for stereo input + -dtx : enable SILK DTX + -loss : simulate packet loss, in percent (0-100); default: 0 + +input and output are little-endian signed 16-bit PCM files or opus +bitstreams with simple opus_demo proprietary framing. + +== Testing == + +This package includes a collection of automated unit and system tests +which SHOULD be run after compiling the package especially the first +time it is run on a new platform. + +To run the integrated tests: + + % make check + +There is also collection of standard test vectors which are not +included in this package for size reasons but can be obtained from: +https://opus-codec.org/docs/opus_testvectors-rfc8251.tar.gz + +To run compare the code to these test vectors: + + % curl -OL https://opus-codec.org/docs/opus_testvectors-rfc8251.tar.gz + % tar -zxf opus_testvectors-rfc8251.tar.gz + % ./tests/run_vectors.sh ./ opus_newvectors 48000 + +== Portability notes == + +This implementation uses floating-point by default but can be compiled to +use only fixed-point arithmetic by setting --enable-fixed-point (if using +autoconf) or by defining the FIXED_POINT macro (if building manually). +The fixed point implementation has somewhat lower audio quality and is +slower on platforms with fast FPUs, it is normally only used in embedded +environments. + +The implementation can be compiled with either a C89 or a C99 compiler. +While it does not rely on any _undefined behavior_ as defined by C89 or +C99, it relies on common _implementation-defined behavior_ for two's +complement architectures: + +o Right shifts of negative values are consistent with two's + complement arithmetic, so that a>>b is equivalent to + floor(a/(2^b)), + +o For conversion to a signed integer of N bits, the value is reduced + modulo 2^N to be within range of the type, + +o The result of integer division of a negative value is truncated + towards zero, and + +o The compiler provides a 64-bit integer type (a C99 requirement + which is supported by most C89 compilers). diff --git a/lib/opus/README.draft b/lib/opus/README.draft new file mode 100644 index 00000000..9c31bd02 --- /dev/null +++ b/lib/opus/README.draft @@ -0,0 +1,54 @@ +To build this source code, simply type: + +% make + +If this does not work, or if you want to change the default configuration +(e.g., to compile for a fixed-point architecture), simply edit the options +in the Makefile. + +An up-to-date implementation conforming to this standard is available in a +Git repository at https://gitlab.xiph.org/xiph/opus.git or on a website at: +https://opus-codec.org/ +However, although that implementation is expected to remain conformant +with the standard, it is the code in this RFC that shall remain normative. +To build from the git repository instead of using this RFC, follow these +steps: + +1) Clone the repository (latest implementation of this standard at the time +of publication) + +% git clone https://gitlab.xiph.org/xiph/opus.git +% cd opus + +2) Compile + +% ./autogen.sh +% ./configure +% make + +Once you have compiled the codec, there will be a opus_demo executable in +the top directory. + +Usage: opus_demo [-e] + [options] + opus_demo -d [options] + + +mode: voip | audio | restricted-lowdelay +options: +-e : only runs the encoder (output the bit-stream) +-d : only runs the decoder (reads the bit-stream as input) +-cbr : enable constant bitrate; default: variable bitrate +-cvbr : enable constrained variable bitrate; default: unconstrained +-bandwidth : audio bandwidth (from narrowband to fullband); + default: sampling rate +-framesize <2.5|5|10|20|40|60> : frame size in ms; default: 20 +-max_payload : maximum payload size in bytes, default: 1024 +-complexity : complexity, 0 (lowest) ... 10 (highest); default: 10 +-inbandfec : enable SILK inband FEC +-forcemono : force mono encoding, even for stereo input +-dtx : enable SILK DTX +-loss : simulate packet loss, in percent (0-100); default: 0 + +input and output are little endian signed 16-bit PCM files or opus bitstreams +with simple opus_demo proprietary framing. diff --git a/lib/opus/autogen.sh b/lib/opus/autogen.sh new file mode 100755 index 00000000..380d1f3c --- /dev/null +++ b/lib/opus/autogen.sh @@ -0,0 +1,14 @@ +#!/bin/sh +# Copyright (c) 2010-2015 Xiph.Org Foundation and contributors. +# Use of this source code is governed by a BSD-style license that can be +# found in the COPYING file. + +# Run this to set up the build system: configure, makefiles, etc. +set -e + +srcdir=`dirname $0` +test -n "$srcdir" && cd "$srcdir" + +echo "Updating build configuration files, please wait...." + +autoreconf -isf diff --git a/lib/opus/celt/_kiss_fft_guts.h b/lib/opus/celt/_kiss_fft_guts.h new file mode 100644 index 00000000..17392b3e --- /dev/null +++ b/lib/opus/celt/_kiss_fft_guts.h @@ -0,0 +1,182 @@ +/*Copyright (c) 2003-2004, Mark Borgerding + + All rights reserved. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE.*/ + +#ifndef KISS_FFT_GUTS_H +#define KISS_FFT_GUTS_H + +#define MIN(a,b) ((a)<(b) ? (a):(b)) +#define MAX(a,b) ((a)>(b) ? (a):(b)) + +/* kiss_fft.h + defines kiss_fft_scalar as either short or a float type + and defines + typedef struct { kiss_fft_scalar r; kiss_fft_scalar i; }kiss_fft_cpx; */ +#include "kiss_fft.h" + +/* + Explanation of macros dealing with complex math: + + C_MUL(m,a,b) : m = a*b + C_FIXDIV( c , div ) : if a fixed point impl., c /= div. noop otherwise + C_SUB( res, a,b) : res = a - b + C_SUBFROM( res , a) : res -= a + C_ADDTO( res , a) : res += a + * */ +#ifdef FIXED_POINT +#include "arch.h" + + +#define SAMP_MAX 2147483647 +#define TWID_MAX 32767 +#define TRIG_UPSCALE 1 + +#define SAMP_MIN -SAMP_MAX + + +# define S_MUL(a,b) MULT16_32_Q15(b, a) + +# define C_MUL(m,a,b) \ + do{ (m).r = SUB32_ovflw(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \ + (m).i = ADD32_ovflw(S_MUL((a).r,(b).i) , S_MUL((a).i,(b).r)); }while(0) + +# define C_MULC(m,a,b) \ + do{ (m).r = ADD32_ovflw(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \ + (m).i = SUB32_ovflw(S_MUL((a).i,(b).r) , S_MUL((a).r,(b).i)); }while(0) + +# define C_MULBYSCALAR( c, s ) \ + do{ (c).r = S_MUL( (c).r , s ) ;\ + (c).i = S_MUL( (c).i , s ) ; }while(0) + +# define DIVSCALAR(x,k) \ + (x) = S_MUL( x, (TWID_MAX-((k)>>1))/(k)+1 ) + +# define C_FIXDIV(c,div) \ + do { DIVSCALAR( (c).r , div); \ + DIVSCALAR( (c).i , div); }while (0) + +#define C_ADD( res, a,b)\ + do {(res).r=ADD32_ovflw((a).r,(b).r); (res).i=ADD32_ovflw((a).i,(b).i); \ + }while(0) +#define C_SUB( res, a,b)\ + do {(res).r=SUB32_ovflw((a).r,(b).r); (res).i=SUB32_ovflw((a).i,(b).i); \ + }while(0) +#define C_ADDTO( res , a)\ + do {(res).r = ADD32_ovflw((res).r, (a).r); (res).i = ADD32_ovflw((res).i,(a).i);\ + }while(0) + +#define C_SUBFROM( res , a)\ + do {(res).r = ADD32_ovflw((res).r,(a).r); (res).i = SUB32_ovflw((res).i,(a).i); \ + }while(0) + +#if defined(OPUS_ARM_INLINE_ASM) +#include "arm/kiss_fft_armv4.h" +#endif + +#if defined(OPUS_ARM_INLINE_EDSP) +#include "arm/kiss_fft_armv5e.h" +#endif +#if defined(MIPSr1_ASM) +#include "mips/kiss_fft_mipsr1.h" +#endif + +#else /* not FIXED_POINT*/ + +# define S_MUL(a,b) ( (a)*(b) ) +#define C_MUL(m,a,b) \ + do{ (m).r = (a).r*(b).r - (a).i*(b).i;\ + (m).i = (a).r*(b).i + (a).i*(b).r; }while(0) +#define C_MULC(m,a,b) \ + do{ (m).r = (a).r*(b).r + (a).i*(b).i;\ + (m).i = (a).i*(b).r - (a).r*(b).i; }while(0) + +#define C_MUL4(m,a,b) C_MUL(m,a,b) + +# define C_FIXDIV(c,div) /* NOOP */ +# define C_MULBYSCALAR( c, s ) \ + do{ (c).r *= (s);\ + (c).i *= (s); }while(0) +#endif + +#ifndef CHECK_OVERFLOW_OP +# define CHECK_OVERFLOW_OP(a,op,b) /* noop */ +#endif + +#ifndef C_ADD +#define C_ADD( res, a,b)\ + do { \ + CHECK_OVERFLOW_OP((a).r,+,(b).r)\ + CHECK_OVERFLOW_OP((a).i,+,(b).i)\ + (res).r=(a).r+(b).r; (res).i=(a).i+(b).i; \ + }while(0) +#define C_SUB( res, a,b)\ + do { \ + CHECK_OVERFLOW_OP((a).r,-,(b).r)\ + CHECK_OVERFLOW_OP((a).i,-,(b).i)\ + (res).r=(a).r-(b).r; (res).i=(a).i-(b).i; \ + }while(0) +#define C_ADDTO( res , a)\ + do { \ + CHECK_OVERFLOW_OP((res).r,+,(a).r)\ + CHECK_OVERFLOW_OP((res).i,+,(a).i)\ + (res).r += (a).r; (res).i += (a).i;\ + }while(0) + +#define C_SUBFROM( res , a)\ + do {\ + CHECK_OVERFLOW_OP((res).r,-,(a).r)\ + CHECK_OVERFLOW_OP((res).i,-,(a).i)\ + (res).r -= (a).r; (res).i -= (a).i; \ + }while(0) +#endif /* C_ADD defined */ + +#ifdef FIXED_POINT +/*# define KISS_FFT_COS(phase) TRIG_UPSCALE*floor(MIN(32767,MAX(-32767,.5+32768 * cos (phase)))) +# define KISS_FFT_SIN(phase) TRIG_UPSCALE*floor(MIN(32767,MAX(-32767,.5+32768 * sin (phase))))*/ +# define KISS_FFT_COS(phase) floor(.5+TWID_MAX*cos (phase)) +# define KISS_FFT_SIN(phase) floor(.5+TWID_MAX*sin (phase)) +# define HALF_OF(x) ((x)>>1) +#elif defined(USE_SIMD) +# define KISS_FFT_COS(phase) _mm_set1_ps( cos(phase) ) +# define KISS_FFT_SIN(phase) _mm_set1_ps( sin(phase) ) +# define HALF_OF(x) ((x)*_mm_set1_ps(.5f)) +#else +# define KISS_FFT_COS(phase) (kiss_fft_scalar) cos(phase) +# define KISS_FFT_SIN(phase) (kiss_fft_scalar) sin(phase) +# define HALF_OF(x) ((x)*.5f) +#endif + +#define kf_cexp(x,phase) \ + do{ \ + (x)->r = KISS_FFT_COS(phase);\ + (x)->i = KISS_FFT_SIN(phase);\ + }while(0) + +#define kf_cexp2(x,phase) \ + do{ \ + (x)->r = TRIG_UPSCALE*celt_cos_norm((phase));\ + (x)->i = TRIG_UPSCALE*celt_cos_norm((phase)-32768);\ +}while(0) + +#endif /* KISS_FFT_GUTS_H */ diff --git a/lib/opus/celt/arch.h b/lib/opus/celt/arch.h new file mode 100644 index 00000000..3845c3a0 --- /dev/null +++ b/lib/opus/celt/arch.h @@ -0,0 +1,291 @@ +/* Copyright (c) 2003-2008 Jean-Marc Valin + Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/** + @file arch.h + @brief Various architecture definitions for CELT +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef ARCH_H +#define ARCH_H + +#include "opus_types.h" +#include "opus_defines.h" + +# if !defined(__GNUC_PREREQ) +# if defined(__GNUC__)&&defined(__GNUC_MINOR__) +# define __GNUC_PREREQ(_maj,_min) \ + ((__GNUC__<<16)+__GNUC_MINOR__>=((_maj)<<16)+(_min)) +# else +# define __GNUC_PREREQ(_maj,_min) 0 +# endif +# endif + +#if OPUS_GNUC_PREREQ(3, 0) +#define opus_likely(x) (__builtin_expect(!!(x), 1)) +#define opus_unlikely(x) (__builtin_expect(!!(x), 0)) +#else +#define opus_likely(x) (!!(x)) +#define opus_unlikely(x) (!!(x)) +#endif + +#define CELT_SIG_SCALE 32768.f + +#define CELT_FATAL(str) celt_fatal(str, __FILE__, __LINE__); + +#if defined(ENABLE_ASSERTIONS) || defined(ENABLE_HARDENING) +#ifdef __GNUC__ +__attribute__((noreturn)) +#endif +void celt_fatal(const char *str, const char *file, int line); + +#if defined(CELT_C) && !defined(OVERRIDE_celt_fatal) +#include +#include +#ifdef __GNUC__ +__attribute__((noreturn)) +#endif +void celt_fatal(const char *str, const char *file, int line) +{ + fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str); +#if defined(_MSC_VER) + _set_abort_behavior( 0, _WRITE_ABORT_MSG); +#endif + abort(); +} +#endif + +#define celt_assert(cond) {if (!(cond)) {CELT_FATAL("assertion failed: " #cond);}} +#define celt_assert2(cond, message) {if (!(cond)) {CELT_FATAL("assertion failed: " #cond "\n" message);}} +#define MUST_SUCCEED(call) celt_assert((call) == OPUS_OK) +#else +#define celt_assert(cond) +#define celt_assert2(cond, message) +#define MUST_SUCCEED(call) do {if((call) != OPUS_OK) {RESTORE_STACK; return OPUS_INTERNAL_ERROR;} } while (0) +#endif + +#if defined(ENABLE_ASSERTIONS) +#define celt_sig_assert(cond) {if (!(cond)) {CELT_FATAL("signal assertion failed: " #cond);}} +#else +#define celt_sig_assert(cond) +#endif + +#define IMUL32(a,b) ((a)*(b)) + +#define MIN16(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum 16-bit value. */ +#define MAX16(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum 16-bit value. */ +#define MIN32(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum 32-bit value. */ +#define MAX32(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum 32-bit value. */ +#define IMIN(a,b) ((a) < (b) ? (a) : (b)) /**< Minimum int value. */ +#define IMAX(a,b) ((a) > (b) ? (a) : (b)) /**< Maximum int value. */ +#define UADD32(a,b) ((a)+(b)) +#define USUB32(a,b) ((a)-(b)) + +/* Set this if opus_int64 is a native type of the CPU. */ +/* Assume that all LP64 architectures have fast 64-bit types; also x86_64 + (which can be ILP32 for x32) and Win64 (which is LLP64). */ +#if defined(__x86_64__) || defined(__LP64__) || defined(_WIN64) +#define OPUS_FAST_INT64 1 +#else +#define OPUS_FAST_INT64 0 +#endif + +#define PRINT_MIPS(file) + +#ifdef FIXED_POINT + +typedef opus_int16 opus_val16; +typedef opus_int32 opus_val32; +typedef opus_int64 opus_val64; + +typedef opus_val32 celt_sig; +typedef opus_val16 celt_norm; +typedef opus_val32 celt_ener; + +#define celt_isnan(x) 0 + +#define Q15ONE 32767 + +#define SIG_SHIFT 12 +/* Safe saturation value for 32-bit signals. Should be less than + 2^31*(1-0.85) to avoid blowing up on DC at deemphasis.*/ +#define SIG_SAT (300000000) + +#define NORM_SCALING 16384 + +#define DB_SHIFT 10 + +#define EPSILON 1 +#define VERY_SMALL 0 +#define VERY_LARGE16 ((opus_val16)32767) +#define Q15_ONE ((opus_val16)32767) + +#define SCALEIN(a) (a) +#define SCALEOUT(a) (a) + +#define ABS16(x) ((x) < 0 ? (-(x)) : (x)) +#define ABS32(x) ((x) < 0 ? (-(x)) : (x)) + +static OPUS_INLINE opus_int16 SAT16(opus_int32 x) { + return x > 32767 ? 32767 : x < -32768 ? -32768 : (opus_int16)x; +} + +#ifdef FIXED_DEBUG +#include "fixed_debug.h" +#else + +#include "fixed_generic.h" + +#ifdef OPUS_ARM_PRESUME_AARCH64_NEON_INTR +#include "arm/fixed_arm64.h" +#elif defined (OPUS_ARM_INLINE_EDSP) +#include "arm/fixed_armv5e.h" +#elif defined (OPUS_ARM_INLINE_ASM) +#include "arm/fixed_armv4.h" +#elif defined (BFIN_ASM) +#include "fixed_bfin.h" +#elif defined (TI_C5X_ASM) +#include "fixed_c5x.h" +#elif defined (TI_C6X_ASM) +#include "fixed_c6x.h" +#endif + +#endif + +#else /* FIXED_POINT */ + +typedef float opus_val16; +typedef float opus_val32; +typedef float opus_val64; + +typedef float celt_sig; +typedef float celt_norm; +typedef float celt_ener; + +#ifdef FLOAT_APPROX +/* This code should reliably detect NaN/inf even when -ffast-math is used. + Assumes IEEE 754 format. */ +static OPUS_INLINE int celt_isnan(float x) +{ + union {float f; opus_uint32 i;} in; + in.f = x; + return ((in.i>>23)&0xFF)==0xFF && (in.i&0x007FFFFF)!=0; +} +#else +#ifdef __FAST_MATH__ +#error Cannot build libopus with -ffast-math unless FLOAT_APPROX is defined. This could result in crashes on extreme (e.g. NaN) input +#endif +#define celt_isnan(x) ((x)!=(x)) +#endif + +#define Q15ONE 1.0f + +#define NORM_SCALING 1.f + +#define EPSILON 1e-15f +#define VERY_SMALL 1e-30f +#define VERY_LARGE16 1e15f +#define Q15_ONE ((opus_val16)1.f) + +/* This appears to be the same speed as C99's fabsf() but it's more portable. */ +#define ABS16(x) ((float)fabs(x)) +#define ABS32(x) ((float)fabs(x)) + +#define QCONST16(x,bits) (x) +#define QCONST32(x,bits) (x) + +#define NEG16(x) (-(x)) +#define NEG32(x) (-(x)) +#define NEG32_ovflw(x) (-(x)) +#define EXTRACT16(x) (x) +#define EXTEND32(x) (x) +#define SHR16(a,shift) (a) +#define SHL16(a,shift) (a) +#define SHR32(a,shift) (a) +#define SHL32(a,shift) (a) +#define PSHR32(a,shift) (a) +#define VSHR32(a,shift) (a) + +#define PSHR(a,shift) (a) +#define SHR(a,shift) (a) +#define SHL(a,shift) (a) +#define SATURATE(x,a) (x) +#define SATURATE16(x) (x) + +#define ROUND16(a,shift) (a) +#define SROUND16(a,shift) (a) +#define HALF16(x) (.5f*(x)) +#define HALF32(x) (.5f*(x)) + +#define ADD16(a,b) ((a)+(b)) +#define SUB16(a,b) ((a)-(b)) +#define ADD32(a,b) ((a)+(b)) +#define SUB32(a,b) ((a)-(b)) +#define ADD32_ovflw(a,b) ((a)+(b)) +#define SUB32_ovflw(a,b) ((a)-(b)) +#define MULT16_16_16(a,b) ((a)*(b)) +#define MULT16_16(a,b) ((opus_val32)(a)*(opus_val32)(b)) +#define MAC16_16(c,a,b) ((c)+(opus_val32)(a)*(opus_val32)(b)) + +#define MULT16_32_Q15(a,b) ((a)*(b)) +#define MULT16_32_Q16(a,b) ((a)*(b)) + +#define MULT32_32_Q31(a,b) ((a)*(b)) + +#define MAC16_32_Q15(c,a,b) ((c)+(a)*(b)) +#define MAC16_32_Q16(c,a,b) ((c)+(a)*(b)) + +#define MULT16_16_Q11_32(a,b) ((a)*(b)) +#define MULT16_16_Q11(a,b) ((a)*(b)) +#define MULT16_16_Q13(a,b) ((a)*(b)) +#define MULT16_16_Q14(a,b) ((a)*(b)) +#define MULT16_16_Q15(a,b) ((a)*(b)) +#define MULT16_16_P15(a,b) ((a)*(b)) +#define MULT16_16_P13(a,b) ((a)*(b)) +#define MULT16_16_P14(a,b) ((a)*(b)) +#define MULT16_32_P16(a,b) ((a)*(b)) + +#define DIV32_16(a,b) (((opus_val32)(a))/(opus_val16)(b)) +#define DIV32(a,b) (((opus_val32)(a))/(opus_val32)(b)) + +#define SCALEIN(a) ((a)*CELT_SIG_SCALE) +#define SCALEOUT(a) ((a)*(1/CELT_SIG_SCALE)) + +#define SIG2WORD16(x) (x) + +#endif /* !FIXED_POINT */ + +#ifndef GLOBAL_STACK_SIZE +#ifdef FIXED_POINT +#define GLOBAL_STACK_SIZE 120000 +#else +#define GLOBAL_STACK_SIZE 120000 +#endif +#endif + +#endif /* ARCH_H */ diff --git a/lib/opus/celt/arm/arm2gnu.pl b/lib/opus/celt/arm/arm2gnu.pl new file mode 100755 index 00000000..a2895f74 --- /dev/null +++ b/lib/opus/celt/arm/arm2gnu.pl @@ -0,0 +1,353 @@ +#!/usr/bin/perl +# Copyright (C) 2002-2013 Xiph.org Foundation +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions +# are met: +# +# - Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# - Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in the +# documentation and/or other materials provided with the distribution. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +# ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +# OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +my $bigend; # little/big endian +my $nxstack; +my $apple = 0; +my $symprefix = ""; + +$nxstack = 0; + +eval 'exec /usr/local/bin/perl -S $0 ${1+"$@"}' + if $running_under_some_shell; + +while ($ARGV[0] =~ /^-/) { + $_ = shift; + last if /^--$/; + if (/^-n$/) { + $nflag++; + next; + } + if (/^--apple$/) { + $apple = 1; + $symprefix = "_"; + next; + } + die "I don't recognize this switch: $_\\n"; +} +$printit++ unless $nflag; + +$\ = "\n"; # automatically add newline on print +$n=0; + +$thumb = 0; # ARM mode by default, not Thumb. +@proc_stack = (); + +printf (" .syntax unified\n"); + +LINE: +while (<>) { + + # For ADRLs we need to add a new line after the substituted one. + $addPadding = 0; + + # First, we do not dare to touch *anything* inside double quotes, do we? + # Second, if you want a dollar character in the string, + # insert two of them -- that's how ARM C and assembler treat strings. + s/^([A-Za-z_]\w*)[ \t]+DCB[ \t]*\"/$1: .ascii \"/ && do { s/\$\$/\$/g; next }; + s/\bDCB\b[ \t]*\"/.ascii \"/ && do { s/\$\$/\$/g; next }; + s/^(\S+)\s+RN\s+(\S+)/$1 .req r$2/ && do { s/\$\$/\$/g; next }; + # If there's nothing on a line but a comment, don't try to apply any further + # substitutions (this is a cheap hack to avoid mucking up the license header) + s/^([ \t]*);/$1@/ && do { s/\$\$/\$/g; next }; + # If substituted -- leave immediately ! + + s/@/,:/; + s/;/@/; + while ( /@.*'/ ) { + s/(@.*)'/$1/g; + } + s/\{FALSE\}/0/g; + s/\{TRUE\}/1/g; + s/\{(\w\w\w\w+)\}/$1/g; + s/\bINCLUDE[ \t]*([^ \t\n]+)/.include \"$1\"/; + s/\bGET[ \t]*([^ \t\n]+)/.include \"${ my $x=$1; $x =~ s|\.s|-gnu.S|; \$x }\"/; + s/\bIMPORT\b/.extern/; + s/\bEXPORT\b\s*/.global $symprefix/; + s/^(\s+)\[/$1IF/; + s/^(\s+)\|/$1ELSE/; + s/^(\s+)\]/$1ENDIF/; + s/IF *:DEF:/ .ifdef/; + s/IF *:LNOT: *:DEF:/ .ifndef/; + s/ELSE/ .else/; + s/ENDIF/ .endif/; + + if( /\bIF\b/ ) { + s/\bIF\b/ .if/; + s/=/==/; + } + if ( $n == 2) { + s/\$/\\/g; + } + if ($n == 1) { + s/\$//g; + s/label//g; + $n = 2; + } + if ( /MACRO/ ) { + s/MACRO *\n/.macro/; + $n=1; + } + if ( /\bMEND\b/ ) { + s/\bMEND\b/.endm/; + $n=0; + } + + # ".rdata" doesn't work in 'as' version 2.13.2, as it is ".rodata" there. + # + if ( /\bAREA\b/ ) { + my $align; + $align = "2"; + if ( /ALIGN=(\d+)/ ) { + $align = $1; + } + if ( /CODE/ ) { + $nxstack = 1; + } + s/^(.+)CODE(.+)READONLY(.*)/ .text/; + s/^(.+)DATA(.+)READONLY(.*)/ .section .rdata/; + s/^(.+)\|\|\.data\|\|(.+)/ .data/; + s/^(.+)\|\|\.bss\|\|(.+)/ .bss/; + s/$/; .p2align $align/; + # Enable NEON instructions but don't produce a binary that requires + # ARMv7. RVCT does not have equivalent directives, so we just do this + # for all CODE areas. + if ( /.text/ ) { + # Separating .arch, .fpu, etc., by semicolons does not work (gas + # thinks the semicolon is part of the arch name, even when there's + # whitespace separating them). Sadly this means our line numbers + # won't match the original source file (we could use the .line + # directive, which is documented to be obsolete, but then gdb will + # show the wrong line in the translated source file). + s/$/; .arch armv7-a\n .fpu neon\n .object_arch armv4t/ unless ($apple); + } + } + + s/\|\|\.constdata\$(\d+)\|\|/.L_CONST$1/; # ||.constdata$3|| + s/\|\|\.bss\$(\d+)\|\|/.L_BSS$1/; # ||.bss$2|| + s/\|\|\.data\$(\d+)\|\|/.L_DATA$1/; # ||.data$2|| + s/\|\|([a-zA-Z0-9_]+)\@([a-zA-Z0-9_]+)\|\|/@ $&/; + s/^(\s+)\%(\s)/ .space $1/; + + s/\|(.+)\.(\d+)\|/\.$1_$2/; # |L80.123| -> .L80_123 + s/\bCODE32\b/.code 32/ && do {$thumb = 0}; + s/\bCODE16\b/.code 16/ && do {$thumb = 1}; + if (/\bPROC\b/) + { + my $prefix; + my $proc; + /^([A-Za-z_\.]\w+)\b/; + $proc = $1; + $prefix = ""; + if ($proc) + { + $prefix = $prefix.sprintf("\t.type\t%s, %%function", $proc) unless ($apple); + # Make sure we $prefix isn't empty here (for the $apple case). + # We handle mangling the label here, make sure it doesn't match + # the label handling below (if $prefix would be empty). + $prefix = $prefix."; "; + push(@proc_stack, $proc); + s/^[A-Za-z_\.]\w+/$symprefix$&:/; + } + $prefix = $prefix."\t.thumb_func; " if ($thumb); + s/\bPROC\b/@ $&/; + $_ = $prefix.$_; + } + s/^(\s*)(S|Q|SH|U|UQ|UH)ASX\b/$1$2ADDSUBX/; + s/^(\s*)(S|Q|SH|U|UQ|UH)SAX\b/$1$2SUBADDX/; + if (/\bENDP\b/) + { + my $proc; + s/\bENDP\b/@ $&/; + $proc = pop(@proc_stack); + $_ = "\t.size $proc, .-$proc".$_ if ($proc && !$apple); + } + s/\bSUBT\b/@ $&/; + s/\bDATA\b/@ $&/; # DATA directive is deprecated -- Asm guide, p.7-25 + s/\bKEEP\b/@ $&/; + s/\bEXPORTAS\b/@ $&/; + s/\|\|(.)+\bEQU\b/@ $&/; + s/\|\|([\w\$]+)\|\|/$1/; + s/\bENTRY\b/@ $&/; + s/\bASSERT\b/@ $&/; + s/\bGBLL\b/@ $&/; + s/\bGBLA\b/@ $&/; + s/^\W+OPT\b/@ $&/; + s/:OR:/|/g; + s/:SHL:/<>/g; + s/:AND:/&/g; + s/:LAND:/&&/g; + s/CPSR/cpsr/; + s/SPSR/spsr/; + s/ALIGN$/.balign 4/; + s/ALIGN\s+([0-9x]+)$/.balign $1/; + s/psr_cxsf/psr_all/; + s/LTORG/.ltorg/; + s/^([A-Za-z_]\w*)[ \t]+EQU/ .set $1,/; + s/^([A-Za-z_]\w*)[ \t]+SETL/ .set $1,/; + s/^([A-Za-z_]\w*)[ \t]+SETA/ .set $1,/; + s/^([A-Za-z_]\w*)[ \t]+\*/ .set $1,/; + + # {PC} + 0xdeadfeed --> . + 0xdeadfeed + s/\{PC\} \+/ \. +/; + + # Single hex constant on the line ! + # + # >>> NOTE <<< + # Double-precision floats in gcc are always mixed-endian, which means + # bytes in two words are little-endian, but words are big-endian. + # So, 0x0000deadfeed0000 would be stored as 0x0000dead at low address + # and 0xfeed0000 at high address. + # + s/\bDCFD\b[ \t]+0x([a-fA-F0-9]{8})([a-fA-F0-9]{8})/.long 0x$1, 0x$2/; + # Only decimal constants on the line, no hex ! + s/\bDCFD\b[ \t]+([0-9\.\-]+)/.double $1/; + + # Single hex constant on the line ! +# s/\bDCFS\b[ \t]+0x([a-f0-9]{8})([a-f0-9]{8})/.long 0x$1, 0x$2/; + # Only decimal constants on the line, no hex ! +# s/\bDCFS\b[ \t]+([0-9\.\-]+)/.double $1/; + s/\bDCFS[ \t]+0x/.word 0x/; + s/\bDCFS\b/.float/; + + s/^([A-Za-z_]\w*)[ \t]+DCD/$1 .word/; + s/\bDCD\b/.word/; + s/^([A-Za-z_]\w*)[ \t]+DCW/$1 .short/; + s/\bDCW\b/.short/; + s/^([A-Za-z_]\w*)[ \t]+DCB/$1 .byte/; + s/\bDCB\b/.byte/; + s/^([A-Za-z_]\w*)[ \t]+\%/.comm $1,/; + s/^[A-Za-z_\.]\w+/$&:/; + s/^(\d+)/$1:/; + s/\%(\d+)/$1b_or_f/; + s/\%[Bb](\d+)/$1b/; + s/\%[Ff](\d+)/$1f/; + s/\%[Ff][Tt](\d+)/$1f/; + s/&([\dA-Fa-f]+)/0x$1/; + if ( /\b2_[01]+\b/ ) { + s/\b2_([01]+)\b/conv$1&&&&/g; + while ( /[01][01][01][01]&&&&/ ) { + s/0000&&&&/&&&&0/g; + s/0001&&&&/&&&&1/g; + s/0010&&&&/&&&&2/g; + s/0011&&&&/&&&&3/g; + s/0100&&&&/&&&&4/g; + s/0101&&&&/&&&&5/g; + s/0110&&&&/&&&&6/g; + s/0111&&&&/&&&&7/g; + s/1000&&&&/&&&&8/g; + s/1001&&&&/&&&&9/g; + s/1010&&&&/&&&&A/g; + s/1011&&&&/&&&&B/g; + s/1100&&&&/&&&&C/g; + s/1101&&&&/&&&&D/g; + s/1110&&&&/&&&&E/g; + s/1111&&&&/&&&&F/g; + } + s/000&&&&/&&&&0/g; + s/001&&&&/&&&&1/g; + s/010&&&&/&&&&2/g; + s/011&&&&/&&&&3/g; + s/100&&&&/&&&&4/g; + s/101&&&&/&&&&5/g; + s/110&&&&/&&&&6/g; + s/111&&&&/&&&&7/g; + s/00&&&&/&&&&0/g; + s/01&&&&/&&&&1/g; + s/10&&&&/&&&&2/g; + s/11&&&&/&&&&3/g; + s/0&&&&/&&&&0/g; + s/1&&&&/&&&&1/g; + s/conv&&&&/0x/g; + } + + if ( /commandline/) + { + if( /-bigend/) + { + $bigend=1; + } + } + + if ( /\bDCDU\b/ ) + { + my $cmd=$_; + my $value; + my $prefix; + my $w1; + my $w2; + my $w3; + my $w4; + + s/\s+DCDU\b/@ $&/; + + $cmd =~ /\bDCDU\b\s+0x(\d+)/; + $value = $1; + $value =~ /(\w\w)(\w\w)(\w\w)(\w\w)/; + $w1 = $1; + $w2 = $2; + $w3 = $3; + $w4 = $4; + + if( $bigend ne "") + { + # big endian + $prefix = "\t.byte\t0x".$w1.";". + "\t.byte\t0x".$w2.";". + "\t.byte\t0x".$w3.";". + "\t.byte\t0x".$w4."; "; + } + else + { + # little endian + $prefix = "\t.byte\t0x".$w4.";". + "\t.byte\t0x".$w3.";". + "\t.byte\t0x".$w2.";". + "\t.byte\t0x".$w1."; "; + } + $_=$prefix.$_; + } + + if ( /\badrl\b/i ) + { + s/\badrl\s+(\w+)\s*,\s*(\w+)/ldr $1,=$2/i; + $addPadding = 1; + } + s/\bEND\b/@ END/; +} continue { + printf ("%s", $_) if $printit; + if ($addPadding != 0) + { + printf (" mov r0,r0\n"); + $addPadding = 0; + } +} +#If we had a code section, mark that this object doesn't need an executable +# stack. +if ($nxstack && !$apple) { + printf (" .section\t.note.GNU-stack,\"\",\%\%progbits\n"); +} diff --git a/lib/opus/celt/arm/arm_celt_map.c b/lib/opus/celt/arm/arm_celt_map.c new file mode 100644 index 00000000..ca988b66 --- /dev/null +++ b/lib/opus/celt/arm/arm_celt_map.c @@ -0,0 +1,160 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "pitch.h" +#include "kiss_fft.h" +#include "mdct.h" + +#if defined(OPUS_HAVE_RTCD) + +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR) +opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x, const opus_val16 *y, int N) = { + celt_inner_prod_c, /* ARMv4 */ + celt_inner_prod_c, /* EDSP */ + celt_inner_prod_c, /* Media */ + celt_inner_prod_neon /* NEON */ +}; + +void (*const DUAL_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02, + int N, opus_val32 *xy1, opus_val32 *xy2) = { + dual_inner_prod_c, /* ARMv4 */ + dual_inner_prod_c, /* EDSP */ + dual_inner_prod_c, /* Media */ + dual_inner_prod_neon /* NEON */ +}; +# endif + +# if defined(FIXED_POINT) +# if ((defined(OPUS_ARM_MAY_HAVE_NEON) && !defined(OPUS_ARM_PRESUME_NEON)) || \ + (defined(OPUS_ARM_MAY_HAVE_MEDIA) && !defined(OPUS_ARM_PRESUME_MEDIA)) || \ + (defined(OPUS_ARM_MAY_HAVE_EDSP) && !defined(OPUS_ARM_PRESUME_EDSP))) +opus_val32 (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *, + const opus_val16 *, opus_val32 *, int, int, int) = { + celt_pitch_xcorr_c, /* ARMv4 */ + MAY_HAVE_EDSP(celt_pitch_xcorr), /* EDSP */ + MAY_HAVE_MEDIA(celt_pitch_xcorr), /* Media */ + MAY_HAVE_NEON(celt_pitch_xcorr) /* NEON */ +}; + +# endif +# else /* !FIXED_POINT */ +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR) +void (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *, + const opus_val16 *, opus_val32 *, int, int, int) = { + celt_pitch_xcorr_c, /* ARMv4 */ + celt_pitch_xcorr_c, /* EDSP */ + celt_pitch_xcorr_c, /* Media */ + celt_pitch_xcorr_float_neon /* Neon */ +}; +# endif +# endif /* FIXED_POINT */ + +#if defined(FIXED_POINT) && defined(OPUS_HAVE_RTCD) && \ + defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR) + +void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])( + const opus_val16 *x, + const opus_val16 *y, + opus_val32 sum[4], + int len +) = { + xcorr_kernel_c, /* ARMv4 */ + xcorr_kernel_c, /* EDSP */ + xcorr_kernel_c, /* Media */ + xcorr_kernel_neon_fixed, /* Neon */ +}; + +#endif + +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +# if defined(HAVE_ARM_NE10) +# if defined(CUSTOM_MODES) +int (*const OPUS_FFT_ALLOC_ARCH_IMPL[OPUS_ARCHMASK+1])(kiss_fft_state *st) = { + opus_fft_alloc_arch_c, /* ARMv4 */ + opus_fft_alloc_arch_c, /* EDSP */ + opus_fft_alloc_arch_c, /* Media */ + opus_fft_alloc_arm_neon /* Neon with NE10 library support */ +}; + +void (*const OPUS_FFT_FREE_ARCH_IMPL[OPUS_ARCHMASK+1])(kiss_fft_state *st) = { + opus_fft_free_arch_c, /* ARMv4 */ + opus_fft_free_arch_c, /* EDSP */ + opus_fft_free_arch_c, /* Media */ + opus_fft_free_arm_neon /* Neon with NE10 */ +}; +# endif /* CUSTOM_MODES */ + +void (*const OPUS_FFT[OPUS_ARCHMASK+1])(const kiss_fft_state *cfg, + const kiss_fft_cpx *fin, + kiss_fft_cpx *fout) = { + opus_fft_c, /* ARMv4 */ + opus_fft_c, /* EDSP */ + opus_fft_c, /* Media */ + opus_fft_neon /* Neon with NE10 */ +}; + +void (*const OPUS_IFFT[OPUS_ARCHMASK+1])(const kiss_fft_state *cfg, + const kiss_fft_cpx *fin, + kiss_fft_cpx *fout) = { + opus_ifft_c, /* ARMv4 */ + opus_ifft_c, /* EDSP */ + opus_ifft_c, /* Media */ + opus_ifft_neon /* Neon with NE10 */ +}; + +void (*const CLT_MDCT_FORWARD_IMPL[OPUS_ARCHMASK+1])(const mdct_lookup *l, + kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, + int overlap, int shift, + int stride, int arch) = { + clt_mdct_forward_c, /* ARMv4 */ + clt_mdct_forward_c, /* EDSP */ + clt_mdct_forward_c, /* Media */ + clt_mdct_forward_neon /* Neon with NE10 */ +}; + +void (*const CLT_MDCT_BACKWARD_IMPL[OPUS_ARCHMASK+1])(const mdct_lookup *l, + kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, + int overlap, int shift, + int stride, int arch) = { + clt_mdct_backward_c, /* ARMv4 */ + clt_mdct_backward_c, /* EDSP */ + clt_mdct_backward_c, /* Media */ + clt_mdct_backward_neon /* Neon with NE10 */ +}; + +# endif /* HAVE_ARM_NE10 */ +# endif /* OPUS_ARM_MAY_HAVE_NEON_INTR */ + +#endif /* OPUS_HAVE_RTCD */ diff --git a/lib/opus/celt/arm/armcpu.c b/lib/opus/celt/arm/armcpu.c new file mode 100644 index 00000000..c7d16e6d --- /dev/null +++ b/lib/opus/celt/arm/armcpu.c @@ -0,0 +1,194 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* Original code from libtheora modified to suit to Opus */ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#ifdef OPUS_HAVE_RTCD + +#include "armcpu.h" +#include "cpu_support.h" +#include "os_support.h" +#include "opus_types.h" +#include "arch.h" + +#define OPUS_CPU_ARM_V4_FLAG (1< + +static OPUS_INLINE opus_uint32 opus_cpu_capabilities(void){ + opus_uint32 flags; + flags=0; + /* MSVC has no OPUS_INLINE __asm support for ARM, but it does let you __emit + * instructions via their assembled hex code. + * All of these instructions should be essentially nops. */ +# if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_MEDIA) \ + || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + __try{ + /*PLD [r13]*/ + __emit(0xF5DDF000); + flags|=OPUS_CPU_ARM_EDSP_FLAG; + } + __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ + /*Ignore exception.*/ + } +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) \ + || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + __try{ + /*SHADD8 r3,r3,r3*/ + __emit(0xE6333F93); + flags|=OPUS_CPU_ARM_MEDIA_FLAG; + } + __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ + /*Ignore exception.*/ + } +# if defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + __try{ + /*VORR q0,q0,q0*/ + __emit(0xF2200150); + flags|=OPUS_CPU_ARM_NEON_FLAG; + } + __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ + /*Ignore exception.*/ + } +# endif +# endif +# endif + return flags; +} + +#elif defined(__linux__) +/* Linux based */ +#include + +opus_uint32 opus_cpu_capabilities(void) +{ + opus_uint32 flags = 0; + FILE *cpuinfo; + + /* Reading /proc/self/auxv would be easier, but that doesn't work reliably on + * Android */ + cpuinfo = fopen("/proc/cpuinfo", "r"); + + if(cpuinfo != NULL) + { + /* 512 should be enough for anybody (it's even enough for all the flags that + * x86 has accumulated... so far). */ + char buf[512]; + + while(fgets(buf, 512, cpuinfo) != NULL) + { +# if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_MEDIA) \ + || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + /* Search for edsp and neon flag */ + if(memcmp(buf, "Features", 8) == 0) + { + char *p; + p = strstr(buf, " edsp"); + if(p != NULL && (p[5] == ' ' || p[5] == '\n')) + flags |= OPUS_CPU_ARM_EDSP_FLAG; + +# if defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + p = strstr(buf, " neon"); + if(p != NULL && (p[5] == ' ' || p[5] == '\n')) + flags |= OPUS_CPU_ARM_NEON_FLAG; +# endif + } +# endif + +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) \ + || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + /* Search for media capabilities (>= ARMv6) */ + if(memcmp(buf, "CPU architecture:", 17) == 0) + { + int version; + version = atoi(buf+17); + + if(version >= 6) + flags |= OPUS_CPU_ARM_MEDIA_FLAG; + } +# endif + } + + fclose(cpuinfo); + } + return flags; +} +#else +/* The feature registers which can tell us what the processor supports are + * accessible in priveleged modes only, so we can't have a general user-space + * detection method like on x86.*/ +# error "Configured to use ARM asm but no CPU detection method available for " \ + "your platform. Reconfigure with --disable-rtcd (or send patches)." +#endif + +static int opus_select_arch_impl(void) +{ + opus_uint32 flags = opus_cpu_capabilities(); + int arch = 0; + + if(!(flags & OPUS_CPU_ARM_EDSP_FLAG)) { + /* Asserts ensure arch values are sequential */ + celt_assert(arch == OPUS_ARCH_ARM_V4); + return arch; + } + arch++; + + if(!(flags & OPUS_CPU_ARM_MEDIA_FLAG)) { + celt_assert(arch == OPUS_ARCH_ARM_EDSP); + return arch; + } + arch++; + + if(!(flags & OPUS_CPU_ARM_NEON_FLAG)) { + celt_assert(arch == OPUS_ARCH_ARM_MEDIA); + return arch; + } + arch++; + + celt_assert(arch == OPUS_ARCH_ARM_NEON); + return arch; +} + +int opus_select_arch(void) { + int arch = opus_select_arch_impl(); +#ifdef FUZZING + arch = rand()%(arch+1); +#endif + return arch; +} +#endif diff --git a/lib/opus/celt/arm/armcpu.h b/lib/opus/celt/arm/armcpu.h new file mode 100644 index 00000000..820262ff --- /dev/null +++ b/lib/opus/celt/arm/armcpu.h @@ -0,0 +1,77 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(ARMCPU_H) +# define ARMCPU_H + +# if defined(OPUS_ARM_MAY_HAVE_EDSP) +# define MAY_HAVE_EDSP(name) name ## _edsp +# else +# define MAY_HAVE_EDSP(name) name ## _c +# endif + +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) +# define MAY_HAVE_MEDIA(name) name ## _media +# else +# define MAY_HAVE_MEDIA(name) MAY_HAVE_EDSP(name) +# endif + +# if defined(OPUS_ARM_MAY_HAVE_NEON) +# define MAY_HAVE_NEON(name) name ## _neon +# else +# define MAY_HAVE_NEON(name) MAY_HAVE_MEDIA(name) +# endif + +# if defined(OPUS_ARM_PRESUME_EDSP) +# define PRESUME_EDSP(name) name ## _edsp +# else +# define PRESUME_EDSP(name) name ## _c +# endif + +# if defined(OPUS_ARM_PRESUME_MEDIA) +# define PRESUME_MEDIA(name) name ## _media +# else +# define PRESUME_MEDIA(name) PRESUME_EDSP(name) +# endif + +# if defined(OPUS_ARM_PRESUME_NEON) +# define PRESUME_NEON(name) name ## _neon +# else +# define PRESUME_NEON(name) PRESUME_MEDIA(name) +# endif + +# if defined(OPUS_HAVE_RTCD) +int opus_select_arch(void); + +#define OPUS_ARCH_ARM_V4 (0) +#define OPUS_ARCH_ARM_EDSP (1) +#define OPUS_ARCH_ARM_MEDIA (2) +#define OPUS_ARCH_ARM_NEON (3) + +# endif + +#endif diff --git a/lib/opus/celt/arm/armopts.s.in b/lib/opus/celt/arm/armopts.s.in new file mode 100644 index 00000000..3d8aaf27 --- /dev/null +++ b/lib/opus/celt/arm/armopts.s.in @@ -0,0 +1,37 @@ +/* Copyright (C) 2013 Mozilla Corporation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +; Set the following to 1 if we have EDSP instructions +; (LDRD/STRD, etc., ARMv5E and later). +OPUS_ARM_MAY_HAVE_EDSP * @OPUS_ARM_MAY_HAVE_EDSP@ + +; Set the following to 1 if we have ARMv6 media instructions. +OPUS_ARM_MAY_HAVE_MEDIA * @OPUS_ARM_MAY_HAVE_MEDIA@ + +; Set the following to 1 if we have NEON (some ARMv7) +OPUS_ARM_MAY_HAVE_NEON * @OPUS_ARM_MAY_HAVE_NEON@ + +END diff --git a/lib/opus/celt/arm/celt_fft_ne10.c b/lib/opus/celt/arm/celt_fft_ne10.c new file mode 100644 index 00000000..ea5fd780 --- /dev/null +++ b/lib/opus/celt/arm/celt_fft_ne10.c @@ -0,0 +1,173 @@ +/* Copyright (c) 2015 Xiph.Org Foundation + Written by Viswanath Puttagunta */ +/** + @file celt_fft_ne10.c + @brief ARM Neon optimizations for fft using NE10 library + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef SKIP_CONFIG_H +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#endif + +#include +#include "os_support.h" +#include "kiss_fft.h" +#include "stack_alloc.h" + +#if !defined(FIXED_POINT) +# define NE10_FFT_ALLOC_C2C_TYPE_NEON ne10_fft_alloc_c2c_float32_neon +# define NE10_FFT_CFG_TYPE_T ne10_fft_cfg_float32_t +# define NE10_FFT_STATE_TYPE_T ne10_fft_state_float32_t +# define NE10_FFT_DESTROY_C2C_TYPE ne10_fft_destroy_c2c_float32 +# define NE10_FFT_CPX_TYPE_T ne10_fft_cpx_float32_t +# define NE10_FFT_C2C_1D_TYPE_NEON ne10_fft_c2c_1d_float32_neon +#else +# define NE10_FFT_ALLOC_C2C_TYPE_NEON(nfft) ne10_fft_alloc_c2c_int32_neon(nfft) +# define NE10_FFT_CFG_TYPE_T ne10_fft_cfg_int32_t +# define NE10_FFT_STATE_TYPE_T ne10_fft_state_int32_t +# define NE10_FFT_DESTROY_C2C_TYPE ne10_fft_destroy_c2c_int32 +# define NE10_FFT_DESTROY_C2C_TYPE ne10_fft_destroy_c2c_int32 +# define NE10_FFT_CPX_TYPE_T ne10_fft_cpx_int32_t +# define NE10_FFT_C2C_1D_TYPE_NEON ne10_fft_c2c_1d_int32_neon +#endif + +#if defined(CUSTOM_MODES) + +/* nfft lengths in NE10 that support scaled fft */ +# define NE10_FFTSCALED_SUPPORT_MAX 4 +static const int ne10_fft_scaled_support[NE10_FFTSCALED_SUPPORT_MAX] = { + 480, 240, 120, 60 +}; + +int opus_fft_alloc_arm_neon(kiss_fft_state *st) +{ + int i; + size_t memneeded = sizeof(struct arch_fft_state); + + st->arch_fft = (arch_fft_state *)opus_alloc(memneeded); + if (!st->arch_fft) + return -1; + + for (i = 0; i < NE10_FFTSCALED_SUPPORT_MAX; i++) { + if(st->nfft == ne10_fft_scaled_support[i]) + break; + } + if (i == NE10_FFTSCALED_SUPPORT_MAX) { + /* This nfft length (scaled fft) is not supported in NE10 */ + st->arch_fft->is_supported = 0; + st->arch_fft->priv = NULL; + } + else { + st->arch_fft->is_supported = 1; + st->arch_fft->priv = (void *)NE10_FFT_ALLOC_C2C_TYPE_NEON(st->nfft); + if (st->arch_fft->priv == NULL) { + return -1; + } + } + return 0; +} + +void opus_fft_free_arm_neon(kiss_fft_state *st) +{ + NE10_FFT_CFG_TYPE_T cfg; + + if (!st->arch_fft) + return; + + cfg = (NE10_FFT_CFG_TYPE_T)st->arch_fft->priv; + if (cfg) + NE10_FFT_DESTROY_C2C_TYPE(cfg); + opus_free(st->arch_fft); +} +#endif + +void opus_fft_neon(const kiss_fft_state *st, + const kiss_fft_cpx *fin, + kiss_fft_cpx *fout) +{ + NE10_FFT_STATE_TYPE_T state; + NE10_FFT_CFG_TYPE_T cfg = &state; + VARDECL(NE10_FFT_CPX_TYPE_T, buffer); + SAVE_STACK; + ALLOC(buffer, st->nfft, NE10_FFT_CPX_TYPE_T); + + if (!st->arch_fft->is_supported) { + /* This nfft length (scaled fft) not supported in NE10 */ + opus_fft_c(st, fin, fout); + } + else { + memcpy((void *)cfg, st->arch_fft->priv, sizeof(NE10_FFT_STATE_TYPE_T)); + state.buffer = (NE10_FFT_CPX_TYPE_T *)&buffer[0]; +#if !defined(FIXED_POINT) + state.is_forward_scaled = 1; + + NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, + (NE10_FFT_CPX_TYPE_T *)fin, + cfg, 0); +#else + NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, + (NE10_FFT_CPX_TYPE_T *)fin, + cfg, 0, 1); +#endif + } + RESTORE_STACK; +} + +void opus_ifft_neon(const kiss_fft_state *st, + const kiss_fft_cpx *fin, + kiss_fft_cpx *fout) +{ + NE10_FFT_STATE_TYPE_T state; + NE10_FFT_CFG_TYPE_T cfg = &state; + VARDECL(NE10_FFT_CPX_TYPE_T, buffer); + SAVE_STACK; + ALLOC(buffer, st->nfft, NE10_FFT_CPX_TYPE_T); + + if (!st->arch_fft->is_supported) { + /* This nfft length (scaled fft) not supported in NE10 */ + opus_ifft_c(st, fin, fout); + } + else { + memcpy((void *)cfg, st->arch_fft->priv, sizeof(NE10_FFT_STATE_TYPE_T)); + state.buffer = (NE10_FFT_CPX_TYPE_T *)&buffer[0]; +#if !defined(FIXED_POINT) + state.is_backward_scaled = 0; + + NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, + (NE10_FFT_CPX_TYPE_T *)fin, + cfg, 1); +#else + NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, + (NE10_FFT_CPX_TYPE_T *)fin, + cfg, 1, 0); +#endif + } + RESTORE_STACK; +} diff --git a/lib/opus/celt/arm/celt_mdct_ne10.c b/lib/opus/celt/arm/celt_mdct_ne10.c new file mode 100644 index 00000000..3531d02d --- /dev/null +++ b/lib/opus/celt/arm/celt_mdct_ne10.c @@ -0,0 +1,258 @@ +/* Copyright (c) 2015 Xiph.Org Foundation + Written by Viswanath Puttagunta */ +/** + @file celt_mdct_ne10.c + @brief ARM Neon optimizations for mdct using NE10 library + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef SKIP_CONFIG_H +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#endif + +#include "kiss_fft.h" +#include "_kiss_fft_guts.h" +#include "mdct.h" +#include "stack_alloc.h" + +void clt_mdct_forward_neon(const mdct_lookup *l, + kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, + int overlap, int shift, int stride, int arch) +{ + int i; + int N, N2, N4; + VARDECL(kiss_fft_scalar, f); + VARDECL(kiss_fft_cpx, f2); + const kiss_fft_state *st = l->kfft[shift]; + const kiss_twiddle_scalar *trig; + + SAVE_STACK; + + N = l->n; + trig = l->trig; + for (i=0;i>= 1; + trig += N; + } + N2 = N>>1; + N4 = N>>2; + + ALLOC(f, N2, kiss_fft_scalar); + ALLOC(f2, N4, kiss_fft_cpx); + + /* Consider the input to be composed of four blocks: [a, b, c, d] */ + /* Window, shuffle, fold */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_scalar * OPUS_RESTRICT xp1 = in+(overlap>>1); + const kiss_fft_scalar * OPUS_RESTRICT xp2 = in+N2-1+(overlap>>1); + kiss_fft_scalar * OPUS_RESTRICT yp = f; + const opus_val16 * OPUS_RESTRICT wp1 = window+(overlap>>1); + const opus_val16 * OPUS_RESTRICT wp2 = window+(overlap>>1)-1; + for(i=0;i<((overlap+3)>>2);i++) + { + /* Real part arranged as -d-cR, Imag part arranged as -b+aR*/ + *yp++ = MULT16_32_Q15(*wp2, xp1[N2]) + MULT16_32_Q15(*wp1,*xp2); + *yp++ = MULT16_32_Q15(*wp1, *xp1) - MULT16_32_Q15(*wp2, xp2[-N2]); + xp1+=2; + xp2-=2; + wp1+=2; + wp2-=2; + } + wp1 = window; + wp2 = window+overlap-1; + for(;i>2);i++) + { + /* Real part arranged as a-bR, Imag part arranged as -c-dR */ + *yp++ = *xp2; + *yp++ = *xp1; + xp1+=2; + xp2-=2; + } + for(;ii,t[N4+i]) - S_MUL(fp->r,t[i]); + yi = S_MUL(fp->r,t[N4+i]) + S_MUL(fp->i,t[i]); + *yp1 = yr; + *yp2 = yi; + fp++; + yp1 += 2*stride; + yp2 -= 2*stride; + } + } + RESTORE_STACK; +} + +void clt_mdct_backward_neon(const mdct_lookup *l, + kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 * OPUS_RESTRICT window, + int overlap, int shift, int stride, int arch) +{ + int i; + int N, N2, N4; + VARDECL(kiss_fft_scalar, f); + const kiss_twiddle_scalar *trig; + const kiss_fft_state *st = l->kfft[shift]; + + N = l->n; + trig = l->trig; + for (i=0;i>= 1; + trig += N; + } + N2 = N>>1; + N4 = N>>2; + + ALLOC(f, N2, kiss_fft_scalar); + + /* Pre-rotate */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_scalar * OPUS_RESTRICT xp1 = in; + const kiss_fft_scalar * OPUS_RESTRICT xp2 = in+stride*(N2-1); + kiss_fft_scalar * OPUS_RESTRICT yp = f; + const kiss_twiddle_scalar * OPUS_RESTRICT t = &trig[0]; + for(i=0;i>1)), arch); + + /* Post-rotate and de-shuffle from both ends of the buffer at once to make + it in-place. */ + { + kiss_fft_scalar * yp0 = out+(overlap>>1); + kiss_fft_scalar * yp1 = out+(overlap>>1)+N2-2; + const kiss_twiddle_scalar *t = &trig[0]; + /* Loop to (N4+1)>>1 to handle odd N4. When N4 is odd, the + middle pair will be computed twice. */ + for(i=0;i<(N4+1)>>1;i++) + { + kiss_fft_scalar re, im, yr, yi; + kiss_twiddle_scalar t0, t1; + re = yp0[0]; + im = yp0[1]; + t0 = t[i]; + t1 = t[N4+i]; + /* We'd scale up by 2 here, but instead it's done when mixing the windows */ + yr = S_MUL(re,t0) + S_MUL(im,t1); + yi = S_MUL(re,t1) - S_MUL(im,t0); + re = yp1[0]; + im = yp1[1]; + yp0[0] = yr; + yp1[1] = yi; + + t0 = t[(N4-i-1)]; + t1 = t[(N2-i-1)]; + /* We'd scale up by 2 here, but instead it's done when mixing the windows */ + yr = S_MUL(re,t0) + S_MUL(im,t1); + yi = S_MUL(re,t1) - S_MUL(im,t0); + yp1[0] = yr; + yp0[1] = yi; + yp0 += 2; + yp1 -= 2; + } + } + + /* Mirror on both sides for TDAC */ + { + kiss_fft_scalar * OPUS_RESTRICT xp1 = out+overlap-1; + kiss_fft_scalar * OPUS_RESTRICT yp1 = out; + const opus_val16 * OPUS_RESTRICT wp1 = window; + const opus_val16 * OPUS_RESTRICT wp2 = window+overlap-1; + + for(i = 0; i < overlap/2; i++) + { + kiss_fft_scalar x1, x2; + x1 = *xp1; + x2 = *yp1; + *yp1++ = MULT16_32_Q15(*wp2, x2) - MULT16_32_Q15(*wp1, x1); + *xp1-- = MULT16_32_Q15(*wp1, x2) + MULT16_32_Q15(*wp2, x1); + wp1++; + wp2--; + } + } + RESTORE_STACK; +} diff --git a/lib/opus/celt/arm/celt_neon_intr.c b/lib/opus/celt/arm/celt_neon_intr.c new file mode 100644 index 00000000..effda769 --- /dev/null +++ b/lib/opus/celt/arm/celt_neon_intr.c @@ -0,0 +1,211 @@ +/* Copyright (c) 2014-2015 Xiph.Org Foundation + Written by Viswanath Puttagunta */ +/** + @file celt_neon_intr.c + @brief ARM Neon Intrinsic optimizations for celt + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "../pitch.h" + +#if defined(FIXED_POINT) +void xcorr_kernel_neon_fixed(const opus_val16 * x, const opus_val16 * y, opus_val32 sum[4], int len) +{ + int j; + int32x4_t a = vld1q_s32(sum); + /* Load y[0...3] */ + /* This requires len>0 to always be valid (which we assert in the C code). */ + int16x4_t y0 = vld1_s16(y); + y += 4; + + for (j = 0; j + 8 <= len; j += 8) + { + /* Load x[0...7] */ + int16x8_t xx = vld1q_s16(x); + int16x4_t x0 = vget_low_s16(xx); + int16x4_t x4 = vget_high_s16(xx); + /* Load y[4...11] */ + int16x8_t yy = vld1q_s16(y); + int16x4_t y4 = vget_low_s16(yy); + int16x4_t y8 = vget_high_s16(yy); + int32x4_t a0 = vmlal_lane_s16(a, y0, x0, 0); + int32x4_t a1 = vmlal_lane_s16(a0, y4, x4, 0); + + int16x4_t y1 = vext_s16(y0, y4, 1); + int16x4_t y5 = vext_s16(y4, y8, 1); + int32x4_t a2 = vmlal_lane_s16(a1, y1, x0, 1); + int32x4_t a3 = vmlal_lane_s16(a2, y5, x4, 1); + + int16x4_t y2 = vext_s16(y0, y4, 2); + int16x4_t y6 = vext_s16(y4, y8, 2); + int32x4_t a4 = vmlal_lane_s16(a3, y2, x0, 2); + int32x4_t a5 = vmlal_lane_s16(a4, y6, x4, 2); + + int16x4_t y3 = vext_s16(y0, y4, 3); + int16x4_t y7 = vext_s16(y4, y8, 3); + int32x4_t a6 = vmlal_lane_s16(a5, y3, x0, 3); + int32x4_t a7 = vmlal_lane_s16(a6, y7, x4, 3); + + y0 = y8; + a = a7; + x += 8; + y += 8; + } + + for (; j < len; j++) + { + int16x4_t x0 = vld1_dup_s16(x); /* load next x */ + int32x4_t a0 = vmlal_s16(a, y0, x0); + + int16x4_t y4 = vld1_dup_s16(y); /* load next y */ + y0 = vext_s16(y0, y4, 1); + a = a0; + x++; + y++; + } + + vst1q_s32(sum, a); +} + +#else +/* + * Function: xcorr_kernel_neon_float + * --------------------------------- + * Computes 4 correlation values and stores them in sum[4] + */ +static void xcorr_kernel_neon_float(const float32_t *x, const float32_t *y, + float32_t sum[4], int len) { + float32x4_t YY[3]; + float32x4_t YEXT[3]; + float32x4_t XX[2]; + float32x2_t XX_2; + float32x4_t SUMM; + const float32_t *xi = x; + const float32_t *yi = y; + + celt_assert(len>0); + + YY[0] = vld1q_f32(yi); + SUMM = vdupq_n_f32(0); + + /* Consume 8 elements in x vector and 12 elements in y + * vector. However, the 12'th element never really gets + * touched in this loop. So, if len == 8, then we only + * must access y[0] to y[10]. y[11] must not be accessed + * hence make sure len > 8 and not len >= 8 + */ + while (len > 8) { + yi += 4; + YY[1] = vld1q_f32(yi); + yi += 4; + YY[2] = vld1q_f32(yi); + + XX[0] = vld1q_f32(xi); + xi += 4; + XX[1] = vld1q_f32(xi); + xi += 4; + + SUMM = vmlaq_lane_f32(SUMM, YY[0], vget_low_f32(XX[0]), 0); + YEXT[0] = vextq_f32(YY[0], YY[1], 1); + SUMM = vmlaq_lane_f32(SUMM, YEXT[0], vget_low_f32(XX[0]), 1); + YEXT[1] = vextq_f32(YY[0], YY[1], 2); + SUMM = vmlaq_lane_f32(SUMM, YEXT[1], vget_high_f32(XX[0]), 0); + YEXT[2] = vextq_f32(YY[0], YY[1], 3); + SUMM = vmlaq_lane_f32(SUMM, YEXT[2], vget_high_f32(XX[0]), 1); + + SUMM = vmlaq_lane_f32(SUMM, YY[1], vget_low_f32(XX[1]), 0); + YEXT[0] = vextq_f32(YY[1], YY[2], 1); + SUMM = vmlaq_lane_f32(SUMM, YEXT[0], vget_low_f32(XX[1]), 1); + YEXT[1] = vextq_f32(YY[1], YY[2], 2); + SUMM = vmlaq_lane_f32(SUMM, YEXT[1], vget_high_f32(XX[1]), 0); + YEXT[2] = vextq_f32(YY[1], YY[2], 3); + SUMM = vmlaq_lane_f32(SUMM, YEXT[2], vget_high_f32(XX[1]), 1); + + YY[0] = YY[2]; + len -= 8; + } + + /* Consume 4 elements in x vector and 8 elements in y + * vector. However, the 8'th element in y never really gets + * touched in this loop. So, if len == 4, then we only + * must access y[0] to y[6]. y[7] must not be accessed + * hence make sure len>4 and not len>=4 + */ + if (len > 4) { + yi += 4; + YY[1] = vld1q_f32(yi); + + XX[0] = vld1q_f32(xi); + xi += 4; + + SUMM = vmlaq_lane_f32(SUMM, YY[0], vget_low_f32(XX[0]), 0); + YEXT[0] = vextq_f32(YY[0], YY[1], 1); + SUMM = vmlaq_lane_f32(SUMM, YEXT[0], vget_low_f32(XX[0]), 1); + YEXT[1] = vextq_f32(YY[0], YY[1], 2); + SUMM = vmlaq_lane_f32(SUMM, YEXT[1], vget_high_f32(XX[0]), 0); + YEXT[2] = vextq_f32(YY[0], YY[1], 3); + SUMM = vmlaq_lane_f32(SUMM, YEXT[2], vget_high_f32(XX[0]), 1); + + YY[0] = YY[1]; + len -= 4; + } + + while (--len > 0) { + XX_2 = vld1_dup_f32(xi++); + SUMM = vmlaq_lane_f32(SUMM, YY[0], XX_2, 0); + YY[0]= vld1q_f32(++yi); + } + + XX_2 = vld1_dup_f32(xi); + SUMM = vmlaq_lane_f32(SUMM, YY[0], XX_2, 0); + + vst1q_f32(sum, SUMM); +} + +void celt_pitch_xcorr_float_neon(const opus_val16 *_x, const opus_val16 *_y, + opus_val32 *xcorr, int len, int max_pitch, int arch) { + int i; + (void)arch; + celt_assert(max_pitch > 0); + celt_sig_assert((((unsigned char *)_x-(unsigned char *)NULL)&3)==0); + + for (i = 0; i < (max_pitch-3); i += 4) { + xcorr_kernel_neon_float((const float32_t *)_x, (const float32_t *)_y+i, + (float32_t *)xcorr+i, len); + } + + /* In case max_pitch isn't a multiple of 4, do non-unrolled version. */ + for (; i < max_pitch; i++) { + xcorr[i] = celt_inner_prod_neon(_x, _y+i, len); + } +} +#endif diff --git a/lib/opus/celt/arm/celt_pitch_xcorr_arm.s b/lib/opus/celt/arm/celt_pitch_xcorr_arm.s new file mode 100644 index 00000000..6e873afc --- /dev/null +++ b/lib/opus/celt/arm/celt_pitch_xcorr_arm.s @@ -0,0 +1,551 @@ +; Copyright (c) 2007-2008 CSIRO +; Copyright (c) 2007-2009 Xiph.Org Foundation +; Copyright (c) 2013 Parrot +; Written by Aurélien Zanelli +; +; Redistribution and use in source and binary forms, with or without +; modification, are permitted provided that the following conditions +; are met: +; +; - Redistributions of source code must retain the above copyright +; notice, this list of conditions and the following disclaimer. +; +; - Redistributions in binary form must reproduce the above copyright +; notice, this list of conditions and the following disclaimer in the +; documentation and/or other materials provided with the distribution. +; +; THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +; ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +; LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +; A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +; OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +; EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +; PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +; PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +; LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +; NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +; SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + + AREA |.text|, CODE, READONLY + + GET celt/arm/armopts.s + +IF OPUS_ARM_MAY_HAVE_EDSP + EXPORT celt_pitch_xcorr_edsp +ENDIF + +IF OPUS_ARM_MAY_HAVE_NEON + EXPORT celt_pitch_xcorr_neon +ENDIF + +IF OPUS_ARM_MAY_HAVE_NEON + +; Compute sum[k]=sum(x[j]*y[j+k],j=0...len-1), k=0...3 +xcorr_kernel_neon PROC +xcorr_kernel_neon_start + ; input: + ; r3 = int len + ; r4 = opus_val16 *x + ; r5 = opus_val16 *y + ; q0 = opus_val32 sum[4] + ; output: + ; q0 = opus_val32 sum[4] + ; preserved: r0-r3, r6-r11, d2, q4-q7, q9-q15 + ; internal usage: + ; r12 = int j + ; d3 = y_3|y_2|y_1|y_0 + ; q2 = y_B|y_A|y_9|y_8|y_7|y_6|y_5|y_4 + ; q3 = x_7|x_6|x_5|x_4|x_3|x_2|x_1|x_0 + ; q8 = scratch + ; + ; Load y[0...3] + ; This requires len>0 to always be valid (which we assert in the C code). + VLD1.16 {d5}, [r5]! + SUBS r12, r3, #8 + BLE xcorr_kernel_neon_process4 +; Process 8 samples at a time. +; This loop loads one y value more than we actually need. Therefore we have to +; stop as soon as there are 8 or fewer samples left (instead of 7), to avoid +; reading past the end of the array. +xcorr_kernel_neon_process8 + ; This loop has 19 total instructions (10 cycles to issue, minimum), with + ; - 2 cycles of ARM insrtuctions, + ; - 10 cycles of load/store/byte permute instructions, and + ; - 9 cycles of data processing instructions. + ; On a Cortex A8, we dual-issue the maximum amount (9 cycles) between the + ; latter two categories, meaning the whole loop should run in 10 cycles per + ; iteration, barring cache misses. + ; + ; Load x[0...7] + VLD1.16 {d6, d7}, [r4]! + ; Unlike VMOV, VAND is a data processsing instruction (and doesn't get + ; assembled to VMOV, like VORR would), so it dual-issues with the prior VLD1. + VAND d3, d5, d5 + SUBS r12, r12, #8 + ; Load y[4...11] + VLD1.16 {d4, d5}, [r5]! + VMLAL.S16 q0, d3, d6[0] + VEXT.16 d16, d3, d4, #1 + VMLAL.S16 q0, d4, d7[0] + VEXT.16 d17, d4, d5, #1 + VMLAL.S16 q0, d16, d6[1] + VEXT.16 d16, d3, d4, #2 + VMLAL.S16 q0, d17, d7[1] + VEXT.16 d17, d4, d5, #2 + VMLAL.S16 q0, d16, d6[2] + VEXT.16 d16, d3, d4, #3 + VMLAL.S16 q0, d17, d7[2] + VEXT.16 d17, d4, d5, #3 + VMLAL.S16 q0, d16, d6[3] + VMLAL.S16 q0, d17, d7[3] + BGT xcorr_kernel_neon_process8 +; Process 4 samples here if we have > 4 left (still reading one extra y value). +xcorr_kernel_neon_process4 + ADDS r12, r12, #4 + BLE xcorr_kernel_neon_process2 + ; Load x[0...3] + VLD1.16 d6, [r4]! + ; Use VAND since it's a data processing instruction again. + VAND d4, d5, d5 + SUB r12, r12, #4 + ; Load y[4...7] + VLD1.16 d5, [r5]! + VMLAL.S16 q0, d4, d6[0] + VEXT.16 d16, d4, d5, #1 + VMLAL.S16 q0, d16, d6[1] + VEXT.16 d16, d4, d5, #2 + VMLAL.S16 q0, d16, d6[2] + VEXT.16 d16, d4, d5, #3 + VMLAL.S16 q0, d16, d6[3] +; Process 2 samples here if we have > 2 left (still reading one extra y value). +xcorr_kernel_neon_process2 + ADDS r12, r12, #2 + BLE xcorr_kernel_neon_process1 + ; Load x[0...1] + VLD2.16 {d6[],d7[]}, [r4]! + ; Use VAND since it's a data processing instruction again. + VAND d4, d5, d5 + SUB r12, r12, #2 + ; Load y[4...5] + VLD1.32 {d5[]}, [r5]! + VMLAL.S16 q0, d4, d6 + VEXT.16 d16, d4, d5, #1 + ; Replace bottom copy of {y5,y4} in d5 with {y3,y2} from d4, using VSRI + ; instead of VEXT, since it's a data-processing instruction. + VSRI.64 d5, d4, #32 + VMLAL.S16 q0, d16, d7 +; Process 1 sample using the extra y value we loaded above. +xcorr_kernel_neon_process1 + ; Load next *x + VLD1.16 {d6[]}, [r4]! + ADDS r12, r12, #1 + ; y[0...3] are left in d5 from prior iteration(s) (if any) + VMLAL.S16 q0, d5, d6 + MOVLE pc, lr +; Now process 1 last sample, not reading ahead. + ; Load last *y + VLD1.16 {d4[]}, [r5]! + VSRI.64 d4, d5, #16 + ; Load last *x + VLD1.16 {d6[]}, [r4]! + VMLAL.S16 q0, d4, d6 + MOV pc, lr + ENDP + +; opus_val32 celt_pitch_xcorr_neon(opus_val16 *_x, opus_val16 *_y, +; opus_val32 *xcorr, int len, int max_pitch, int arch) +celt_pitch_xcorr_neon PROC + ; input: + ; r0 = opus_val16 *_x + ; r1 = opus_val16 *_y + ; r2 = opus_val32 *xcorr + ; r3 = int len + ; output: + ; r0 = int maxcorr + ; internal usage: + ; r4 = opus_val16 *x (for xcorr_kernel_neon()) + ; r5 = opus_val16 *y (for xcorr_kernel_neon()) + ; r6 = int max_pitch + ; r12 = int j + ; q15 = int maxcorr[4] (q15 is not used by xcorr_kernel_neon()) + ; ignored: + ; int arch + STMFD sp!, {r4-r6, lr} + LDR r6, [sp, #16] + VMOV.S32 q15, #1 + ; if (max_pitch < 4) goto celt_pitch_xcorr_neon_process4_done + SUBS r6, r6, #4 + BLT celt_pitch_xcorr_neon_process4_done +celt_pitch_xcorr_neon_process4 + ; xcorr_kernel_neon parameters: + ; r3 = len, r4 = _x, r5 = _y, q0 = {0, 0, 0, 0} + MOV r4, r0 + MOV r5, r1 + VEOR q0, q0, q0 + ; xcorr_kernel_neon only modifies r4, r5, r12, and q0...q3. + ; So we don't save/restore any other registers. + BL xcorr_kernel_neon_start + SUBS r6, r6, #4 + VST1.32 {q0}, [r2]! + ; _y += 4 + ADD r1, r1, #8 + VMAX.S32 q15, q15, q0 + ; if (max_pitch < 4) goto celt_pitch_xcorr_neon_process4_done + BGE celt_pitch_xcorr_neon_process4 +; We have less than 4 sums left to compute. +celt_pitch_xcorr_neon_process4_done + ADDS r6, r6, #4 + ; Reduce maxcorr to a single value + VMAX.S32 d30, d30, d31 + VPMAX.S32 d30, d30, d30 + ; if (max_pitch <= 0) goto celt_pitch_xcorr_neon_done + BLE celt_pitch_xcorr_neon_done +; Now compute each remaining sum one at a time. +celt_pitch_xcorr_neon_process_remaining + MOV r4, r0 + MOV r5, r1 + VMOV.I32 q0, #0 + SUBS r12, r3, #8 + BLT celt_pitch_xcorr_neon_process_remaining4 +; Sum terms 8 at a time. +celt_pitch_xcorr_neon_process_remaining_loop8 + ; Load x[0...7] + VLD1.16 {q1}, [r4]! + ; Load y[0...7] + VLD1.16 {q2}, [r5]! + SUBS r12, r12, #8 + VMLAL.S16 q0, d4, d2 + VMLAL.S16 q0, d5, d3 + BGE celt_pitch_xcorr_neon_process_remaining_loop8 +; Sum terms 4 at a time. +celt_pitch_xcorr_neon_process_remaining4 + ADDS r12, r12, #4 + BLT celt_pitch_xcorr_neon_process_remaining4_done + ; Load x[0...3] + VLD1.16 {d2}, [r4]! + ; Load y[0...3] + VLD1.16 {d3}, [r5]! + SUB r12, r12, #4 + VMLAL.S16 q0, d3, d2 +celt_pitch_xcorr_neon_process_remaining4_done + ; Reduce the sum to a single value. + VADD.S32 d0, d0, d1 + VPADDL.S32 d0, d0 + ADDS r12, r12, #4 + BLE celt_pitch_xcorr_neon_process_remaining_loop_done +; Sum terms 1 at a time. +celt_pitch_xcorr_neon_process_remaining_loop1 + VLD1.16 {d2[]}, [r4]! + VLD1.16 {d3[]}, [r5]! + SUBS r12, r12, #1 + VMLAL.S16 q0, d2, d3 + BGT celt_pitch_xcorr_neon_process_remaining_loop1 +celt_pitch_xcorr_neon_process_remaining_loop_done + VST1.32 {d0[0]}, [r2]! + VMAX.S32 d30, d30, d0 + SUBS r6, r6, #1 + ; _y++ + ADD r1, r1, #2 + ; if (--max_pitch > 0) goto celt_pitch_xcorr_neon_process_remaining + BGT celt_pitch_xcorr_neon_process_remaining +celt_pitch_xcorr_neon_done + VMOV.32 r0, d30[0] + LDMFD sp!, {r4-r6, pc} + ENDP + +ENDIF + +IF OPUS_ARM_MAY_HAVE_EDSP + +; This will get used on ARMv7 devices without NEON, so it has been optimized +; to take advantage of dual-issuing where possible. +xcorr_kernel_edsp PROC +xcorr_kernel_edsp_start + ; input: + ; r3 = int len + ; r4 = opus_val16 *_x (must be 32-bit aligned) + ; r5 = opus_val16 *_y (must be 32-bit aligned) + ; r6...r9 = opus_val32 sum[4] + ; output: + ; r6...r9 = opus_val32 sum[4] + ; preserved: r0-r5 + ; internal usage + ; r2 = int j + ; r12,r14 = opus_val16 x[4] + ; r10,r11 = opus_val16 y[4] + STMFD sp!, {r2,r4,r5,lr} + LDR r10, [r5], #4 ; Load y[0...1] + SUBS r2, r3, #4 ; j = len-4 + LDR r11, [r5], #4 ; Load y[2...3] + BLE xcorr_kernel_edsp_process4_done + LDR r12, [r4], #4 ; Load x[0...1] + ; Stall +xcorr_kernel_edsp_process4 + ; The multiplies must issue from pipeline 0, and can't dual-issue with each + ; other. Every other instruction here dual-issues with a multiply, and is + ; thus "free". There should be no stalls in the body of the loop. + SMLABB r6, r12, r10, r6 ; sum[0] = MAC16_16(sum[0],x_0,y_0) + LDR r14, [r4], #4 ; Load x[2...3] + SMLABT r7, r12, r10, r7 ; sum[1] = MAC16_16(sum[1],x_0,y_1) + SUBS r2, r2, #4 ; j-=4 + SMLABB r8, r12, r11, r8 ; sum[2] = MAC16_16(sum[2],x_0,y_2) + SMLABT r9, r12, r11, r9 ; sum[3] = MAC16_16(sum[3],x_0,y_3) + SMLATT r6, r12, r10, r6 ; sum[0] = MAC16_16(sum[0],x_1,y_1) + LDR r10, [r5], #4 ; Load y[4...5] + SMLATB r7, r12, r11, r7 ; sum[1] = MAC16_16(sum[1],x_1,y_2) + SMLATT r8, r12, r11, r8 ; sum[2] = MAC16_16(sum[2],x_1,y_3) + SMLATB r9, r12, r10, r9 ; sum[3] = MAC16_16(sum[3],x_1,y_4) + LDRGT r12, [r4], #4 ; Load x[0...1] + SMLABB r6, r14, r11, r6 ; sum[0] = MAC16_16(sum[0],x_2,y_2) + SMLABT r7, r14, r11, r7 ; sum[1] = MAC16_16(sum[1],x_2,y_3) + SMLABB r8, r14, r10, r8 ; sum[2] = MAC16_16(sum[2],x_2,y_4) + SMLABT r9, r14, r10, r9 ; sum[3] = MAC16_16(sum[3],x_2,y_5) + SMLATT r6, r14, r11, r6 ; sum[0] = MAC16_16(sum[0],x_3,y_3) + LDR r11, [r5], #4 ; Load y[6...7] + SMLATB r7, r14, r10, r7 ; sum[1] = MAC16_16(sum[1],x_3,y_4) + SMLATT r8, r14, r10, r8 ; sum[2] = MAC16_16(sum[2],x_3,y_5) + SMLATB r9, r14, r11, r9 ; sum[3] = MAC16_16(sum[3],x_3,y_6) + BGT xcorr_kernel_edsp_process4 +xcorr_kernel_edsp_process4_done + ADDS r2, r2, #4 + BLE xcorr_kernel_edsp_done + LDRH r12, [r4], #2 ; r12 = *x++ + SUBS r2, r2, #1 ; j-- + ; Stall + SMLABB r6, r12, r10, r6 ; sum[0] = MAC16_16(sum[0],x,y_0) + LDRHGT r14, [r4], #2 ; r14 = *x++ + SMLABT r7, r12, r10, r7 ; sum[1] = MAC16_16(sum[1],x,y_1) + SMLABB r8, r12, r11, r8 ; sum[2] = MAC16_16(sum[2],x,y_2) + SMLABT r9, r12, r11, r9 ; sum[3] = MAC16_16(sum[3],x,y_3) + BLE xcorr_kernel_edsp_done + SMLABT r6, r14, r10, r6 ; sum[0] = MAC16_16(sum[0],x,y_1) + SUBS r2, r2, #1 ; j-- + SMLABB r7, r14, r11, r7 ; sum[1] = MAC16_16(sum[1],x,y_2) + LDRH r10, [r5], #2 ; r10 = y_4 = *y++ + SMLABT r8, r14, r11, r8 ; sum[2] = MAC16_16(sum[2],x,y_3) + LDRHGT r12, [r4], #2 ; r12 = *x++ + SMLABB r9, r14, r10, r9 ; sum[3] = MAC16_16(sum[3],x,y_4) + BLE xcorr_kernel_edsp_done + SMLABB r6, r12, r11, r6 ; sum[0] = MAC16_16(sum[0],tmp,y_2) + CMP r2, #1 ; j-- + SMLABT r7, r12, r11, r7 ; sum[1] = MAC16_16(sum[1],tmp,y_3) + LDRH r2, [r5], #2 ; r2 = y_5 = *y++ + SMLABB r8, r12, r10, r8 ; sum[2] = MAC16_16(sum[2],tmp,y_4) + LDRHGT r14, [r4] ; r14 = *x + SMLABB r9, r12, r2, r9 ; sum[3] = MAC16_16(sum[3],tmp,y_5) + BLE xcorr_kernel_edsp_done + SMLABT r6, r14, r11, r6 ; sum[0] = MAC16_16(sum[0],tmp,y_3) + LDRH r11, [r5] ; r11 = y_6 = *y + SMLABB r7, r14, r10, r7 ; sum[1] = MAC16_16(sum[1],tmp,y_4) + SMLABB r8, r14, r2, r8 ; sum[2] = MAC16_16(sum[2],tmp,y_5) + SMLABB r9, r14, r11, r9 ; sum[3] = MAC16_16(sum[3],tmp,y_6) +xcorr_kernel_edsp_done + LDMFD sp!, {r2,r4,r5,pc} + ENDP + +celt_pitch_xcorr_edsp PROC + ; input: + ; r0 = opus_val16 *_x (must be 32-bit aligned) + ; r1 = opus_val16 *_y (only needs to be 16-bit aligned) + ; r2 = opus_val32 *xcorr + ; r3 = int len + ; output: + ; r0 = maxcorr + ; internal usage + ; r4 = opus_val16 *x + ; r5 = opus_val16 *y + ; r6 = opus_val32 sum0 + ; r7 = opus_val32 sum1 + ; r8 = opus_val32 sum2 + ; r9 = opus_val32 sum3 + ; r1 = int max_pitch + ; r12 = int j + ; ignored: + ; int arch + STMFD sp!, {r4-r11, lr} + MOV r5, r1 + LDR r1, [sp, #36] + MOV r4, r0 + TST r5, #3 + ; maxcorr = 1 + MOV r0, #1 + BEQ celt_pitch_xcorr_edsp_process1u_done +; Compute one sum at the start to make y 32-bit aligned. + SUBS r12, r3, #4 + ; r14 = sum = 0 + MOV r14, #0 + LDRH r8, [r5], #2 + BLE celt_pitch_xcorr_edsp_process1u_loop4_done + LDR r6, [r4], #4 + MOV r8, r8, LSL #16 +celt_pitch_xcorr_edsp_process1u_loop4 + LDR r9, [r5], #4 + SMLABT r14, r6, r8, r14 ; sum = MAC16_16(sum, x_0, y_0) + LDR r7, [r4], #4 + SMLATB r14, r6, r9, r14 ; sum = MAC16_16(sum, x_1, y_1) + LDR r8, [r5], #4 + SMLABT r14, r7, r9, r14 ; sum = MAC16_16(sum, x_2, y_2) + SUBS r12, r12, #4 ; j-=4 + SMLATB r14, r7, r8, r14 ; sum = MAC16_16(sum, x_3, y_3) + LDRGT r6, [r4], #4 + BGT celt_pitch_xcorr_edsp_process1u_loop4 + MOV r8, r8, LSR #16 +celt_pitch_xcorr_edsp_process1u_loop4_done + ADDS r12, r12, #4 +celt_pitch_xcorr_edsp_process1u_loop1 + LDRHGE r6, [r4], #2 + ; Stall + SMLABBGE r14, r6, r8, r14 ; sum = MAC16_16(sum, *x, *y) + SUBSGE r12, r12, #1 + LDRHGT r8, [r5], #2 + BGT celt_pitch_xcorr_edsp_process1u_loop1 + ; Restore _x + SUB r4, r4, r3, LSL #1 + ; Restore and advance _y + SUB r5, r5, r3, LSL #1 + ; maxcorr = max(maxcorr, sum) + CMP r0, r14 + ADD r5, r5, #2 + MOVLT r0, r14 + SUBS r1, r1, #1 + ; xcorr[i] = sum + STR r14, [r2], #4 + BLE celt_pitch_xcorr_edsp_done +celt_pitch_xcorr_edsp_process1u_done + ; if (max_pitch < 4) goto celt_pitch_xcorr_edsp_process2 + SUBS r1, r1, #4 + BLT celt_pitch_xcorr_edsp_process2 +celt_pitch_xcorr_edsp_process4 + ; xcorr_kernel_edsp parameters: + ; r3 = len, r4 = _x, r5 = _y, r6...r9 = sum[4] = {0, 0, 0, 0} + MOV r6, #0 + MOV r7, #0 + MOV r8, #0 + MOV r9, #0 + BL xcorr_kernel_edsp_start ; xcorr_kernel_edsp(_x, _y+i, xcorr+i, len) + ; maxcorr = max(maxcorr, sum0, sum1, sum2, sum3) + CMP r0, r6 + ; _y+=4 + ADD r5, r5, #8 + MOVLT r0, r6 + CMP r0, r7 + MOVLT r0, r7 + CMP r0, r8 + MOVLT r0, r8 + CMP r0, r9 + MOVLT r0, r9 + STMIA r2!, {r6-r9} + SUBS r1, r1, #4 + BGE celt_pitch_xcorr_edsp_process4 +celt_pitch_xcorr_edsp_process2 + ADDS r1, r1, #2 + BLT celt_pitch_xcorr_edsp_process1a + SUBS r12, r3, #4 + ; {r10, r11} = {sum0, sum1} = {0, 0} + MOV r10, #0 + MOV r11, #0 + LDR r8, [r5], #4 + BLE celt_pitch_xcorr_edsp_process2_loop_done + LDR r6, [r4], #4 + LDR r9, [r5], #4 +celt_pitch_xcorr_edsp_process2_loop4 + SMLABB r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_0) + LDR r7, [r4], #4 + SMLABT r11, r6, r8, r11 ; sum1 = MAC16_16(sum1, x_0, y_1) + SUBS r12, r12, #4 ; j-=4 + SMLATT r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_1, y_1) + LDR r8, [r5], #4 + SMLATB r11, r6, r9, r11 ; sum1 = MAC16_16(sum1, x_1, y_2) + LDRGT r6, [r4], #4 + SMLABB r10, r7, r9, r10 ; sum0 = MAC16_16(sum0, x_2, y_2) + SMLABT r11, r7, r9, r11 ; sum1 = MAC16_16(sum1, x_2, y_3) + SMLATT r10, r7, r9, r10 ; sum0 = MAC16_16(sum0, x_3, y_3) + LDRGT r9, [r5], #4 + SMLATB r11, r7, r8, r11 ; sum1 = MAC16_16(sum1, x_3, y_4) + BGT celt_pitch_xcorr_edsp_process2_loop4 +celt_pitch_xcorr_edsp_process2_loop_done + ADDS r12, r12, #2 + BLE celt_pitch_xcorr_edsp_process2_1 + LDR r6, [r4], #4 + ; Stall + SMLABB r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_0) + LDR r9, [r5], #4 + SMLABT r11, r6, r8, r11 ; sum1 = MAC16_16(sum1, x_0, y_1) + SUB r12, r12, #2 + SMLATT r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_1, y_1) + MOV r8, r9 + SMLATB r11, r6, r9, r11 ; sum1 = MAC16_16(sum1, x_1, y_2) +celt_pitch_xcorr_edsp_process2_1 + LDRH r6, [r4], #2 + ADDS r12, r12, #1 + ; Stall + SMLABB r10, r6, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_0) + LDRHGT r7, [r4], #2 + SMLABT r11, r6, r8, r11 ; sum1 = MAC16_16(sum1, x_0, y_1) + BLE celt_pitch_xcorr_edsp_process2_done + LDRH r9, [r5], #2 + SMLABT r10, r7, r8, r10 ; sum0 = MAC16_16(sum0, x_0, y_1) + SMLABB r11, r7, r9, r11 ; sum1 = MAC16_16(sum1, x_0, y_2) +celt_pitch_xcorr_edsp_process2_done + ; Restore _x + SUB r4, r4, r3, LSL #1 + ; Restore and advance _y + SUB r5, r5, r3, LSL #1 + ; maxcorr = max(maxcorr, sum0) + CMP r0, r10 + ADD r5, r5, #2 + MOVLT r0, r10 + SUB r1, r1, #2 + ; maxcorr = max(maxcorr, sum1) + CMP r0, r11 + ; xcorr[i] = sum + STR r10, [r2], #4 + MOVLT r0, r11 + STR r11, [r2], #4 +celt_pitch_xcorr_edsp_process1a + ADDS r1, r1, #1 + BLT celt_pitch_xcorr_edsp_done + SUBS r12, r3, #4 + ; r14 = sum = 0 + MOV r14, #0 + BLT celt_pitch_xcorr_edsp_process1a_loop_done + LDR r6, [r4], #4 + LDR r8, [r5], #4 + LDR r7, [r4], #4 + LDR r9, [r5], #4 +celt_pitch_xcorr_edsp_process1a_loop4 + SMLABB r14, r6, r8, r14 ; sum = MAC16_16(sum, x_0, y_0) + SUBS r12, r12, #4 ; j-=4 + SMLATT r14, r6, r8, r14 ; sum = MAC16_16(sum, x_1, y_1) + LDRGE r6, [r4], #4 + SMLABB r14, r7, r9, r14 ; sum = MAC16_16(sum, x_2, y_2) + LDRGE r8, [r5], #4 + SMLATT r14, r7, r9, r14 ; sum = MAC16_16(sum, x_3, y_3) + LDRGE r7, [r4], #4 + LDRGE r9, [r5], #4 + BGE celt_pitch_xcorr_edsp_process1a_loop4 +celt_pitch_xcorr_edsp_process1a_loop_done + ADDS r12, r12, #2 + LDRGE r6, [r4], #4 + LDRGE r8, [r5], #4 + ; Stall + SMLABBGE r14, r6, r8, r14 ; sum = MAC16_16(sum, x_0, y_0) + SUBGE r12, r12, #2 + SMLATTGE r14, r6, r8, r14 ; sum = MAC16_16(sum, x_1, y_1) + ADDS r12, r12, #1 + LDRHGE r6, [r4], #2 + LDRHGE r8, [r5], #2 + ; Stall + SMLABBGE r14, r6, r8, r14 ; sum = MAC16_16(sum, *x, *y) + ; maxcorr = max(maxcorr, sum) + CMP r0, r14 + ; xcorr[i] = sum + STR r14, [r2], #4 + MOVLT r0, r14 +celt_pitch_xcorr_edsp_done + LDMFD sp!, {r4-r11, pc} + ENDP + +ENDIF + +END diff --git a/lib/opus/celt/arm/fft_arm.h b/lib/opus/celt/arm/fft_arm.h new file mode 100644 index 00000000..0b78175f --- /dev/null +++ b/lib/opus/celt/arm/fft_arm.h @@ -0,0 +1,71 @@ +/* Copyright (c) 2015 Xiph.Org Foundation + Written by Viswanath Puttagunta */ +/** + @file fft_arm.h + @brief ARM Neon Intrinsic optimizations for fft using NE10 library + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + + +#if !defined(FFT_ARM_H) +#define FFT_ARM_H + +#include "kiss_fft.h" + +#if defined(HAVE_ARM_NE10) + +int opus_fft_alloc_arm_neon(kiss_fft_state *st); +void opus_fft_free_arm_neon(kiss_fft_state *st); + +void opus_fft_neon(const kiss_fft_state *st, + const kiss_fft_cpx *fin, + kiss_fft_cpx *fout); + +void opus_ifft_neon(const kiss_fft_state *st, + const kiss_fft_cpx *fin, + kiss_fft_cpx *fout); + +#if !defined(OPUS_HAVE_RTCD) +#define OVERRIDE_OPUS_FFT (1) + +#define opus_fft_alloc_arch(_st, arch) \ + ((void)(arch), opus_fft_alloc_arm_neon(_st)) + +#define opus_fft_free_arch(_st, arch) \ + ((void)(arch), opus_fft_free_arm_neon(_st)) + +#define opus_fft(_st, _fin, _fout, arch) \ + ((void)(arch), opus_fft_neon(_st, _fin, _fout)) + +#define opus_ifft(_st, _fin, _fout, arch) \ + ((void)(arch), opus_ifft_neon(_st, _fin, _fout)) + +#endif /* OPUS_HAVE_RTCD */ + +#endif /* HAVE_ARM_NE10 */ + +#endif diff --git a/lib/opus/celt/arm/fixed_arm64.h b/lib/opus/celt/arm/fixed_arm64.h new file mode 100644 index 00000000..c6fbd3db --- /dev/null +++ b/lib/opus/celt/arm/fixed_arm64.h @@ -0,0 +1,35 @@ +/* Copyright (C) 2015 Vidyo */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_ARM64_H +#define FIXED_ARM64_H + +#include + +#undef SIG2WORD16 +#define SIG2WORD16(x) (vqmovns_s32(PSHR32((x), SIG_SHIFT))) + +#endif diff --git a/lib/opus/celt/arm/fixed_armv4.h b/lib/opus/celt/arm/fixed_armv4.h new file mode 100644 index 00000000..d84888a7 --- /dev/null +++ b/lib/opus/celt/arm/fixed_armv4.h @@ -0,0 +1,80 @@ +/* Copyright (C) 2013 Xiph.Org Foundation and contributors */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_ARMv4_H +#define FIXED_ARMv4_H + +/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q16 +static OPUS_INLINE opus_val32 MULT16_32_Q16_armv4(opus_val16 a, opus_val32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#MULT16_32_Q16\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(b),"r"(SHL32(a,16)) + ); + return rd_hi; +} +#define MULT16_32_Q16(a, b) (MULT16_32_Q16_armv4(a, b)) + + +/** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q15 +static OPUS_INLINE opus_val32 MULT16_32_Q15_armv4(opus_val16 a, opus_val32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#MULT16_32_Q15\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(b), "r"(SHL32(a,16)) + ); + /*We intentionally don't OR in the high bit of rd_lo for speed.*/ + return SHL32(rd_hi,1); +} +#define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv4(a, b)) + + +/** 16x32 multiply, followed by a 15-bit shift right and 32-bit add. + b must fit in 31 bits. + Result fits in 32 bits. */ +#undef MAC16_32_Q15 +#define MAC16_32_Q15(c, a, b) ADD32(c, MULT16_32_Q15(a, b)) + +/** 16x32 multiply, followed by a 16-bit shift right and 32-bit add. + Result fits in 32 bits. */ +#undef MAC16_32_Q16 +#define MAC16_32_Q16(c, a, b) ADD32(c, MULT16_32_Q16(a, b)) + +/** 32x32 multiplication, followed by a 31-bit shift right. Results fits in 32 bits */ +#undef MULT32_32_Q31 +#define MULT32_32_Q31(a,b) (opus_val32)((((opus_int64)(a)) * ((opus_int64)(b)))>>31) + +#endif diff --git a/lib/opus/celt/arm/fixed_armv5e.h b/lib/opus/celt/arm/fixed_armv5e.h new file mode 100644 index 00000000..6bf73cba --- /dev/null +++ b/lib/opus/celt/arm/fixed_armv5e.h @@ -0,0 +1,151 @@ +/* Copyright (C) 2007-2009 Xiph.Org Foundation + Copyright (C) 2003-2008 Jean-Marc Valin + Copyright (C) 2007-2008 CSIRO + Copyright (C) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_ARMv5E_H +#define FIXED_ARMv5E_H + +#include "fixed_armv4.h" + +/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q16 +static OPUS_INLINE opus_val32 MULT16_32_Q16_armv5e(opus_val16 a, opus_val32 b) +{ + int res; + __asm__( + "#MULT16_32_Q16\n\t" + "smulwb %0, %1, %2\n\t" + : "=r"(res) + : "r"(b),"r"(a) + ); + return res; +} +#define MULT16_32_Q16(a, b) (MULT16_32_Q16_armv5e(a, b)) + + +/** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */ +#undef MULT16_32_Q15 +static OPUS_INLINE opus_val32 MULT16_32_Q15_armv5e(opus_val16 a, opus_val32 b) +{ + int res; + __asm__( + "#MULT16_32_Q15\n\t" + "smulwb %0, %1, %2\n\t" + : "=r"(res) + : "r"(b), "r"(a) + ); + return SHL32(res,1); +} +#define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv5e(a, b)) + + +/** 16x32 multiply, followed by a 15-bit shift right and 32-bit add. + b must fit in 31 bits. + Result fits in 32 bits. */ +#undef MAC16_32_Q15 +static OPUS_INLINE opus_val32 MAC16_32_Q15_armv5e(opus_val32 c, opus_val16 a, + opus_val32 b) +{ + int res; + __asm__( + "#MAC16_32_Q15\n\t" + "smlawb %0, %1, %2, %3;\n" + : "=r"(res) + : "r"(SHL32(b,1)), "r"(a), "r"(c) + ); + return res; +} +#define MAC16_32_Q15(c, a, b) (MAC16_32_Q15_armv5e(c, a, b)) + +/** 16x32 multiply, followed by a 16-bit shift right and 32-bit add. + Result fits in 32 bits. */ +#undef MAC16_32_Q16 +static OPUS_INLINE opus_val32 MAC16_32_Q16_armv5e(opus_val32 c, opus_val16 a, + opus_val32 b) +{ + int res; + __asm__( + "#MAC16_32_Q16\n\t" + "smlawb %0, %1, %2, %3;\n" + : "=r"(res) + : "r"(b), "r"(a), "r"(c) + ); + return res; +} +#define MAC16_32_Q16(c, a, b) (MAC16_32_Q16_armv5e(c, a, b)) + +/** 16x16 multiply-add where the result fits in 32 bits */ +#undef MAC16_16 +static OPUS_INLINE opus_val32 MAC16_16_armv5e(opus_val32 c, opus_val16 a, + opus_val16 b) +{ + int res; + __asm__( + "#MAC16_16\n\t" + "smlabb %0, %1, %2, %3;\n" + : "=r"(res) + : "r"(a), "r"(b), "r"(c) + ); + return res; +} +#define MAC16_16(c, a, b) (MAC16_16_armv5e(c, a, b)) + +/** 16x16 multiplication where the result fits in 32 bits */ +#undef MULT16_16 +static OPUS_INLINE opus_val32 MULT16_16_armv5e(opus_val16 a, opus_val16 b) +{ + int res; + __asm__( + "#MULT16_16\n\t" + "smulbb %0, %1, %2;\n" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define MULT16_16(a, b) (MULT16_16_armv5e(a, b)) + +#ifdef OPUS_ARM_INLINE_MEDIA + +#undef SIG2WORD16 +static OPUS_INLINE opus_val16 SIG2WORD16_armv6(opus_val32 x) +{ + celt_sig res; + __asm__( + "#SIG2WORD16\n\t" + "ssat %0, #16, %1, ASR #12\n\t" + : "=r"(res) + : "r"(x+2048) + ); + return EXTRACT16(res); +} +#define SIG2WORD16(x) (SIG2WORD16_armv6(x)) + +#endif /* OPUS_ARM_INLINE_MEDIA */ + +#endif diff --git a/lib/opus/celt/arm/kiss_fft_armv4.h b/lib/opus/celt/arm/kiss_fft_armv4.h new file mode 100644 index 00000000..e4faad6f --- /dev/null +++ b/lib/opus/celt/arm/kiss_fft_armv4.h @@ -0,0 +1,121 @@ +/*Copyright (c) 2013, Xiph.Org Foundation and contributors. + + All rights reserved. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE.*/ + +#ifndef KISS_FFT_ARMv4_H +#define KISS_FFT_ARMv4_H + +#if !defined(KISS_FFT_GUTS_H) +#error "This file should only be included from _kiss_fft_guts.h" +#endif + +#ifdef FIXED_POINT + +#undef C_MUL +#define C_MUL(m,a,b) \ + do{ \ + int br__; \ + int bi__; \ + int tt__; \ + __asm__ __volatile__( \ + "#C_MUL\n\t" \ + "ldrsh %[br], [%[bp], #0]\n\t" \ + "ldm %[ap], {r0,r1}\n\t" \ + "ldrsh %[bi], [%[bp], #2]\n\t" \ + "smull %[tt], %[mi], r1, %[br]\n\t" \ + "smlal %[tt], %[mi], r0, %[bi]\n\t" \ + "rsb %[bi], %[bi], #0\n\t" \ + "smull %[br], %[mr], r0, %[br]\n\t" \ + "mov %[tt], %[tt], lsr #15\n\t" \ + "smlal %[br], %[mr], r1, %[bi]\n\t" \ + "orr %[mi], %[tt], %[mi], lsl #17\n\t" \ + "mov %[br], %[br], lsr #15\n\t" \ + "orr %[mr], %[br], %[mr], lsl #17\n\t" \ + : [mr]"=r"((m).r), [mi]"=r"((m).i), \ + [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \ + : [ap]"r"(&(a)), [bp]"r"(&(b)) \ + : "r0", "r1" \ + ); \ + } \ + while(0) + +#undef C_MUL4 +#define C_MUL4(m,a,b) \ + do{ \ + int br__; \ + int bi__; \ + int tt__; \ + __asm__ __volatile__( \ + "#C_MUL4\n\t" \ + "ldrsh %[br], [%[bp], #0]\n\t" \ + "ldm %[ap], {r0,r1}\n\t" \ + "ldrsh %[bi], [%[bp], #2]\n\t" \ + "smull %[tt], %[mi], r1, %[br]\n\t" \ + "smlal %[tt], %[mi], r0, %[bi]\n\t" \ + "rsb %[bi], %[bi], #0\n\t" \ + "smull %[br], %[mr], r0, %[br]\n\t" \ + "mov %[tt], %[tt], lsr #17\n\t" \ + "smlal %[br], %[mr], r1, %[bi]\n\t" \ + "orr %[mi], %[tt], %[mi], lsl #15\n\t" \ + "mov %[br], %[br], lsr #17\n\t" \ + "orr %[mr], %[br], %[mr], lsl #15\n\t" \ + : [mr]"=r"((m).r), [mi]"=r"((m).i), \ + [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \ + : [ap]"r"(&(a)), [bp]"r"(&(b)) \ + : "r0", "r1" \ + ); \ + } \ + while(0) + +#undef C_MULC +#define C_MULC(m,a,b) \ + do{ \ + int br__; \ + int bi__; \ + int tt__; \ + __asm__ __volatile__( \ + "#C_MULC\n\t" \ + "ldrsh %[br], [%[bp], #0]\n\t" \ + "ldm %[ap], {r0,r1}\n\t" \ + "ldrsh %[bi], [%[bp], #2]\n\t" \ + "smull %[tt], %[mr], r0, %[br]\n\t" \ + "smlal %[tt], %[mr], r1, %[bi]\n\t" \ + "rsb %[bi], %[bi], #0\n\t" \ + "smull %[br], %[mi], r1, %[br]\n\t" \ + "mov %[tt], %[tt], lsr #15\n\t" \ + "smlal %[br], %[mi], r0, %[bi]\n\t" \ + "orr %[mr], %[tt], %[mr], lsl #17\n\t" \ + "mov %[br], %[br], lsr #15\n\t" \ + "orr %[mi], %[br], %[mi], lsl #17\n\t" \ + : [mr]"=r"((m).r), [mi]"=r"((m).i), \ + [br]"=&r"(br__), [bi]"=r"(bi__), [tt]"=r"(tt__) \ + : [ap]"r"(&(a)), [bp]"r"(&(b)) \ + : "r0", "r1" \ + ); \ + } \ + while(0) + +#endif /* FIXED_POINT */ + +#endif /* KISS_FFT_ARMv4_H */ diff --git a/lib/opus/celt/arm/kiss_fft_armv5e.h b/lib/opus/celt/arm/kiss_fft_armv5e.h new file mode 100644 index 00000000..9eca183d --- /dev/null +++ b/lib/opus/celt/arm/kiss_fft_armv5e.h @@ -0,0 +1,118 @@ +/*Copyright (c) 2013, Xiph.Org Foundation and contributors. + + All rights reserved. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE.*/ + +#ifndef KISS_FFT_ARMv5E_H +#define KISS_FFT_ARMv5E_H + +#if !defined(KISS_FFT_GUTS_H) +#error "This file should only be included from _kiss_fft_guts.h" +#endif + +#ifdef FIXED_POINT + +#if defined(__thumb__)||defined(__thumb2__) +#define LDRD_CONS "Q" +#else +#define LDRD_CONS "Uq" +#endif + +#undef C_MUL +#define C_MUL(m,a,b) \ + do{ \ + int mr1__; \ + int mr2__; \ + int mi__; \ + long long aval__; \ + int bval__; \ + __asm__( \ + "#C_MUL\n\t" \ + "ldrd %[aval], %H[aval], %[ap]\n\t" \ + "ldr %[bval], %[bp]\n\t" \ + "smulwb %[mi], %H[aval], %[bval]\n\t" \ + "smulwb %[mr1], %[aval], %[bval]\n\t" \ + "smulwt %[mr2], %H[aval], %[bval]\n\t" \ + "smlawt %[mi], %[aval], %[bval], %[mi]\n\t" \ + : [mr1]"=r"(mr1__), [mr2]"=r"(mr2__), [mi]"=r"(mi__), \ + [aval]"=&r"(aval__), [bval]"=r"(bval__) \ + : [ap]LDRD_CONS(a), [bp]"m"(b) \ + ); \ + (m).r = SHL32(SUB32(mr1__, mr2__), 1); \ + (m).i = SHL32(mi__, 1); \ + } \ + while(0) + +#undef C_MUL4 +#define C_MUL4(m,a,b) \ + do{ \ + int mr1__; \ + int mr2__; \ + int mi__; \ + long long aval__; \ + int bval__; \ + __asm__( \ + "#C_MUL4\n\t" \ + "ldrd %[aval], %H[aval], %[ap]\n\t" \ + "ldr %[bval], %[bp]\n\t" \ + "smulwb %[mi], %H[aval], %[bval]\n\t" \ + "smulwb %[mr1], %[aval], %[bval]\n\t" \ + "smulwt %[mr2], %H[aval], %[bval]\n\t" \ + "smlawt %[mi], %[aval], %[bval], %[mi]\n\t" \ + : [mr1]"=r"(mr1__), [mr2]"=r"(mr2__), [mi]"=r"(mi__), \ + [aval]"=&r"(aval__), [bval]"=r"(bval__) \ + : [ap]LDRD_CONS(a), [bp]"m"(b) \ + ); \ + (m).r = SHR32(SUB32(mr1__, mr2__), 1); \ + (m).i = SHR32(mi__, 1); \ + } \ + while(0) + +#undef C_MULC +#define C_MULC(m,a,b) \ + do{ \ + int mr__; \ + int mi1__; \ + int mi2__; \ + long long aval__; \ + int bval__; \ + __asm__( \ + "#C_MULC\n\t" \ + "ldrd %[aval], %H[aval], %[ap]\n\t" \ + "ldr %[bval], %[bp]\n\t" \ + "smulwb %[mr], %[aval], %[bval]\n\t" \ + "smulwb %[mi1], %H[aval], %[bval]\n\t" \ + "smulwt %[mi2], %[aval], %[bval]\n\t" \ + "smlawt %[mr], %H[aval], %[bval], %[mr]\n\t" \ + : [mr]"=r"(mr__), [mi1]"=r"(mi1__), [mi2]"=r"(mi2__), \ + [aval]"=&r"(aval__), [bval]"=r"(bval__) \ + : [ap]LDRD_CONS(a), [bp]"m"(b) \ + ); \ + (m).r = SHL32(mr__, 1); \ + (m).i = SHL32(SUB32(mi1__, mi2__), 1); \ + } \ + while(0) + +#endif /* FIXED_POINT */ + +#endif /* KISS_FFT_GUTS_H */ diff --git a/lib/opus/celt/arm/mdct_arm.h b/lib/opus/celt/arm/mdct_arm.h new file mode 100644 index 00000000..14200bac --- /dev/null +++ b/lib/opus/celt/arm/mdct_arm.h @@ -0,0 +1,59 @@ +/* Copyright (c) 2015 Xiph.Org Foundation + Written by Viswanath Puttagunta */ +/** + @file arm_mdct.h + @brief ARM Neon Intrinsic optimizations for mdct using NE10 library + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(MDCT_ARM_H) +#define MDCT_ARM_H + +#include "mdct.h" + +#if defined(HAVE_ARM_NE10) +/** Compute a forward MDCT and scale by 4/N, trashes the input array */ +void clt_mdct_forward_neon(const mdct_lookup *l, kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, int overlap, + int shift, int stride, int arch); + +void clt_mdct_backward_neon(const mdct_lookup *l, kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, int overlap, + int shift, int stride, int arch); + +#if !defined(OPUS_HAVE_RTCD) +#define OVERRIDE_OPUS_MDCT (1) +#define clt_mdct_forward(_l, _in, _out, _window, _int, _shift, _stride, _arch) \ + clt_mdct_forward_neon(_l, _in, _out, _window, _int, _shift, _stride, _arch) +#define clt_mdct_backward(_l, _in, _out, _window, _int, _shift, _stride, _arch) \ + clt_mdct_backward_neon(_l, _in, _out, _window, _int, _shift, _stride, _arch) +#endif /* OPUS_HAVE_RTCD */ +#endif /* HAVE_ARM_NE10 */ + +#endif diff --git a/lib/opus/celt/arm/pitch_arm.h b/lib/opus/celt/arm/pitch_arm.h new file mode 100644 index 00000000..bed8b04e --- /dev/null +++ b/lib/opus/celt/arm/pitch_arm.h @@ -0,0 +1,160 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(PITCH_ARM_H) +# define PITCH_ARM_H + +# include "armcpu.h" + +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +opus_val32 celt_inner_prod_neon(const opus_val16 *x, const opus_val16 *y, int N); +void dual_inner_prod_neon(const opus_val16 *x, const opus_val16 *y01, + const opus_val16 *y02, int N, opus_val32 *xy1, opus_val32 *xy2); + +# if !defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_PRESUME_NEON) +# define OVERRIDE_CELT_INNER_PROD (1) +# define OVERRIDE_DUAL_INNER_PROD (1) +# define celt_inner_prod(x, y, N, arch) ((void)(arch), PRESUME_NEON(celt_inner_prod)(x, y, N)) +# define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) ((void)(arch), PRESUME_NEON(dual_inner_prod)(x, y01, y02, N, xy1, xy2)) +# endif +# endif + +# if !defined(OVERRIDE_CELT_INNER_PROD) +# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)) +extern opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x, const opus_val16 *y, int N); +# define OVERRIDE_CELT_INNER_PROD (1) +# define celt_inner_prod(x, y, N, arch) ((*CELT_INNER_PROD_IMPL[(arch)&OPUS_ARCHMASK])(x, y, N)) +# elif defined(OPUS_ARM_PRESUME_NEON_INTR) +# define OVERRIDE_CELT_INNER_PROD (1) +# define celt_inner_prod(x, y, N, arch) ((void)(arch), celt_inner_prod_neon(x, y, N)) +# endif +# endif + +# if !defined(OVERRIDE_DUAL_INNER_PROD) +# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)) +extern void (*const DUAL_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x, + const opus_val16 *y01, const opus_val16 *y02, int N, opus_val32 *xy1, opus_val32 *xy2); +# define OVERRIDE_DUAL_INNER_PROD (1) +# define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) ((*DUAL_INNER_PROD_IMPL[(arch)&OPUS_ARCHMASK])(x, y01, y02, N, xy1, xy2)) +# elif defined(OPUS_ARM_PRESUME_NEON_INTR) +# define OVERRIDE_DUAL_INNER_PROD (1) +# define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) ((void)(arch), dual_inner_prod_neon(x, y01, y02, N, xy1, xy2)) +# endif +# endif + +# if defined(FIXED_POINT) + +# if defined(OPUS_ARM_MAY_HAVE_NEON) +opus_val32 celt_pitch_xcorr_neon(const opus_val16 *_x, const opus_val16 *_y, + opus_val32 *xcorr, int len, int max_pitch, int arch); +# endif + +# if defined(OPUS_ARM_MAY_HAVE_MEDIA) +# define celt_pitch_xcorr_media MAY_HAVE_EDSP(celt_pitch_xcorr) +# endif + +# if defined(OPUS_ARM_MAY_HAVE_EDSP) +opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y, + opus_val32 *xcorr, int len, int max_pitch, int arch); +# endif + +# if defined(OPUS_HAVE_RTCD) && \ + ((defined(OPUS_ARM_MAY_HAVE_NEON) && !defined(OPUS_ARM_PRESUME_NEON)) || \ + (defined(OPUS_ARM_MAY_HAVE_MEDIA) && !defined(OPUS_ARM_PRESUME_MEDIA)) || \ + (defined(OPUS_ARM_MAY_HAVE_EDSP) && !defined(OPUS_ARM_PRESUME_EDSP))) +extern opus_val32 +(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *, + const opus_val16 *, opus_val32 *, int, int, int); +# define OVERRIDE_PITCH_XCORR (1) +# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \ + ((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \ + xcorr, len, max_pitch, arch)) + +# elif defined(OPUS_ARM_PRESUME_EDSP) || \ + defined(OPUS_ARM_PRESUME_MEDIA) || \ + defined(OPUS_ARM_PRESUME_NEON) +# define OVERRIDE_PITCH_XCORR (1) +# define celt_pitch_xcorr (PRESUME_NEON(celt_pitch_xcorr)) + +# endif + +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +void xcorr_kernel_neon_fixed( + const opus_val16 *x, + const opus_val16 *y, + opus_val32 sum[4], + int len); +# endif + +# if defined(OPUS_HAVE_RTCD) && \ + (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)) + +extern void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])( + const opus_val16 *x, + const opus_val16 *y, + opus_val32 sum[4], + int len); + +# define OVERRIDE_XCORR_KERNEL (1) +# define xcorr_kernel(x, y, sum, len, arch) \ + ((*XCORR_KERNEL_IMPL[(arch) & OPUS_ARCHMASK])(x, y, sum, len)) + +# elif defined(OPUS_ARM_PRESUME_NEON_INTR) +# define OVERRIDE_XCORR_KERNEL (1) +# define xcorr_kernel(x, y, sum, len, arch) \ + ((void)arch, xcorr_kernel_neon_fixed(x, y, sum, len)) + +# endif + +#else /* Start !FIXED_POINT */ +/* Float case */ +#if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +void celt_pitch_xcorr_float_neon(const opus_val16 *_x, const opus_val16 *_y, + opus_val32 *xcorr, int len, int max_pitch, int arch); +#endif + +# if defined(OPUS_HAVE_RTCD) && \ + (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)) +extern void +(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *, + const opus_val16 *, opus_val32 *, int, int, int); + +# define OVERRIDE_PITCH_XCORR (1) +# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \ + ((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \ + xcorr, len, max_pitch, arch)) + +# elif defined(OPUS_ARM_PRESUME_NEON_INTR) + +# define OVERRIDE_PITCH_XCORR (1) +# define celt_pitch_xcorr celt_pitch_xcorr_float_neon + +# endif + +#endif /* end !FIXED_POINT */ + +#endif diff --git a/lib/opus/celt/arm/pitch_neon_intr.c b/lib/opus/celt/arm/pitch_neon_intr.c new file mode 100644 index 00000000..35cc46e2 --- /dev/null +++ b/lib/opus/celt/arm/pitch_neon_intr.c @@ -0,0 +1,281 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "pitch.h" + +#ifdef FIXED_POINT + +opus_val32 celt_inner_prod_neon(const opus_val16 *x, const opus_val16 *y, int N) +{ + int i; + opus_val32 xy; + int16x8_t x_s16x8, y_s16x8; + int32x4_t xy_s32x4 = vdupq_n_s32(0); + int64x2_t xy_s64x2; + int64x1_t xy_s64x1; + + for (i = 0; i < N - 7; i += 8) { + x_s16x8 = vld1q_s16(&x[i]); + y_s16x8 = vld1q_s16(&y[i]); + xy_s32x4 = vmlal_s16(xy_s32x4, vget_low_s16 (x_s16x8), vget_low_s16 (y_s16x8)); + xy_s32x4 = vmlal_s16(xy_s32x4, vget_high_s16(x_s16x8), vget_high_s16(y_s16x8)); + } + + if (N - i >= 4) { + const int16x4_t x_s16x4 = vld1_s16(&x[i]); + const int16x4_t y_s16x4 = vld1_s16(&y[i]); + xy_s32x4 = vmlal_s16(xy_s32x4, x_s16x4, y_s16x4); + i += 4; + } + + xy_s64x2 = vpaddlq_s32(xy_s32x4); + xy_s64x1 = vadd_s64(vget_low_s64(xy_s64x2), vget_high_s64(xy_s64x2)); + xy = vget_lane_s32(vreinterpret_s32_s64(xy_s64x1), 0); + + for (; i < N; i++) { + xy = MAC16_16(xy, x[i], y[i]); + } + +#ifdef OPUS_CHECK_ASM + celt_assert(celt_inner_prod_c(x, y, N) == xy); +#endif + + return xy; +} + +void dual_inner_prod_neon(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02, + int N, opus_val32 *xy1, opus_val32 *xy2) +{ + int i; + opus_val32 xy01, xy02; + int16x8_t x_s16x8, y01_s16x8, y02_s16x8; + int32x4_t xy01_s32x4 = vdupq_n_s32(0); + int32x4_t xy02_s32x4 = vdupq_n_s32(0); + int64x2_t xy01_s64x2, xy02_s64x2; + int64x1_t xy01_s64x1, xy02_s64x1; + + for (i = 0; i < N - 7; i += 8) { + x_s16x8 = vld1q_s16(&x[i]); + y01_s16x8 = vld1q_s16(&y01[i]); + y02_s16x8 = vld1q_s16(&y02[i]); + xy01_s32x4 = vmlal_s16(xy01_s32x4, vget_low_s16 (x_s16x8), vget_low_s16 (y01_s16x8)); + xy02_s32x4 = vmlal_s16(xy02_s32x4, vget_low_s16 (x_s16x8), vget_low_s16 (y02_s16x8)); + xy01_s32x4 = vmlal_s16(xy01_s32x4, vget_high_s16(x_s16x8), vget_high_s16(y01_s16x8)); + xy02_s32x4 = vmlal_s16(xy02_s32x4, vget_high_s16(x_s16x8), vget_high_s16(y02_s16x8)); + } + + if (N - i >= 4) { + const int16x4_t x_s16x4 = vld1_s16(&x[i]); + const int16x4_t y01_s16x4 = vld1_s16(&y01[i]); + const int16x4_t y02_s16x4 = vld1_s16(&y02[i]); + xy01_s32x4 = vmlal_s16(xy01_s32x4, x_s16x4, y01_s16x4); + xy02_s32x4 = vmlal_s16(xy02_s32x4, x_s16x4, y02_s16x4); + i += 4; + } + + xy01_s64x2 = vpaddlq_s32(xy01_s32x4); + xy02_s64x2 = vpaddlq_s32(xy02_s32x4); + xy01_s64x1 = vadd_s64(vget_low_s64(xy01_s64x2), vget_high_s64(xy01_s64x2)); + xy02_s64x1 = vadd_s64(vget_low_s64(xy02_s64x2), vget_high_s64(xy02_s64x2)); + xy01 = vget_lane_s32(vreinterpret_s32_s64(xy01_s64x1), 0); + xy02 = vget_lane_s32(vreinterpret_s32_s64(xy02_s64x1), 0); + + for (; i < N; i++) { + xy01 = MAC16_16(xy01, x[i], y01[i]); + xy02 = MAC16_16(xy02, x[i], y02[i]); + } + *xy1 = xy01; + *xy2 = xy02; + +#ifdef OPUS_CHECK_ASM + { + opus_val32 xy1_c, xy2_c; + dual_inner_prod_c(x, y01, y02, N, &xy1_c, &xy2_c); + celt_assert(xy1_c == *xy1); + celt_assert(xy2_c == *xy2); + } +#endif +} + +#else /* !FIXED_POINT */ + +/* ========================================================================== */ + +#ifdef OPUS_CHECK_ASM + +/* This part of code simulates floating-point NEON operations. */ + +/* celt_inner_prod_neon_float_c_simulation() simulates the floating-point */ +/* operations of celt_inner_prod_neon(), and both functions should have bit */ +/* exact output. */ +static opus_val32 celt_inner_prod_neon_float_c_simulation(const opus_val16 *x, const opus_val16 *y, float *err, int N) +{ + int i; + *err = 0; + opus_val32 xy, xy0 = 0, xy1 = 0, xy2 = 0, xy3 = 0; + for (i = 0; i < N - 3; i += 4) { + xy0 = MAC16_16(xy0, x[i + 0], y[i + 0]); + xy1 = MAC16_16(xy1, x[i + 1], y[i + 1]); + xy2 = MAC16_16(xy2, x[i + 2], y[i + 2]); + xy3 = MAC16_16(xy3, x[i + 3], y[i + 3]); + *err += ABS32(xy0)+ABS32(xy1)+ABS32(xy2)+ABS32(xy3); + } + xy0 += xy2; + xy1 += xy3; + xy = xy0 + xy1; + *err += ABS32(xy1)+ABS32(xy0)+ABS32(xy); + for (; i < N; i++) { + xy = MAC16_16(xy, x[i], y[i]); + *err += ABS32(xy); + } + *err = *err*2e-7 + N*1e-37; + return xy; +} + +/* dual_inner_prod_neon_float_c_simulation() simulates the floating-point */ +/* operations of dual_inner_prod_neon(), and both functions should have bit */ +/* exact output. */ +static void dual_inner_prod_neon_float_c_simulation(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02, + int N, opus_val32 *xy1, opus_val32 *xy2, float *err) +{ + *xy1 = celt_inner_prod_neon_float_c_simulation(x, y01, &err[0], N); + *xy2 = celt_inner_prod_neon_float_c_simulation(x, y02, &err[1], N); +} + +#endif /* OPUS_CHECK_ASM */ + +/* ========================================================================== */ + +opus_val32 celt_inner_prod_neon(const opus_val16 *x, const opus_val16 *y, int N) +{ + int i; + opus_val32 xy; + float32x4_t xy_f32x4 = vdupq_n_f32(0); + float32x2_t xy_f32x2; + + for (i = 0; i < N - 7; i += 8) { + float32x4_t x_f32x4, y_f32x4; + x_f32x4 = vld1q_f32(&x[i]); + y_f32x4 = vld1q_f32(&y[i]); + xy_f32x4 = vmlaq_f32(xy_f32x4, x_f32x4, y_f32x4); + x_f32x4 = vld1q_f32(&x[i + 4]); + y_f32x4 = vld1q_f32(&y[i + 4]); + xy_f32x4 = vmlaq_f32(xy_f32x4, x_f32x4, y_f32x4); + } + + if (N - i >= 4) { + const float32x4_t x_f32x4 = vld1q_f32(&x[i]); + const float32x4_t y_f32x4 = vld1q_f32(&y[i]); + xy_f32x4 = vmlaq_f32(xy_f32x4, x_f32x4, y_f32x4); + i += 4; + } + + xy_f32x2 = vadd_f32(vget_low_f32(xy_f32x4), vget_high_f32(xy_f32x4)); + xy_f32x2 = vpadd_f32(xy_f32x2, xy_f32x2); + xy = vget_lane_f32(xy_f32x2, 0); + + for (; i < N; i++) { + xy = MAC16_16(xy, x[i], y[i]); + } + +#ifdef OPUS_CHECK_ASM + { + float err, res; + res = celt_inner_prod_neon_float_c_simulation(x, y, &err, N); + /*if (ABS32(res - xy) > err) fprintf(stderr, "%g %g %g\n", res, xy, err);*/ + celt_assert(ABS32(res - xy) <= err); + } +#endif + + return xy; +} + +void dual_inner_prod_neon(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02, + int N, opus_val32 *xy1, opus_val32 *xy2) +{ + int i; + opus_val32 xy01, xy02; + float32x4_t xy01_f32x4 = vdupq_n_f32(0); + float32x4_t xy02_f32x4 = vdupq_n_f32(0); + float32x2_t xy01_f32x2, xy02_f32x2; + + for (i = 0; i < N - 7; i += 8) { + float32x4_t x_f32x4, y01_f32x4, y02_f32x4; + x_f32x4 = vld1q_f32(&x[i]); + y01_f32x4 = vld1q_f32(&y01[i]); + y02_f32x4 = vld1q_f32(&y02[i]); + xy01_f32x4 = vmlaq_f32(xy01_f32x4, x_f32x4, y01_f32x4); + xy02_f32x4 = vmlaq_f32(xy02_f32x4, x_f32x4, y02_f32x4); + x_f32x4 = vld1q_f32(&x[i + 4]); + y01_f32x4 = vld1q_f32(&y01[i + 4]); + y02_f32x4 = vld1q_f32(&y02[i + 4]); + xy01_f32x4 = vmlaq_f32(xy01_f32x4, x_f32x4, y01_f32x4); + xy02_f32x4 = vmlaq_f32(xy02_f32x4, x_f32x4, y02_f32x4); + } + + if (N - i >= 4) { + const float32x4_t x_f32x4 = vld1q_f32(&x[i]); + const float32x4_t y01_f32x4 = vld1q_f32(&y01[i]); + const float32x4_t y02_f32x4 = vld1q_f32(&y02[i]); + xy01_f32x4 = vmlaq_f32(xy01_f32x4, x_f32x4, y01_f32x4); + xy02_f32x4 = vmlaq_f32(xy02_f32x4, x_f32x4, y02_f32x4); + i += 4; + } + + xy01_f32x2 = vadd_f32(vget_low_f32(xy01_f32x4), vget_high_f32(xy01_f32x4)); + xy02_f32x2 = vadd_f32(vget_low_f32(xy02_f32x4), vget_high_f32(xy02_f32x4)); + xy01_f32x2 = vpadd_f32(xy01_f32x2, xy01_f32x2); + xy02_f32x2 = vpadd_f32(xy02_f32x2, xy02_f32x2); + xy01 = vget_lane_f32(xy01_f32x2, 0); + xy02 = vget_lane_f32(xy02_f32x2, 0); + + for (; i < N; i++) { + xy01 = MAC16_16(xy01, x[i], y01[i]); + xy02 = MAC16_16(xy02, x[i], y02[i]); + } + *xy1 = xy01; + *xy2 = xy02; + +#ifdef OPUS_CHECK_ASM + { + opus_val32 xy1_c, xy2_c; + float err[2]; + dual_inner_prod_neon_float_c_simulation(x, y01, y02, N, &xy1_c, &xy2_c, err); + /*if (ABS32(xy1_c - *xy1) > err[0]) fprintf(stderr, "dual1 fail: %g %g %g\n", xy1_c, *xy1, err[0]); + if (ABS32(xy2_c - *xy2) > err[1]) fprintf(stderr, "dual2 fail: %g %g %g\n", xy2_c, *xy2, err[1]);*/ + celt_assert(ABS32(xy1_c - *xy1) <= err[0]); + celt_assert(ABS32(xy2_c - *xy2) <= err[1]); + } +#endif +} + +#endif /* FIXED_POINT */ diff --git a/lib/opus/celt/bands.c b/lib/opus/celt/bands.c new file mode 100644 index 00000000..5320ffab --- /dev/null +++ b/lib/opus/celt/bands.c @@ -0,0 +1,1673 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2008-2009 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "bands.h" +#include "modes.h" +#include "vq.h" +#include "cwrs.h" +#include "stack_alloc.h" +#include "os_support.h" +#include "mathops.h" +#include "rate.h" +#include "quant_bands.h" +#include "pitch.h" + +int hysteresis_decision(opus_val16 val, const opus_val16 *thresholds, const opus_val16 *hysteresis, int N, int prev) +{ + int i; + for (i=0;iprev && val < thresholds[prev]+hysteresis[prev]) + i=prev; + if (i thresholds[prev-1]-hysteresis[prev-1]) + i=prev; + return i; +} + +opus_uint32 celt_lcg_rand(opus_uint32 seed) +{ + return 1664525 * seed + 1013904223; +} + +/* This is a cos() approximation designed to be bit-exact on any platform. Bit exactness + with this approximation is important because it has an impact on the bit allocation */ +opus_int16 bitexact_cos(opus_int16 x) +{ + opus_int32 tmp; + opus_int16 x2; + tmp = (4096+((opus_int32)(x)*(x)))>>13; + celt_sig_assert(tmp<=32767); + x2 = tmp; + x2 = (32767-x2) + FRAC_MUL16(x2, (-7651 + FRAC_MUL16(x2, (8277 + FRAC_MUL16(-626, x2))))); + celt_sig_assert(x2<=32766); + return 1+x2; +} + +int bitexact_log2tan(int isin,int icos) +{ + int lc; + int ls; + lc=EC_ILOG(icos); + ls=EC_ILOG(isin); + icos<<=15-lc; + isin<<=15-ls; + return (ls-lc)*(1<<11) + +FRAC_MUL16(isin, FRAC_MUL16(isin, -2597) + 7932) + -FRAC_MUL16(icos, FRAC_MUL16(icos, -2597) + 7932); +} + +#ifdef FIXED_POINT +/* Compute the amplitude (sqrt energy) in each of the bands */ +void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM, int arch) +{ + int i, c, N; + const opus_int16 *eBands = m->eBands; + (void)arch; + N = m->shortMdctSize< 0) + { + int shift = celt_ilog2(maxval) - 14 + (((m->logN[i]>>BITRES)+LM+1)>>1); + j=eBands[i]<0) + { + do { + sum = MAC16_16(sum, EXTRACT16(SHR32(X[j+c*N],shift)), + EXTRACT16(SHR32(X[j+c*N],shift))); + } while (++jnbEBands] = EPSILON+VSHR32(EXTEND32(celt_sqrt(sum)),-shift); + } else { + bandE[i+c*m->nbEBands] = EPSILON; + } + /*printf ("%f ", bandE[i+c*m->nbEBands]);*/ + } + } while (++ceBands; + N = M*m->shortMdctSize; + c=0; do { + i=0; do { + opus_val16 g; + int j,shift; + opus_val16 E; + shift = celt_zlog2(bandE[i+c*m->nbEBands])-13; + E = VSHR32(bandE[i+c*m->nbEBands], shift); + g = EXTRACT16(celt_rcp(SHL32(E,3))); + j=M*eBands[i]; do { + X[j+c*N] = MULT16_16_Q15(VSHR32(freq[j+c*N],shift-1),g); + } while (++jeBands; + N = m->shortMdctSize<nbEBands] = celt_sqrt(sum); + /*printf ("%f ", bandE[i+c*m->nbEBands]);*/ + } + } while (++ceBands; + N = M*m->shortMdctSize; + c=0; do { + for (i=0;inbEBands]); + for (j=M*eBands[i];jeBands; + N = M*m->shortMdctSize; + bound = M*eBands[end]; + if (downsample!=1) + bound = IMIN(bound, N/downsample); + if (silence) + { + bound = 0; + start = end = 0; + } + f = freq; + x = X+M*eBands[start]; + for (i=0;i>DB_SHIFT); + if (shift>31) + { + shift=0; + g=0; + } else { + /* Handle the fractional part. */ + g = celt_exp2_frac(lg&((1< 16384 we'd be likely to overflow, so we're + capping the gain here, which is equivalent to a cap of 18 on lg. + This shouldn't trigger unless the bitstream is already corrupted. */ + if (shift <= -2) + { + g = 16384; + shift = -2; + } + do { + *f++ = SHL32(MULT16_16(*x++, g), -shift); + } while (++jeBands[i+1]-m->eBands[i]; + /* depth in 1/8 bits */ + celt_sig_assert(pulses[i]>=0); + depth = celt_udiv(1+pulses[i], (m->eBands[i+1]-m->eBands[i]))>>LM; + +#ifdef FIXED_POINT + thresh32 = SHR32(celt_exp2(-SHL16(depth, 10-BITRES)),1); + thresh = MULT16_32_Q15(QCONST16(0.5f, 15), MIN32(32767,thresh32)); + { + opus_val32 t; + t = N0<>1; + t = SHL32(t, (7-shift)<<1); + sqrt_1 = celt_rsqrt_norm(t); + } +#else + thresh = .5f*celt_exp2(-.125f*depth); + sqrt_1 = celt_rsqrt(N0<nbEBands+i]; + prev2 = prev2logE[c*m->nbEBands+i]; + if (C==1) + { + prev1 = MAX16(prev1,prev1logE[m->nbEBands+i]); + prev2 = MAX16(prev2,prev2logE[m->nbEBands+i]); + } + Ediff = EXTEND32(logE[c*m->nbEBands+i])-EXTEND32(MIN16(prev1,prev2)); + Ediff = MAX32(0, Ediff); + +#ifdef FIXED_POINT + if (Ediff < 16384) + { + opus_val32 r32 = SHR32(celt_exp2(-EXTRACT16(Ediff)),1); + r = 2*MIN16(16383,r32); + } else { + r = 0; + } + if (LM==3) + r = MULT16_16_Q14(23170, MIN32(23169, r)); + r = SHR16(MIN16(thresh, r),1); + r = SHR32(MULT16_16_Q15(sqrt_1, r),shift); +#else + /* r needs to be multiplied by 2 or 2*sqrt(2) depending on LM because + short blocks don't have the same energy as long */ + r = 2.f*celt_exp2(-Ediff); + if (LM==3) + r *= 1.41421356f; + r = MIN16(thresh, r); + r = r*sqrt_1; +#endif + X = X_+c*size+(m->eBands[i]<nbEBands]))-13; +#endif + left = VSHR32(bandE[i],shift); + right = VSHR32(bandE[i+m->nbEBands],shift); + norm = EPSILON + celt_sqrt(EPSILON+MULT16_16(left,left)+MULT16_16(right,right)); + a1 = DIV32_16(SHL32(EXTEND32(left),14),norm); + a2 = DIV32_16(SHL32(EXTEND32(right),14),norm); + for (j=0;j>1; + kr = celt_ilog2(Er)>>1; +#endif + t = VSHR32(El, (kl-7)<<1); + lgain = celt_rsqrt_norm(t); + t = VSHR32(Er, (kr-7)<<1); + rgain = celt_rsqrt_norm(t); + +#ifdef FIXED_POINT + if (kl < 7) + kl = 7; + if (kr < 7) + kr = 7; +#endif + + for (j=0;jeBands; + int decision; + int hf_sum=0; + + celt_assert(end>0); + + N0 = M*m->shortMdctSize; + + if (M*(eBands[end]-eBands[end-1]) <= 8) + return SPREAD_NONE; + c=0; do { + for (i=0;im->nbEBands-4) + hf_sum += celt_udiv(32*(tcount[1]+tcount[0]), N); + tmp = (2*tcount[2] >= N) + (2*tcount[1] >= N) + (2*tcount[0] >= N); + sum += tmp*spread_weight[i]; + nbBands+=spread_weight[i]; + } + } while (++cnbEBands+end)); + *hf_average = (*hf_average+hf_sum)>>1; + hf_sum = *hf_average; + if (*tapset_decision==2) + hf_sum += 4; + else if (*tapset_decision==0) + hf_sum -= 4; + if (hf_sum > 22) + *tapset_decision=2; + else if (hf_sum > 18) + *tapset_decision=1; + else + *tapset_decision=0; + } + /*printf("%d %d %d\n", hf_sum, *hf_average, *tapset_decision);*/ + celt_assert(nbBands>0); /* end has to be non-zero */ + celt_assert(sum>=0); + sum = celt_udiv((opus_int32)sum<<8, nbBands); + /* Recursive averaging */ + sum = (sum+*average)>>1; + *average = sum; + /* Hysteresis */ + sum = (3*sum + (((3-last_decision)<<7) + 64) + 2)>>2; + if (sum < 80) + { + decision = SPREAD_AGGRESSIVE; + } else if (sum < 256) + { + decision = SPREAD_NORMAL; + } else if (sum < 384) + { + decision = SPREAD_LIGHT; + } else { + decision = SPREAD_NONE; + } +#ifdef FUZZING + decision = rand()&0x3; + *tapset_decision=rand()%3; +#endif + return decision; +} + +/* Indexing table for converting from natural Hadamard to ordery Hadamard + This is essentially a bit-reversed Gray, on top of which we've added + an inversion of the order because we want the DC at the end rather than + the beginning. The lines are for N=2, 4, 8, 16 */ +static const int ordery_table[] = { + 1, 0, + 3, 0, 2, 1, + 7, 0, 4, 3, 6, 1, 5, 2, + 15, 0, 8, 7, 12, 3, 11, 4, 14, 1, 9, 6, 13, 2, 10, 5, +}; + +static void deinterleave_hadamard(celt_norm *X, int N0, int stride, int hadamard) +{ + int i,j; + VARDECL(celt_norm, tmp); + int N; + SAVE_STACK; + N = N0*stride; + ALLOC(tmp, N, celt_norm); + celt_assert(stride>0); + if (hadamard) + { + const int *ordery = ordery_table+stride-2; + for (i=0;i>= 1; + for (i=0;i>1)) { + qn = 1; + } else { + qn = exp2_table8[qb&0x7]>>(14-(qb>>BITRES)); + qn = (qn+1)>>1<<1; + } + celt_assert(qn <= 256); + return qn; +} + +struct band_ctx { + int encode; + int resynth; + const CELTMode *m; + int i; + int intensity; + int spread; + int tf_change; + ec_ctx *ec; + opus_int32 remaining_bits; + const celt_ener *bandE; + opus_uint32 seed; + int arch; + int theta_round; + int disable_inv; + int avoid_split_noise; +}; + +struct split_ctx { + int inv; + int imid; + int iside; + int delta; + int itheta; + int qalloc; +}; + +static void compute_theta(struct band_ctx *ctx, struct split_ctx *sctx, + celt_norm *X, celt_norm *Y, int N, int *b, int B, int B0, + int LM, + int stereo, int *fill) +{ + int qn; + int itheta=0; + int delta; + int imid, iside; + int qalloc; + int pulse_cap; + int offset; + opus_int32 tell; + int inv=0; + int encode; + const CELTMode *m; + int i; + int intensity; + ec_ctx *ec; + const celt_ener *bandE; + + encode = ctx->encode; + m = ctx->m; + i = ctx->i; + intensity = ctx->intensity; + ec = ctx->ec; + bandE = ctx->bandE; + + /* Decide on the resolution to give to the split parameter theta */ + pulse_cap = m->logN[i]+LM*(1<>1) - (stereo&&N==2 ? QTHETA_OFFSET_TWOPHASE : QTHETA_OFFSET); + qn = compute_qn(N, *b, offset, pulse_cap, stereo); + if (stereo && i>=intensity) + qn = 1; + if (encode) + { + /* theta is the atan() of the ratio between the (normalized) + side and mid. With just that parameter, we can re-scale both + mid and side because we know that 1) they have unit norm and + 2) they are orthogonal. */ + itheta = stereo_itheta(X, Y, stereo, N, ctx->arch); + } + tell = ec_tell_frac(ec); + if (qn!=1) + { + if (encode) + { + if (!stereo || ctx->theta_round == 0) + { + itheta = (itheta*(opus_int32)qn+8192)>>14; + if (!stereo && ctx->avoid_split_noise && itheta > 0 && itheta < qn) + { + /* Check if the selected value of theta will cause the bit allocation + to inject noise on one side. If so, make sure the energy of that side + is zero. */ + int unquantized = celt_udiv((opus_int32)itheta*16384, qn); + imid = bitexact_cos((opus_int16)unquantized); + iside = bitexact_cos((opus_int16)(16384-unquantized)); + delta = FRAC_MUL16((N-1)<<7,bitexact_log2tan(iside,imid)); + if (delta > *b) + itheta = qn; + else if (delta < -*b) + itheta = 0; + } + } else { + int down; + /* Bias quantization towards itheta=0 and itheta=16384. */ + int bias = itheta > 8192 ? 32767/qn : -32767/qn; + down = IMIN(qn-1, IMAX(0, (itheta*(opus_int32)qn + bias)>>14)); + if (ctx->theta_round < 0) + itheta = down; + else + itheta = down+1; + } + } + /* Entropy coding of the angle. We use a uniform pdf for the + time split, a step for stereo, and a triangular one for the rest. */ + if (stereo && N>2) + { + int p0 = 3; + int x = itheta; + int x0 = qn/2; + int ft = p0*(x0+1) + x0; + /* Use a probability of p0 up to itheta=8192 and then use 1 after */ + if (encode) + { + ec_encode(ec,x<=x0?p0*x:(x-1-x0)+(x0+1)*p0,x<=x0?p0*(x+1):(x-x0)+(x0+1)*p0,ft); + } else { + int fs; + fs=ec_decode(ec,ft); + if (fs<(x0+1)*p0) + x=fs/p0; + else + x=x0+1+(fs-(x0+1)*p0); + ec_dec_update(ec,x<=x0?p0*x:(x-1-x0)+(x0+1)*p0,x<=x0?p0*(x+1):(x-x0)+(x0+1)*p0,ft); + itheta = x; + } + } else if (B0>1 || stereo) { + /* Uniform pdf */ + if (encode) + ec_enc_uint(ec, itheta, qn+1); + else + itheta = ec_dec_uint(ec, qn+1); + } else { + int fs=1, ft; + ft = ((qn>>1)+1)*((qn>>1)+1); + if (encode) + { + int fl; + + fs = itheta <= (qn>>1) ? itheta + 1 : qn + 1 - itheta; + fl = itheta <= (qn>>1) ? itheta*(itheta + 1)>>1 : + ft - ((qn + 1 - itheta)*(qn + 2 - itheta)>>1); + + ec_encode(ec, fl, fl+fs, ft); + } else { + /* Triangular pdf */ + int fl=0; + int fm; + fm = ec_decode(ec, ft); + + if (fm < ((qn>>1)*((qn>>1) + 1)>>1)) + { + itheta = (isqrt32(8*(opus_uint32)fm + 1) - 1)>>1; + fs = itheta + 1; + fl = itheta*(itheta + 1)>>1; + } + else + { + itheta = (2*(qn + 1) + - isqrt32(8*(opus_uint32)(ft - fm - 1) + 1))>>1; + fs = qn + 1 - itheta; + fl = ft - ((qn + 1 - itheta)*(qn + 2 - itheta)>>1); + } + + ec_dec_update(ec, fl, fl+fs, ft); + } + } + celt_assert(itheta>=0); + itheta = celt_udiv((opus_int32)itheta*16384, qn); + if (encode && stereo) + { + if (itheta==0) + intensity_stereo(m, X, Y, bandE, i, N); + else + stereo_split(X, Y, N); + } + /* NOTE: Renormalising X and Y *may* help fixed-point a bit at very high rate. + Let's do that at higher complexity */ + } else if (stereo) { + if (encode) + { + inv = itheta > 8192 && !ctx->disable_inv; + if (inv) + { + int j; + for (j=0;j2<remaining_bits > 2<disable_inv) + inv = 0; + itheta = 0; + } + qalloc = ec_tell_frac(ec) - tell; + *b -= qalloc; + + if (itheta == 0) + { + imid = 32767; + iside = 0; + *fill &= (1<inv = inv; + sctx->imid = imid; + sctx->iside = iside; + sctx->delta = delta; + sctx->itheta = itheta; + sctx->qalloc = qalloc; +} +static unsigned quant_band_n1(struct band_ctx *ctx, celt_norm *X, celt_norm *Y, + celt_norm *lowband_out) +{ + int c; + int stereo; + celt_norm *x = X; + int encode; + ec_ctx *ec; + + encode = ctx->encode; + ec = ctx->ec; + + stereo = Y != NULL; + c=0; do { + int sign=0; + if (ctx->remaining_bits>=1<remaining_bits -= 1<resynth) + x[0] = sign ? -NORM_SCALING : NORM_SCALING; + x = Y; + } while (++c<1+stereo); + if (lowband_out) + lowband_out[0] = SHR16(X[0],4); + return 1; +} + +/* This function is responsible for encoding and decoding a mono partition. + It can split the band in two and transmit the energy difference with + the two half-bands. It can be called recursively so bands can end up being + split in 8 parts. */ +static unsigned quant_partition(struct band_ctx *ctx, celt_norm *X, + int N, int b, int B, celt_norm *lowband, + int LM, + opus_val16 gain, int fill) +{ + const unsigned char *cache; + int q; + int curr_bits; + int imid=0, iside=0; + int B0=B; + opus_val16 mid=0, side=0; + unsigned cm=0; + celt_norm *Y=NULL; + int encode; + const CELTMode *m; + int i; + int spread; + ec_ctx *ec; + + encode = ctx->encode; + m = ctx->m; + i = ctx->i; + spread = ctx->spread; + ec = ctx->ec; + + /* If we need 1.5 more bit than we can produce, split the band in two. */ + cache = m->cache.bits + m->cache.index[(LM+1)*m->nbEBands+i]; + if (LM != -1 && b > cache[cache[0]]+12 && N>2) + { + int mbits, sbits, delta; + int itheta; + int qalloc; + struct split_ctx sctx; + celt_norm *next_lowband2=NULL; + opus_int32 rebalance; + + N >>= 1; + Y = X+N; + LM -= 1; + if (B==1) + fill = (fill&1)|(fill<<1); + B = (B+1)>>1; + + compute_theta(ctx, &sctx, X, Y, N, &b, B, B0, LM, 0, &fill); + imid = sctx.imid; + iside = sctx.iside; + delta = sctx.delta; + itheta = sctx.itheta; + qalloc = sctx.qalloc; +#ifdef FIXED_POINT + mid = imid; + side = iside; +#else + mid = (1.f/32768)*imid; + side = (1.f/32768)*iside; +#endif + + /* Give more bits to low-energy MDCTs than they would otherwise deserve */ + if (B0>1 && (itheta&0x3fff)) + { + if (itheta > 8192) + /* Rough approximation for pre-echo masking */ + delta -= delta>>(4-LM); + else + /* Corresponds to a forward-masking slope of 1.5 dB per 10 ms */ + delta = IMIN(0, delta + (N<>(5-LM))); + } + mbits = IMAX(0, IMIN(b, (b-delta)/2)); + sbits = b-mbits; + ctx->remaining_bits -= qalloc; + + if (lowband) + next_lowband2 = lowband+N; /* >32-bit split case */ + + rebalance = ctx->remaining_bits; + if (mbits >= sbits) + { + cm = quant_partition(ctx, X, N, mbits, B, lowband, LM, + MULT16_16_P15(gain,mid), fill); + rebalance = mbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<>B)<<(B0>>1); + } else { + cm = quant_partition(ctx, Y, N, sbits, B, next_lowband2, LM, + MULT16_16_P15(gain,side), fill>>B)<<(B0>>1); + rebalance = sbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<remaining_bits -= curr_bits; + + /* Ensures we can never bust the budget */ + while (ctx->remaining_bits < 0 && q > 0) + { + ctx->remaining_bits += curr_bits; + q--; + curr_bits = pulses2bits(m, i, LM, q); + ctx->remaining_bits -= curr_bits; + } + + if (q!=0) + { + int K = get_pulses(q); + + /* Finally do the actual quantization */ + if (encode) + { + cm = alg_quant(X, N, K, spread, B, ec, gain, ctx->resynth, ctx->arch); + } else { + cm = alg_unquant(X, N, K, spread, B, ec, gain); + } + } else { + /* If there's no pulse, fill the band anyway */ + int j; + if (ctx->resynth) + { + unsigned cm_mask; + /* B can be as large as 16, so this shift might overflow an int on a + 16-bit platform; use a long to get defined behavior.*/ + cm_mask = (unsigned)(1UL<seed = celt_lcg_rand(ctx->seed); + X[j] = (celt_norm)((opus_int32)ctx->seed>>20); + } + cm = cm_mask; + } else { + /* Folded spectrum */ + for (j=0;jseed = celt_lcg_rand(ctx->seed); + /* About 48 dB below the "normal" folding level */ + tmp = QCONST16(1.0f/256, 10); + tmp = (ctx->seed)&0x8000 ? tmp : -tmp; + X[j] = lowband[j]+tmp; + } + cm = fill; + } + renormalise_vector(X, N, gain, ctx->arch); + } + } + } + } + + return cm; +} + + +/* This function is responsible for encoding and decoding a band for the mono case. */ +static unsigned quant_band(struct band_ctx *ctx, celt_norm *X, + int N, int b, int B, celt_norm *lowband, + int LM, celt_norm *lowband_out, + opus_val16 gain, celt_norm *lowband_scratch, int fill) +{ + int N0=N; + int N_B=N; + int N_B0; + int B0=B; + int time_divide=0; + int recombine=0; + int longBlocks; + unsigned cm=0; + int k; + int encode; + int tf_change; + + encode = ctx->encode; + tf_change = ctx->tf_change; + + longBlocks = B0==1; + + N_B = celt_udiv(N_B, B); + + /* Special case for one sample */ + if (N==1) + { + return quant_band_n1(ctx, X, NULL, lowband_out); + } + + if (tf_change>0) + recombine = tf_change; + /* Band recombining to increase frequency resolution */ + + if (lowband_scratch && lowband && (recombine || ((N_B&1) == 0 && tf_change<0) || B0>1)) + { + OPUS_COPY(lowband_scratch, lowband, N); + lowband = lowband_scratch; + } + + for (k=0;k>k, 1<>k, 1<>4]<<2; + } + B>>=recombine; + N_B<<=recombine; + + /* Increasing the time resolution */ + while ((N_B&1) == 0 && tf_change<0) + { + if (encode) + haar1(X, N_B, B); + if (lowband) + haar1(lowband, N_B, B); + fill |= fill<>= 1; + time_divide++; + tf_change++; + } + B0=B; + N_B0 = N_B; + + /* Reorganize the samples in time order instead of frequency order */ + if (B0>1) + { + if (encode) + deinterleave_hadamard(X, N_B>>recombine, B0<>recombine, B0<resynth) + { + /* Undo the sample reorganization going from time order to frequency order */ + if (B0>1) + interleave_hadamard(X, N_B>>recombine, B0<>= 1; + N_B <<= 1; + cm |= cm>>B; + haar1(X, N_B, B); + } + + for (k=0;k>k, 1<encode; + ec = ctx->ec; + + /* Special case for one sample */ + if (N==1) + { + return quant_band_n1(ctx, X, Y, lowband_out); + } + + orig_fill = fill; + + compute_theta(ctx, &sctx, X, Y, N, &b, B, B, LM, 1, &fill); + inv = sctx.inv; + imid = sctx.imid; + iside = sctx.iside; + delta = sctx.delta; + itheta = sctx.itheta; + qalloc = sctx.qalloc; +#ifdef FIXED_POINT + mid = imid; + side = iside; +#else + mid = (1.f/32768)*imid; + side = (1.f/32768)*iside; +#endif + + /* This is a special case for N=2 that only works for stereo and takes + advantage of the fact that mid and side are orthogonal to encode + the side with just one bit. */ + if (N==2) + { + int c; + int sign=0; + celt_norm *x2, *y2; + mbits = b; + sbits = 0; + /* Only need one bit for the side. */ + if (itheta != 0 && itheta != 16384) + sbits = 1< 8192; + ctx->remaining_bits -= qalloc+sbits; + + x2 = c ? Y : X; + y2 = c ? X : Y; + if (sbits) + { + if (encode) + { + /* Here we only need to encode a sign for the side. */ + sign = x2[0]*y2[1] - x2[1]*y2[0] < 0; + ec_enc_bits(ec, sign, 1); + } else { + sign = ec_dec_bits(ec, 1); + } + } + sign = 1-2*sign; + /* We use orig_fill here because we want to fold the side, but if + itheta==16384, we'll have cleared the low bits of fill. */ + cm = quant_band(ctx, x2, N, mbits, B, lowband, LM, lowband_out, Q15ONE, + lowband_scratch, orig_fill); + /* We don't split N=2 bands, so cm is either 1 or 0 (for a fold-collapse), + and there's no need to worry about mixing with the other channel. */ + y2[0] = -sign*x2[1]; + y2[1] = sign*x2[0]; + if (ctx->resynth) + { + celt_norm tmp; + X[0] = MULT16_16_Q15(mid, X[0]); + X[1] = MULT16_16_Q15(mid, X[1]); + Y[0] = MULT16_16_Q15(side, Y[0]); + Y[1] = MULT16_16_Q15(side, Y[1]); + tmp = X[0]; + X[0] = SUB16(tmp,Y[0]); + Y[0] = ADD16(tmp,Y[0]); + tmp = X[1]; + X[1] = SUB16(tmp,Y[1]); + Y[1] = ADD16(tmp,Y[1]); + } + } else { + /* "Normal" split code */ + opus_int32 rebalance; + + mbits = IMAX(0, IMIN(b, (b-delta)/2)); + sbits = b-mbits; + ctx->remaining_bits -= qalloc; + + rebalance = ctx->remaining_bits; + if (mbits >= sbits) + { + /* In stereo mode, we do not apply a scaling to the mid because we need the normalized + mid for folding later. */ + cm = quant_band(ctx, X, N, mbits, B, lowband, LM, lowband_out, Q15ONE, + lowband_scratch, fill); + rebalance = mbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<>B); + } else { + /* For a stereo split, the high bits of fill are always zero, so no + folding will be done to the side. */ + cm = quant_band(ctx, Y, N, sbits, B, NULL, LM, NULL, side, NULL, fill>>B); + rebalance = sbits - (rebalance-ctx->remaining_bits); + if (rebalance > 3<resynth) + { + if (N!=2) + stereo_merge(X, Y, mid, N, ctx->arch); + if (inv) + { + int j; + for (j=0;jeBands; + n1 = M*(eBands[start+1]-eBands[start]); + n2 = M*(eBands[start+2]-eBands[start+1]); + /* Duplicate enough of the first band folding data to be able to fold the second band. + Copies no data for CELT-only mode. */ + OPUS_COPY(&norm[n1], &norm[2*n1 - n2], n2-n1); + if (dual_stereo) + OPUS_COPY(&norm2[n1], &norm2[2*n1 - n2], n2-n1); +} +#endif + +void quant_all_bands(int encode, const CELTMode *m, int start, int end, + celt_norm *X_, celt_norm *Y_, unsigned char *collapse_masks, + const celt_ener *bandE, int *pulses, int shortBlocks, int spread, + int dual_stereo, int intensity, int *tf_res, opus_int32 total_bits, + opus_int32 balance, ec_ctx *ec, int LM, int codedBands, + opus_uint32 *seed, int complexity, int arch, int disable_inv) +{ + int i; + opus_int32 remaining_bits; + const opus_int16 * OPUS_RESTRICT eBands = m->eBands; + celt_norm * OPUS_RESTRICT norm, * OPUS_RESTRICT norm2; + VARDECL(celt_norm, _norm); + VARDECL(celt_norm, _lowband_scratch); + VARDECL(celt_norm, X_save); + VARDECL(celt_norm, Y_save); + VARDECL(celt_norm, X_save2); + VARDECL(celt_norm, Y_save2); + VARDECL(celt_norm, norm_save2); + int resynth_alloc; + celt_norm *lowband_scratch; + int B; + int M; + int lowband_offset; + int update_lowband = 1; + int C = Y_ != NULL ? 2 : 1; + int norm_offset; + int theta_rdo = encode && Y_!=NULL && !dual_stereo && complexity>=8; +#ifdef RESYNTH + int resynth = 1; +#else + int resynth = !encode || theta_rdo; +#endif + struct band_ctx ctx; + SAVE_STACK; + + M = 1<nbEBands-1]-norm_offset), celt_norm); + norm = _norm; + norm2 = norm + M*eBands[m->nbEBands-1]-norm_offset; + + /* For decoding, we can use the last band as scratch space because we don't need that + scratch space for the last band and we don't care about the data there until we're + decoding the last band. */ + if (encode && resynth) + resynth_alloc = M*(eBands[m->nbEBands]-eBands[m->nbEBands-1]); + else + resynth_alloc = ALLOC_NONE; + ALLOC(_lowband_scratch, resynth_alloc, celt_norm); + if (encode && resynth) + lowband_scratch = _lowband_scratch; + else + lowband_scratch = X_+M*eBands[m->nbEBands-1]; + ALLOC(X_save, resynth_alloc, celt_norm); + ALLOC(Y_save, resynth_alloc, celt_norm); + ALLOC(X_save2, resynth_alloc, celt_norm); + ALLOC(Y_save2, resynth_alloc, celt_norm); + ALLOC(norm_save2, resynth_alloc, celt_norm); + + lowband_offset = 0; + ctx.bandE = bandE; + ctx.ec = ec; + ctx.encode = encode; + ctx.intensity = intensity; + ctx.m = m; + ctx.seed = *seed; + ctx.spread = spread; + ctx.arch = arch; + ctx.disable_inv = disable_inv; + ctx.resynth = resynth; + ctx.theta_round = 0; + /* Avoid injecting noise in the first band on transients. */ + ctx.avoid_split_noise = B > 1; + for (i=start;i 0); + tell = ec_tell_frac(ec); + + /* Compute how many bits we want to allocate to this band */ + if (i != start) + balance -= tell; + remaining_bits = total_bits-tell-1; + ctx.remaining_bits = remaining_bits; + if (i <= codedBands-1) + { + curr_balance = celt_sudiv(balance, IMIN(3, codedBands-i)); + b = IMAX(0, IMIN(16383, IMIN(remaining_bits+1,pulses[i]+curr_balance))); + } else { + b = 0; + } + +#ifndef DISABLE_UPDATE_DRAFT + if (resynth && (M*eBands[i]-N >= M*eBands[start] || i==start+1) && (update_lowband || lowband_offset==0)) + lowband_offset = i; + if (i == start+1) + special_hybrid_folding(m, norm, norm2, start, M, dual_stereo); +#else + if (resynth && M*eBands[i]-N >= M*eBands[start] && (update_lowband || lowband_offset==0)) + lowband_offset = i; +#endif + + tf_change = tf_res[i]; + ctx.tf_change = tf_change; + if (i>=m->effEBands) + { + X=norm; + if (Y_!=NULL) + Y = norm; + lowband_scratch = NULL; + } + if (last && !theta_rdo) + lowband_scratch = NULL; + + /* Get a conservative estimate of the collapse_mask's for the bands we're + going to be folding from. */ + if (lowband_offset != 0 && (spread!=SPREAD_AGGRESSIVE || B>1 || tf_change<0)) + { + int fold_start; + int fold_end; + int fold_i; + /* This ensures we never repeat spectral content within one band */ + effective_lowband = IMAX(0, M*eBands[lowband_offset]-norm_offset-N); + fold_start = lowband_offset; + while(M*eBands[--fold_start] > effective_lowband+norm_offset); + fold_end = lowband_offset-1; +#ifndef DISABLE_UPDATE_DRAFT + while(++fold_end < i && M*eBands[fold_end] < effective_lowband+norm_offset+N); +#else + while(M*eBands[++fold_end] < effective_lowband+norm_offset+N); +#endif + x_cm = y_cm = 0; + fold_i = fold_start; do { + x_cm |= collapse_masks[fold_i*C+0]; + y_cm |= collapse_masks[fold_i*C+C-1]; + } while (++fold_inbEBands], w); + /* Make a copy. */ + cm = x_cm|y_cm; + ec_save = *ec; + ctx_save = ctx; + OPUS_COPY(X_save, X, N); + OPUS_COPY(Y_save, Y, N); + /* Encode and round down. */ + ctx.theta_round = -1; + x_cm = quant_band_stereo(&ctx, X, Y, N, b, B, + effective_lowband != -1 ? norm+effective_lowband : NULL, LM, + last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, cm); + dist0 = MULT16_32_Q15(w[0], celt_inner_prod(X_save, X, N, arch)) + MULT16_32_Q15(w[1], celt_inner_prod(Y_save, Y, N, arch)); + + /* Save first result. */ + cm2 = x_cm; + ec_save2 = *ec; + ctx_save2 = ctx; + OPUS_COPY(X_save2, X, N); + OPUS_COPY(Y_save2, Y, N); + if (!last) + OPUS_COPY(norm_save2, norm+M*eBands[i]-norm_offset, N); + nstart_bytes = ec_save.offs; + nend_bytes = ec_save.storage; + bytes_buf = ec_save.buf+nstart_bytes; + save_bytes = nend_bytes-nstart_bytes; + OPUS_COPY(bytes_save, bytes_buf, save_bytes); + + /* Restore */ + *ec = ec_save; + ctx = ctx_save; + OPUS_COPY(X, X_save, N); + OPUS_COPY(Y, Y_save, N); +#ifndef DISABLE_UPDATE_DRAFT + if (i == start+1) + special_hybrid_folding(m, norm, norm2, start, M, dual_stereo); +#endif + /* Encode and round up. */ + ctx.theta_round = 1; + x_cm = quant_band_stereo(&ctx, X, Y, N, b, B, + effective_lowband != -1 ? norm+effective_lowband : NULL, LM, + last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, cm); + dist1 = MULT16_32_Q15(w[0], celt_inner_prod(X_save, X, N, arch)) + MULT16_32_Q15(w[1], celt_inner_prod(Y_save, Y, N, arch)); + if (dist0 >= dist1) { + x_cm = cm2; + *ec = ec_save2; + ctx = ctx_save2; + OPUS_COPY(X, X_save2, N); + OPUS_COPY(Y, Y_save2, N); + if (!last) + OPUS_COPY(norm+M*eBands[i]-norm_offset, norm_save2, N); + OPUS_COPY(bytes_buf, bytes_save, save_bytes); + } + } else { + ctx.theta_round = 0; + x_cm = quant_band_stereo(&ctx, X, Y, N, b, B, + effective_lowband != -1 ? norm+effective_lowband : NULL, LM, + last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, x_cm|y_cm); + } + } else { + x_cm = quant_band(&ctx, X, N, b, B, + effective_lowband != -1 ? norm+effective_lowband : NULL, LM, + last?NULL:norm+M*eBands[i]-norm_offset, Q15ONE, lowband_scratch, x_cm|y_cm); + } + y_cm = x_cm; + } + collapse_masks[i*C+0] = (unsigned char)x_cm; + collapse_masks[i*C+C-1] = (unsigned char)y_cm; + balance += pulses[i] + tell; + + /* Update the folding position only as long as we have 1 bit/sample depth. */ + update_lowband = b>(N< +#include "celt.h" +#include "pitch.h" +#include "bands.h" +#include "modes.h" +#include "entcode.h" +#include "quant_bands.h" +#include "rate.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "float_cast.h" +#include +#include "celt_lpc.h" +#include "vq.h" + +#ifndef PACKAGE_VERSION +#define PACKAGE_VERSION "unknown" +#endif + +#if defined(MIPSr1_ASM) +#include "mips/celt_mipsr1.h" +#endif + + +int resampling_factor(opus_int32 rate) +{ + int ret; + switch (rate) + { + case 48000: + ret = 1; + break; + case 24000: + ret = 2; + break; + case 16000: + ret = 3; + break; + case 12000: + ret = 4; + break; + case 8000: + ret = 6; + break; + default: +#ifndef CUSTOM_MODES + celt_assert(0); +#endif + ret = 0; + break; + } + return ret; +} + +#if !defined(OVERRIDE_COMB_FILTER_CONST) || defined(NON_STATIC_COMB_FILTER_CONST_C) +/* This version should be faster on ARM */ +#ifdef OPUS_ARM_ASM +#ifndef NON_STATIC_COMB_FILTER_CONST_C +static +#endif +void comb_filter_const_c(opus_val32 *y, opus_val32 *x, int T, int N, + opus_val16 g10, opus_val16 g11, opus_val16 g12) +{ + opus_val32 x0, x1, x2, x3, x4; + int i; + x4 = SHL32(x[-T-2], 1); + x3 = SHL32(x[-T-1], 1); + x2 = SHL32(x[-T], 1); + x1 = SHL32(x[-T+1], 1); + for (i=0;inbEBands;i++) + { + int N; + N=(m->eBands[i+1]-m->eBands[i])<cache.caps[m->nbEBands*(2*LM+C-1)+i]+64)*C*N>>2; + } +} + + + +const char *opus_strerror(int error) +{ + static const char * const error_strings[8] = { + "success", + "invalid argument", + "buffer too small", + "internal error", + "corrupted stream", + "request not implemented", + "invalid state", + "memory allocation failed" + }; + if (error > 0 || error < -7) + return "unknown error"; + else + return error_strings[-error]; +} + +const char *opus_get_version_string(void) +{ + return "libopus " PACKAGE_VERSION + /* Applications may rely on the presence of this substring in the version + string to determine if they have a fixed-point or floating-point build + at runtime. */ +#ifdef FIXED_POINT + "-fixed" +#endif +#ifdef FUZZING + "-fuzzing" +#endif + ; +} diff --git a/lib/opus/celt/celt.h b/lib/opus/celt/celt.h new file mode 100644 index 00000000..24b6b2b5 --- /dev/null +++ b/lib/opus/celt/celt.h @@ -0,0 +1,251 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2008 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/** + @file celt.h + @brief Contains all the functions for encoding and decoding audio + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef CELT_H +#define CELT_H + +#include "opus_types.h" +#include "opus_defines.h" +#include "opus_custom.h" +#include "entenc.h" +#include "entdec.h" +#include "arch.h" + +#ifdef __cplusplus +extern "C" { +#endif + +#define CELTEncoder OpusCustomEncoder +#define CELTDecoder OpusCustomDecoder +#define CELTMode OpusCustomMode + +#define LEAK_BANDS 19 + +typedef struct { + int valid; + float tonality; + float tonality_slope; + float noisiness; + float activity; + float music_prob; + float music_prob_min; + float music_prob_max; + int bandwidth; + float activity_probability; + float max_pitch_ratio; + /* Store as Q6 char to save space. */ + unsigned char leak_boost[LEAK_BANDS]; +} AnalysisInfo; + +typedef struct { + int signalType; + int offset; +} SILKInfo; + +#define __celt_check_mode_ptr_ptr(ptr) ((ptr) + ((ptr) - (const CELTMode**)(ptr))) + +#define __celt_check_analysis_ptr(ptr) ((ptr) + ((ptr) - (const AnalysisInfo*)(ptr))) + +#define __celt_check_silkinfo_ptr(ptr) ((ptr) + ((ptr) - (const SILKInfo*)(ptr))) + +/* Encoder/decoder Requests */ + + +#define CELT_SET_PREDICTION_REQUEST 10002 +/** Controls the use of interframe prediction. + 0=Independent frames + 1=Short term interframe prediction allowed + 2=Long term prediction allowed + */ +#define CELT_SET_PREDICTION(x) CELT_SET_PREDICTION_REQUEST, __opus_check_int(x) + +#define CELT_SET_INPUT_CLIPPING_REQUEST 10004 +#define CELT_SET_INPUT_CLIPPING(x) CELT_SET_INPUT_CLIPPING_REQUEST, __opus_check_int(x) + +#define CELT_GET_AND_CLEAR_ERROR_REQUEST 10007 +#define CELT_GET_AND_CLEAR_ERROR(x) CELT_GET_AND_CLEAR_ERROR_REQUEST, __opus_check_int_ptr(x) + +#define CELT_SET_CHANNELS_REQUEST 10008 +#define CELT_SET_CHANNELS(x) CELT_SET_CHANNELS_REQUEST, __opus_check_int(x) + + +/* Internal */ +#define CELT_SET_START_BAND_REQUEST 10010 +#define CELT_SET_START_BAND(x) CELT_SET_START_BAND_REQUEST, __opus_check_int(x) + +#define CELT_SET_END_BAND_REQUEST 10012 +#define CELT_SET_END_BAND(x) CELT_SET_END_BAND_REQUEST, __opus_check_int(x) + +#define CELT_GET_MODE_REQUEST 10015 +/** Get the CELTMode used by an encoder or decoder */ +#define CELT_GET_MODE(x) CELT_GET_MODE_REQUEST, __celt_check_mode_ptr_ptr(x) + +#define CELT_SET_SIGNALLING_REQUEST 10016 +#define CELT_SET_SIGNALLING(x) CELT_SET_SIGNALLING_REQUEST, __opus_check_int(x) + +#define CELT_SET_TONALITY_REQUEST 10018 +#define CELT_SET_TONALITY(x) CELT_SET_TONALITY_REQUEST, __opus_check_int(x) +#define CELT_SET_TONALITY_SLOPE_REQUEST 10020 +#define CELT_SET_TONALITY_SLOPE(x) CELT_SET_TONALITY_SLOPE_REQUEST, __opus_check_int(x) + +#define CELT_SET_ANALYSIS_REQUEST 10022 +#define CELT_SET_ANALYSIS(x) CELT_SET_ANALYSIS_REQUEST, __celt_check_analysis_ptr(x) + +#define OPUS_SET_LFE_REQUEST 10024 +#define OPUS_SET_LFE(x) OPUS_SET_LFE_REQUEST, __opus_check_int(x) + +#define OPUS_SET_ENERGY_MASK_REQUEST 10026 +#define OPUS_SET_ENERGY_MASK(x) OPUS_SET_ENERGY_MASK_REQUEST, __opus_check_val16_ptr(x) + +#define CELT_SET_SILK_INFO_REQUEST 10028 +#define CELT_SET_SILK_INFO(x) CELT_SET_SILK_INFO_REQUEST, __celt_check_silkinfo_ptr(x) + +/* Encoder stuff */ + +int celt_encoder_get_size(int channels); + +int celt_encode_with_ec(OpusCustomEncoder * OPUS_RESTRICT st, const opus_val16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes, ec_enc *enc); + +int celt_encoder_init(CELTEncoder *st, opus_int32 sampling_rate, int channels, + int arch); + + + +/* Decoder stuff */ + +int celt_decoder_get_size(int channels); + + +int celt_decoder_init(CELTDecoder *st, opus_int32 sampling_rate, int channels); + +int celt_decode_with_ec(OpusCustomDecoder * OPUS_RESTRICT st, const unsigned char *data, + int len, opus_val16 * OPUS_RESTRICT pcm, int frame_size, ec_dec *dec, int accum); + +#define celt_encoder_ctl opus_custom_encoder_ctl +#define celt_decoder_ctl opus_custom_decoder_ctl + + +#ifdef CUSTOM_MODES +#define OPUS_CUSTOM_NOSTATIC +#else +#define OPUS_CUSTOM_NOSTATIC static OPUS_INLINE +#endif + +static const unsigned char trim_icdf[11] = {126, 124, 119, 109, 87, 41, 19, 9, 4, 2, 0}; +/* Probs: NONE: 21.875%, LIGHT: 6.25%, NORMAL: 65.625%, AGGRESSIVE: 6.25% */ +static const unsigned char spread_icdf[4] = {25, 23, 2, 0}; + +static const unsigned char tapset_icdf[3]={2,1,0}; + +#ifdef CUSTOM_MODES +static const unsigned char toOpusTable[20] = { + 0xE0, 0xE8, 0xF0, 0xF8, + 0xC0, 0xC8, 0xD0, 0xD8, + 0xA0, 0xA8, 0xB0, 0xB8, + 0x00, 0x00, 0x00, 0x00, + 0x80, 0x88, 0x90, 0x98, +}; + +static const unsigned char fromOpusTable[16] = { + 0x80, 0x88, 0x90, 0x98, + 0x40, 0x48, 0x50, 0x58, + 0x20, 0x28, 0x30, 0x38, + 0x00, 0x08, 0x10, 0x18 +}; + +static OPUS_INLINE int toOpus(unsigned char c) +{ + int ret=0; + if (c<0xA0) + ret = toOpusTable[c>>3]; + if (ret == 0) + return -1; + else + return ret|(c&0x7); +} + +static OPUS_INLINE int fromOpus(unsigned char c) +{ + if (c<0x80) + return -1; + else + return fromOpusTable[(c>>3)-16] | (c&0x7); +} +#endif /* CUSTOM_MODES */ + +#define COMBFILTER_MAXPERIOD 1024 +#define COMBFILTER_MINPERIOD 15 + +extern const signed char tf_select_table[4][8]; + +#if defined(ENABLE_HARDENING) || defined(ENABLE_ASSERTIONS) +void validate_celt_decoder(CELTDecoder *st); +#define VALIDATE_CELT_DECODER(st) validate_celt_decoder(st) +#else +#define VALIDATE_CELT_DECODER(st) +#endif + +int resampling_factor(opus_int32 rate); + +void celt_preemphasis(const opus_val16 * OPUS_RESTRICT pcmp, celt_sig * OPUS_RESTRICT inp, + int N, int CC, int upsample, const opus_val16 *coef, celt_sig *mem, int clip); + +void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N, + opus_val16 g0, opus_val16 g1, int tapset0, int tapset1, + const opus_val16 *window, int overlap, int arch); + +#ifdef NON_STATIC_COMB_FILTER_CONST_C +void comb_filter_const_c(opus_val32 *y, opus_val32 *x, int T, int N, + opus_val16 g10, opus_val16 g11, opus_val16 g12); +#endif + +#ifndef OVERRIDE_COMB_FILTER_CONST +# define comb_filter_const(y, x, T, N, g10, g11, g12, arch) \ + ((void)(arch),comb_filter_const_c(y, x, T, N, g10, g11, g12)) +#endif + +void init_caps(const CELTMode *m,int *cap,int LM,int C); + +#ifdef RESYNTH +void deemphasis(celt_sig *in[], opus_val16 *pcm, int N, int C, int downsample, const opus_val16 *coef, celt_sig *mem); +void celt_synthesis(const CELTMode *mode, celt_norm *X, celt_sig * out_syn[], + opus_val16 *oldBandE, int start, int effEnd, int C, int CC, int isTransient, + int LM, int downsample, int silence); +#endif + +#ifdef __cplusplus +} +#endif + +#endif /* CELT_H */ diff --git a/lib/opus/celt/celt_decoder.c b/lib/opus/celt/celt_decoder.c new file mode 100644 index 00000000..883dae15 --- /dev/null +++ b/lib/opus/celt/celt_decoder.c @@ -0,0 +1,1375 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2010 Xiph.Org Foundation + Copyright (c) 2008 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#define CELT_DECODER_C + +#include "cpu_support.h" +#include "os_support.h" +#include "mdct.h" +#include +#include "celt.h" +#include "pitch.h" +#include "bands.h" +#include "modes.h" +#include "entcode.h" +#include "quant_bands.h" +#include "rate.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "float_cast.h" +#include +#include "celt_lpc.h" +#include "vq.h" + +/* The maximum pitch lag to allow in the pitch-based PLC. It's possible to save + CPU time in the PLC pitch search by making this smaller than MAX_PERIOD. The + current value corresponds to a pitch of 66.67 Hz. */ +#define PLC_PITCH_LAG_MAX (720) +/* The minimum pitch lag to allow in the pitch-based PLC. This corresponds to a + pitch of 480 Hz. */ +#define PLC_PITCH_LAG_MIN (100) + +#if defined(SMALL_FOOTPRINT) && defined(FIXED_POINT) +#define NORM_ALIASING_HACK +#endif +/**********************************************************************/ +/* */ +/* DECODER */ +/* */ +/**********************************************************************/ +#define DECODE_BUFFER_SIZE 2048 + +/** Decoder state + @brief Decoder state + */ +struct OpusCustomDecoder { + const OpusCustomMode *mode; + int overlap; + int channels; + int stream_channels; + + int downsample; + int start, end; + int signalling; + int disable_inv; + int arch; + + /* Everything beyond this point gets cleared on a reset */ +#define DECODER_RESET_START rng + + opus_uint32 rng; + int error; + int last_pitch_index; + int loss_duration; + int skip_plc; + int postfilter_period; + int postfilter_period_old; + opus_val16 postfilter_gain; + opus_val16 postfilter_gain_old; + int postfilter_tapset; + int postfilter_tapset_old; + + celt_sig preemph_memD[2]; + + celt_sig _decode_mem[1]; /* Size = channels*(DECODE_BUFFER_SIZE+mode->overlap) */ + /* opus_val16 lpc[], Size = channels*LPC_ORDER */ + /* opus_val16 oldEBands[], Size = 2*mode->nbEBands */ + /* opus_val16 oldLogE[], Size = 2*mode->nbEBands */ + /* opus_val16 oldLogE2[], Size = 2*mode->nbEBands */ + /* opus_val16 backgroundLogE[], Size = 2*mode->nbEBands */ +}; + +#if defined(ENABLE_HARDENING) || defined(ENABLE_ASSERTIONS) +/* Make basic checks on the CELT state to ensure we don't end + up writing all over memory. */ +void validate_celt_decoder(CELTDecoder *st) +{ +#ifndef CUSTOM_MODES + celt_assert(st->mode == opus_custom_mode_create(48000, 960, NULL)); + celt_assert(st->overlap == 120); + celt_assert(st->end <= 21); +#else +/* From Section 4.3 in the spec: "The normal CELT layer uses 21 of those bands, + though Opus Custom (see Section 6.2) may use a different number of bands" + + Check if it's within the maximum number of Bark frequency bands instead */ + celt_assert(st->end <= 25); +#endif + celt_assert(st->channels == 1 || st->channels == 2); + celt_assert(st->stream_channels == 1 || st->stream_channels == 2); + celt_assert(st->downsample > 0); + celt_assert(st->start == 0 || st->start == 17); + celt_assert(st->start < st->end); +#ifdef OPUS_ARCHMASK + celt_assert(st->arch >= 0); + celt_assert(st->arch <= OPUS_ARCHMASK); +#endif + celt_assert(st->last_pitch_index <= PLC_PITCH_LAG_MAX); + celt_assert(st->last_pitch_index >= PLC_PITCH_LAG_MIN || st->last_pitch_index == 0); + celt_assert(st->postfilter_period < MAX_PERIOD); + celt_assert(st->postfilter_period >= COMBFILTER_MINPERIOD || st->postfilter_period == 0); + celt_assert(st->postfilter_period_old < MAX_PERIOD); + celt_assert(st->postfilter_period_old >= COMBFILTER_MINPERIOD || st->postfilter_period_old == 0); + celt_assert(st->postfilter_tapset <= 2); + celt_assert(st->postfilter_tapset >= 0); + celt_assert(st->postfilter_tapset_old <= 2); + celt_assert(st->postfilter_tapset_old >= 0); +} +#endif + +int celt_decoder_get_size(int channels) +{ + const CELTMode *mode = opus_custom_mode_create(48000, 960, NULL); + return opus_custom_decoder_get_size(mode, channels); +} + +OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_get_size(const CELTMode *mode, int channels) +{ + int size = sizeof(struct CELTDecoder) + + (channels*(DECODE_BUFFER_SIZE+mode->overlap)-1)*sizeof(celt_sig) + + channels*LPC_ORDER*sizeof(opus_val16) + + 4*2*mode->nbEBands*sizeof(opus_val16); + return size; +} + +#ifdef CUSTOM_MODES +CELTDecoder *opus_custom_decoder_create(const CELTMode *mode, int channels, int *error) +{ + int ret; + CELTDecoder *st = (CELTDecoder *)opus_alloc(opus_custom_decoder_get_size(mode, channels)); + ret = opus_custom_decoder_init(st, mode, channels); + if (ret != OPUS_OK) + { + opus_custom_decoder_destroy(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} +#endif /* CUSTOM_MODES */ + +int celt_decoder_init(CELTDecoder *st, opus_int32 sampling_rate, int channels) +{ + int ret; + ret = opus_custom_decoder_init(st, opus_custom_mode_create(48000, 960, NULL), channels); + if (ret != OPUS_OK) + return ret; + st->downsample = resampling_factor(sampling_rate); + if (st->downsample==0) + return OPUS_BAD_ARG; + else + return OPUS_OK; +} + +OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_init(CELTDecoder *st, const CELTMode *mode, int channels) +{ + if (channels < 0 || channels > 2) + return OPUS_BAD_ARG; + + if (st==NULL) + return OPUS_ALLOC_FAIL; + + OPUS_CLEAR((char*)st, opus_custom_decoder_get_size(mode, channels)); + + st->mode = mode; + st->overlap = mode->overlap; + st->stream_channels = st->channels = channels; + + st->downsample = 1; + st->start = 0; + st->end = st->mode->effEBands; + st->signalling = 1; +#ifndef DISABLE_UPDATE_DRAFT + st->disable_inv = channels == 1; +#else + st->disable_inv = 0; +#endif + st->arch = opus_select_arch(); + + opus_custom_decoder_ctl(st, OPUS_RESET_STATE); + + return OPUS_OK; +} + +#ifdef CUSTOM_MODES +void opus_custom_decoder_destroy(CELTDecoder *st) +{ + opus_free(st); +} +#endif /* CUSTOM_MODES */ + +#ifndef CUSTOM_MODES +/* Special case for stereo with no downsampling and no accumulation. This is + quite common and we can make it faster by processing both channels in the + same loop, reducing overhead due to the dependency loop in the IIR filter. */ +static void deemphasis_stereo_simple(celt_sig *in[], opus_val16 *pcm, int N, const opus_val16 coef0, + celt_sig *mem) +{ + celt_sig * OPUS_RESTRICT x0; + celt_sig * OPUS_RESTRICT x1; + celt_sig m0, m1; + int j; + x0=in[0]; + x1=in[1]; + m0 = mem[0]; + m1 = mem[1]; + for (j=0;j1) + { + /* Shortcut for the standard (non-custom modes) case */ + for (j=0;joverlap; + nbEBands = mode->nbEBands; + N = mode->shortMdctSize<shortMdctSize; + shift = mode->maxLM; + } else { + B = 1; + NB = mode->shortMdctSize<maxLM-LM; + } + + if (CC==2&&C==1) + { + /* Copying a mono streams to two channels */ + celt_sig *freq2; + denormalise_bands(mode, X, freq, oldBandE, start, effEnd, M, + downsample, silence); + /* Store a temporary copy in the output buffer because the IMDCT destroys its input. */ + freq2 = out_syn[1]+overlap/2; + OPUS_COPY(freq2, freq, N); + for (b=0;bmdct, &freq2[b], out_syn[0]+NB*b, mode->window, overlap, shift, B, arch); + for (b=0;bmdct, &freq[b], out_syn[1]+NB*b, mode->window, overlap, shift, B, arch); + } else if (CC==1&&C==2) + { + /* Downmixing a stereo stream to mono */ + celt_sig *freq2; + freq2 = out_syn[0]+overlap/2; + denormalise_bands(mode, X, freq, oldBandE, start, effEnd, M, + downsample, silence); + /* Use the output buffer as temp array before downmixing. */ + denormalise_bands(mode, X+N, freq2, oldBandE+nbEBands, start, effEnd, M, + downsample, silence); + for (i=0;imdct, &freq[b], out_syn[0]+NB*b, mode->window, overlap, shift, B, arch); + } else { + /* Normal case (mono or stereo) */ + c=0; do { + denormalise_bands(mode, X+c*N, freq, oldBandE+c*nbEBands, start, effEnd, M, + downsample, silence); + for (b=0;bmdct, &freq[b], out_syn[c]+NB*b, mode->window, overlap, shift, B, arch); + } while (++cstorage*8; + tell = ec_tell(dec); + logp = isTransient ? 2 : 4; + tf_select_rsv = LM>0 && tell+logp+1<=budget; + budget -= tf_select_rsv; + tf_changed = curr = 0; + for (i=start;i>1, opus_val16 ); + pitch_downsample(decode_mem, lp_pitch_buf, + DECODE_BUFFER_SIZE, C, arch); + pitch_search(lp_pitch_buf+(PLC_PITCH_LAG_MAX>>1), lp_pitch_buf, + DECODE_BUFFER_SIZE-PLC_PITCH_LAG_MAX, + PLC_PITCH_LAG_MAX-PLC_PITCH_LAG_MIN, &pitch_index, arch); + pitch_index = PLC_PITCH_LAG_MAX-pitch_index; + RESTORE_STACK; + return pitch_index; +} + +static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM) +{ + int c; + int i; + const int C = st->channels; + celt_sig *decode_mem[2]; + celt_sig *out_syn[2]; + opus_val16 *lpc; + opus_val16 *oldBandE, *oldLogE, *oldLogE2, *backgroundLogE; + const OpusCustomMode *mode; + int nbEBands; + int overlap; + int start; + int loss_duration; + int noise_based; + const opus_int16 *eBands; + SAVE_STACK; + + mode = st->mode; + nbEBands = mode->nbEBands; + overlap = mode->overlap; + eBands = mode->eBands; + + c=0; do { + decode_mem[c] = st->_decode_mem + c*(DECODE_BUFFER_SIZE+overlap); + out_syn[c] = decode_mem[c]+DECODE_BUFFER_SIZE-N; + } while (++c_decode_mem+(DECODE_BUFFER_SIZE+overlap)*C); + oldBandE = lpc+C*LPC_ORDER; + oldLogE = oldBandE + 2*nbEBands; + oldLogE2 = oldLogE + 2*nbEBands; + backgroundLogE = oldLogE2 + 2*nbEBands; + + loss_duration = st->loss_duration; + start = st->start; + noise_based = loss_duration >= 40 || start != 0 || st->skip_plc; + if (noise_based) + { + /* Noise-based PLC/CNG */ +#ifdef NORM_ALIASING_HACK + celt_norm *X; +#else + VARDECL(celt_norm, X); +#endif + opus_uint32 seed; + int end; + int effEnd; + opus_val16 decay; + end = st->end; + effEnd = IMAX(start, IMIN(end, mode->effEBands)); + +#ifdef NORM_ALIASING_HACK + /* This is an ugly hack that breaks aliasing rules and would be easily broken, + but it saves almost 4kB of stack. */ + X = (celt_norm*)(out_syn[C-1]+overlap/2); +#else + ALLOC(X, C*N, celt_norm); /**< Interleaved normalised MDCTs */ +#endif + c=0; do { + OPUS_MOVE(decode_mem[c], decode_mem[c]+N, + DECODE_BUFFER_SIZE-N+(overlap>>1)); + } while (++crng; + for (c=0;c>20); + } + renormalise_vector(X+boffs, blen, Q15ONE, st->arch); + } + } + st->rng = seed; + + celt_synthesis(mode, X, out_syn, oldBandE, start, effEnd, C, C, 0, LM, st->downsample, 0, st->arch); + } else { + int exc_length; + /* Pitch-based PLC */ + const opus_val16 *window; + opus_val16 *exc; + opus_val16 fade = Q15ONE; + int pitch_index; + VARDECL(opus_val32, etmp); + VARDECL(opus_val16, _exc); + VARDECL(opus_val16, fir_tmp); + + if (loss_duration == 0) + { + st->last_pitch_index = pitch_index = celt_plc_pitch_search(decode_mem, C, st->arch); + } else { + pitch_index = st->last_pitch_index; + fade = QCONST16(.8f,15); + } + + /* We want the excitation for 2 pitch periods in order to look for a + decaying signal, but we can't get more than MAX_PERIOD. */ + exc_length = IMIN(2*pitch_index, MAX_PERIOD); + + ALLOC(etmp, overlap, opus_val32); + ALLOC(_exc, MAX_PERIOD+LPC_ORDER, opus_val16); + ALLOC(fir_tmp, exc_length, opus_val16); + exc = _exc+LPC_ORDER; + window = mode->window; + c=0; do { + opus_val16 decay; + opus_val16 attenuation; + opus_val32 S1=0; + celt_sig *buf; + int extrapolation_offset; + int extrapolation_len; + int j; + + buf = decode_mem[c]; + for (i=0;iarch); + /* Add a noise floor of -40 dB. */ +#ifdef FIXED_POINT + ac[0] += SHR32(ac[0],13); +#else + ac[0] *= 1.0001f; +#endif + /* Use lag windowing to stabilize the Levinson-Durbin recursion. */ + for (i=1;i<=LPC_ORDER;i++) + { + /*ac[i] *= exp(-.5*(2*M_PI*.002*i)*(2*M_PI*.002*i));*/ +#ifdef FIXED_POINT + ac[i] -= MULT16_32_Q15(2*i*i, ac[i]); +#else + ac[i] -= ac[i]*(0.008f*0.008f)*i*i; +#endif + } + _celt_lpc(lpc+c*LPC_ORDER, ac, LPC_ORDER); +#ifdef FIXED_POINT + /* For fixed-point, apply bandwidth expansion until we can guarantee that + no overflow can happen in the IIR filter. This means: + 32768*sum(abs(filter)) < 2^31 */ + while (1) { + opus_val16 tmp=Q15ONE; + opus_val32 sum=QCONST16(1., SIG_SHIFT); + for (i=0;iarch); + OPUS_COPY(exc+MAX_PERIOD-exc_length, fir_tmp, exc_length); + } + + /* Check if the waveform is decaying, and if so how fast. + We do this to avoid adding energy when concealing in a segment + with decaying energy. */ + { + opus_val32 E1=1, E2=1; + int decay_length; +#ifdef FIXED_POINT + int shift = IMAX(0,2*celt_zlog2(celt_maxabs16(&exc[MAX_PERIOD-exc_length], exc_length))-20); +#endif + decay_length = exc_length>>1; + for (i=0;i= pitch_index) { + j -= pitch_index; + attenuation = MULT16_16_Q15(attenuation, decay); + } + buf[DECODE_BUFFER_SIZE-N+i] = + SHL32(EXTEND32(MULT16_16_Q15(attenuation, + exc[extrapolation_offset+j])), SIG_SHIFT); + /* Compute the energy of the previously decoded signal whose + excitation we're copying. */ + tmp = SROUND16( + buf[DECODE_BUFFER_SIZE-MAX_PERIOD-N+extrapolation_offset+j], + SIG_SHIFT); + S1 += SHR32(MULT16_16(tmp, tmp), 10); + } + { + opus_val16 lpc_mem[LPC_ORDER]; + /* Copy the last decoded samples (prior to the overlap region) to + synthesis filter memory so we can have a continuous signal. */ + for (i=0;iarch); +#ifdef FIXED_POINT + for (i=0; i < extrapolation_len; i++) + buf[DECODE_BUFFER_SIZE-N+i] = SATURATE(buf[DECODE_BUFFER_SIZE-N+i], SIG_SAT); +#endif + } + + /* Check if the synthesis energy is higher than expected, which can + happen with the signal changes during our window. If so, + attenuate. */ + { + opus_val32 S2=0; + for (i=0;i SHR32(S2,2))) +#else + /* The float test is written this way to catch NaNs in the output + of the IIR filter at the same time. */ + if (!(S1 > 0.2f*S2)) +#endif + { + for (i=0;ipostfilter_period, st->postfilter_period, overlap, + -st->postfilter_gain, -st->postfilter_gain, + st->postfilter_tapset, st->postfilter_tapset, NULL, 0, st->arch); + + /* Simulate TDAC on the concealed audio so that it blends with the + MDCT of the next frame. */ + for (i=0;iloss_duration = IMIN(10000, loss_duration+(1<channels; + int LM, M; + int start; + int end; + int effEnd; + int codedBands; + int alloc_trim; + int postfilter_pitch; + opus_val16 postfilter_gain; + int intensity=0; + int dual_stereo=0; + opus_int32 total_bits; + opus_int32 balance; + opus_int32 tell; + int dynalloc_logp; + int postfilter_tapset; + int anti_collapse_rsv; + int anti_collapse_on=0; + int silence; + int C = st->stream_channels; + const OpusCustomMode *mode; + int nbEBands; + int overlap; + const opus_int16 *eBands; + opus_val16 max_background_increase; + ALLOC_STACK; + + VALIDATE_CELT_DECODER(st); + mode = st->mode; + nbEBands = mode->nbEBands; + overlap = mode->overlap; + eBands = mode->eBands; + start = st->start; + end = st->end; + frame_size *= st->downsample; + + lpc = (opus_val16*)(st->_decode_mem+(DECODE_BUFFER_SIZE+overlap)*CC); + oldBandE = lpc+CC*LPC_ORDER; + oldLogE = oldBandE + 2*nbEBands; + oldLogE2 = oldLogE + 2*nbEBands; + backgroundLogE = oldLogE2 + 2*nbEBands; + +#ifdef CUSTOM_MODES + if (st->signalling && data!=NULL) + { + int data0=data[0]; + /* Convert "standard mode" to Opus header */ + if (mode->Fs==48000 && mode->shortMdctSize==120) + { + data0 = fromOpus(data0); + if (data0<0) + return OPUS_INVALID_PACKET; + } + st->end = end = IMAX(1, mode->effEBands-2*(data0>>5)); + LM = (data0>>3)&0x3; + C = 1 + ((data0>>2)&0x1); + data++; + len--; + if (LM>mode->maxLM) + return OPUS_INVALID_PACKET; + if (frame_size < mode->shortMdctSize<shortMdctSize<maxLM;LM++) + if (mode->shortMdctSize<mode->maxLM) + return OPUS_BAD_ARG; + } + M=1<1275 || pcm==NULL) + return OPUS_BAD_ARG; + + N = M*mode->shortMdctSize; + c=0; do { + decode_mem[c] = st->_decode_mem + c*(DECODE_BUFFER_SIZE+overlap); + out_syn[c] = decode_mem[c]+DECODE_BUFFER_SIZE-N; + } while (++c mode->effEBands) + effEnd = mode->effEBands; + + if (data == NULL || len<=1) + { + celt_decode_lost(st, N, LM); + deemphasis(out_syn, pcm, N, CC, st->downsample, mode->preemph, st->preemph_memD, accum); + RESTORE_STACK; + return frame_size/st->downsample; + } + + /* Check if there are at least two packets received consecutively before + * turning on the pitch-based PLC */ + st->skip_plc = st->loss_duration != 0; + + if (dec == NULL) + { + ec_dec_init(&_dec,(unsigned char*)data,len); + dec = &_dec; + } + + if (C==1) + { + for (i=0;i= total_bits) + silence = 1; + else if (tell==1) + silence = ec_dec_bit_logp(dec, 15); + else + silence = 0; + if (silence) + { + /* Pretend we've read all the remaining bits */ + tell = len*8; + dec->nbits_total+=tell-ec_tell(dec); + } + + postfilter_gain = 0; + postfilter_pitch = 0; + postfilter_tapset = 0; + if (start==0 && tell+16 <= total_bits) + { + if(ec_dec_bit_logp(dec, 1)) + { + int qg, octave; + octave = ec_dec_uint(dec, 6); + postfilter_pitch = (16< 0 && tell+3 <= total_bits) + { + isTransient = ec_dec_bit_logp(dec, 3); + tell = ec_tell(dec); + } + else + isTransient = 0; + + if (isTransient) + shortBlocks = M; + else + shortBlocks = 0; + + /* Decode the global flags (first symbols in the stream) */ + intra_ener = tell+3<=total_bits ? ec_dec_bit_logp(dec, 3) : 0; + /* Get band energies */ + unquant_coarse_energy(mode, start, end, oldBandE, + intra_ener, dec, C, LM); + + ALLOC(tf_res, nbEBands, int); + tf_decode(start, end, isTransient, tf_res, LM, dec); + + tell = ec_tell(dec); + spread_decision = SPREAD_NORMAL; + if (tell+4 <= total_bits) + spread_decision = ec_dec_icdf(dec, spread_icdf, 5); + + ALLOC(cap, nbEBands, int); + + init_caps(mode,cap,LM,C); + + ALLOC(offsets, nbEBands, int); + + dynalloc_logp = 6; + total_bits<<=BITRES; + tell = ec_tell_frac(dec); + for (i=start;i0) + dynalloc_logp = IMAX(2, dynalloc_logp-1); + } + + ALLOC(fine_quant, nbEBands, int); + alloc_trim = tell+(6<=2&&bits>=((LM+2)<rng, 0, + st->arch, st->disable_inv); + + if (anti_collapse_rsv > 0) + { + anti_collapse_on = ec_dec_bits(dec, 1); + } + + unquant_energy_finalise(mode, start, end, oldBandE, + fine_quant, fine_priority, len*8-ec_tell(dec), dec, C); + + if (anti_collapse_on) + anti_collapse(mode, X, collapse_masks, LM, C, N, + start, end, oldBandE, oldLogE, oldLogE2, pulses, st->rng, st->arch); + + if (silence) + { + for (i=0;idownsample, silence, st->arch); + + c=0; do { + st->postfilter_period=IMAX(st->postfilter_period, COMBFILTER_MINPERIOD); + st->postfilter_period_old=IMAX(st->postfilter_period_old, COMBFILTER_MINPERIOD); + comb_filter(out_syn[c], out_syn[c], st->postfilter_period_old, st->postfilter_period, mode->shortMdctSize, + st->postfilter_gain_old, st->postfilter_gain, st->postfilter_tapset_old, st->postfilter_tapset, + mode->window, overlap, st->arch); + if (LM!=0) + comb_filter(out_syn[c]+mode->shortMdctSize, out_syn[c]+mode->shortMdctSize, st->postfilter_period, postfilter_pitch, N-mode->shortMdctSize, + st->postfilter_gain, postfilter_gain, st->postfilter_tapset, postfilter_tapset, + mode->window, overlap, st->arch); + + } while (++cpostfilter_period_old = st->postfilter_period; + st->postfilter_gain_old = st->postfilter_gain; + st->postfilter_tapset_old = st->postfilter_tapset; + st->postfilter_period = postfilter_pitch; + st->postfilter_gain = postfilter_gain; + st->postfilter_tapset = postfilter_tapset; + if (LM!=0) + { + st->postfilter_period_old = st->postfilter_period; + st->postfilter_gain_old = st->postfilter_gain; + st->postfilter_tapset_old = st->postfilter_tapset; + } + + if (C==1) + OPUS_COPY(&oldBandE[nbEBands], oldBandE, nbEBands); + + if (!isTransient) + { + OPUS_COPY(oldLogE2, oldLogE, 2*nbEBands); + OPUS_COPY(oldLogE, oldBandE, 2*nbEBands); + } else { + for (i=0;i<2*nbEBands;i++) + oldLogE[i] = MIN16(oldLogE[i], oldBandE[i]); + } + /* In normal circumstances, we only allow the noise floor to increase by + up to 2.4 dB/second, but when we're in DTX we give the weight of + all missing packets to the update packet. */ + max_background_increase = IMIN(160, st->loss_duration+M)*QCONST16(0.001f,DB_SHIFT); + for (i=0;i<2*nbEBands;i++) + backgroundLogE[i] = MIN16(backgroundLogE[i] + max_background_increase, oldBandE[i]); + /* In case start or end were to change */ + c=0; do + { + for (i=0;irng = dec->rng; + + deemphasis(out_syn, pcm, N, CC, st->downsample, mode->preemph, st->preemph_memD, accum); + st->loss_duration = 0; + RESTORE_STACK; + if (ec_tell(dec) > 8*len) + return OPUS_INTERNAL_ERROR; + if(ec_get_error(dec)) + st->error = 1; + return frame_size/st->downsample; +} + + +#ifdef CUSTOM_MODES + +#ifdef FIXED_POINT +int opus_custom_decode(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, opus_int16 * OPUS_RESTRICT pcm, int frame_size) +{ + return celt_decode_with_ec(st, data, len, pcm, frame_size, NULL, 0); +} + +#ifndef DISABLE_FLOAT_API +int opus_custom_decode_float(CELTDecoder * OPUS_RESTRICT st, const unsigned char *data, int len, float * OPUS_RESTRICT pcm, int frame_size) +{ + int j, ret, C, N; + VARDECL(opus_int16, out); + ALLOC_STACK; + + if (pcm==NULL) + return OPUS_BAD_ARG; + + C = st->channels; + N = frame_size; + + ALLOC(out, C*N, opus_int16); + ret=celt_decode_with_ec(st, data, len, out, frame_size, NULL, 0); + if (ret>0) + for (j=0;jchannels; + N = frame_size; + ALLOC(out, C*N, celt_sig); + + ret=celt_decode_with_ec(st, data, len, out, frame_size, NULL, 0); + + if (ret>0) + for (j=0;j=st->mode->nbEBands) + goto bad_arg; + st->start = value; + } + break; + case CELT_SET_END_BAND_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>st->mode->nbEBands) + goto bad_arg; + st->end = value; + } + break; + case CELT_SET_CHANNELS_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>2) + goto bad_arg; + st->stream_channels = value; + } + break; + case CELT_GET_AND_CLEAR_ERROR_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (value==NULL) + goto bad_arg; + *value=st->error; + st->error = 0; + } + break; + case OPUS_GET_LOOKAHEAD_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (value==NULL) + goto bad_arg; + *value = st->overlap/st->downsample; + } + break; + case OPUS_RESET_STATE: + { + int i; + opus_val16 *lpc, *oldBandE, *oldLogE, *oldLogE2; + lpc = (opus_val16*)(st->_decode_mem+(DECODE_BUFFER_SIZE+st->overlap)*st->channels); + oldBandE = lpc+st->channels*LPC_ORDER; + oldLogE = oldBandE + 2*st->mode->nbEBands; + oldLogE2 = oldLogE + 2*st->mode->nbEBands; + OPUS_CLEAR((char*)&st->DECODER_RESET_START, + opus_custom_decoder_get_size(st->mode, st->channels)- + ((char*)&st->DECODER_RESET_START - (char*)st)); + for (i=0;i<2*st->mode->nbEBands;i++) + oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT); + st->skip_plc = 1; + } + break; + case OPUS_GET_PITCH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (value==NULL) + goto bad_arg; + *value = st->postfilter_period; + } + break; + case CELT_GET_MODE_REQUEST: + { + const CELTMode ** value = va_arg(ap, const CELTMode**); + if (value==0) + goto bad_arg; + *value=st->mode; + } + break; + case CELT_SET_SIGNALLING_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->signalling = value; + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + opus_uint32 * value = va_arg(ap, opus_uint32 *); + if (value==0) + goto bad_arg; + *value=st->rng; + } + break; + case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>1) + { + goto bad_arg; + } + st->disable_inv = value; + } + break; + case OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->disable_inv; + } + break; + default: + goto bad_request; + } + va_end(ap); + return OPUS_OK; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +bad_request: + va_end(ap); + return OPUS_UNIMPLEMENTED; +} diff --git a/lib/opus/celt/celt_encoder.c b/lib/opus/celt/celt_encoder.c new file mode 100644 index 00000000..637d442c --- /dev/null +++ b/lib/opus/celt/celt_encoder.c @@ -0,0 +1,2613 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2010 Xiph.Org Foundation + Copyright (c) 2008 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#define CELT_ENCODER_C + +#include "cpu_support.h" +#include "os_support.h" +#include "mdct.h" +#include +#include "celt.h" +#include "pitch.h" +#include "bands.h" +#include "modes.h" +#include "entcode.h" +#include "quant_bands.h" +#include "rate.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "float_cast.h" +#include +#include "celt_lpc.h" +#include "vq.h" + + +/** Encoder state + @brief Encoder state + */ +struct OpusCustomEncoder { + const OpusCustomMode *mode; /**< Mode used by the encoder */ + int channels; + int stream_channels; + + int force_intra; + int clip; + int disable_pf; + int complexity; + int upsample; + int start, end; + + opus_int32 bitrate; + int vbr; + int signalling; + int constrained_vbr; /* If zero, VBR can do whatever it likes with the rate */ + int loss_rate; + int lsb_depth; + int lfe; + int disable_inv; + int arch; + + /* Everything beyond this point gets cleared on a reset */ +#define ENCODER_RESET_START rng + + opus_uint32 rng; + int spread_decision; + opus_val32 delayedIntra; + int tonal_average; + int lastCodedBands; + int hf_average; + int tapset_decision; + + int prefilter_period; + opus_val16 prefilter_gain; + int prefilter_tapset; +#ifdef RESYNTH + int prefilter_period_old; + opus_val16 prefilter_gain_old; + int prefilter_tapset_old; +#endif + int consec_transient; + AnalysisInfo analysis; + SILKInfo silk_info; + + opus_val32 preemph_memE[2]; + opus_val32 preemph_memD[2]; + + /* VBR-related parameters */ + opus_int32 vbr_reservoir; + opus_int32 vbr_drift; + opus_int32 vbr_offset; + opus_int32 vbr_count; + opus_val32 overlap_max; + opus_val16 stereo_saving; + int intensity; + opus_val16 *energy_mask; + opus_val16 spec_avg; + +#ifdef RESYNTH + /* +MAX_PERIOD/2 to make space for overlap */ + celt_sig syn_mem[2][2*MAX_PERIOD+MAX_PERIOD/2]; +#endif + + celt_sig in_mem[1]; /* Size = channels*mode->overlap */ + /* celt_sig prefilter_mem[], Size = channels*COMBFILTER_MAXPERIOD */ + /* opus_val16 oldBandE[], Size = channels*mode->nbEBands */ + /* opus_val16 oldLogE[], Size = channels*mode->nbEBands */ + /* opus_val16 oldLogE2[], Size = channels*mode->nbEBands */ + /* opus_val16 energyError[], Size = channels*mode->nbEBands */ +}; + +int celt_encoder_get_size(int channels) +{ + CELTMode *mode = opus_custom_mode_create(48000, 960, NULL); + return opus_custom_encoder_get_size(mode, channels); +} + +OPUS_CUSTOM_NOSTATIC int opus_custom_encoder_get_size(const CELTMode *mode, int channels) +{ + int size = sizeof(struct CELTEncoder) + + (channels*mode->overlap-1)*sizeof(celt_sig) /* celt_sig in_mem[channels*mode->overlap]; */ + + channels*COMBFILTER_MAXPERIOD*sizeof(celt_sig) /* celt_sig prefilter_mem[channels*COMBFILTER_MAXPERIOD]; */ + + 4*channels*mode->nbEBands*sizeof(opus_val16); /* opus_val16 oldBandE[channels*mode->nbEBands]; */ + /* opus_val16 oldLogE[channels*mode->nbEBands]; */ + /* opus_val16 oldLogE2[channels*mode->nbEBands]; */ + /* opus_val16 energyError[channels*mode->nbEBands]; */ + return size; +} + +#ifdef CUSTOM_MODES +CELTEncoder *opus_custom_encoder_create(const CELTMode *mode, int channels, int *error) +{ + int ret; + CELTEncoder *st = (CELTEncoder *)opus_alloc(opus_custom_encoder_get_size(mode, channels)); + /* init will handle the NULL case */ + ret = opus_custom_encoder_init(st, mode, channels); + if (ret != OPUS_OK) + { + opus_custom_encoder_destroy(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} +#endif /* CUSTOM_MODES */ + +static int opus_custom_encoder_init_arch(CELTEncoder *st, const CELTMode *mode, + int channels, int arch) +{ + if (channels < 0 || channels > 2) + return OPUS_BAD_ARG; + + if (st==NULL || mode==NULL) + return OPUS_ALLOC_FAIL; + + OPUS_CLEAR((char*)st, opus_custom_encoder_get_size(mode, channels)); + + st->mode = mode; + st->stream_channels = st->channels = channels; + + st->upsample = 1; + st->start = 0; + st->end = st->mode->effEBands; + st->signalling = 1; + st->arch = arch; + + st->constrained_vbr = 1; + st->clip = 1; + + st->bitrate = OPUS_BITRATE_MAX; + st->vbr = 0; + st->force_intra = 0; + st->complexity = 5; + st->lsb_depth=24; + + opus_custom_encoder_ctl(st, OPUS_RESET_STATE); + + return OPUS_OK; +} + +#ifdef CUSTOM_MODES +int opus_custom_encoder_init(CELTEncoder *st, const CELTMode *mode, int channels) +{ + return opus_custom_encoder_init_arch(st, mode, channels, opus_select_arch()); +} +#endif + +int celt_encoder_init(CELTEncoder *st, opus_int32 sampling_rate, int channels, + int arch) +{ + int ret; + ret = opus_custom_encoder_init_arch(st, + opus_custom_mode_create(48000, 960, NULL), channels, arch); + if (ret != OPUS_OK) + return ret; + st->upsample = resampling_factor(sampling_rate); + return OPUS_OK; +} + +#ifdef CUSTOM_MODES +void opus_custom_encoder_destroy(CELTEncoder *st) +{ + opus_free(st); +} +#endif /* CUSTOM_MODES */ + + +static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int C, + opus_val16 *tf_estimate, int *tf_chan, int allow_weak_transients, + int *weak_transient) +{ + int i; + VARDECL(opus_val16, tmp); + opus_val32 mem0,mem1; + int is_transient = 0; + opus_int32 mask_metric = 0; + int c; + opus_val16 tf_max; + int len2; + /* Forward masking: 6.7 dB/ms. */ +#ifdef FIXED_POINT + int forward_shift = 4; +#else + opus_val16 forward_decay = QCONST16(.0625f,15); +#endif + /* Table of 6*64/x, trained on real data to minimize the average error */ + static const unsigned char inv_table[128] = { + 255,255,156,110, 86, 70, 59, 51, 45, 40, 37, 33, 31, 28, 26, 25, + 23, 22, 21, 20, 19, 18, 17, 16, 16, 15, 15, 14, 13, 13, 12, 12, + 12, 12, 11, 11, 11, 10, 10, 10, 9, 9, 9, 9, 9, 9, 8, 8, + 8, 8, 8, 7, 7, 7, 7, 7, 7, 6, 6, 6, 6, 6, 6, 6, + 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 5, 5, 5, 5, 5, + 5, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, + 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, + 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, + }; + SAVE_STACK; + ALLOC(tmp, len, opus_val16); + + *weak_transient = 0; + /* For lower bitrates, let's be more conservative and have a forward masking + decay of 3.3 dB/ms. This avoids having to code transients at very low + bitrate (mostly for hybrid), which can result in unstable energy and/or + partial collapse. */ + if (allow_weak_transients) + { +#ifdef FIXED_POINT + forward_shift = 5; +#else + forward_decay = QCONST16(.03125f,15); +#endif + } + len2=len/2; + for (c=0;c=0;i--) + { + /* Backward masking: 13.9 dB/ms. */ +#ifdef FIXED_POINT + /* FIXME: Use PSHR16() instead */ + tmp[i] = mem0 + PSHR32(tmp[i]-mem0,3); +#else + tmp[i] = mem0 + MULT16_16_P15(QCONST16(0.125f,15),tmp[i]-mem0); +#endif + mem0 = tmp[i]; + maxE = MAX16(maxE, mem0); + } + /*for (i=0;i>1))); +#else + mean = celt_sqrt(mean * maxE*.5*len2); +#endif + /* Inverse of the mean energy in Q15+6 */ + norm = SHL32(EXTEND32(len2),6+14)/ADD32(EPSILON,SHR32(mean,1)); + /* Compute harmonic mean discarding the unreliable boundaries + The data is smooth, so we only take 1/4th of the samples */ + unmask=0; + /* We should never see NaNs here. If we find any, then something really bad happened and we better abort + before it does any damage later on. If these asserts are disabled (no hardening), then the table + lookup a few lines below (id = ...) is likely to crash dur to an out-of-bounds read. DO NOT FIX + that crash on NaN since it could result in a worse issue later on. */ + celt_assert(!celt_isnan(tmp[0])); + celt_assert(!celt_isnan(norm)); + for (i=12;imask_metric) + { + *tf_chan = c; + mask_metric = unmask; + } + } + is_transient = mask_metric>200; + /* For low bitrates, define "weak transients" that need to be + handled differently to avoid partial collapse. */ + if (allow_weak_transients && is_transient && mask_metric<600) { + is_transient = 0; + *weak_transient = 1; + } + /* Arbitrary metric for VBR boost */ + tf_max = MAX16(0,celt_sqrt(27*mask_metric)-42); + /* *tf_estimate = 1 + MIN16(1, sqrt(MAX16(0, tf_max-30))/20); */ + *tf_estimate = celt_sqrt(MAX32(0, SHL32(MULT16_16(QCONST16(0.0069,14),MIN16(163,tf_max)),14)-QCONST32(0.139,28))); + /*printf("%d %f\n", tf_max, mask_metric);*/ + RESTORE_STACK; +#ifdef FUZZING + is_transient = rand()&0x1; +#endif + /*printf("%d %f %d\n", is_transient, (float)*tf_estimate, tf_max);*/ + return is_transient; +} + +/* Looks for sudden increases of energy to decide whether we need to patch + the transient decision */ +static int patch_transient_decision(opus_val16 *newE, opus_val16 *oldE, int nbEBands, + int start, int end, int C) +{ + int i, c; + opus_val32 mean_diff=0; + opus_val16 spread_old[26]; + /* Apply an aggressive (-6 dB/Bark) spreading function to the old frame to + avoid false detection caused by irrelevant bands */ + if (C==1) + { + spread_old[start] = oldE[start]; + for (i=start+1;i=start;i--) + spread_old[i] = MAX16(spread_old[i], spread_old[i+1]-QCONST16(1.0f, DB_SHIFT)); + /* Compute mean increase */ + c=0; do { + for (i=IMAX(2,start);i QCONST16(1.f, DB_SHIFT); +} + +/** Apply window and compute the MDCT for all sub-frames and + all channels in a frame */ +static void compute_mdcts(const CELTMode *mode, int shortBlocks, celt_sig * OPUS_RESTRICT in, + celt_sig * OPUS_RESTRICT out, int C, int CC, int LM, int upsample, + int arch) +{ + const int overlap = mode->overlap; + int N; + int B; + int shift; + int i, b, c; + if (shortBlocks) + { + B = shortBlocks; + N = mode->shortMdctSize; + shift = mode->maxLM; + } else { + B = 1; + N = mode->shortMdctSize<maxLM-LM; + } + c=0; do { + for (b=0;bmdct, in+c*(B*N+overlap)+b*N, + &out[b+c*N*B], mode->window, overlap, shift, B, + arch); + } + } while (++ceBands[len]-m->eBands[len-1])<eBands[len]-m->eBands[len-1])<eBands[i+1]-m->eBands[i])<eBands[i+1]-m->eBands[i])==1; + OPUS_COPY(tmp, &X[tf_chan*N0 + (m->eBands[i]<eBands[i]<>LM, 1<>k, 1<=0;i--) + { + if (tf_res[i+1] == 1) + tf_res[i] = path1[i+1]; + else + tf_res[i] = path0[i+1]; + } + /*printf("%d %f\n", *tf_sum, tf_estimate);*/ + RESTORE_STACK; +#ifdef FUZZING + tf_select = rand()&0x1; + tf_res[0] = rand()&0x1; + for (i=1;istorage*8; + tell = ec_tell(enc); + logp = isTransient ? 2 : 4; + /* Reserve space to code the tf_select decision. */ + tf_select_rsv = LM>0 && tell+logp+1 <= budget; + budget -= tf_select_rsv; + curr = tf_changed = 0; + for (i=start;i> 10; + trim = QCONST16(4.f, 8) + QCONST16(1.f/16.f, 8)*frac; + } + if (C==2) + { + opus_val16 sum = 0; /* Q10 */ + opus_val16 minXC; /* Q10 */ + /* Compute inter-channel correlation for low frequencies */ + for (i=0;i<8;i++) + { + opus_val32 partial; + partial = celt_inner_prod(&X[m->eBands[i]<eBands[i]<eBands[i+1]-m->eBands[i])<eBands[i]<eBands[i]<eBands[i+1]-m->eBands[i])<nbEBands]*(opus_int32)(2+2*i-end); + } + } while (++cvalid) + { + trim -= MAX16(-QCONST16(2.f, 8), MIN16(QCONST16(2.f, 8), + (opus_val16)(QCONST16(2.f, 8)*(analysis->tonality_slope+.05f)))); + } +#else + (void)analysis; +#endif + +#ifdef FIXED_POINT + trim_index = PSHR32(trim, 8); +#else + trim_index = (int)floor(.5f+trim); +#endif + trim_index = IMAX(0, IMIN(10, trim_index)); + /*printf("%d\n", trim_index);*/ +#ifdef FUZZING + trim_index = rand()%11; +#endif + return trim_index; +} + +static int stereo_analysis(const CELTMode *m, const celt_norm *X, + int LM, int N0) +{ + int i; + int thetas; + opus_val32 sumLR = EPSILON, sumMS = EPSILON; + + /* Use the L1 norm to model the entropy of the L/R signal vs the M/S signal */ + for (i=0;i<13;i++) + { + int j; + for (j=m->eBands[i]<eBands[i+1]<eBands[13]<<(LM+1))+thetas, sumMS) + > MULT16_32_Q15(m->eBands[13]<<(LM+1), sumLR); +} + +#define MSWAP(a,b) do {opus_val16 tmp = a;a=b;b=tmp;} while(0) +static opus_val16 median_of_5(const opus_val16 *x) +{ + opus_val16 t0, t1, t2, t3, t4; + t2 = x[2]; + if (x[0] > x[1]) + { + t0 = x[1]; + t1 = x[0]; + } else { + t0 = x[0]; + t1 = x[1]; + } + if (x[3] > x[4]) + { + t3 = x[4]; + t4 = x[3]; + } else { + t3 = x[3]; + t4 = x[4]; + } + if (t0 > t3) + { + MSWAP(t0, t3); + MSWAP(t1, t4); + } + if (t2 > t1) + { + if (t1 < t3) + return MIN16(t2, t3); + else + return MIN16(t4, t1); + } else { + if (t2 < t3) + return MIN16(t1, t3); + else + return MIN16(t2, t4); + } +} + +static opus_val16 median_of_3(const opus_val16 *x) +{ + opus_val16 t0, t1, t2; + if (x[0] > x[1]) + { + t0 = x[1]; + t1 = x[0]; + } else { + t0 = x[0]; + t1 = x[1]; + } + t2 = x[2]; + if (t1 < t2) + return t1; + else if (t0 < t2) + return t2; + else + return t0; +} + +static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16 *bandLogE2, + int nbEBands, int start, int end, int C, int *offsets, int lsb_depth, const opus_int16 *logN, + int isTransient, int vbr, int constrained_vbr, const opus_int16 *eBands, int LM, + int effectiveBytes, opus_int32 *tot_boost_, int lfe, opus_val16 *surround_dynalloc, + AnalysisInfo *analysis, int *importance, int *spread_weight) +{ + int i, c; + opus_int32 tot_boost=0; + opus_val16 maxDepth; + VARDECL(opus_val16, follower); + VARDECL(opus_val16, noise_floor); + SAVE_STACK; + ALLOC(follower, C*nbEBands, opus_val16); + ALLOC(noise_floor, C*nbEBands, opus_val16); + OPUS_CLEAR(offsets, nbEBands); + /* Dynamic allocation code */ + maxDepth=-QCONST16(31.9f, DB_SHIFT); + for (i=0;i=0;i--) + mask[i] = MAX16(mask[i], mask[i+1] - QCONST16(3.f, DB_SHIFT)); + for (i=0;i> shift; + } + /*for (i=0;i 50 && LM>=1 && !lfe) + { + int last=0; + c=0;do + { + opus_val16 offset; + opus_val16 tmp; + opus_val16 *f; + f = &follower[c*nbEBands]; + f[0] = bandLogE2[c*nbEBands]; + for (i=1;i bandLogE2[c*nbEBands+i-1]+QCONST16(.5f,DB_SHIFT)) + last=i; + f[i] = MIN16(f[i-1]+QCONST16(1.5f,DB_SHIFT), bandLogE2[c*nbEBands+i]); + } + for (i=last-1;i>=0;i--) + f[i] = MIN16(f[i], MIN16(f[i+1]+QCONST16(2.f,DB_SHIFT), bandLogE2[c*nbEBands+i])); + + /* Combine with a median filter to avoid dynalloc triggering unnecessarily. + The "offset" value controls how conservative we are -- a higher offset + reduces the impact of the median filter and makes dynalloc use more bits. */ + offset = QCONST16(1.f, DB_SHIFT); + for (i=2;i=12) + follower[i] = HALF16(follower[i]); + } +#ifdef DISABLE_FLOAT_API + (void)analysis; +#else + if (analysis->valid) + { + for (i=start;ileak_boost[i]; + } +#endif + for (i=start;i 48) { + boost = (int)SHR32(EXTEND32(follower[i])*8,DB_SHIFT); + boost_bits = (boost*width<>BITRES>>3 > 2*effectiveBytes/3) + { + opus_int32 cap = ((2*effectiveBytes/3)<mode; + overlap = mode->overlap; + ALLOC(_pre, CC*(N+COMBFILTER_MAXPERIOD), celt_sig); + + pre[0] = _pre; + pre[1] = _pre + (N+COMBFILTER_MAXPERIOD); + + + c=0; do { + OPUS_COPY(pre[c], prefilter_mem+c*COMBFILTER_MAXPERIOD, COMBFILTER_MAXPERIOD); + OPUS_COPY(pre[c]+COMBFILTER_MAXPERIOD, in+c*(N+overlap)+overlap, N); + } while (++c>1, opus_val16); + + pitch_downsample(pre, pitch_buf, COMBFILTER_MAXPERIOD+N, CC, st->arch); + /* Don't search for the fir last 1.5 octave of the range because + there's too many false-positives due to short-term correlation */ + pitch_search(pitch_buf+(COMBFILTER_MAXPERIOD>>1), pitch_buf, N, + COMBFILTER_MAXPERIOD-3*COMBFILTER_MINPERIOD, &pitch_index, + st->arch); + pitch_index = COMBFILTER_MAXPERIOD-pitch_index; + + gain1 = remove_doubling(pitch_buf, COMBFILTER_MAXPERIOD, COMBFILTER_MINPERIOD, + N, &pitch_index, st->prefilter_period, st->prefilter_gain, st->arch); + if (pitch_index > COMBFILTER_MAXPERIOD-2) + pitch_index = COMBFILTER_MAXPERIOD-2; + gain1 = MULT16_16_Q15(QCONST16(.7f,15),gain1); + /*printf("%d %d %f %f\n", pitch_change, pitch_index, gain1, st->analysis.tonality);*/ + if (st->loss_rate>2) + gain1 = HALF32(gain1); + if (st->loss_rate>4) + gain1 = HALF32(gain1); + if (st->loss_rate>8) + gain1 = 0; + } else { + gain1 = 0; + pitch_index = COMBFILTER_MINPERIOD; + } +#ifndef DISABLE_FLOAT_API + if (analysis->valid) + gain1 = (opus_val16)(gain1 * analysis->max_pitch_ratio); +#else + (void)analysis; +#endif + /* Gain threshold for enabling the prefilter/postfilter */ + pf_threshold = QCONST16(.2f,15); + + /* Adjusting the threshold based on rate and continuity */ + if (abs(pitch_index-st->prefilter_period)*10>pitch_index) + pf_threshold += QCONST16(.2f,15); + if (nbAvailableBytes<25) + pf_threshold += QCONST16(.1f,15); + if (nbAvailableBytes<35) + pf_threshold += QCONST16(.1f,15); + if (st->prefilter_gain > QCONST16(.4f,15)) + pf_threshold -= QCONST16(.1f,15); + if (st->prefilter_gain > QCONST16(.55f,15)) + pf_threshold -= QCONST16(.1f,15); + + /* Hard threshold at 0.2 */ + pf_threshold = MAX16(pf_threshold, QCONST16(.2f,15)); + if (gain1prefilter_gain)prefilter_gain; + +#ifdef FIXED_POINT + qg = ((gain1+1536)>>10)/3-1; +#else + qg = (int)floor(.5f+gain1*32/3)-1; +#endif + qg = IMAX(0, IMIN(7, qg)); + gain1 = QCONST16(0.09375f,15)*(qg+1); + pf_on = 1; + } + /*printf("%d %f\n", pitch_index, gain1);*/ + + c=0; do { + int offset = mode->shortMdctSize-overlap; + st->prefilter_period=IMAX(st->prefilter_period, COMBFILTER_MINPERIOD); + OPUS_COPY(in+c*(N+overlap), st->in_mem+c*(overlap), overlap); + if (offset) + comb_filter(in+c*(N+overlap)+overlap, pre[c]+COMBFILTER_MAXPERIOD, + st->prefilter_period, st->prefilter_period, offset, -st->prefilter_gain, -st->prefilter_gain, + st->prefilter_tapset, st->prefilter_tapset, NULL, 0, st->arch); + + comb_filter(in+c*(N+overlap)+overlap+offset, pre[c]+COMBFILTER_MAXPERIOD+offset, + st->prefilter_period, pitch_index, N-offset, -st->prefilter_gain, -gain1, + st->prefilter_tapset, prefilter_tapset, mode->window, overlap, st->arch); + OPUS_COPY(st->in_mem+c*(overlap), in+c*(N+overlap)+N, overlap); + + if (N>COMBFILTER_MAXPERIOD) + { + OPUS_COPY(prefilter_mem+c*COMBFILTER_MAXPERIOD, pre[c]+N, COMBFILTER_MAXPERIOD); + } else { + OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, prefilter_mem+c*COMBFILTER_MAXPERIOD+N, COMBFILTER_MAXPERIOD-N); + OPUS_COPY(prefilter_mem+c*COMBFILTER_MAXPERIOD+COMBFILTER_MAXPERIOD-N, pre[c]+COMBFILTER_MAXPERIOD, N); + } + } while (++cnbEBands; + eBands = mode->eBands; + + coded_bands = lastCodedBands ? lastCodedBands : nbEBands; + coded_bins = eBands[coded_bands]<analysis.activity, st->analysis.tonality, tf_estimate, st->stereo_saving, tot_boost, coded_bands);*/ +#ifndef DISABLE_FLOAT_API + if (analysis->valid && analysis->activity<.4) + target -= (opus_int32)((coded_bins<activity)); +#endif + /* Stereo savings */ + if (C==2) + { + int coded_stereo_bands; + int coded_stereo_dof; + opus_val16 max_frac; + coded_stereo_bands = IMIN(intensity, coded_bands); + coded_stereo_dof = (eBands[coded_stereo_bands]<valid && !lfe) + { + opus_int32 tonal_target; + float tonal; + + /* Tonality boost (compensating for the average). */ + tonal = MAX16(0.f,analysis->tonality-.15f)-0.12f; + tonal_target = target + (opus_int32)((coded_bins<tonality, tonal);*/ + target = tonal_target; + } +#else + (void)analysis; + (void)pitch_change; +#endif + + if (has_surround_mask&&!lfe) + { + opus_int32 surround_target = target + (opus_int32)SHR32(MULT16_16(surround_masking,coded_bins<end, st->intensity, surround_target, target, st->bitrate);*/ + target = IMAX(target/4, surround_target); + } + + { + opus_int32 floor_depth; + int bins; + bins = eBands[nbEBands-2]<>2); + target = IMIN(target, floor_depth); + /*printf("%f %d\n", maxDepth, floor_depth);*/ + } + + /* Make VBR less aggressive for constrained VBR because we can't keep a higher bitrate + for long. Needs tuning. */ + if ((!has_surround_mask||lfe) && constrained_vbr) + { + target = base_target + (opus_int32)MULT16_32_Q15(QCONST16(0.67f, 15), target-base_target); + } + + if (!has_surround_mask && tf_estimate < QCONST16(.2f, 14)) + { + opus_val16 amount; + opus_val16 tvbr_factor; + amount = MULT16_16_Q15(QCONST16(.0000031f, 30), IMAX(0, IMIN(32000, 96000-bitrate))); + tvbr_factor = SHR32(MULT16_16(temporal_vbr, amount), DB_SHIFT); + target += (opus_int32)MULT16_32_Q15(tvbr_factor, target); + } + + /* Don't allow more than doubling the rate */ + target = IMIN(2*base_target, target); + + return target; +} + +int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes, ec_enc *enc) +{ + int i, c, N; + opus_int32 bits; + ec_enc _enc; + VARDECL(celt_sig, in); + VARDECL(celt_sig, freq); + VARDECL(celt_norm, X); + VARDECL(celt_ener, bandE); + VARDECL(opus_val16, bandLogE); + VARDECL(opus_val16, bandLogE2); + VARDECL(int, fine_quant); + VARDECL(opus_val16, error); + VARDECL(int, pulses); + VARDECL(int, cap); + VARDECL(int, offsets); + VARDECL(int, importance); + VARDECL(int, spread_weight); + VARDECL(int, fine_priority); + VARDECL(int, tf_res); + VARDECL(unsigned char, collapse_masks); + celt_sig *prefilter_mem; + opus_val16 *oldBandE, *oldLogE, *oldLogE2, *energyError; + int shortBlocks=0; + int isTransient=0; + const int CC = st->channels; + const int C = st->stream_channels; + int LM, M; + int tf_select; + int nbFilledBytes, nbAvailableBytes; + int start; + int end; + int effEnd; + int codedBands; + int alloc_trim; + int pitch_index=COMBFILTER_MINPERIOD; + opus_val16 gain1 = 0; + int dual_stereo=0; + int effectiveBytes; + int dynalloc_logp; + opus_int32 vbr_rate; + opus_int32 total_bits; + opus_int32 total_boost; + opus_int32 balance; + opus_int32 tell; + opus_int32 tell0_frac; + int prefilter_tapset=0; + int pf_on; + int anti_collapse_rsv; + int anti_collapse_on=0; + int silence=0; + int tf_chan = 0; + opus_val16 tf_estimate; + int pitch_change=0; + opus_int32 tot_boost; + opus_val32 sample_max; + opus_val16 maxDepth; + const OpusCustomMode *mode; + int nbEBands; + int overlap; + const opus_int16 *eBands; + int secondMdct; + int signalBandwidth; + int transient_got_disabled=0; + opus_val16 surround_masking=0; + opus_val16 temporal_vbr=0; + opus_val16 surround_trim = 0; + opus_int32 equiv_rate; + int hybrid; + int weak_transient = 0; + int enable_tf_analysis; + VARDECL(opus_val16, surround_dynalloc); + ALLOC_STACK; + + mode = st->mode; + nbEBands = mode->nbEBands; + overlap = mode->overlap; + eBands = mode->eBands; + start = st->start; + end = st->end; + hybrid = start != 0; + tf_estimate = 0; + if (nbCompressedBytes<2 || pcm==NULL) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + + frame_size *= st->upsample; + for (LM=0;LM<=mode->maxLM;LM++) + if (mode->shortMdctSize<mode->maxLM) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + M=1<shortMdctSize; + + prefilter_mem = st->in_mem+CC*(overlap); + oldBandE = (opus_val16*)(st->in_mem+CC*(overlap+COMBFILTER_MAXPERIOD)); + oldLogE = oldBandE + CC*nbEBands; + oldLogE2 = oldLogE + CC*nbEBands; + energyError = oldLogE2 + CC*nbEBands; + + if (enc==NULL) + { + tell0_frac=tell=1; + nbFilledBytes=0; + } else { + tell0_frac=ec_tell_frac(enc); + tell=ec_tell(enc); + nbFilledBytes=(tell+4)>>3; + } + +#ifdef CUSTOM_MODES + if (st->signalling && enc==NULL) + { + int tmp = (mode->effEBands-end)>>1; + end = st->end = IMAX(1, mode->effEBands-tmp); + compressed[0] = tmp<<5; + compressed[0] |= LM<<3; + compressed[0] |= (C==2)<<2; + /* Convert "standard mode" to Opus header */ + if (mode->Fs==48000 && mode->shortMdctSize==120) + { + int c0 = toOpus(compressed[0]); + if (c0<0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + compressed[0] = c0; + } + compressed++; + nbCompressedBytes--; + } +#else + celt_assert(st->signalling==0); +#endif + + /* Can't produce more than 1275 output bytes */ + nbCompressedBytes = IMIN(nbCompressedBytes,1275); + nbAvailableBytes = nbCompressedBytes - nbFilledBytes; + + if (st->vbr && st->bitrate!=OPUS_BITRATE_MAX) + { + opus_int32 den=mode->Fs>>BITRES; + vbr_rate=(st->bitrate*frame_size+(den>>1))/den; +#ifdef CUSTOM_MODES + if (st->signalling) + vbr_rate -= 8<>(3+BITRES); + } else { + opus_int32 tmp; + vbr_rate = 0; + tmp = st->bitrate*frame_size; + if (tell>1) + tmp += tell; + if (st->bitrate!=OPUS_BITRATE_MAX) + nbCompressedBytes = IMAX(2, IMIN(nbCompressedBytes, + (tmp+4*mode->Fs)/(8*mode->Fs)-!!st->signalling)); + effectiveBytes = nbCompressedBytes - nbFilledBytes; + } + equiv_rate = ((opus_int32)nbCompressedBytes*8*50 << (3-LM)) - (40*C+20)*((400>>LM) - 50); + if (st->bitrate != OPUS_BITRATE_MAX) + equiv_rate = IMIN(equiv_rate, st->bitrate - (40*C+20)*((400>>LM) - 50)); + + if (enc==NULL) + { + ec_enc_init(&_enc, compressed, nbCompressedBytes); + enc = &_enc; + } + + if (vbr_rate>0) + { + /* Computes the max bit-rate allowed in VBR mode to avoid violating the + target rate and buffering. + We must do this up front so that bust-prevention logic triggers + correctly if we don't have enough bits. */ + if (st->constrained_vbr) + { + opus_int32 vbr_bound; + opus_int32 max_allowed; + /* We could use any multiple of vbr_rate as bound (depending on the + delay). + This is clamped to ensure we use at least two bytes if the encoder + was entirely empty, but to allow 0 in hybrid mode. */ + vbr_bound = vbr_rate; + max_allowed = IMIN(IMAX(tell==1?2:0, + (vbr_rate+vbr_bound-st->vbr_reservoir)>>(BITRES+3)), + nbAvailableBytes); + if(max_allowed < nbAvailableBytes) + { + nbCompressedBytes = nbFilledBytes+max_allowed; + nbAvailableBytes = max_allowed; + ec_enc_shrink(enc, nbCompressedBytes); + } + } + } + total_bits = nbCompressedBytes*8; + + effEnd = end; + if (effEnd > mode->effEBands) + effEnd = mode->effEBands; + + ALLOC(in, CC*(N+overlap), celt_sig); + + sample_max=MAX32(st->overlap_max, celt_maxabs16(pcm, C*(N-overlap)/st->upsample)); + st->overlap_max=celt_maxabs16(pcm+C*(N-overlap)/st->upsample, C*overlap/st->upsample); + sample_max=MAX32(sample_max, st->overlap_max); +#ifdef FIXED_POINT + silence = (sample_max==0); +#else + silence = (sample_max <= (opus_val16)1/(1<lsb_depth)); +#endif +#ifdef FUZZING + if ((rand()&0x3F)==0) + silence = 1; +#endif + if (tell==1) + ec_enc_bit_logp(enc, silence, 15); + else + silence=0; + if (silence) + { + /*In VBR mode there is no need to send more than the minimum. */ + if (vbr_rate>0) + { + effectiveBytes=nbCompressedBytes=IMIN(nbCompressedBytes, nbFilledBytes+2); + total_bits=nbCompressedBytes*8; + nbAvailableBytes=2; + ec_enc_shrink(enc, nbCompressedBytes); + } + /* Pretend we've filled all the remaining bits with zeros + (that's what the initialiser did anyway) */ + tell = nbCompressedBytes*8; + enc->nbits_total+=tell-ec_tell(enc); + } + c=0; do { + int need_clip=0; +#ifndef FIXED_POINT + need_clip = st->clip && sample_max>65536.f; +#endif + celt_preemphasis(pcm+c, in+c*(N+overlap)+overlap, N, CC, st->upsample, + mode->preemph, st->preemph_memE+c, need_clip); + } while (++clfe&&nbAvailableBytes>3) || nbAvailableBytes>12*C) && !hybrid && !silence && !st->disable_pf + && st->complexity >= 5; + + prefilter_tapset = st->tapset_decision; + pf_on = run_prefilter(st, in, prefilter_mem, CC, N, prefilter_tapset, &pitch_index, &gain1, &qg, enabled, nbAvailableBytes, &st->analysis); + if ((gain1 > QCONST16(.4f,15) || st->prefilter_gain > QCONST16(.4f,15)) && (!st->analysis.valid || st->analysis.tonality > .3) + && (pitch_index > 1.26*st->prefilter_period || pitch_index < .79*st->prefilter_period)) + pitch_change = 1; + if (pf_on==0) + { + if(!hybrid && tell+16<=total_bits) + ec_enc_bit_logp(enc, 0, 1); + } else { + /*This block is not gated by a total bits check only because + of the nbAvailableBytes check above.*/ + int octave; + ec_enc_bit_logp(enc, 1, 1); + pitch_index += 1; + octave = EC_ILOG(pitch_index)-5; + ec_enc_uint(enc, octave, 6); + ec_enc_bits(enc, pitch_index-(16<complexity >= 1 && !st->lfe) + { + /* Reduces the likelihood of energy instability on fricatives at low bitrate + in hybrid mode. It seems like we still want to have real transients on vowels + though (small SILK quantization offset value). */ + int allow_weak_transients = hybrid && effectiveBytes<15 && st->silk_info.signalType != 2; + isTransient = transient_analysis(in, N+overlap, CC, + &tf_estimate, &tf_chan, allow_weak_transients, &weak_transient); + } + if (LM>0 && ec_tell(enc)+3<=total_bits) + { + if (isTransient) + shortBlocks = M; + } else { + isTransient = 0; + transient_got_disabled=1; + } + + ALLOC(freq, CC*N, celt_sig); /**< Interleaved signal MDCTs */ + ALLOC(bandE,nbEBands*CC, celt_ener); + ALLOC(bandLogE,nbEBands*CC, opus_val16); + + secondMdct = shortBlocks && st->complexity>=8; + ALLOC(bandLogE2, C*nbEBands, opus_val16); + if (secondMdct) + { + compute_mdcts(mode, 0, in, freq, C, CC, LM, st->upsample, st->arch); + compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch); + amp2Log2(mode, effEnd, end, bandE, bandLogE2, C); + for (c=0;cupsample, st->arch); + /* This should catch any NaN in the CELT input. Since we're not supposed to see any (they're filtered + at the Opus layer), just abort. */ + celt_assert(!celt_isnan(freq[0]) && (C==1 || !celt_isnan(freq[N]))); + if (CC==2&&C==1) + tf_chan = 0; + compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch); + + if (st->lfe) + { + for (i=2;ienergy_mask&&!st->lfe) + { + int mask_end; + int midband; + int count_dynalloc; + opus_val32 mask_avg=0; + opus_val32 diff=0; + int count=0; + mask_end = IMAX(2,st->lastCodedBands); + for (c=0;cenergy_mask[nbEBands*c+i], + QCONST16(.25f, DB_SHIFT)), -QCONST16(2.0f, DB_SHIFT)); + if (mask > 0) + mask = HALF16(mask); + mask_avg += MULT16_16(mask, eBands[i+1]-eBands[i]); + count += eBands[i+1]-eBands[i]; + diff += MULT16_16(mask, 1+2*i-mask_end); + } + } + celt_assert(count>0); + mask_avg = DIV32_16(mask_avg,count); + mask_avg += QCONST16(.2f, DB_SHIFT); + diff = diff*6/(C*(mask_end-1)*(mask_end+1)*mask_end); + /* Again, being conservative */ + diff = HALF32(diff); + diff = MAX32(MIN32(diff, QCONST32(.031f, DB_SHIFT)), -QCONST32(.031f, DB_SHIFT)); + /* Find the band that's in the middle of the coded spectrum */ + for (midband=0;eBands[midband+1] < eBands[mask_end]/2;midband++); + count_dynalloc=0; + for(i=0;ienergy_mask[i], st->energy_mask[nbEBands+i]); + else + unmask = st->energy_mask[i]; + unmask = MIN16(unmask, QCONST16(.0f, DB_SHIFT)); + unmask -= lin; + if (unmask > QCONST16(.25f, DB_SHIFT)) + { + surround_dynalloc[i] = unmask - QCONST16(.25f, DB_SHIFT); + count_dynalloc++; + } + } + if (count_dynalloc>=3) + { + /* If we need dynalloc in many bands, it's probably because our + initial masking rate was too low. */ + mask_avg += QCONST16(.25f, DB_SHIFT); + if (mask_avg>0) + { + /* Something went really wrong in the original calculations, + disabling masking. */ + mask_avg = 0; + diff = 0; + OPUS_CLEAR(surround_dynalloc, mask_end); + } else { + for(i=0;ilfe) + { + opus_val16 follow=-QCONST16(10.0f,DB_SHIFT); + opus_val32 frame_avg=0; + opus_val16 offset = shortBlocks?HALF16(SHL16(LM, DB_SHIFT)):0; + for(i=start;ispec_avg); + temporal_vbr = MIN16(QCONST16(3.f, DB_SHIFT), MAX16(-QCONST16(1.5f, DB_SHIFT), temporal_vbr)); + st->spec_avg += MULT16_16_Q15(QCONST16(.02f, 15), temporal_vbr); + } + /*for (i=0;i<21;i++) + printf("%f ", bandLogE[i]); + printf("\n");*/ + + if (!secondMdct) + { + OPUS_COPY(bandLogE2, bandLogE, C*nbEBands); + } + + /* Last chance to catch any transient we might have missed in the + time-domain analysis */ + if (LM>0 && ec_tell(enc)+3<=total_bits && !isTransient && st->complexity>=5 && !st->lfe && !hybrid) + { + if (patch_transient_decision(bandLogE, oldBandE, nbEBands, start, end, C)) + { + isTransient = 1; + shortBlocks = M; + compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample, st->arch); + compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch); + amp2Log2(mode, effEnd, end, bandE, bandLogE, C); + /* Compensate for the scaling of short vs long mdcts */ + for (c=0;c0 && ec_tell(enc)+3<=total_bits) + ec_enc_bit_logp(enc, isTransient, 3); + + ALLOC(X, C*N, celt_norm); /**< Interleaved normalised MDCTs */ + + /* Band normalisation */ + normalise_bands(mode, freq, X, bandE, effEnd, C, M); + + enable_tf_analysis = effectiveBytes>=15*C && !hybrid && st->complexity>=2 && !st->lfe; + + ALLOC(offsets, nbEBands, int); + ALLOC(importance, nbEBands, int); + ALLOC(spread_weight, nbEBands, int); + + maxDepth = dynalloc_analysis(bandLogE, bandLogE2, nbEBands, start, end, C, offsets, + st->lsb_depth, mode->logN, isTransient, st->vbr, st->constrained_vbr, + eBands, LM, effectiveBytes, &tot_boost, st->lfe, surround_dynalloc, &st->analysis, importance, spread_weight); + + ALLOC(tf_res, nbEBands, int); + /* Disable variable tf resolution for hybrid and at very low bitrate */ + if (enable_tf_analysis) + { + int lambda; + lambda = IMAX(80, 20480/effectiveBytes + 2); + tf_select = tf_analysis(mode, effEnd, isTransient, tf_res, lambda, X, N, LM, tf_estimate, tf_chan, importance); + for (i=effEnd;isilk_info.signalType != 2) + { + /* For low bitrate hybrid, we force temporal resolution to 5 ms rather than 2.5 ms. */ + for (i=0;iforce_intra, + &st->delayedIntra, st->complexity >= 4, st->loss_rate, st->lfe); + + tf_encode(start, end, isTransient, tf_res, LM, tf_select, enc); + + if (ec_tell(enc)+4<=total_bits) + { + if (st->lfe) + { + st->tapset_decision = 0; + st->spread_decision = SPREAD_NORMAL; + } else if (hybrid) + { + if (st->complexity == 0) + st->spread_decision = SPREAD_NONE; + else if (isTransient) + st->spread_decision = SPREAD_NORMAL; + else + st->spread_decision = SPREAD_AGGRESSIVE; + } else if (shortBlocks || st->complexity < 3 || nbAvailableBytes < 10*C) + { + if (st->complexity == 0) + st->spread_decision = SPREAD_NONE; + else + st->spread_decision = SPREAD_NORMAL; + } else { + /* Disable new spreading+tapset estimator until we can show it works + better than the old one. So far it seems like spreading_decision() + works best. */ +#if 0 + if (st->analysis.valid) + { + static const opus_val16 spread_thresholds[3] = {-QCONST16(.6f, 15), -QCONST16(.2f, 15), -QCONST16(.07f, 15)}; + static const opus_val16 spread_histeresis[3] = {QCONST16(.15f, 15), QCONST16(.07f, 15), QCONST16(.02f, 15)}; + static const opus_val16 tapset_thresholds[2] = {QCONST16(.0f, 15), QCONST16(.15f, 15)}; + static const opus_val16 tapset_histeresis[2] = {QCONST16(.1f, 15), QCONST16(.05f, 15)}; + st->spread_decision = hysteresis_decision(-st->analysis.tonality, spread_thresholds, spread_histeresis, 3, st->spread_decision); + st->tapset_decision = hysteresis_decision(st->analysis.tonality_slope, tapset_thresholds, tapset_histeresis, 2, st->tapset_decision); + } else +#endif + { + st->spread_decision = spreading_decision(mode, X, + &st->tonal_average, st->spread_decision, &st->hf_average, + &st->tapset_decision, pf_on&&!shortBlocks, effEnd, C, M, spread_weight); + } + /*printf("%d %d\n", st->tapset_decision, st->spread_decision);*/ + /*printf("%f %d %f %d\n\n", st->analysis.tonality, st->spread_decision, st->analysis.tonality_slope, st->tapset_decision);*/ + } + ec_enc_icdf(enc, st->spread_decision, spread_icdf, 5); + } + + /* For LFE, everything interesting is in the first band */ + if (st->lfe) + offsets[0] = IMIN(8, effectiveBytes/3); + ALLOC(cap, nbEBands, int); + init_caps(mode,cap,LM,C); + + dynalloc_logp = 6; + total_bits<<=BITRES; + total_boost = 0; + tell = ec_tell_frac(enc); + for (i=start;iintensity = hysteresis_decision((opus_val16)(equiv_rate/1000), + intensity_thresholds, intensity_histeresis, 21, st->intensity); + st->intensity = IMIN(end,IMAX(start, st->intensity)); + } + + alloc_trim = 5; + if (tell+(6< 0 || st->lfe) + { + st->stereo_saving = 0; + alloc_trim = 5; + } else { + alloc_trim = alloc_trim_analysis(mode, X, bandLogE, + end, LM, C, N, &st->analysis, &st->stereo_saving, tf_estimate, + st->intensity, surround_trim, equiv_rate, st->arch); + } + ec_enc_icdf(enc, alloc_trim, trim_icdf, 7); + tell = ec_tell_frac(enc); + } + + /* Variable bitrate */ + if (vbr_rate>0) + { + opus_val16 alpha; + opus_int32 delta; + /* The target rate in 8th bits per frame */ + opus_int32 target, base_target; + opus_int32 min_allowed; + int lm_diff = mode->maxLM - LM; + + /* Don't attempt to use more than 510 kb/s, even for frames smaller than 20 ms. + The CELT allocator will just not be able to use more than that anyway. */ + nbCompressedBytes = IMIN(nbCompressedBytes,1275>>(3-LM)); + if (!hybrid) + { + base_target = vbr_rate - ((40*C+20)<constrained_vbr) + base_target += (st->vbr_offset>>lm_diff); + + if (!hybrid) + { + target = compute_vbr(mode, &st->analysis, base_target, LM, equiv_rate, + st->lastCodedBands, C, st->intensity, st->constrained_vbr, + st->stereo_saving, tot_boost, tf_estimate, pitch_change, maxDepth, + st->lfe, st->energy_mask!=NULL, surround_masking, + temporal_vbr); + } else { + target = base_target; + /* Tonal frames (offset<100) need more bits than noisy (offset>100) ones. */ + if (st->silk_info.offset < 100) target += 12 << BITRES >> (3-LM); + if (st->silk_info.offset > 100) target -= 18 << BITRES >> (3-LM); + /* Boosting bitrate on transients and vowels with significant temporal + spikes. */ + target += (opus_int32)MULT16_16_Q14(tf_estimate-QCONST16(.25f,14), (50< QCONST16(.7f,14)) + target = IMAX(target, 50<>(BITRES+3)) + 2; + /* Take into account the 37 bits we need to have left in the packet to + signal a redundant frame in hybrid mode. Creating a shorter packet would + create an entropy coder desync. */ + if (hybrid) + min_allowed = IMAX(min_allowed, (tell0_frac+(37<>(BITRES+3)); + + nbAvailableBytes = (target+(1<<(BITRES+2)))>>(BITRES+3); + nbAvailableBytes = IMAX(min_allowed,nbAvailableBytes); + nbAvailableBytes = IMIN(nbCompressedBytes,nbAvailableBytes); + + /* By how much did we "miss" the target on that frame */ + delta = target - vbr_rate; + + target=nbAvailableBytes<<(BITRES+3); + + /*If the frame is silent we don't adjust our drift, otherwise + the encoder will shoot to very high rates after hitting a + span of silence, but we do allow the bitres to refill. + This means that we'll undershoot our target in CVBR/VBR modes + on files with lots of silence. */ + if(silence) + { + nbAvailableBytes = 2; + target = 2*8<vbr_count < 970) + { + st->vbr_count++; + alpha = celt_rcp(SHL32(EXTEND32(st->vbr_count+20),16)); + } else + alpha = QCONST16(.001f,15); + /* How many bits have we used in excess of what we're allowed */ + if (st->constrained_vbr) + st->vbr_reservoir += target - vbr_rate; + /*printf ("%d\n", st->vbr_reservoir);*/ + + /* Compute the offset we need to apply in order to reach the target */ + if (st->constrained_vbr) + { + st->vbr_drift += (opus_int32)MULT16_32_Q15(alpha,(delta*(1<vbr_offset-st->vbr_drift); + st->vbr_offset = -st->vbr_drift; + } + /*printf ("%d\n", st->vbr_drift);*/ + + if (st->constrained_vbr && st->vbr_reservoir < 0) + { + /* We're under the min value -- increase rate */ + int adjust = (-st->vbr_reservoir)/(8<vbr_reservoir = 0; + /*printf ("+%d\n", adjust);*/ + } + nbCompressedBytes = IMIN(nbCompressedBytes,nbAvailableBytes); + /*printf("%d\n", nbCompressedBytes*50*8);*/ + /* This moves the raw bits to take into account the new compressed size */ + ec_enc_shrink(enc, nbCompressedBytes); + } + + /* Bit allocation */ + ALLOC(fine_quant, nbEBands, int); + ALLOC(pulses, nbEBands, int); + ALLOC(fine_priority, nbEBands, int); + + /* bits = packet size - where we are - safety*/ + bits = (((opus_int32)nbCompressedBytes*8)<=2&&bits>=((LM+2)<analysis.valid) + { + int min_bandwidth; + if (equiv_rate < (opus_int32)32000*C) + min_bandwidth = 13; + else if (equiv_rate < (opus_int32)48000*C) + min_bandwidth = 16; + else if (equiv_rate < (opus_int32)60000*C) + min_bandwidth = 18; + else if (equiv_rate < (opus_int32)80000*C) + min_bandwidth = 19; + else + min_bandwidth = 20; + signalBandwidth = IMAX(st->analysis.bandwidth, min_bandwidth); + } +#endif + if (st->lfe) + signalBandwidth = 1; + codedBands = clt_compute_allocation(mode, start, end, offsets, cap, + alloc_trim, &st->intensity, &dual_stereo, bits, &balance, pulses, + fine_quant, fine_priority, C, LM, enc, 1, st->lastCodedBands, signalBandwidth); + if (st->lastCodedBands) + st->lastCodedBands = IMIN(st->lastCodedBands+1,IMAX(st->lastCodedBands-1,codedBands)); + else + st->lastCodedBands = codedBands; + + quant_fine_energy(mode, start, end, oldBandE, error, fine_quant, enc, C); + + /* Residual quantisation */ + ALLOC(collapse_masks, C*nbEBands, unsigned char); + quant_all_bands(1, mode, start, end, X, C==2 ? X+N : NULL, collapse_masks, + bandE, pulses, shortBlocks, st->spread_decision, + dual_stereo, st->intensity, tf_res, nbCompressedBytes*(8<rng, st->complexity, st->arch, st->disable_inv); + + if (anti_collapse_rsv > 0) + { + anti_collapse_on = st->consec_transient<2; +#ifdef FUZZING + anti_collapse_on = rand()&0x1; +#endif + ec_enc_bits(enc, anti_collapse_on, 1); + } + quant_energy_finalise(mode, start, end, oldBandE, error, fine_quant, fine_priority, nbCompressedBytes*8-ec_tell(enc), enc, C); + OPUS_CLEAR(energyError, nbEBands*CC); + c=0; + do { + for (i=start;irng); + } + + c=0; do { + OPUS_MOVE(st->syn_mem[c], st->syn_mem[c]+N, 2*MAX_PERIOD-N+overlap/2); + } while (++csyn_mem[c]+2*MAX_PERIOD-N; + } while (++cupsample, silence, st->arch); + + c=0; do { + st->prefilter_period=IMAX(st->prefilter_period, COMBFILTER_MINPERIOD); + st->prefilter_period_old=IMAX(st->prefilter_period_old, COMBFILTER_MINPERIOD); + comb_filter(out_mem[c], out_mem[c], st->prefilter_period_old, st->prefilter_period, mode->shortMdctSize, + st->prefilter_gain_old, st->prefilter_gain, st->prefilter_tapset_old, st->prefilter_tapset, + mode->window, overlap); + if (LM!=0) + comb_filter(out_mem[c]+mode->shortMdctSize, out_mem[c]+mode->shortMdctSize, st->prefilter_period, pitch_index, N-mode->shortMdctSize, + st->prefilter_gain, gain1, st->prefilter_tapset, prefilter_tapset, + mode->window, overlap); + } while (++cupsample, mode->preemph, st->preemph_memD); + st->prefilter_period_old = st->prefilter_period; + st->prefilter_gain_old = st->prefilter_gain; + st->prefilter_tapset_old = st->prefilter_tapset; + } +#endif + + st->prefilter_period = pitch_index; + st->prefilter_gain = gain1; + st->prefilter_tapset = prefilter_tapset; +#ifdef RESYNTH + if (LM!=0) + { + st->prefilter_period_old = st->prefilter_period; + st->prefilter_gain_old = st->prefilter_gain; + st->prefilter_tapset_old = st->prefilter_tapset; + } +#endif + + if (CC==2&&C==1) { + OPUS_COPY(&oldBandE[nbEBands], oldBandE, nbEBands); + } + + if (!isTransient) + { + OPUS_COPY(oldLogE2, oldLogE, CC*nbEBands); + OPUS_COPY(oldLogE, oldBandE, CC*nbEBands); + } else { + for (i=0;iconsec_transient++; + else + st->consec_transient=0; + st->rng = enc->rng; + + /* If there's any room left (can only happen for very high rates), + it's already filled with zeros */ + ec_enc_done(enc); + +#ifdef CUSTOM_MODES + if (st->signalling) + nbCompressedBytes++; +#endif + + RESTORE_STACK; + if (ec_get_error(enc)) + return OPUS_INTERNAL_ERROR; + else + return nbCompressedBytes; +} + + +#ifdef CUSTOM_MODES + +#ifdef FIXED_POINT +int opus_custom_encode(CELTEncoder * OPUS_RESTRICT st, const opus_int16 * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes) +{ + return celt_encode_with_ec(st, pcm, frame_size, compressed, nbCompressedBytes, NULL); +} + +#ifndef DISABLE_FLOAT_API +int opus_custom_encode_float(CELTEncoder * OPUS_RESTRICT st, const float * pcm, int frame_size, unsigned char *compressed, int nbCompressedBytes) +{ + int j, ret, C, N; + VARDECL(opus_int16, in); + ALLOC_STACK; + + if (pcm==NULL) + return OPUS_BAD_ARG; + + C = st->channels; + N = frame_size; + ALLOC(in, C*N, opus_int16); + + for (j=0;jchannels; + N=frame_size; + ALLOC(in, C*N, celt_sig); + for (j=0;j10) + goto bad_arg; + st->complexity = value; + } + break; + case CELT_SET_START_BAND_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<0 || value>=st->mode->nbEBands) + goto bad_arg; + st->start = value; + } + break; + case CELT_SET_END_BAND_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>st->mode->nbEBands) + goto bad_arg; + st->end = value; + } + break; + case CELT_SET_PREDICTION_REQUEST: + { + int value = va_arg(ap, opus_int32); + if (value<0 || value>2) + goto bad_arg; + st->disable_pf = value<=1; + st->force_intra = value==0; + } + break; + case OPUS_SET_PACKET_LOSS_PERC_REQUEST: + { + int value = va_arg(ap, opus_int32); + if (value<0 || value>100) + goto bad_arg; + st->loss_rate = value; + } + break; + case OPUS_SET_VBR_CONSTRAINT_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->constrained_vbr = value; + } + break; + case OPUS_SET_VBR_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->vbr = value; + } + break; + case OPUS_SET_BITRATE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<=500 && value!=OPUS_BITRATE_MAX) + goto bad_arg; + value = IMIN(value, 260000*st->channels); + st->bitrate = value; + } + break; + case CELT_SET_CHANNELS_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<1 || value>2) + goto bad_arg; + st->stream_channels = value; + } + break; + case OPUS_SET_LSB_DEPTH_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<8 || value>24) + goto bad_arg; + st->lsb_depth=value; + } + break; + case OPUS_GET_LSB_DEPTH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + *value=st->lsb_depth; + } + break; + case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>1) + { + goto bad_arg; + } + st->disable_inv = value; + } + break; + case OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->disable_inv; + } + break; + case OPUS_RESET_STATE: + { + int i; + opus_val16 *oldBandE, *oldLogE, *oldLogE2; + oldBandE = (opus_val16*)(st->in_mem+st->channels*(st->mode->overlap+COMBFILTER_MAXPERIOD)); + oldLogE = oldBandE + st->channels*st->mode->nbEBands; + oldLogE2 = oldLogE + st->channels*st->mode->nbEBands; + OPUS_CLEAR((char*)&st->ENCODER_RESET_START, + opus_custom_encoder_get_size(st->mode, st->channels)- + ((char*)&st->ENCODER_RESET_START - (char*)st)); + for (i=0;ichannels*st->mode->nbEBands;i++) + oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT); + st->vbr_offset = 0; + st->delayedIntra = 1; + st->spread_decision = SPREAD_NORMAL; + st->tonal_average = 256; + st->hf_average = 0; + st->tapset_decision = 0; + } + break; +#ifdef CUSTOM_MODES + case CELT_SET_INPUT_CLIPPING_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->clip = value; + } + break; +#endif + case CELT_SET_SIGNALLING_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->signalling = value; + } + break; + case CELT_SET_ANALYSIS_REQUEST: + { + AnalysisInfo *info = va_arg(ap, AnalysisInfo *); + if (info) + OPUS_COPY(&st->analysis, info, 1); + } + break; + case CELT_SET_SILK_INFO_REQUEST: + { + SILKInfo *info = va_arg(ap, SILKInfo *); + if (info) + OPUS_COPY(&st->silk_info, info, 1); + } + break; + case CELT_GET_MODE_REQUEST: + { + const CELTMode ** value = va_arg(ap, const CELTMode**); + if (value==0) + goto bad_arg; + *value=st->mode; + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + opus_uint32 * value = va_arg(ap, opus_uint32 *); + if (value==0) + goto bad_arg; + *value=st->rng; + } + break; + case OPUS_SET_LFE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->lfe = value; + } + break; + case OPUS_SET_ENERGY_MASK_REQUEST: + { + opus_val16 *value = va_arg(ap, opus_val16*); + st->energy_mask = value; + } + break; + default: + goto bad_request; + } + va_end(ap); + return OPUS_OK; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +bad_request: + va_end(ap); + return OPUS_UNIMPLEMENTED; +} diff --git a/lib/opus/celt/celt_lpc.c b/lib/opus/celt/celt_lpc.c new file mode 100644 index 00000000..f91721bc --- /dev/null +++ b/lib/opus/celt/celt_lpc.c @@ -0,0 +1,344 @@ +/* Copyright (c) 2009-2010 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "celt_lpc.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "pitch.h" + +void _celt_lpc( + opus_val16 *_lpc, /* out: [0...p-1] LPC coefficients */ +const opus_val32 *ac, /* in: [0...p] autocorrelation values */ +int p +) +{ + int i, j; + opus_val32 r; + opus_val32 error = ac[0]; +#ifdef FIXED_POINT + opus_val32 lpc[LPC_ORDER]; +#else + float *lpc = _lpc; +#endif + + OPUS_CLEAR(lpc, p); +#ifdef FIXED_POINT + if (ac[0] != 0) +#else + if (ac[0] > 1e-10f) +#endif + { + for (i = 0; i < p; i++) { + /* Sum up this iteration's reflection coefficient */ + opus_val32 rr = 0; + for (j = 0; j < i; j++) + rr += MULT32_32_Q31(lpc[j],ac[i - j]); + rr += SHR32(ac[i + 1],6); + r = -frac_div32(SHL32(rr,6), error); + /* Update LPC coefficients and total error */ + lpc[i] = SHR32(r,6); + for (j = 0; j < (i+1)>>1; j++) + { + opus_val32 tmp1, tmp2; + tmp1 = lpc[j]; + tmp2 = lpc[i-1-j]; + lpc[j] = tmp1 + MULT32_32_Q31(r,tmp2); + lpc[i-1-j] = tmp2 + MULT32_32_Q31(r,tmp1); + } + + error = error - MULT32_32_Q31(MULT32_32_Q31(r,r),error); + /* Bail out once we get 30 dB gain */ +#ifdef FIXED_POINT + if (error<=SHR32(ac[0],10)) + break; +#else + if (error<=.001f*ac[0]) + break; +#endif + } + } +#ifdef FIXED_POINT + { + /* Convert the int32 lpcs to int16 and ensure there are no wrap-arounds. + This reuses the logic in silk_LPC_fit() and silk_bwexpander_32(). Any bug + fixes should also be applied there. */ + int iter, idx = 0; + opus_val32 maxabs, absval, chirp_Q16, chirp_minus_one_Q16; + + for (iter = 0; iter < 10; iter++) { + maxabs = 0; + for (i = 0; i < p; i++) { + absval = ABS32(lpc[i]); + if (absval > maxabs) { + maxabs = absval; + idx = i; + } + } + maxabs = PSHR32(maxabs, 13); /* Q25->Q12 */ + + if (maxabs > 32767) { + maxabs = MIN32(maxabs, 163838); + chirp_Q16 = QCONST32(0.999, 16) - DIV32(SHL32(maxabs - 32767, 14), + SHR32(MULT32_32_32(maxabs, idx + 1), 2)); + chirp_minus_one_Q16 = chirp_Q16 - 65536; + + /* Apply bandwidth expansion. */ + for (i = 0; i < p - 1; i++) { + lpc[i] = MULT32_32_Q16(chirp_Q16, lpc[i]); + chirp_Q16 += PSHR32(MULT32_32_32(chirp_Q16, chirp_minus_one_Q16), 16); + } + lpc[p - 1] = MULT32_32_Q16(chirp_Q16, lpc[p - 1]); + } else { + break; + } + } + + if (iter == 10) { + /* If the coeffs still do not fit into the 16 bit range after 10 iterations, + fall back to the A(z)=1 filter. */ + OPUS_CLEAR(lpc, p); + _lpc[0] = 4096; /* Q12 */ + } else { + for (i = 0; i < p; i++) { + _lpc[i] = EXTRACT16(PSHR32(lpc[i], 13)); /* Q25->Q12 */ + } + } + } +#endif +} + + +void celt_fir_c( + const opus_val16 *x, + const opus_val16 *num, + opus_val16 *y, + int N, + int ord, + int arch) +{ + int i,j; + VARDECL(opus_val16, rnum); + SAVE_STACK; + celt_assert(x != y); + ALLOC(rnum, ord, opus_val16); + for(i=0;i=1;j--) + { + mem[j]=mem[j-1]; + } + mem[0] = SROUND16(sum, SIG_SHIFT); + _y[i] = sum; + } +#else + int i,j; + VARDECL(opus_val16, rden); + VARDECL(opus_val16, y); + SAVE_STACK; + + celt_assert((ord&3)==0); + ALLOC(rden, ord, opus_val16); + ALLOC(y, N+ord, opus_val16); + for(i=0;i0); + celt_assert(overlap>=0); + if (overlap == 0) + { + xptr = x; + } else { + for (i=0;i0) + { + for(i=0;i= 536870912) + { + int shift2=1; + if (ac[0] >= 1073741824) + shift2++; + for (i=0;i<=lag;i++) + ac[i] = SHR32(ac[i], shift2); + shift += shift2; + } +#endif + + RESTORE_STACK; + return shift; +} diff --git a/lib/opus/celt/celt_lpc.h b/lib/opus/celt/celt_lpc.h new file mode 100644 index 00000000..a4c5fd6e --- /dev/null +++ b/lib/opus/celt/celt_lpc.h @@ -0,0 +1,66 @@ +/* Copyright (c) 2009-2010 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef PLC_H +#define PLC_H + +#include "arch.h" +#include "cpu_support.h" + +#if defined(OPUS_X86_MAY_HAVE_SSE4_1) +#include "x86/celt_lpc_sse.h" +#endif + +#define LPC_ORDER 24 + +void _celt_lpc(opus_val16 *_lpc, const opus_val32 *ac, int p); + +void celt_fir_c( + const opus_val16 *x, + const opus_val16 *num, + opus_val16 *y, + int N, + int ord, + int arch); + +#if !defined(OVERRIDE_CELT_FIR) +#define celt_fir(x, num, y, N, ord, arch) \ + (celt_fir_c(x, num, y, N, ord, arch)) +#endif + +void celt_iir(const opus_val32 *x, + const opus_val16 *den, + opus_val32 *y, + int N, + int ord, + opus_val16 *mem, + int arch); + +int _celt_autocorr(const opus_val16 *x, opus_val32 *ac, + const opus_val16 *window, int overlap, int lag, int n, int arch); + +#endif /* PLC_H */ diff --git a/lib/opus/celt/cpu_support.h b/lib/opus/celt/cpu_support.h new file mode 100644 index 00000000..7b5c56ca --- /dev/null +++ b/lib/opus/celt/cpu_support.h @@ -0,0 +1,71 @@ +/* Copyright (c) 2010 Xiph.Org Foundation + * Copyright (c) 2013 Parrot */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef CPU_SUPPORT_H +#define CPU_SUPPORT_H + +#include "opus_types.h" +#include "opus_defines.h" + +#if defined(OPUS_HAVE_RTCD) && \ + (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#include "arm/armcpu.h" + +/* We currently support 4 ARM variants: + * arch[0] -> ARMv4 + * arch[1] -> ARMv5E + * arch[2] -> ARMv6 + * arch[3] -> NEON + */ +#define OPUS_ARCHMASK 3 + +#elif defined(OPUS_HAVE_RTCD) && \ + ((defined(OPUS_X86_MAY_HAVE_SSE) && !defined(OPUS_X86_PRESUME_SSE)) || \ + (defined(OPUS_X86_MAY_HAVE_SSE2) && !defined(OPUS_X86_PRESUME_SSE2)) || \ + (defined(OPUS_X86_MAY_HAVE_SSE4_1) && !defined(OPUS_X86_PRESUME_SSE4_1)) || \ + (defined(OPUS_X86_MAY_HAVE_AVX) && !defined(OPUS_X86_PRESUME_AVX))) + +#include "x86/x86cpu.h" +/* We currently support 5 x86 variants: + * arch[0] -> non-sse + * arch[1] -> sse + * arch[2] -> sse2 + * arch[3] -> sse4.1 + * arch[4] -> avx + */ +#define OPUS_ARCHMASK 7 +int opus_select_arch(void); + +#else +#define OPUS_ARCHMASK 0 + +static OPUS_INLINE int opus_select_arch(void) +{ + return 0; +} +#endif +#endif diff --git a/lib/opus/celt/cwrs.c b/lib/opus/celt/cwrs.c new file mode 100644 index 00000000..a552e4f0 --- /dev/null +++ b/lib/opus/celt/cwrs.c @@ -0,0 +1,715 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2007-2009 Timothy B. Terriberry + Written by Timothy B. Terriberry and Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "os_support.h" +#include "cwrs.h" +#include "mathops.h" +#include "arch.h" + +#ifdef CUSTOM_MODES + +/*Guaranteed to return a conservatively large estimate of the binary logarithm + with frac bits of fractional precision. + Tested for all possible 32-bit inputs with frac=4, where the maximum + overestimation is 0.06254243 bits.*/ +int log2_frac(opus_uint32 val, int frac) +{ + int l; + l=EC_ILOG(val); + if(val&(val-1)){ + /*This is (val>>l-16), but guaranteed to round up, even if adding a bias + before the shift would cause overflow (e.g., for 0xFFFFxxxx). + Doesn't work for val=0, but that case fails the test above.*/ + if(l>16)val=((val-1)>>(l-16))+1; + else val<<=16-l; + l=(l-1)<>16); + l+=b<>b; + val=(val*val+0x7FFF)>>15; + } + while(frac-->0); + /*If val is not exactly 0x8000, then we have to round up the remainder.*/ + return l+(val>0x8000); + } + /*Exact powers of two require no rounding.*/ + else return (l-1)<0 ? sum(k=1...K,2**k*choose(N,k)*choose(K-1,k-1)) : 1, + where choose() is the binomial function. + A table of values for N<10 and K<10 looks like: + V[10][10] = { + {1, 0, 0, 0, 0, 0, 0, 0, 0, 0}, + {1, 2, 2, 2, 2, 2, 2, 2, 2, 2}, + {1, 4, 8, 12, 16, 20, 24, 28, 32, 36}, + {1, 6, 18, 38, 66, 102, 146, 198, 258, 326}, + {1, 8, 32, 88, 192, 360, 608, 952, 1408, 1992}, + {1, 10, 50, 170, 450, 1002, 1970, 3530, 5890, 9290}, + {1, 12, 72, 292, 912, 2364, 5336, 10836, 20256, 35436}, + {1, 14, 98, 462, 1666, 4942, 12642, 28814, 59906, 115598}, + {1, 16, 128, 688, 2816, 9424, 27008, 68464, 157184, 332688}, + {1, 18, 162, 978, 4482, 16722, 53154, 148626, 374274, 864146} + }; + + U(N,K) = the number of such combinations wherein N-1 objects are taken at + most K-1 at a time. + This is given by + U(N,K) = sum(k=0...K-1,V(N-1,k)) + = K>0 ? (V(N-1,K-1) + V(N,K-1))/2 : 0. + The latter expression also makes clear that U(N,K) is half the number of such + combinations wherein the first object is taken at least once. + Although it may not be clear from either of these definitions, U(N,K) is the + natural function to work with when enumerating the pulse vector codebooks, + not V(N,K). + U(N,K) is not well-defined for N=0, but with the extension + U(0,K) = K>0 ? 0 : 1, + the function becomes symmetric: U(N,K) = U(K,N), with a similar table: + U[10][10] = { + {1, 0, 0, 0, 0, 0, 0, 0, 0, 0}, + {0, 1, 1, 1, 1, 1, 1, 1, 1, 1}, + {0, 1, 3, 5, 7, 9, 11, 13, 15, 17}, + {0, 1, 5, 13, 25, 41, 61, 85, 113, 145}, + {0, 1, 7, 25, 63, 129, 231, 377, 575, 833}, + {0, 1, 9, 41, 129, 321, 681, 1289, 2241, 3649}, + {0, 1, 11, 61, 231, 681, 1683, 3653, 7183, 13073}, + {0, 1, 13, 85, 377, 1289, 3653, 8989, 19825, 40081}, + {0, 1, 15, 113, 575, 2241, 7183, 19825, 48639, 108545}, + {0, 1, 17, 145, 833, 3649, 13073, 40081, 108545, 265729} + }; + + With this extension, V(N,K) may be written in terms of U(N,K): + V(N,K) = U(N,K) + U(N,K+1) + for all N>=0, K>=0. + Thus U(N,K+1) represents the number of combinations where the first element + is positive or zero, and U(N,K) represents the number of combinations where + it is negative. + With a large enough table of U(N,K) values, we could write O(N) encoding + and O(min(N*log(K),N+K)) decoding routines, but such a table would be + prohibitively large for small embedded devices (K may be as large as 32767 + for small N, and N may be as large as 200). + + Both functions obey the same recurrence relation: + V(N,K) = V(N-1,K) + V(N,K-1) + V(N-1,K-1), + U(N,K) = U(N-1,K) + U(N,K-1) + U(N-1,K-1), + for all N>0, K>0, with different initial conditions at N=0 or K=0. + This allows us to construct a row of one of the tables above given the + previous row or the next row. + Thus we can derive O(NK) encoding and decoding routines with O(K) memory + using only addition and subtraction. + + When encoding, we build up from the U(2,K) row and work our way forwards. + When decoding, we need to start at the U(N,K) row and work our way backwards, + which requires a means of computing U(N,K). + U(N,K) may be computed from two previous values with the same N: + U(N,K) = ((2*N-1)*U(N,K-1) - U(N,K-2))/(K-1) + U(N,K-2) + for all N>1, and since U(N,K) is symmetric, a similar relation holds for two + previous values with the same K: + U(N,K>1) = ((2*K-1)*U(N-1,K) - U(N-2,K))/(N-1) + U(N-2,K) + for all K>1. + This allows us to construct an arbitrary row of the U(N,K) table by starting + with the first two values, which are constants. + This saves roughly 2/3 the work in our O(NK) decoding routine, but costs O(K) + multiplications. + Similar relations can be derived for V(N,K), but are not used here. + + For N>0 and K>0, U(N,K) and V(N,K) take on the form of an (N-1)-degree + polynomial for fixed N. + The first few are + U(1,K) = 1, + U(2,K) = 2*K-1, + U(3,K) = (2*K-2)*K+1, + U(4,K) = (((4*K-6)*K+8)*K-3)/3, + U(5,K) = ((((2*K-4)*K+10)*K-8)*K+3)/3, + and + V(1,K) = 2, + V(2,K) = 4*K, + V(3,K) = 4*K*K+2, + V(4,K) = 8*(K*K+2)*K/3, + V(5,K) = ((4*K*K+20)*K*K+6)/3, + for all K>0. + This allows us to derive O(N) encoding and O(N*log(K)) decoding routines for + small N (and indeed decoding is also O(N) for N<3). + + @ARTICLE{Fis86, + author="Thomas R. Fischer", + title="A Pyramid Vector Quantizer", + journal="IEEE Transactions on Information Theory", + volume="IT-32", + number=4, + pages="568--583", + month=Jul, + year=1986 + }*/ + +#if !defined(SMALL_FOOTPRINT) + +/*U(N,K) = U(K,N) := N>0?K>0?U(N-1,K)+U(N,K-1)+U(N-1,K-1):0:K>0?1:0*/ +# define CELT_PVQ_U(_n,_k) (CELT_PVQ_U_ROW[IMIN(_n,_k)][IMAX(_n,_k)]) +/*V(N,K) := U(N,K)+U(N,K+1) = the number of PVQ codewords for a band of size N + with K pulses allocated to it.*/ +# define CELT_PVQ_V(_n,_k) (CELT_PVQ_U(_n,_k)+CELT_PVQ_U(_n,(_k)+1)) + +/*For each V(N,K) supported, we will access element U(min(N,K+1),max(N,K+1)). + Thus, the number of entries in row I is the larger of the maximum number of + pulses we will ever allocate for a given N=I (K=128, or however many fit in + 32 bits, whichever is smaller), plus one, and the maximum N for which + K=I-1 pulses fit in 32 bits. + The largest band size in an Opus Custom mode is 208. + Otherwise, we can limit things to the set of N which can be achieved by + splitting a band from a standard Opus mode: 176, 144, 96, 88, 72, 64, 48, + 44, 36, 32, 24, 22, 18, 16, 8, 4, 2).*/ +#if defined(CUSTOM_MODES) +static const opus_uint32 CELT_PVQ_U_DATA[1488]={ +#else +static const opus_uint32 CELT_PVQ_U_DATA[1272]={ +#endif + /*N=0, K=0...176:*/ + 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +#if defined(CUSTOM_MODES) + /*...208:*/ + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, +#endif + /*N=1, K=1...176:*/ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, +#if defined(CUSTOM_MODES) + /*...208:*/ + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, +#endif + /*N=2, K=2...176:*/ + 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, + 43, 45, 47, 49, 51, 53, 55, 57, 59, 61, 63, 65, 67, 69, 71, 73, 75, 77, 79, + 81, 83, 85, 87, 89, 91, 93, 95, 97, 99, 101, 103, 105, 107, 109, 111, 113, + 115, 117, 119, 121, 123, 125, 127, 129, 131, 133, 135, 137, 139, 141, 143, + 145, 147, 149, 151, 153, 155, 157, 159, 161, 163, 165, 167, 169, 171, 173, + 175, 177, 179, 181, 183, 185, 187, 189, 191, 193, 195, 197, 199, 201, 203, + 205, 207, 209, 211, 213, 215, 217, 219, 221, 223, 225, 227, 229, 231, 233, + 235, 237, 239, 241, 243, 245, 247, 249, 251, 253, 255, 257, 259, 261, 263, + 265, 267, 269, 271, 273, 275, 277, 279, 281, 283, 285, 287, 289, 291, 293, + 295, 297, 299, 301, 303, 305, 307, 309, 311, 313, 315, 317, 319, 321, 323, + 325, 327, 329, 331, 333, 335, 337, 339, 341, 343, 345, 347, 349, 351, +#if defined(CUSTOM_MODES) + /*...208:*/ + 353, 355, 357, 359, 361, 363, 365, 367, 369, 371, 373, 375, 377, 379, 381, + 383, 385, 387, 389, 391, 393, 395, 397, 399, 401, 403, 405, 407, 409, 411, + 413, 415, +#endif + /*N=3, K=3...176:*/ + 13, 25, 41, 61, 85, 113, 145, 181, 221, 265, 313, 365, 421, 481, 545, 613, + 685, 761, 841, 925, 1013, 1105, 1201, 1301, 1405, 1513, 1625, 1741, 1861, + 1985, 2113, 2245, 2381, 2521, 2665, 2813, 2965, 3121, 3281, 3445, 3613, 3785, + 3961, 4141, 4325, 4513, 4705, 4901, 5101, 5305, 5513, 5725, 5941, 6161, 6385, + 6613, 6845, 7081, 7321, 7565, 7813, 8065, 8321, 8581, 8845, 9113, 9385, 9661, + 9941, 10225, 10513, 10805, 11101, 11401, 11705, 12013, 12325, 12641, 12961, + 13285, 13613, 13945, 14281, 14621, 14965, 15313, 15665, 16021, 16381, 16745, + 17113, 17485, 17861, 18241, 18625, 19013, 19405, 19801, 20201, 20605, 21013, + 21425, 21841, 22261, 22685, 23113, 23545, 23981, 24421, 24865, 25313, 25765, + 26221, 26681, 27145, 27613, 28085, 28561, 29041, 29525, 30013, 30505, 31001, + 31501, 32005, 32513, 33025, 33541, 34061, 34585, 35113, 35645, 36181, 36721, + 37265, 37813, 38365, 38921, 39481, 40045, 40613, 41185, 41761, 42341, 42925, + 43513, 44105, 44701, 45301, 45905, 46513, 47125, 47741, 48361, 48985, 49613, + 50245, 50881, 51521, 52165, 52813, 53465, 54121, 54781, 55445, 56113, 56785, + 57461, 58141, 58825, 59513, 60205, 60901, 61601, +#if defined(CUSTOM_MODES) + /*...208:*/ + 62305, 63013, 63725, 64441, 65161, 65885, 66613, 67345, 68081, 68821, 69565, + 70313, 71065, 71821, 72581, 73345, 74113, 74885, 75661, 76441, 77225, 78013, + 78805, 79601, 80401, 81205, 82013, 82825, 83641, 84461, 85285, 86113, +#endif + /*N=4, K=4...176:*/ + 63, 129, 231, 377, 575, 833, 1159, 1561, 2047, 2625, 3303, 4089, 4991, 6017, + 7175, 8473, 9919, 11521, 13287, 15225, 17343, 19649, 22151, 24857, 27775, + 30913, 34279, 37881, 41727, 45825, 50183, 54809, 59711, 64897, 70375, 76153, + 82239, 88641, 95367, 102425, 109823, 117569, 125671, 134137, 142975, 152193, + 161799, 171801, 182207, 193025, 204263, 215929, 228031, 240577, 253575, + 267033, 280959, 295361, 310247, 325625, 341503, 357889, 374791, 392217, + 410175, 428673, 447719, 467321, 487487, 508225, 529543, 551449, 573951, + 597057, 620775, 645113, 670079, 695681, 721927, 748825, 776383, 804609, + 833511, 863097, 893375, 924353, 956039, 988441, 1021567, 1055425, 1090023, + 1125369, 1161471, 1198337, 1235975, 1274393, 1313599, 1353601, 1394407, + 1436025, 1478463, 1521729, 1565831, 1610777, 1656575, 1703233, 1750759, + 1799161, 1848447, 1898625, 1949703, 2001689, 2054591, 2108417, 2163175, + 2218873, 2275519, 2333121, 2391687, 2451225, 2511743, 2573249, 2635751, + 2699257, 2763775, 2829313, 2895879, 2963481, 3032127, 3101825, 3172583, + 3244409, 3317311, 3391297, 3466375, 3542553, 3619839, 3698241, 3777767, + 3858425, 3940223, 4023169, 4107271, 4192537, 4278975, 4366593, 4455399, + 4545401, 4636607, 4729025, 4822663, 4917529, 5013631, 5110977, 5209575, + 5309433, 5410559, 5512961, 5616647, 5721625, 5827903, 5935489, 6044391, + 6154617, 6266175, 6379073, 6493319, 6608921, 6725887, 6844225, 6963943, + 7085049, 7207551, +#if defined(CUSTOM_MODES) + /*...208:*/ + 7331457, 7456775, 7583513, 7711679, 7841281, 7972327, 8104825, 8238783, + 8374209, 8511111, 8649497, 8789375, 8930753, 9073639, 9218041, 9363967, + 9511425, 9660423, 9810969, 9963071, 10116737, 10271975, 10428793, 10587199, + 10747201, 10908807, 11072025, 11236863, 11403329, 11571431, 11741177, + 11912575, +#endif + /*N=5, K=5...176:*/ + 321, 681, 1289, 2241, 3649, 5641, 8361, 11969, 16641, 22569, 29961, 39041, + 50049, 63241, 78889, 97281, 118721, 143529, 172041, 204609, 241601, 283401, + 330409, 383041, 441729, 506921, 579081, 658689, 746241, 842249, 947241, + 1061761, 1186369, 1321641, 1468169, 1626561, 1797441, 1981449, 2179241, + 2391489, 2618881, 2862121, 3121929, 3399041, 3694209, 4008201, 4341801, + 4695809, 5071041, 5468329, 5888521, 6332481, 6801089, 7295241, 7815849, + 8363841, 8940161, 9545769, 10181641, 10848769, 11548161, 12280841, 13047849, + 13850241, 14689089, 15565481, 16480521, 17435329, 18431041, 19468809, + 20549801, 21675201, 22846209, 24064041, 25329929, 26645121, 28010881, + 29428489, 30899241, 32424449, 34005441, 35643561, 37340169, 39096641, + 40914369, 42794761, 44739241, 46749249, 48826241, 50971689, 53187081, + 55473921, 57833729, 60268041, 62778409, 65366401, 68033601, 70781609, + 73612041, 76526529, 79526721, 82614281, 85790889, 89058241, 92418049, + 95872041, 99421961, 103069569, 106816641, 110664969, 114616361, 118672641, + 122835649, 127107241, 131489289, 135983681, 140592321, 145317129, 150160041, + 155123009, 160208001, 165417001, 170752009, 176215041, 181808129, 187533321, + 193392681, 199388289, 205522241, 211796649, 218213641, 224775361, 231483969, + 238341641, 245350569, 252512961, 259831041, 267307049, 274943241, 282741889, + 290705281, 298835721, 307135529, 315607041, 324252609, 333074601, 342075401, + 351257409, 360623041, 370174729, 379914921, 389846081, 399970689, 410291241, + 420810249, 431530241, 442453761, 453583369, 464921641, 476471169, 488234561, + 500214441, 512413449, 524834241, 537479489, 550351881, 563454121, 576788929, + 590359041, 604167209, 618216201, 632508801, +#if defined(CUSTOM_MODES) + /*...208:*/ + 647047809, 661836041, 676876329, 692171521, 707724481, 723538089, 739615241, + 755958849, 772571841, 789457161, 806617769, 824056641, 841776769, 859781161, + 878072841, 896654849, 915530241, 934702089, 954173481, 973947521, 994027329, + 1014416041, 1035116809, 1056132801, 1077467201, 1099123209, 1121104041, + 1143412929, 1166053121, 1189027881, 1212340489, 1235994241, +#endif + /*N=6, K=6...96:*/ + 1683, 3653, 7183, 13073, 22363, 36365, 56695, 85305, 124515, 177045, 246047, + 335137, 448427, 590557, 766727, 982729, 1244979, 1560549, 1937199, 2383409, + 2908411, 3522221, 4235671, 5060441, 6009091, 7095093, 8332863, 9737793, + 11326283, 13115773, 15124775, 17372905, 19880915, 22670725, 25765455, + 29189457, 32968347, 37129037, 41699767, 46710137, 52191139, 58175189, + 64696159, 71789409, 79491819, 87841821, 96879431, 106646281, 117185651, + 128542501, 140763503, 153897073, 167993403, 183104493, 199284183, 216588185, + 235074115, 254801525, 275831935, 298228865, 322057867, 347386557, 374284647, + 402823977, 433078547, 465124549, 499040399, 534906769, 572806619, 612825229, + 655050231, 699571641, 746481891, 795875861, 847850911, 902506913, 959946283, + 1020274013, 1083597703, 1150027593, 1219676595, 1292660325, 1369097135, + 1449108145, 1532817275, 1620351277, 1711839767, 1807415257, 1907213187, + 2011371957, 2120032959, +#if defined(CUSTOM_MODES) + /*...109:*/ + 2233340609U, 2351442379U, 2474488829U, 2602633639U, 2736033641U, 2874848851U, + 3019242501U, 3169381071U, 3325434321U, 3487575323U, 3655980493U, 3830829623U, + 4012305913U, +#endif + /*N=7, K=7...54*/ + 8989, 19825, 40081, 75517, 134245, 227305, 369305, 579125, 880685, 1303777, + 1884961, 2668525, 3707509, 5064793, 6814249, 9041957, 11847485, 15345233, + 19665841, 24957661, 31388293, 39146185, 48442297, 59511829, 72616013, + 88043969, 106114625, 127178701, 151620757, 179861305, 212358985, 249612805, + 292164445, 340600625, 395555537, 457713341, 527810725, 606639529, 695049433, + 793950709, 904317037, 1027188385, 1163673953, 1314955181, 1482288821, + 1667010073, 1870535785, 2094367717, +#if defined(CUSTOM_MODES) + /*...60:*/ + 2340095869U, 2609401873U, 2904062449U, 3225952925U, 3577050821U, 3959439497U, +#endif + /*N=8, K=8...37*/ + 48639, 108545, 224143, 433905, 795455, 1392065, 2340495, 3800305, 5984767, + 9173505, 13726991, 20103025, 28875327, 40754369, 56610575, 77500017, + 104692735, 139703809, 184327311, 240673265, 311207743, 398796225, 506750351, + 638878193, 799538175, 993696769, 1226990095, 1505789553, 1837271615, + 2229491905U, +#if defined(CUSTOM_MODES) + /*...40:*/ + 2691463695U, 3233240945U, 3866006015U, +#endif + /*N=9, K=9...28:*/ + 265729, 598417, 1256465, 2485825, 4673345, 8405905, 14546705, 24331777, + 39490049, 62390545, 96220561, 145198913, 214828609, 312193553, 446304145, + 628496897, 872893441, 1196924561, 1621925137, 2173806145U, +#if defined(CUSTOM_MODES) + /*...29:*/ + 2883810113U, +#endif + /*N=10, K=10...24:*/ + 1462563, 3317445, 7059735, 14218905, 27298155, 50250765, 89129247, 152951073, + 254831667, 413442773, 654862247, 1014889769, 1541911931, 2300409629U, + 3375210671U, + /*N=11, K=11...19:*/ + 8097453, 18474633, 39753273, 81270333, 158819253, 298199265, 540279585, + 948062325, 1616336765, +#if defined(CUSTOM_MODES) + /*...20:*/ + 2684641785U, +#endif + /*N=12, K=12...18:*/ + 45046719, 103274625, 224298231, 464387817, 921406335, 1759885185, + 3248227095U, + /*N=13, K=13...16:*/ + 251595969, 579168825, 1267854873, 2653649025U, + /*N=14, K=14:*/ + 1409933619 +}; + +#if defined(CUSTOM_MODES) +static const opus_uint32 *const CELT_PVQ_U_ROW[15]={ + CELT_PVQ_U_DATA+ 0,CELT_PVQ_U_DATA+ 208,CELT_PVQ_U_DATA+ 415, + CELT_PVQ_U_DATA+ 621,CELT_PVQ_U_DATA+ 826,CELT_PVQ_U_DATA+1030, + CELT_PVQ_U_DATA+1233,CELT_PVQ_U_DATA+1336,CELT_PVQ_U_DATA+1389, + CELT_PVQ_U_DATA+1421,CELT_PVQ_U_DATA+1441,CELT_PVQ_U_DATA+1455, + CELT_PVQ_U_DATA+1464,CELT_PVQ_U_DATA+1470,CELT_PVQ_U_DATA+1473 +}; +#else +static const opus_uint32 *const CELT_PVQ_U_ROW[15]={ + CELT_PVQ_U_DATA+ 0,CELT_PVQ_U_DATA+ 176,CELT_PVQ_U_DATA+ 351, + CELT_PVQ_U_DATA+ 525,CELT_PVQ_U_DATA+ 698,CELT_PVQ_U_DATA+ 870, + CELT_PVQ_U_DATA+1041,CELT_PVQ_U_DATA+1131,CELT_PVQ_U_DATA+1178, + CELT_PVQ_U_DATA+1207,CELT_PVQ_U_DATA+1226,CELT_PVQ_U_DATA+1240, + CELT_PVQ_U_DATA+1248,CELT_PVQ_U_DATA+1254,CELT_PVQ_U_DATA+1257 +}; +#endif + +#if defined(CUSTOM_MODES) +void get_required_bits(opus_int16 *_bits,int _n,int _maxk,int _frac){ + int k; + /*_maxk==0 => there's nothing to do.*/ + celt_assert(_maxk>0); + _bits[0]=0; + for(k=1;k<=_maxk;k++)_bits[k]=log2_frac(CELT_PVQ_V(_n,k),_frac); +} +#endif + +static opus_uint32 icwrs(int _n,const int *_y){ + opus_uint32 i; + int j; + int k; + celt_assert(_n>=2); + j=_n-1; + i=_y[j]<0; + k=abs(_y[j]); + do{ + j--; + i+=CELT_PVQ_U(_n-j,k); + k+=abs(_y[j]); + if(_y[j]<0)i+=CELT_PVQ_U(_n-j,k+1); + } + while(j>0); + return i; +} + +void encode_pulses(const int *_y,int _n,int _k,ec_enc *_enc){ + celt_assert(_k>0); + ec_enc_uint(_enc,icwrs(_n,_y),CELT_PVQ_V(_n,_k)); +} + +static opus_val32 cwrsi(int _n,int _k,opus_uint32 _i,int *_y){ + opus_uint32 p; + int s; + int k0; + opus_int16 val; + opus_val32 yy=0; + celt_assert(_k>0); + celt_assert(_n>1); + while(_n>2){ + opus_uint32 q; + /*Lots of pulses case:*/ + if(_k>=_n){ + const opus_uint32 *row; + row=CELT_PVQ_U_ROW[_n]; + /*Are the pulses in this dimension negative?*/ + p=row[_k+1]; + s=-(_i>=p); + _i-=p&s; + /*Count how many pulses were placed in this dimension.*/ + k0=_k; + q=row[_n]; + if(q>_i){ + celt_sig_assert(p>q); + _k=_n; + do p=CELT_PVQ_U_ROW[--_k][_n]; + while(p>_i); + } + else for(p=row[_k];p>_i;p=row[_k])_k--; + _i-=p; + val=(k0-_k+s)^s; + *_y++=val; + yy=MAC16_16(yy,val,val); + } + /*Lots of dimensions case:*/ + else{ + /*Are there any pulses in this dimension at all?*/ + p=CELT_PVQ_U_ROW[_k][_n]; + q=CELT_PVQ_U_ROW[_k+1][_n]; + if(p<=_i&&_i=q); + _i-=q&s; + /*Count how many pulses were placed in this dimension.*/ + k0=_k; + do p=CELT_PVQ_U_ROW[--_k][_n]; + while(p>_i); + _i-=p; + val=(k0-_k+s)^s; + *_y++=val; + yy=MAC16_16(yy,val,val); + } + } + _n--; + } + /*_n==2*/ + p=2*_k+1; + s=-(_i>=p); + _i-=p&s; + k0=_k; + _k=(_i+1)>>1; + if(_k)_i-=2*_k-1; + val=(k0-_k+s)^s; + *_y++=val; + yy=MAC16_16(yy,val,val); + /*_n==1*/ + s=-(int)_i; + val=(_k+s)^s; + *_y=val; + yy=MAC16_16(yy,val,val); + return yy; +} + +opus_val32 decode_pulses(int *_y,int _n,int _k,ec_dec *_dec){ + return cwrsi(_n,_k,ec_dec_uint(_dec,CELT_PVQ_V(_n,_k)),_y); +} + +#else /* SMALL_FOOTPRINT */ + +/*Computes the next row/column of any recurrence that obeys the relation + u[i][j]=u[i-1][j]+u[i][j-1]+u[i-1][j-1]. + _ui0 is the base case for the new row/column.*/ +static OPUS_INLINE void unext(opus_uint32 *_ui,unsigned _len,opus_uint32 _ui0){ + opus_uint32 ui1; + unsigned j; + /*This do-while will overrun the array if we don't have storage for at least + 2 values.*/ + j=1; do { + ui1=UADD32(UADD32(_ui[j],_ui[j-1]),_ui0); + _ui[j-1]=_ui0; + _ui0=ui1; + } while (++j<_len); + _ui[j-1]=_ui0; +} + +/*Computes the previous row/column of any recurrence that obeys the relation + u[i-1][j]=u[i][j]-u[i][j-1]-u[i-1][j-1]. + _ui0 is the base case for the new row/column.*/ +static OPUS_INLINE void uprev(opus_uint32 *_ui,unsigned _n,opus_uint32 _ui0){ + opus_uint32 ui1; + unsigned j; + /*This do-while will overrun the array if we don't have storage for at least + 2 values.*/ + j=1; do { + ui1=USUB32(USUB32(_ui[j],_ui[j-1]),_ui0); + _ui[j-1]=_ui0; + _ui0=ui1; + } while (++j<_n); + _ui[j-1]=_ui0; +} + +/*Compute V(_n,_k), as well as U(_n,0..._k+1). + _u: On exit, _u[i] contains U(_n,i) for i in [0..._k+1].*/ +static opus_uint32 ncwrs_urow(unsigned _n,unsigned _k,opus_uint32 *_u){ + opus_uint32 um2; + unsigned len; + unsigned k; + len=_k+2; + /*We require storage at least 3 values (e.g., _k>0).*/ + celt_assert(len>=3); + _u[0]=0; + _u[1]=um2=1; + /*If _n==0, _u[0] should be 1 and the rest should be 0.*/ + /*If _n==1, _u[i] should be 1 for i>1.*/ + celt_assert(_n>=2); + /*If _k==0, the following do-while loop will overflow the buffer.*/ + celt_assert(_k>0); + k=2; + do _u[k]=(k<<1)-1; + while(++k0); + j=0; + do{ + opus_uint32 p; + int s; + int yj; + p=_u[_k+1]; + s=-(_i>=p); + _i-=p&s; + yj=_k; + p=_u[_k]; + while(p>_i)p=_u[--_k]; + _i-=p; + yj-=_k; + val=(yj+s)^s; + _y[j]=val; + yy=MAC16_16(yy,val,val); + uprev(_u,_k+2,0); + } + while(++j<_n); + return yy; +} + +/*Returns the index of the given combination of K elements chosen from a set + of size 1 with associated sign bits. + _y: The vector of pulses, whose sum of absolute values is K. + _k: Returns K.*/ +static OPUS_INLINE opus_uint32 icwrs1(const int *_y,int *_k){ + *_k=abs(_y[0]); + return _y[0]<0; +} + +/*Returns the index of the given combination of K elements chosen from a set + of size _n with associated sign bits. + _y: The vector of pulses, whose sum of absolute values must be _k. + _nc: Returns V(_n,_k).*/ +static OPUS_INLINE opus_uint32 icwrs(int _n,int _k,opus_uint32 *_nc,const int *_y, + opus_uint32 *_u){ + opus_uint32 i; + int j; + int k; + /*We can't unroll the first two iterations of the loop unless _n>=2.*/ + celt_assert(_n>=2); + _u[0]=0; + for(k=1;k<=_k+1;k++)_u[k]=(k<<1)-1; + i=icwrs1(_y+_n-1,&k); + j=_n-2; + i+=_u[k]; + k+=abs(_y[j]); + if(_y[j]<0)i+=_u[k+1]; + while(j-->0){ + unext(_u,_k+2,0); + i+=_u[k]; + k+=abs(_y[j]); + if(_y[j]<0)i+=_u[k+1]; + } + *_nc=_u[k]+_u[k+1]; + return i; +} + +#ifdef CUSTOM_MODES +void get_required_bits(opus_int16 *_bits,int _n,int _maxk,int _frac){ + int k; + /*_maxk==0 => there's nothing to do.*/ + celt_assert(_maxk>0); + _bits[0]=0; + if (_n==1) + { + for (k=1;k<=_maxk;k++) + _bits[k] = 1<<_frac; + } + else { + VARDECL(opus_uint32,u); + SAVE_STACK; + ALLOC(u,_maxk+2U,opus_uint32); + ncwrs_urow(_n,_maxk,u); + for(k=1;k<=_maxk;k++) + _bits[k]=log2_frac(u[k]+u[k+1],_frac); + RESTORE_STACK; + } +} +#endif /* CUSTOM_MODES */ + +void encode_pulses(const int *_y,int _n,int _k,ec_enc *_enc){ + opus_uint32 i; + VARDECL(opus_uint32,u); + opus_uint32 nc; + SAVE_STACK; + celt_assert(_k>0); + ALLOC(u,_k+2U,opus_uint32); + i=icwrs(_n,_k,&nc,_y,u); + ec_enc_uint(_enc,i,nc); + RESTORE_STACK; +} + +opus_val32 decode_pulses(int *_y,int _n,int _k,ec_dec *_dec){ + VARDECL(opus_uint32,u); + int ret; + SAVE_STACK; + celt_assert(_k>0); + ALLOC(u,_k+2U,opus_uint32); + ret = cwrsi(_n,_k,ec_dec_uint(_dec,ncwrs_urow(_n,_k,u)),_y,u); + RESTORE_STACK; + return ret; +} + +#endif /* SMALL_FOOTPRINT */ diff --git a/lib/opus/celt/cwrs.h b/lib/opus/celt/cwrs.h new file mode 100644 index 00000000..7cd47174 --- /dev/null +++ b/lib/opus/celt/cwrs.h @@ -0,0 +1,48 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2007-2009 Timothy B. Terriberry + Written by Timothy B. Terriberry and Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef CWRS_H +#define CWRS_H + +#include "arch.h" +#include "stack_alloc.h" +#include "entenc.h" +#include "entdec.h" + +#ifdef CUSTOM_MODES +int log2_frac(opus_uint32 val, int frac); +#endif + +void get_required_bits(opus_int16 *bits, int N, int K, int frac); + +void encode_pulses(const int *_y, int N, int K, ec_enc *enc); + +opus_val32 decode_pulses(int *_y, int N, int K, ec_dec *dec); + +#endif /* CWRS_H */ diff --git a/lib/opus/celt/dump_modes/Makefile b/lib/opus/celt/dump_modes/Makefile new file mode 100644 index 00000000..93f599fb --- /dev/null +++ b/lib/opus/celt/dump_modes/Makefile @@ -0,0 +1,32 @@ + +CFLAGS=-O2 -Wall -Wextra -DHAVE_CONFIG_H +INCLUDES=-I. -I../ -I../.. -I../../include + +SOURCES = dump_modes.c \ + ../modes.c \ + ../cwrs.c \ + ../rate.c \ + ../entcode.c \ + ../entenc.c \ + ../entdec.c \ + ../mathops.c \ + ../mdct.c \ + ../kiss_fft.c + +ifdef HAVE_ARM_NE10 +CC = gcc +CFLAGS += -mfpu=neon +INCLUDES += -I$(NE10_INCDIR) -DHAVE_ARM_NE10 -DOPUS_ARM_PRESUME_NEON_INTR +LIBS = -L$(NE10_LIBDIR) -lNE10 +SOURCES += ../arm/celt_ne10_fft.c \ + dump_modes_arm_ne10.c \ + ../arm/armcpu.c +endif + +all: dump_modes + +dump_modes: + $(PREFIX)$(CC) $(CFLAGS) $(INCLUDES) -DCUSTOM_MODES_ONLY -DCUSTOM_MODES $(SOURCES) -o $@ $(LIBS) -lm + +clean: + rm -f dump_modes diff --git a/lib/opus/celt/dump_modes/dump_modes.c b/lib/opus/celt/dump_modes/dump_modes.c new file mode 100644 index 00000000..9105a534 --- /dev/null +++ b/lib/opus/celt/dump_modes/dump_modes.c @@ -0,0 +1,353 @@ +/* Copyright (c) 2008 CSIRO + Copyright (c) 2008-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include "modes.h" +#include "celt.h" +#include "rate.h" +#include "dump_modes_arch.h" + +#define INT16 "%d" +#define INT32 "%d" +#define FLOAT "%#0.8gf" + +#ifdef FIXED_POINT +#define WORD16 INT16 +#define WORD32 INT32 +#else +#define WORD16 FLOAT +#define WORD32 FLOAT +#endif + +void dump_modes(FILE *file, CELTMode **modes, int nb_modes) +{ + int i, j, k; + int mdct_twiddles_size; + fprintf(file, "/* The contents of this file was automatically generated by dump_modes.c\n"); + fprintf(file, " with arguments:"); + for (i=0;iFs,mode->shortMdctSize*mode->nbShortMdcts); + } + fprintf(file, "\n It contains static definitions for some pre-defined modes. */\n"); + fprintf(file, "#include \"modes.h\"\n"); + fprintf(file, "#include \"rate.h\"\n"); + fprintf(file, "\n#ifdef HAVE_ARM_NE10\n"); + fprintf(file, "#define OVERRIDE_FFT 1\n"); + fprintf(file, "#include \"%s\"\n", ARM_NE10_ARCH_FILE_NAME); + fprintf(file, "#endif\n"); + + fprintf(file, "\n"); + + for (i=0;ishortMdctSize*mode->nbShortMdcts; + standard = (mode->Fs == 400*(opus_int32)mode->shortMdctSize); + framerate = mode->Fs/mode->shortMdctSize; + + if (!standard) + { + fprintf(file, "#ifndef DEF_EBANDS%d_%d\n", mode->Fs, mdctSize); + fprintf(file, "#define DEF_EBANDS%d_%d\n", mode->Fs, mdctSize); + fprintf (file, "static const opus_int16 eBands%d_%d[%d] = {\n", mode->Fs, mdctSize, mode->nbEBands+2); + for (j=0;jnbEBands+2;j++) + fprintf (file, "%d, ", mode->eBands[j]); + fprintf (file, "};\n"); + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + } + + fprintf(file, "#ifndef DEF_WINDOW%d\n", mode->overlap); + fprintf(file, "#define DEF_WINDOW%d\n", mode->overlap); + fprintf (file, "static const opus_val16 window%d[%d] = {\n", mode->overlap, mode->overlap); + for (j=0;joverlap;j++) + fprintf (file, WORD16 ",%c", mode->window[j],(j+6)%5==0?'\n':' '); + fprintf (file, "};\n"); + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + + if (!standard) + { + fprintf(file, "#ifndef DEF_ALLOC_VECTORS%d_%d\n", mode->Fs, mdctSize); + fprintf(file, "#define DEF_ALLOC_VECTORS%d_%d\n", mode->Fs, mdctSize); + fprintf (file, "static const unsigned char allocVectors%d_%d[%d] = {\n", mode->Fs, mdctSize, mode->nbEBands*mode->nbAllocVectors); + for (j=0;jnbAllocVectors;j++) + { + for (k=0;knbEBands;k++) + fprintf (file, "%2d, ", mode->allocVectors[j*mode->nbEBands+k]); + fprintf (file, "\n"); + } + fprintf (file, "};\n"); + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + } + + fprintf(file, "#ifndef DEF_LOGN%d\n", framerate); + fprintf(file, "#define DEF_LOGN%d\n", framerate); + fprintf (file, "static const opus_int16 logN%d[%d] = {\n", framerate, mode->nbEBands); + for (j=0;jnbEBands;j++) + fprintf (file, "%d, ", mode->logN[j]); + fprintf (file, "};\n"); + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + + /* Pulse cache */ + fprintf(file, "#ifndef DEF_PULSE_CACHE%d\n", mode->Fs/mdctSize); + fprintf(file, "#define DEF_PULSE_CACHE%d\n", mode->Fs/mdctSize); + fprintf (file, "static const opus_int16 cache_index%d[%d] = {\n", mode->Fs/mdctSize, (mode->maxLM+2)*mode->nbEBands); + for (j=0;jnbEBands*(mode->maxLM+2);j++) + fprintf (file, "%d,%c", mode->cache.index[j],(j+16)%15==0?'\n':' '); + fprintf (file, "};\n"); + fprintf (file, "static const unsigned char cache_bits%d[%d] = {\n", mode->Fs/mdctSize, mode->cache.size); + for (j=0;jcache.size;j++) + fprintf (file, "%d,%c", mode->cache.bits[j],(j+16)%15==0?'\n':' '); + fprintf (file, "};\n"); + fprintf (file, "static const unsigned char cache_caps%d[%d] = {\n", mode->Fs/mdctSize, (mode->maxLM+1)*2*mode->nbEBands); + for (j=0;j<(mode->maxLM+1)*2*mode->nbEBands;j++) + fprintf (file, "%d,%c", mode->cache.caps[j],(j+16)%15==0?'\n':' '); + fprintf (file, "};\n"); + + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + + /* FFT twiddles */ + fprintf(file, "#ifndef FFT_TWIDDLES%d_%d\n", mode->Fs, mdctSize); + fprintf(file, "#define FFT_TWIDDLES%d_%d\n", mode->Fs, mdctSize); + fprintf (file, "static const kiss_twiddle_cpx fft_twiddles%d_%d[%d] = {\n", + mode->Fs, mdctSize, mode->mdct.kfft[0]->nfft); + for (j=0;jmdct.kfft[0]->nfft;j++) + fprintf (file, "{" WORD16 ", " WORD16 "},%c", mode->mdct.kfft[0]->twiddles[j].r, mode->mdct.kfft[0]->twiddles[j].i,(j+3)%2==0?'\n':' '); + fprintf (file, "};\n"); + +#ifdef OVERRIDE_FFT + dump_mode_arch(mode); +#endif + /* FFT Bitrev tables */ + for (k=0;k<=mode->mdct.maxshift;k++) + { + fprintf(file, "#ifndef FFT_BITREV%d\n", mode->mdct.kfft[k]->nfft); + fprintf(file, "#define FFT_BITREV%d\n", mode->mdct.kfft[k]->nfft); + fprintf (file, "static const opus_int16 fft_bitrev%d[%d] = {\n", + mode->mdct.kfft[k]->nfft, mode->mdct.kfft[k]->nfft); + for (j=0;jmdct.kfft[k]->nfft;j++) + fprintf (file, "%d,%c", mode->mdct.kfft[k]->bitrev[j],(j+16)%15==0?'\n':' '); + fprintf (file, "};\n"); + + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + } + + /* FFT States */ + for (k=0;k<=mode->mdct.maxshift;k++) + { + fprintf(file, "#ifndef FFT_STATE%d_%d_%d\n", mode->Fs, mdctSize, k); + fprintf(file, "#define FFT_STATE%d_%d_%d\n", mode->Fs, mdctSize, k); + fprintf (file, "static const kiss_fft_state fft_state%d_%d_%d = {\n", + mode->Fs, mdctSize, k); + fprintf (file, "%d, /* nfft */\n", mode->mdct.kfft[k]->nfft); + fprintf (file, WORD16 ", /* scale */\n", mode->mdct.kfft[k]->scale); +#ifdef FIXED_POINT + fprintf (file, "%d, /* scale_shift */\n", mode->mdct.kfft[k]->scale_shift); +#endif + fprintf (file, "%d, /* shift */\n", mode->mdct.kfft[k]->shift); + fprintf (file, "{"); + for (j=0;j<2*MAXFACTORS;j++) + fprintf (file, "%d, ", mode->mdct.kfft[k]->factors[j]); + fprintf (file, "}, /* factors */\n"); + fprintf (file, "fft_bitrev%d, /* bitrev */\n", mode->mdct.kfft[k]->nfft); + fprintf (file, "fft_twiddles%d_%d, /* bitrev */\n", mode->Fs, mdctSize); + + fprintf (file, "#ifdef OVERRIDE_FFT\n"); + fprintf (file, "(arch_fft_state *)&cfg_arch_%d,\n", mode->mdct.kfft[k]->nfft); + fprintf (file, "#else\n"); + fprintf (file, "NULL,\n"); + fprintf(file, "#endif\n"); + + fprintf (file, "};\n"); + + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + } + + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + + /* MDCT twiddles */ + mdct_twiddles_size = mode->mdct.n-(mode->mdct.n/2>>mode->mdct.maxshift); + fprintf(file, "#ifndef MDCT_TWIDDLES%d\n", mdctSize); + fprintf(file, "#define MDCT_TWIDDLES%d\n", mdctSize); + fprintf (file, "static const opus_val16 mdct_twiddles%d[%d] = {\n", + mdctSize, mdct_twiddles_size); + for (j=0;jmdct.trig[j],(j+6)%5==0?'\n':' '); + fprintf (file, "};\n"); + + fprintf(file, "#endif\n"); + fprintf(file, "\n"); + + + /* Print the actual mode data */ + fprintf(file, "static const CELTMode mode%d_%d_%d = {\n", mode->Fs, mdctSize, mode->overlap); + fprintf(file, INT32 ", /* Fs */\n", mode->Fs); + fprintf(file, "%d, /* overlap */\n", mode->overlap); + fprintf(file, "%d, /* nbEBands */\n", mode->nbEBands); + fprintf(file, "%d, /* effEBands */\n", mode->effEBands); + fprintf(file, "{"); + for (j=0;j<4;j++) + fprintf(file, WORD16 ", ", mode->preemph[j]); + fprintf(file, "}, /* preemph */\n"); + if (standard) + fprintf(file, "eband5ms, /* eBands */\n"); + else + fprintf(file, "eBands%d_%d, /* eBands */\n", mode->Fs, mdctSize); + + fprintf(file, "%d, /* maxLM */\n", mode->maxLM); + fprintf(file, "%d, /* nbShortMdcts */\n", mode->nbShortMdcts); + fprintf(file, "%d, /* shortMdctSize */\n", mode->shortMdctSize); + + fprintf(file, "%d, /* nbAllocVectors */\n", mode->nbAllocVectors); + if (standard) + fprintf(file, "band_allocation, /* allocVectors */\n"); + else + fprintf(file, "allocVectors%d_%d, /* allocVectors */\n", mode->Fs, mdctSize); + + fprintf(file, "logN%d, /* logN */\n", framerate); + fprintf(file, "window%d, /* window */\n", mode->overlap); + fprintf(file, "{%d, %d, {", mode->mdct.n, mode->mdct.maxshift); + for (k=0;k<=mode->mdct.maxshift;k++) + fprintf(file, "&fft_state%d_%d_%d, ", mode->Fs, mdctSize, k); + fprintf (file, "}, mdct_twiddles%d}, /* mdct */\n", mdctSize); + + fprintf(file, "{%d, cache_index%d, cache_bits%d, cache_caps%d}, /* cache */\n", + mode->cache.size, mode->Fs/mdctSize, mode->Fs/mdctSize, mode->Fs/mdctSize); + fprintf(file, "};\n"); + } + fprintf(file, "\n"); + fprintf(file, "/* List of all the available modes */\n"); + fprintf(file, "#define TOTAL_MODES %d\n", nb_modes); + fprintf(file, "static const CELTMode * const static_mode_list[TOTAL_MODES] = {\n"); + for (i=0;ishortMdctSize*mode->nbShortMdcts; + fprintf(file, "&mode%d_%d_%d,\n", mode->Fs, mdctSize, mode->overlap); + } + fprintf(file, "};\n"); +} + +void dump_header(FILE *file, CELTMode **modes, int nb_modes) +{ + int i; + int channels = 0; + int frame_size = 0; + int overlap = 0; + fprintf (file, "/* This header file is generated automatically*/\n"); + for (i=0;ishortMdctSize*mode->nbShortMdcts; + else if (frame_size != mode->shortMdctSize*mode->nbShortMdcts) + frame_size = -1; + if (overlap==0) + overlap = mode->overlap; + else if (overlap != mode->overlap) + overlap = -1; + } + if (channels>0) + { + fprintf (file, "#define CHANNELS(mode) %d\n", channels); + if (channels==1) + fprintf (file, "#define DISABLE_STEREO\n"); + } + if (frame_size>0) + { + fprintf (file, "#define FRAMESIZE(mode) %d\n", frame_size); + } + if (overlap>0) + { + fprintf (file, "#define OVERLAP(mode) %d\n", overlap); + } +} + +#ifdef FIXED_POINT +#define BASENAME "static_modes_fixed" +#else +#define BASENAME "static_modes_float" +#endif + +int main(int argc, char **argv) +{ + int i, nb; + FILE *file; + CELTMode **m; + if (argc%2 != 1 || argc<3) + { + fprintf (stderr, "Usage: %s rate frame_size [rate frame_size] [rate frame_size]...\n",argv[0]); + return 1; + } + nb = (argc-1)/2; + m = malloc(nb*sizeof(CELTMode*)); + for (i=0;i +#include +#include "modes.h" +#include "dump_modes_arch.h" +#include + +#if !defined(FIXED_POINT) +# define NE10_FFT_CFG_TYPE_T ne10_fft_cfg_float32_t +# define NE10_FFT_CPX_TYPE_T_STR "ne10_fft_cpx_float32_t" +# define NE10_FFT_STATE_TYPE_T_STR "ne10_fft_state_float32_t" +#else +# define NE10_FFT_CFG_TYPE_T ne10_fft_cfg_int32_t +# define NE10_FFT_CPX_TYPE_T_STR "ne10_fft_cpx_int32_t" +# define NE10_FFT_STATE_TYPE_T_STR "ne10_fft_state_int32_t" +#endif + +static FILE *file; + +void dump_modes_arch_init(CELTMode **modes, int nb_modes) +{ + int i; + + file = fopen(ARM_NE10_ARCH_FILE_NAME, "w"); + fprintf(file, "/* The contents of this file was automatically generated by\n"); + fprintf(file, " * dump_mode_arm_ne10.c with arguments:"); + for (i=0;iFs,mode->shortMdctSize*mode->nbShortMdcts); + } + fprintf(file, "\n * It contains static definitions for some pre-defined modes. */\n"); + fprintf(file, "#include \n\n"); +} + +void dump_modes_arch_finalize() +{ + fclose(file); +} + +void dump_mode_arch(CELTMode *mode) +{ + int k, j; + int mdctSize; + + mdctSize = mode->shortMdctSize*mode->nbShortMdcts; + + fprintf(file, "#ifndef NE10_FFT_PARAMS%d_%d\n", mode->Fs, mdctSize); + fprintf(file, "#define NE10_FFT_PARAMS%d_%d\n", mode->Fs, mdctSize); + /* cfg->factors */ + for(k=0;k<=mode->mdct.maxshift;k++) { + NE10_FFT_CFG_TYPE_T cfg; + cfg = (NE10_FFT_CFG_TYPE_T)mode->mdct.kfft[k]->arch_fft->priv; + if (!cfg) + continue; + fprintf(file, "static const ne10_int32_t ne10_factors_%d[%d] = {\n", + mode->mdct.kfft[k]->nfft, (NE10_MAXFACTORS * 2)); + for(j=0;j<(NE10_MAXFACTORS * 2);j++) { + fprintf(file, "%d,%c", cfg->factors[j],(j+16)%15==0?'\n':' '); + } + fprintf (file, "};\n"); + } + + /* cfg->twiddles */ + for(k=0;k<=mode->mdct.maxshift;k++) { + NE10_FFT_CFG_TYPE_T cfg; + cfg = (NE10_FFT_CFG_TYPE_T)mode->mdct.kfft[k]->arch_fft->priv; + if (!cfg) + continue; + fprintf(file, "static const %s ne10_twiddles_%d[%d] = {\n", + NE10_FFT_CPX_TYPE_T_STR, mode->mdct.kfft[k]->nfft, + mode->mdct.kfft[k]->nfft); + for(j=0;jmdct.kfft[k]->nfft;j++) { +#if !defined(FIXED_POINT) + fprintf(file, "{%#0.8gf,%#0.8gf},%c", + cfg->twiddles[j].r, cfg->twiddles[j].i,(j+4)%3==0?'\n':' '); +#else + fprintf(file, "{%d,%d},%c", + cfg->twiddles[j].r, cfg->twiddles[j].i,(j+4)%3==0?'\n':' '); +#endif + } + fprintf (file, "};\n"); + } + + for(k=0;k<=mode->mdct.maxshift;k++) { + NE10_FFT_CFG_TYPE_T cfg; + cfg = (NE10_FFT_CFG_TYPE_T)mode->mdct.kfft[k]->arch_fft->priv; + if (!cfg) { + fprintf(file, "/* Ne10 does not support scaled FFT for length = %d */\n", + mode->mdct.kfft[k]->nfft); + fprintf(file, "static const arch_fft_state cfg_arch_%d = {\n", mode->mdct.kfft[k]->nfft); + fprintf(file, "0,\n"); + fprintf(file, "NULL\n"); + fprintf(file, "};\n"); + continue; + } + fprintf(file, "static const %s %s_%d = {\n", NE10_FFT_STATE_TYPE_T_STR, + NE10_FFT_STATE_TYPE_T_STR, mode->mdct.kfft[k]->nfft); + fprintf(file, "%d,\n", cfg->nfft); + fprintf(file, "(ne10_int32_t *)ne10_factors_%d,\n", mode->mdct.kfft[k]->nfft); + fprintf(file, "(%s *)ne10_twiddles_%d,\n", + NE10_FFT_CPX_TYPE_T_STR, mode->mdct.kfft[k]->nfft); + fprintf(file, "NULL,\n"); /* buffer */ + fprintf(file, "(%s *)&ne10_twiddles_%d[%d],\n", + NE10_FFT_CPX_TYPE_T_STR, mode->mdct.kfft[k]->nfft, cfg->nfft); +#if !defined(FIXED_POINT) + fprintf(file, "/* is_forward_scaled = true */\n"); + fprintf(file, "(ne10_int32_t) 1,\n"); + fprintf(file, "/* is_backward_scaled = false */\n"); + fprintf(file, "(ne10_int32_t) 0,\n"); +#endif + fprintf(file, "};\n"); + + fprintf(file, "static const arch_fft_state cfg_arch_%d = {\n", + mode->mdct.kfft[k]->nfft); + fprintf(file, "1,\n"); + fprintf(file, "(void *)&%s_%d,\n", + NE10_FFT_STATE_TYPE_T_STR, mode->mdct.kfft[k]->nfft); + fprintf(file, "};\n\n"); + } + fprintf(file, "#endif /* end NE10_FFT_PARAMS%d_%d */\n", mode->Fs, mdctSize); +} diff --git a/lib/opus/celt/ecintrin.h b/lib/opus/celt/ecintrin.h new file mode 100644 index 00000000..66a4c36e --- /dev/null +++ b/lib/opus/celt/ecintrin.h @@ -0,0 +1,91 @@ +/* Copyright (c) 2003-2008 Timothy B. Terriberry + Copyright (c) 2008 Xiph.Org Foundation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/*Some common macros for potential platform-specific optimization.*/ +#include "opus_types.h" +#include +#include +#include "arch.h" +#if !defined(_ecintrin_H) +# define _ecintrin_H (1) + +/*Some specific platforms may have optimized intrinsic or OPUS_INLINE assembly + versions of these functions which can substantially improve performance. + We define macros for them to allow easy incorporation of these non-ANSI + features.*/ + +/*Modern gcc (4.x) can compile the naive versions of min and max with cmov if + given an appropriate architecture, but the branchless bit-twiddling versions + are just as fast, and do not require any special target architecture. + Earlier gcc versions (3.x) compiled both code to the same assembly + instructions, because of the way they represented ((_b)>(_a)) internally.*/ +# define EC_MINI(_a,_b) ((_a)+(((_b)-(_a))&-((_b)<(_a)))) + +/*Count leading zeros. + This macro should only be used for implementing ec_ilog(), if it is defined. + All other code should use EC_ILOG() instead.*/ +#if defined(_MSC_VER) && (_MSC_VER >= 1400) +#if defined(_MSC_VER) && (_MSC_VER >= 1910) +# include /* Improve compiler throughput. */ +#else +# include +#endif +/*In _DEBUG mode this is not an intrinsic by default.*/ +# pragma intrinsic(_BitScanReverse) + +static __inline int ec_bsr(unsigned long _x){ + unsigned long ret; + _BitScanReverse(&ret,_x); + return (int)ret; +} +# define EC_CLZ0 (1) +# define EC_CLZ(_x) (-ec_bsr(_x)) +#elif defined(ENABLE_TI_DSPLIB) +# include "dsplib.h" +# define EC_CLZ0 (31) +# define EC_CLZ(_x) (_lnorm(_x)) +#elif __GNUC_PREREQ(3,4) +# if INT_MAX>=2147483647 +# define EC_CLZ0 ((int)sizeof(unsigned)*CHAR_BIT) +# define EC_CLZ(_x) (__builtin_clz(_x)) +# elif LONG_MAX>=2147483647L +# define EC_CLZ0 ((int)sizeof(unsigned long)*CHAR_BIT) +# define EC_CLZ(_x) (__builtin_clzl(_x)) +# endif +#endif + +#if defined(EC_CLZ) +/*Note that __builtin_clz is not defined when _x==0, according to the gcc + documentation (and that of the BSR instruction that implements it on x86). + The majority of the time we can never pass it zero. + When we need to, it can be special cased.*/ +# define EC_ILOG(_x) (EC_CLZ0-EC_CLZ(_x)) +#else +int ec_ilog(opus_uint32 _v); +# define EC_ILOG(_x) (ec_ilog(_x)) +#endif +#endif diff --git a/lib/opus/celt/entcode.c b/lib/opus/celt/entcode.c new file mode 100644 index 00000000..70f32016 --- /dev/null +++ b/lib/opus/celt/entcode.c @@ -0,0 +1,153 @@ +/* Copyright (c) 2001-2011 Timothy B. Terriberry +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "entcode.h" +#include "arch.h" + +#if !defined(EC_CLZ) +/*This is a fallback for systems where we don't know how to access + a BSR or CLZ instruction (see ecintrin.h). + If you are optimizing Opus on a new platform and it has a native CLZ or + BZR (e.g. cell, MIPS, x86, etc) then making it available to Opus will be + an easy performance win.*/ +int ec_ilog(opus_uint32 _v){ + /*On a Pentium M, this branchless version tested as the fastest on + 1,000,000,000 random 32-bit integers, edging out a similar version with + branches, and a 256-entry LUT version.*/ + int ret; + int m; + ret=!!_v; + m=!!(_v&0xFFFF0000)<<4; + _v>>=m; + ret|=m; + m=!!(_v&0xFF00)<<3; + _v>>=m; + ret|=m; + m=!!(_v&0xF0)<<2; + _v>>=m; + ret|=m; + m=!!(_v&0xC)<<1; + _v>>=m; + ret|=m; + ret+=!!(_v&0x2); + return ret; +} +#endif + +#if 1 +/* This is a faster version of ec_tell_frac() that takes advantage + of the low (1/8 bit) resolution to use just a linear function + followed by a lookup to determine the exact transition thresholds. */ +opus_uint32 ec_tell_frac(ec_ctx *_this){ + static const unsigned correction[8] = + {35733, 38967, 42495, 46340, + 50535, 55109, 60097, 65535}; + opus_uint32 nbits; + opus_uint32 r; + int l; + unsigned b; + nbits=_this->nbits_total<rng); + r=_this->rng>>(l-16); + b = (r>>12)-8; + b += r>correction[b]; + l = (l<<3)+b; + return nbits-l; +} +#else +opus_uint32 ec_tell_frac(ec_ctx *_this){ + opus_uint32 nbits; + opus_uint32 r; + int l; + int i; + /*To handle the non-integral number of bits still left in the encoder/decoder + state, we compute the worst-case number of bits of val that must be + encoded to ensure that the value is inside the range for any possible + subsequent bits. + The computation here is independent of val itself (the decoder does not + even track that value), even though the real number of bits used after + ec_enc_done() may be 1 smaller if rng is a power of two and the + corresponding trailing bits of val are all zeros. + If we did try to track that special case, then coding a value with a + probability of 1/(1<nbits_total<rng); + r=_this->rng>>(l-16); + for(i=BITRES;i-->0;){ + int b; + r=r*r>>15; + b=(int)(r>>16); + l=l<<1|b; + r>>=b; + } + return nbits-l; +} +#endif + +#ifdef USE_SMALL_DIV_TABLE +/* Result of 2^32/(2*i+1), except for i=0. */ +const opus_uint32 SMALL_DIV_TABLE[129] = { + 0xFFFFFFFF, 0x55555555, 0x33333333, 0x24924924, + 0x1C71C71C, 0x1745D174, 0x13B13B13, 0x11111111, + 0x0F0F0F0F, 0x0D79435E, 0x0C30C30C, 0x0B21642C, + 0x0A3D70A3, 0x097B425E, 0x08D3DCB0, 0x08421084, + 0x07C1F07C, 0x07507507, 0x06EB3E45, 0x06906906, + 0x063E7063, 0x05F417D0, 0x05B05B05, 0x0572620A, + 0x05397829, 0x05050505, 0x04D4873E, 0x04A7904A, + 0x047DC11F, 0x0456C797, 0x04325C53, 0x04104104, + 0x03F03F03, 0x03D22635, 0x03B5CC0E, 0x039B0AD1, + 0x0381C0E0, 0x0369D036, 0x03531DEC, 0x033D91D2, + 0x0329161F, 0x03159721, 0x03030303, 0x02F14990, + 0x02E05C0B, 0x02D02D02, 0x02C0B02C, 0x02B1DA46, + 0x02A3A0FD, 0x0295FAD4, 0x0288DF0C, 0x027C4597, + 0x02702702, 0x02647C69, 0x02593F69, 0x024E6A17, + 0x0243F6F0, 0x0239E0D5, 0x02302302, 0x0226B902, + 0x021D9EAD, 0x0214D021, 0x020C49BA, 0x02040810, + 0x01FC07F0, 0x01F44659, 0x01ECC07B, 0x01E573AC, + 0x01DE5D6E, 0x01D77B65, 0x01D0CB58, 0x01CA4B30, + 0x01C3F8F0, 0x01BDD2B8, 0x01B7D6C3, 0x01B20364, + 0x01AC5701, 0x01A6D01A, 0x01A16D3F, 0x019C2D14, + 0x01970E4F, 0x01920FB4, 0x018D3018, 0x01886E5F, + 0x0183C977, 0x017F405F, 0x017AD220, 0x01767DCE, + 0x01724287, 0x016E1F76, 0x016A13CD, 0x01661EC6, + 0x01623FA7, 0x015E75BB, 0x015AC056, 0x01571ED3, + 0x01539094, 0x01501501, 0x014CAB88, 0x0149539E, + 0x01460CBC, 0x0142D662, 0x013FB013, 0x013C995A, + 0x013991C2, 0x013698DF, 0x0133AE45, 0x0130D190, + 0x012E025C, 0x012B404A, 0x01288B01, 0x0125E227, + 0x01234567, 0x0120B470, 0x011E2EF3, 0x011BB4A4, + 0x01194538, 0x0116E068, 0x011485F0, 0x0112358E, + 0x010FEF01, 0x010DB20A, 0x010B7E6E, 0x010953F3, + 0x01073260, 0x0105197F, 0x0103091B, 0x01010101 +}; +#endif diff --git a/lib/opus/celt/entcode.h b/lib/opus/celt/entcode.h new file mode 100644 index 00000000..3763e3f2 --- /dev/null +++ b/lib/opus/celt/entcode.h @@ -0,0 +1,152 @@ +/* Copyright (c) 2001-2011 Timothy B. Terriberry + Copyright (c) 2008-2009 Xiph.Org Foundation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#include "opus_types.h" +#include "opus_defines.h" + +#if !defined(_entcode_H) +# define _entcode_H (1) +# include +# include +# include "ecintrin.h" + +extern const opus_uint32 SMALL_DIV_TABLE[129]; + +#ifdef OPUS_ARM_ASM +#define USE_SMALL_DIV_TABLE +#endif + +/*OPT: ec_window must be at least 32 bits, but if you have fast arithmetic on a + larger type, you can speed up the decoder by using it here.*/ +typedef opus_uint32 ec_window; +typedef struct ec_ctx ec_ctx; +typedef struct ec_ctx ec_enc; +typedef struct ec_ctx ec_dec; + +# define EC_WINDOW_SIZE ((int)sizeof(ec_window)*CHAR_BIT) + +/*The number of bits to use for the range-coded part of unsigned integers.*/ +# define EC_UINT_BITS (8) + +/*The resolution of fractional-precision bit usage measurements, i.e., + 3 => 1/8th bits.*/ +# define BITRES 3 + +/*The entropy encoder/decoder context. + We use the same structure for both, so that common functions like ec_tell() + can be used on either one.*/ +struct ec_ctx{ + /*Buffered input/output.*/ + unsigned char *buf; + /*The size of the buffer.*/ + opus_uint32 storage; + /*The offset at which the last byte containing raw bits was read/written.*/ + opus_uint32 end_offs; + /*Bits that will be read from/written at the end.*/ + ec_window end_window; + /*Number of valid bits in end_window.*/ + int nend_bits; + /*The total number of whole bits read/written. + This does not include partial bits currently in the range coder.*/ + int nbits_total; + /*The offset at which the next range coder byte will be read/written.*/ + opus_uint32 offs; + /*The number of values in the current range.*/ + opus_uint32 rng; + /*In the decoder: the difference between the top of the current range and + the input value, minus one. + In the encoder: the low end of the current range.*/ + opus_uint32 val; + /*In the decoder: the saved normalization factor from ec_decode(). + In the encoder: the number of oustanding carry propagating symbols.*/ + opus_uint32 ext; + /*A buffered input/output symbol, awaiting carry propagation.*/ + int rem; + /*Nonzero if an error occurred.*/ + int error; +}; + +static OPUS_INLINE opus_uint32 ec_range_bytes(ec_ctx *_this){ + return _this->offs; +} + +static OPUS_INLINE unsigned char *ec_get_buffer(ec_ctx *_this){ + return _this->buf; +} + +static OPUS_INLINE int ec_get_error(ec_ctx *_this){ + return _this->error; +} + +/*Returns the number of bits "used" by the encoded or decoded symbols so far. + This same number can be computed in either the encoder or the decoder, and is + suitable for making coding decisions. + Return: The number of bits. + This will always be slightly larger than the exact value (e.g., all + rounding error is in the positive direction).*/ +static OPUS_INLINE int ec_tell(ec_ctx *_this){ + return _this->nbits_total-EC_ILOG(_this->rng); +} + +/*Returns the number of bits "used" by the encoded or decoded symbols so far. + This same number can be computed in either the encoder or the decoder, and is + suitable for making coding decisions. + Return: The number of bits scaled by 2**BITRES. + This will always be slightly larger than the exact value (e.g., all + rounding error is in the positive direction).*/ +opus_uint32 ec_tell_frac(ec_ctx *_this); + +/* Tested exhaustively for all n and for 1<=d<=256 */ +static OPUS_INLINE opus_uint32 celt_udiv(opus_uint32 n, opus_uint32 d) { + celt_sig_assert(d>0); +#ifdef USE_SMALL_DIV_TABLE + if (d>256) + return n/d; + else { + opus_uint32 t, q; + t = EC_ILOG(d&-d); + q = (opus_uint64)SMALL_DIV_TABLE[d>>t]*(n>>(t-1))>>32; + return q+(n-q*d >= d); + } +#else + return n/d; +#endif +} + +static OPUS_INLINE opus_int32 celt_sudiv(opus_int32 n, opus_int32 d) { + celt_sig_assert(d>0); +#ifdef USE_SMALL_DIV_TABLE + if (n<0) + return -(opus_int32)celt_udiv(-n, d); + else + return celt_udiv(n, d); +#else + return n/d; +#endif +} + +#endif diff --git a/lib/opus/celt/entdec.c b/lib/opus/celt/entdec.c new file mode 100644 index 00000000..0b3433ed --- /dev/null +++ b/lib/opus/celt/entdec.c @@ -0,0 +1,245 @@ +/* Copyright (c) 2001-2011 Timothy B. Terriberry + Copyright (c) 2008-2009 Xiph.Org Foundation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "os_support.h" +#include "arch.h" +#include "entdec.h" +#include "mfrngcod.h" + +/*A range decoder. + This is an entropy decoder based upon \cite{Mar79}, which is itself a + rediscovery of the FIFO arithmetic code introduced by \cite{Pas76}. + It is very similar to arithmetic encoding, except that encoding is done with + digits in any base, instead of with bits, and so it is faster when using + larger bases (i.e.: a byte). + The author claims an average waste of $\frac{1}{2}\log_b(2b)$ bits, where $b$ + is the base, longer than the theoretical optimum, but to my knowledge there + is no published justification for this claim. + This only seems true when using near-infinite precision arithmetic so that + the process is carried out with no rounding errors. + + An excellent description of implementation details is available at + http://www.arturocampos.com/ac_range.html + A recent work \cite{MNW98} which proposes several changes to arithmetic + encoding for efficiency actually re-discovers many of the principles + behind range encoding, and presents a good theoretical analysis of them. + + End of stream is handled by writing out the smallest number of bits that + ensures that the stream will be correctly decoded regardless of the value of + any subsequent bits. + ec_tell() can be used to determine how many bits were needed to decode + all the symbols thus far; other data can be packed in the remaining bits of + the input buffer. + @PHDTHESIS{Pas76, + author="Richard Clark Pasco", + title="Source coding algorithms for fast data compression", + school="Dept. of Electrical Engineering, Stanford University", + address="Stanford, CA", + month=May, + year=1976 + } + @INPROCEEDINGS{Mar79, + author="Martin, G.N.N.", + title="Range encoding: an algorithm for removing redundancy from a digitised + message", + booktitle="Video & Data Recording Conference", + year=1979, + address="Southampton", + month=Jul + } + @ARTICLE{MNW98, + author="Alistair Moffat and Radford Neal and Ian H. Witten", + title="Arithmetic Coding Revisited", + journal="{ACM} Transactions on Information Systems", + year=1998, + volume=16, + number=3, + pages="256--294", + month=Jul, + URL="http://www.stanford.edu/class/ee398a/handouts/papers/Moffat98ArithmCoding.pdf" + }*/ + +static int ec_read_byte(ec_dec *_this){ + return _this->offs<_this->storage?_this->buf[_this->offs++]:0; +} + +static int ec_read_byte_from_end(ec_dec *_this){ + return _this->end_offs<_this->storage? + _this->buf[_this->storage-++(_this->end_offs)]:0; +} + +/*Normalizes the contents of val and rng so that rng lies entirely in the + high-order symbol.*/ +static void ec_dec_normalize(ec_dec *_this){ + /*If the range is too small, rescale it and input some bits.*/ + while(_this->rng<=EC_CODE_BOT){ + int sym; + _this->nbits_total+=EC_SYM_BITS; + _this->rng<<=EC_SYM_BITS; + /*Use up the remaining bits from our last symbol.*/ + sym=_this->rem; + /*Read the next value from the input.*/ + _this->rem=ec_read_byte(_this); + /*Take the rest of the bits we need from this new symbol.*/ + sym=(sym<rem)>>(EC_SYM_BITS-EC_CODE_EXTRA); + /*And subtract them from val, capped to be less than EC_CODE_TOP.*/ + _this->val=((_this->val<buf=_buf; + _this->storage=_storage; + _this->end_offs=0; + _this->end_window=0; + _this->nend_bits=0; + /*This is the offset from which ec_tell() will subtract partial bits. + The final value after the ec_dec_normalize() call will be the same as in + the encoder, but we have to compensate for the bits that are added there.*/ + _this->nbits_total=EC_CODE_BITS+1 + -((EC_CODE_BITS-EC_CODE_EXTRA)/EC_SYM_BITS)*EC_SYM_BITS; + _this->offs=0; + _this->rng=1U<rem=ec_read_byte(_this); + _this->val=_this->rng-1-(_this->rem>>(EC_SYM_BITS-EC_CODE_EXTRA)); + _this->error=0; + /*Normalize the interval.*/ + ec_dec_normalize(_this); +} + +unsigned ec_decode(ec_dec *_this,unsigned _ft){ + unsigned s; + _this->ext=celt_udiv(_this->rng,_ft); + s=(unsigned)(_this->val/_this->ext); + return _ft-EC_MINI(s+1,_ft); +} + +unsigned ec_decode_bin(ec_dec *_this,unsigned _bits){ + unsigned s; + _this->ext=_this->rng>>_bits; + s=(unsigned)(_this->val/_this->ext); + return (1U<<_bits)-EC_MINI(s+1U,1U<<_bits); +} + +void ec_dec_update(ec_dec *_this,unsigned _fl,unsigned _fh,unsigned _ft){ + opus_uint32 s; + s=IMUL32(_this->ext,_ft-_fh); + _this->val-=s; + _this->rng=_fl>0?IMUL32(_this->ext,_fh-_fl):_this->rng-s; + ec_dec_normalize(_this); +} + +/*The probability of having a "one" is 1/(1<<_logp).*/ +int ec_dec_bit_logp(ec_dec *_this,unsigned _logp){ + opus_uint32 r; + opus_uint32 d; + opus_uint32 s; + int ret; + r=_this->rng; + d=_this->val; + s=r>>_logp; + ret=dval=d-s; + _this->rng=ret?s:r-s; + ec_dec_normalize(_this); + return ret; +} + +int ec_dec_icdf(ec_dec *_this,const unsigned char *_icdf,unsigned _ftb){ + opus_uint32 r; + opus_uint32 d; + opus_uint32 s; + opus_uint32 t; + int ret; + s=_this->rng; + d=_this->val; + r=s>>_ftb; + ret=-1; + do{ + t=s; + s=IMUL32(r,_icdf[++ret]); + } + while(dval=d-s; + _this->rng=t-s; + ec_dec_normalize(_this); + return ret; +} + +opus_uint32 ec_dec_uint(ec_dec *_this,opus_uint32 _ft){ + unsigned ft; + unsigned s; + int ftb; + /*In order to optimize EC_ILOG(), it is undefined for the value 0.*/ + celt_assert(_ft>1); + _ft--; + ftb=EC_ILOG(_ft); + if(ftb>EC_UINT_BITS){ + opus_uint32 t; + ftb-=EC_UINT_BITS; + ft=(unsigned)(_ft>>ftb)+1; + s=ec_decode(_this,ft); + ec_dec_update(_this,s,s+1,ft); + t=(opus_uint32)s<error=1; + return _ft; + } + else{ + _ft++; + s=ec_decode(_this,(unsigned)_ft); + ec_dec_update(_this,s,s+1,(unsigned)_ft); + return s; + } +} + +opus_uint32 ec_dec_bits(ec_dec *_this,unsigned _bits){ + ec_window window; + int available; + opus_uint32 ret; + window=_this->end_window; + available=_this->nend_bits; + if((unsigned)available<_bits){ + do{ + window|=(ec_window)ec_read_byte_from_end(_this)<>=_bits; + available-=_bits; + _this->end_window=window; + _this->nend_bits=available; + _this->nbits_total+=_bits; + return ret; +} diff --git a/lib/opus/celt/entdec.h b/lib/opus/celt/entdec.h new file mode 100644 index 00000000..025fc187 --- /dev/null +++ b/lib/opus/celt/entdec.h @@ -0,0 +1,100 @@ +/* Copyright (c) 2001-2011 Timothy B. Terriberry + Copyright (c) 2008-2009 Xiph.Org Foundation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(_entdec_H) +# define _entdec_H (1) +# include +# include "entcode.h" + +/*Initializes the decoder. + _buf: The input buffer to use. + Return: 0 on success, or a negative value on error.*/ +void ec_dec_init(ec_dec *_this,unsigned char *_buf,opus_uint32 _storage); + +/*Calculates the cumulative frequency for the next symbol. + This can then be fed into the probability model to determine what that + symbol is, and the additional frequency information required to advance to + the next symbol. + This function cannot be called more than once without a corresponding call to + ec_dec_update(), or decoding will not proceed correctly. + _ft: The total frequency of the symbols in the alphabet the next symbol was + encoded with. + Return: A cumulative frequency representing the encoded symbol. + If the cumulative frequency of all the symbols before the one that + was encoded was fl, and the cumulative frequency of all the symbols + up to and including the one encoded is fh, then the returned value + will fall in the range [fl,fh).*/ +unsigned ec_decode(ec_dec *_this,unsigned _ft); + +/*Equivalent to ec_decode() with _ft==1<<_bits.*/ +unsigned ec_decode_bin(ec_dec *_this,unsigned _bits); + +/*Advance the decoder past the next symbol using the frequency information the + symbol was encoded with. + Exactly one call to ec_decode() must have been made so that all necessary + intermediate calculations are performed. + _fl: The cumulative frequency of all symbols that come before the symbol + decoded. + _fh: The cumulative frequency of all symbols up to and including the symbol + decoded. + Together with _fl, this defines the range [_fl,_fh) in which the value + returned above must fall. + _ft: The total frequency of the symbols in the alphabet the symbol decoded + was encoded in. + This must be the same as passed to the preceding call to ec_decode().*/ +void ec_dec_update(ec_dec *_this,unsigned _fl,unsigned _fh,unsigned _ft); + +/* Decode a bit that has a 1/(1<<_logp) probability of being a one */ +int ec_dec_bit_logp(ec_dec *_this,unsigned _logp); + +/*Decodes a symbol given an "inverse" CDF table. + No call to ec_dec_update() is necessary after this call. + _icdf: The "inverse" CDF, such that symbol s falls in the range + [s>0?ft-_icdf[s-1]:0,ft-_icdf[s]), where ft=1<<_ftb. + The values must be monotonically non-increasing, and the last value + must be 0. + _ftb: The number of bits of precision in the cumulative distribution. + Return: The decoded symbol s.*/ +int ec_dec_icdf(ec_dec *_this,const unsigned char *_icdf,unsigned _ftb); + +/*Extracts a raw unsigned integer with a non-power-of-2 range from the stream. + The bits must have been encoded with ec_enc_uint(). + No call to ec_dec_update() is necessary after this call. + _ft: The number of integers that can be decoded (one more than the max). + This must be at least 2, and no more than 2**32-1. + Return: The decoded bits.*/ +opus_uint32 ec_dec_uint(ec_dec *_this,opus_uint32 _ft); + +/*Extracts a sequence of raw bits from the stream. + The bits must have been encoded with ec_enc_bits(). + No call to ec_dec_update() is necessary after this call. + _ftb: The number of bits to extract. + This must be between 0 and 25, inclusive. + Return: The decoded bits.*/ +opus_uint32 ec_dec_bits(ec_dec *_this,unsigned _ftb); + +#endif diff --git a/lib/opus/celt/entenc.c b/lib/opus/celt/entenc.c new file mode 100644 index 00000000..f1750d25 --- /dev/null +++ b/lib/opus/celt/entenc.c @@ -0,0 +1,294 @@ +/* Copyright (c) 2001-2011 Timothy B. Terriberry + Copyright (c) 2008-2009 Xiph.Org Foundation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if defined(HAVE_CONFIG_H) +# include "config.h" +#endif +#include "os_support.h" +#include "arch.h" +#include "entenc.h" +#include "mfrngcod.h" + +/*A range encoder. + See entdec.c and the references for implementation details \cite{Mar79,MNW98}. + + @INPROCEEDINGS{Mar79, + author="Martin, G.N.N.", + title="Range encoding: an algorithm for removing redundancy from a digitised + message", + booktitle="Video \& Data Recording Conference", + year=1979, + address="Southampton", + month=Jul + } + @ARTICLE{MNW98, + author="Alistair Moffat and Radford Neal and Ian H. Witten", + title="Arithmetic Coding Revisited", + journal="{ACM} Transactions on Information Systems", + year=1998, + volume=16, + number=3, + pages="256--294", + month=Jul, + URL="http://www.stanford.edu/class/ee398/handouts/papers/Moffat98ArithmCoding.pdf" + }*/ + +static int ec_write_byte(ec_enc *_this,unsigned _value){ + if(_this->offs+_this->end_offs>=_this->storage)return -1; + _this->buf[_this->offs++]=(unsigned char)_value; + return 0; +} + +static int ec_write_byte_at_end(ec_enc *_this,unsigned _value){ + if(_this->offs+_this->end_offs>=_this->storage)return -1; + _this->buf[_this->storage-++(_this->end_offs)]=(unsigned char)_value; + return 0; +} + +/*Outputs a symbol, with a carry bit. + If there is a potential to propagate a carry over several symbols, they are + buffered until it can be determined whether or not an actual carry will + occur. + If the counter for the buffered symbols overflows, then the stream becomes + undecodable. + This gives a theoretical limit of a few billion symbols in a single packet on + 32-bit systems. + The alternative is to truncate the range in order to force a carry, but + requires similar carry tracking in the decoder, needlessly slowing it down.*/ +static void ec_enc_carry_out(ec_enc *_this,int _c){ + if(_c!=EC_SYM_MAX){ + /*No further carry propagation possible, flush buffer.*/ + int carry; + carry=_c>>EC_SYM_BITS; + /*Don't output a byte on the first write. + This compare should be taken care of by branch-prediction thereafter.*/ + if(_this->rem>=0)_this->error|=ec_write_byte(_this,_this->rem+carry); + if(_this->ext>0){ + unsigned sym; + sym=(EC_SYM_MAX+carry)&EC_SYM_MAX; + do _this->error|=ec_write_byte(_this,sym); + while(--(_this->ext)>0); + } + _this->rem=_c&EC_SYM_MAX; + } + else _this->ext++; +} + +static OPUS_INLINE void ec_enc_normalize(ec_enc *_this){ + /*If the range is too small, output some bits and rescale it.*/ + while(_this->rng<=EC_CODE_BOT){ + ec_enc_carry_out(_this,(int)(_this->val>>EC_CODE_SHIFT)); + /*Move the next-to-high-order symbol into the high-order position.*/ + _this->val=(_this->val<rng<<=EC_SYM_BITS; + _this->nbits_total+=EC_SYM_BITS; + } +} + +void ec_enc_init(ec_enc *_this,unsigned char *_buf,opus_uint32 _size){ + _this->buf=_buf; + _this->end_offs=0; + _this->end_window=0; + _this->nend_bits=0; + /*This is the offset from which ec_tell() will subtract partial bits.*/ + _this->nbits_total=EC_CODE_BITS+1; + _this->offs=0; + _this->rng=EC_CODE_TOP; + _this->rem=-1; + _this->val=0; + _this->ext=0; + _this->storage=_size; + _this->error=0; +} + +void ec_encode(ec_enc *_this,unsigned _fl,unsigned _fh,unsigned _ft){ + opus_uint32 r; + r=celt_udiv(_this->rng,_ft); + if(_fl>0){ + _this->val+=_this->rng-IMUL32(r,(_ft-_fl)); + _this->rng=IMUL32(r,(_fh-_fl)); + } + else _this->rng-=IMUL32(r,(_ft-_fh)); + ec_enc_normalize(_this); +} + +void ec_encode_bin(ec_enc *_this,unsigned _fl,unsigned _fh,unsigned _bits){ + opus_uint32 r; + r=_this->rng>>_bits; + if(_fl>0){ + _this->val+=_this->rng-IMUL32(r,((1U<<_bits)-_fl)); + _this->rng=IMUL32(r,(_fh-_fl)); + } + else _this->rng-=IMUL32(r,((1U<<_bits)-_fh)); + ec_enc_normalize(_this); +} + +/*The probability of having a "one" is 1/(1<<_logp).*/ +void ec_enc_bit_logp(ec_enc *_this,int _val,unsigned _logp){ + opus_uint32 r; + opus_uint32 s; + opus_uint32 l; + r=_this->rng; + l=_this->val; + s=r>>_logp; + r-=s; + if(_val)_this->val=l+r; + _this->rng=_val?s:r; + ec_enc_normalize(_this); +} + +void ec_enc_icdf(ec_enc *_this,int _s,const unsigned char *_icdf,unsigned _ftb){ + opus_uint32 r; + r=_this->rng>>_ftb; + if(_s>0){ + _this->val+=_this->rng-IMUL32(r,_icdf[_s-1]); + _this->rng=IMUL32(r,_icdf[_s-1]-_icdf[_s]); + } + else _this->rng-=IMUL32(r,_icdf[_s]); + ec_enc_normalize(_this); +} + +void ec_enc_uint(ec_enc *_this,opus_uint32 _fl,opus_uint32 _ft){ + unsigned ft; + unsigned fl; + int ftb; + /*In order to optimize EC_ILOG(), it is undefined for the value 0.*/ + celt_assert(_ft>1); + _ft--; + ftb=EC_ILOG(_ft); + if(ftb>EC_UINT_BITS){ + ftb-=EC_UINT_BITS; + ft=(_ft>>ftb)+1; + fl=(unsigned)(_fl>>ftb); + ec_encode(_this,fl,fl+1,ft); + ec_enc_bits(_this,_fl&(((opus_uint32)1<end_window; + used=_this->nend_bits; + celt_assert(_bits>0); + if(used+_bits>EC_WINDOW_SIZE){ + do{ + _this->error|=ec_write_byte_at_end(_this,(unsigned)window&EC_SYM_MAX); + window>>=EC_SYM_BITS; + used-=EC_SYM_BITS; + } + while(used>=EC_SYM_BITS); + } + window|=(ec_window)_fl<end_window=window; + _this->nend_bits=used; + _this->nbits_total+=_bits; +} + +void ec_enc_patch_initial_bits(ec_enc *_this,unsigned _val,unsigned _nbits){ + int shift; + unsigned mask; + celt_assert(_nbits<=EC_SYM_BITS); + shift=EC_SYM_BITS-_nbits; + mask=((1<<_nbits)-1)<offs>0){ + /*The first byte has been finalized.*/ + _this->buf[0]=(unsigned char)((_this->buf[0]&~mask)|_val<rem>=0){ + /*The first byte is still awaiting carry propagation.*/ + _this->rem=(_this->rem&~mask)|_val<rng<=(EC_CODE_TOP>>_nbits)){ + /*The renormalization loop has never been run.*/ + _this->val=(_this->val&~((opus_uint32)mask<error=-1; +} + +void ec_enc_shrink(ec_enc *_this,opus_uint32 _size){ + celt_assert(_this->offs+_this->end_offs<=_size); + OPUS_MOVE(_this->buf+_size-_this->end_offs, + _this->buf+_this->storage-_this->end_offs,_this->end_offs); + _this->storage=_size; +} + +void ec_enc_done(ec_enc *_this){ + ec_window window; + int used; + opus_uint32 msk; + opus_uint32 end; + int l; + /*We output the minimum number of bits that ensures that the symbols encoded + thus far will be decoded correctly regardless of the bits that follow.*/ + l=EC_CODE_BITS-EC_ILOG(_this->rng); + msk=(EC_CODE_TOP-1)>>l; + end=(_this->val+msk)&~msk; + if((end|msk)>=_this->val+_this->rng){ + l++; + msk>>=1; + end=(_this->val+msk)&~msk; + } + while(l>0){ + ec_enc_carry_out(_this,(int)(end>>EC_CODE_SHIFT)); + end=(end<rem>=0||_this->ext>0)ec_enc_carry_out(_this,0); + /*If we have buffered extra bits, flush them as well.*/ + window=_this->end_window; + used=_this->nend_bits; + while(used>=EC_SYM_BITS){ + _this->error|=ec_write_byte_at_end(_this,(unsigned)window&EC_SYM_MAX); + window>>=EC_SYM_BITS; + used-=EC_SYM_BITS; + } + /*Clear any excess space and add any remaining extra bits to the last byte.*/ + if(!_this->error){ + OPUS_CLEAR(_this->buf+_this->offs, + _this->storage-_this->offs-_this->end_offs); + if(used>0){ + /*If there's no range coder data at all, give up.*/ + if(_this->end_offs>=_this->storage)_this->error=-1; + else{ + l=-l; + /*If we've busted, don't add too many extra bits to the last byte; it + would corrupt the range coder data, and that's more important.*/ + if(_this->offs+_this->end_offs>=_this->storage&&lerror=-1; + } + _this->buf[_this->storage-_this->end_offs-1]|=(unsigned char)window; + } + } + } +} diff --git a/lib/opus/celt/entenc.h b/lib/opus/celt/entenc.h new file mode 100644 index 00000000..f502eaf6 --- /dev/null +++ b/lib/opus/celt/entenc.h @@ -0,0 +1,110 @@ +/* Copyright (c) 2001-2011 Timothy B. Terriberry + Copyright (c) 2008-2009 Xiph.Org Foundation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(_entenc_H) +# define _entenc_H (1) +# include +# include "entcode.h" + +/*Initializes the encoder. + _buf: The buffer to store output bytes in. + _size: The size of the buffer, in chars.*/ +void ec_enc_init(ec_enc *_this,unsigned char *_buf,opus_uint32 _size); +/*Encodes a symbol given its frequency information. + The frequency information must be discernable by the decoder, assuming it + has read only the previous symbols from the stream. + It is allowable to change the frequency information, or even the entire + source alphabet, so long as the decoder can tell from the context of the + previously encoded information that it is supposed to do so as well. + _fl: The cumulative frequency of all symbols that come before the one to be + encoded. + _fh: The cumulative frequency of all symbols up to and including the one to + be encoded. + Together with _fl, this defines the range [_fl,_fh) in which the + decoded value will fall. + _ft: The sum of the frequencies of all the symbols*/ +void ec_encode(ec_enc *_this,unsigned _fl,unsigned _fh,unsigned _ft); + +/*Equivalent to ec_encode() with _ft==1<<_bits.*/ +void ec_encode_bin(ec_enc *_this,unsigned _fl,unsigned _fh,unsigned _bits); + +/* Encode a bit that has a 1/(1<<_logp) probability of being a one */ +void ec_enc_bit_logp(ec_enc *_this,int _val,unsigned _logp); + +/*Encodes a symbol given an "inverse" CDF table. + _s: The index of the symbol to encode. + _icdf: The "inverse" CDF, such that symbol _s falls in the range + [_s>0?ft-_icdf[_s-1]:0,ft-_icdf[_s]), where ft=1<<_ftb. + The values must be monotonically non-increasing, and the last value + must be 0. + _ftb: The number of bits of precision in the cumulative distribution.*/ +void ec_enc_icdf(ec_enc *_this,int _s,const unsigned char *_icdf,unsigned _ftb); + +/*Encodes a raw unsigned integer in the stream. + _fl: The integer to encode. + _ft: The number of integers that can be encoded (one more than the max). + This must be at least 2, and no more than 2**32-1.*/ +void ec_enc_uint(ec_enc *_this,opus_uint32 _fl,opus_uint32 _ft); + +/*Encodes a sequence of raw bits in the stream. + _fl: The bits to encode. + _ftb: The number of bits to encode. + This must be between 1 and 25, inclusive.*/ +void ec_enc_bits(ec_enc *_this,opus_uint32 _fl,unsigned _ftb); + +/*Overwrites a few bits at the very start of an existing stream, after they + have already been encoded. + This makes it possible to have a few flags up front, where it is easy for + decoders to access them without parsing the whole stream, even if their + values are not determined until late in the encoding process, without having + to buffer all the intermediate symbols in the encoder. + In order for this to work, at least _nbits bits must have already been + encoded using probabilities that are an exact power of two. + The encoder can verify the number of encoded bits is sufficient, but cannot + check this latter condition. + _val: The bits to encode (in the least _nbits significant bits). + They will be decoded in order from most-significant to least. + _nbits: The number of bits to overwrite. + This must be no more than 8.*/ +void ec_enc_patch_initial_bits(ec_enc *_this,unsigned _val,unsigned _nbits); + +/*Compacts the data to fit in the target size. + This moves up the raw bits at the end of the current buffer so they are at + the end of the new buffer size. + The caller must ensure that the amount of data that's already been written + will fit in the new size. + _size: The number of bytes in the new buffer. + This must be large enough to contain the bits already written, and + must be no larger than the existing size.*/ +void ec_enc_shrink(ec_enc *_this,opus_uint32 _size); + +/*Indicates that there are no more symbols to encode. + All reamining output bytes are flushed to the output buffer. + ec_enc_init() must be called before the encoder can be used again.*/ +void ec_enc_done(ec_enc *_this); + +#endif diff --git a/lib/opus/celt/fixed_c5x.h b/lib/opus/celt/fixed_c5x.h new file mode 100644 index 00000000..ea95a998 --- /dev/null +++ b/lib/opus/celt/fixed_c5x.h @@ -0,0 +1,79 @@ +/* Copyright (C) 2003 Jean-Marc Valin */ +/** + @file fixed_c5x.h + @brief Fixed-point operations for the TI C5x DSP family +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_C5X_H +#define FIXED_C5X_H + +#include "dsplib.h" + +#undef IMUL32 +static OPUS_INLINE long IMUL32(long i, long j) +{ + long ac0, ac1; + ac0 = _lmpy(i>>16,j); + ac1 = ac0 + _lmpy(i,j>>16); + return _lmpyu(i,j) + (ac1<<16); +} + +#undef MAX16 +#define MAX16(a,b) _max(a,b) + +#undef MIN16 +#define MIN16(a,b) _min(a,b) + +#undef MAX32 +#define MAX32(a,b) _lmax(a,b) + +#undef MIN32 +#define MIN32(a,b) _lmin(a,b) + +#undef VSHR32 +#define VSHR32(a, shift) _lshl(a,-(shift)) + +#undef MULT16_16_Q15 +#define MULT16_16_Q15(a,b) (_smpy(a,b)) + +#undef MULT16_16SU +#define MULT16_16SU(a,b) _lmpysu(a,b) + +#undef MULT_16_16 +#define MULT_16_16(a,b) _lmpy(a,b) + +/* FIXME: This is technically incorrect and is bound to cause problems. Is there any cleaner solution? */ +#undef MULT16_32_Q15 +#define MULT16_32_Q15(a,b) ADD32(SHL(MULT16_16((a),SHR((b),16)),1), SHR(MULT16_16SU((a),(b)),15)) + +#define celt_ilog2(x) (30 - _lnorm(x)) +#define OVERRIDE_CELT_ILOG2 + +#define celt_maxabs16(x, len) MAX32(EXTEND32(maxval((DATA *)x, len)),-EXTEND32(minval((DATA *)x, len))) +#define OVERRIDE_CELT_MAXABS16 + +#endif /* FIXED_C5X_H */ diff --git a/lib/opus/celt/fixed_c6x.h b/lib/opus/celt/fixed_c6x.h new file mode 100644 index 00000000..bb6ad927 --- /dev/null +++ b/lib/opus/celt/fixed_c6x.h @@ -0,0 +1,70 @@ +/* Copyright (C) 2008 CSIRO */ +/** + @file fixed_c6x.h + @brief Fixed-point operations for the TI C6x DSP family +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_C6X_H +#define FIXED_C6X_H + +#undef MULT16_16SU +#define MULT16_16SU(a,b) _mpysu(a,b) + +#undef MULT_16_16 +#define MULT_16_16(a,b) _mpy(a,b) + +#define celt_ilog2(x) (30 - _norm(x)) +#define OVERRIDE_CELT_ILOG2 + +#undef MULT16_32_Q15 +#define MULT16_32_Q15(a,b) (_mpylill(a, b) >> 15) + +#if 0 +#include "dsplib.h" + +#undef MAX16 +#define MAX16(a,b) _max(a,b) + +#undef MIN16 +#define MIN16(a,b) _min(a,b) + +#undef MAX32 +#define MAX32(a,b) _lmax(a,b) + +#undef MIN32 +#define MIN32(a,b) _lmin(a,b) + +#undef VSHR32 +#define VSHR32(a, shift) _lshl(a,-(shift)) + +#undef MULT16_16_Q15 +#define MULT16_16_Q15(a,b) (_smpy(a,b)) + +#define celt_maxabs16(x, len) MAX32(EXTEND32(maxval((DATA *)x, len)),-EXTEND32(minval((DATA *)x, len))) +#define OVERRIDE_CELT_MAXABS16 + +#endif /* FIXED_C6X_H */ diff --git a/lib/opus/celt/fixed_debug.h b/lib/opus/celt/fixed_debug.h new file mode 100644 index 00000000..ef2e5d02 --- /dev/null +++ b/lib/opus/celt/fixed_debug.h @@ -0,0 +1,836 @@ +/* Copyright (C) 2003-2008 Jean-Marc Valin + Copyright (C) 2007-2012 Xiph.Org Foundation */ +/** + @file fixed_debug.h + @brief Fixed-point operations with debugging +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_DEBUG_H +#define FIXED_DEBUG_H + +#include +#include "opus_defines.h" + +#ifdef CELT_C +OPUS_EXPORT opus_int64 celt_mips=0; +#else +extern opus_int64 celt_mips; +#endif + +#define MULT16_16SU(a,b) ((opus_val32)(opus_val16)(a)*(opus_val32)(opus_uint16)(b)) +#define MULT32_32_Q31(a,b) ADD32(ADD32(SHL32(MULT16_16(SHR32((a),16),SHR((b),16)),1), SHR32(MULT16_16SU(SHR32((a),16),((b)&0x0000ffff)),15)), SHR32(MULT16_16SU(SHR32((b),16),((a)&0x0000ffff)),15)) + +/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ +#define MULT16_32_Q16(a,b) ADD32(MULT16_16((a),SHR32((b),16)), SHR32(MULT16_16SU((a),((b)&0x0000ffff)),16)) + +#define MULT16_32_P16(a,b) MULT16_32_PX(a,b,16) + +#define QCONST16(x,bits) ((opus_val16)(.5+(x)*(((opus_val32)1)<<(bits)))) +#define QCONST32(x,bits) ((opus_val32)(.5+(x)*(((opus_val32)1)<<(bits)))) + +#define VERIFY_SHORT(x) ((x)<=32767&&(x)>=-32768) +#define VERIFY_INT(x) ((x)<=2147483647LL&&(x)>=-2147483648LL) +#define VERIFY_UINT(x) ((x)<=(2147483647LLU<<1)) + +#define SHR(a,b) SHR32(a,b) +#define PSHR(a,b) PSHR32(a,b) + +/** Add two 32-bit values, ignore any overflows */ +#define ADD32_ovflw(a,b) (celt_mips+=2,(opus_val32)((opus_uint32)(a)+(opus_uint32)(b))) +/** Subtract two 32-bit values, ignore any overflows */ +#define SUB32_ovflw(a,b) (celt_mips+=2,(opus_val32)((opus_uint32)(a)-(opus_uint32)(b))) +/* Avoid MSVC warning C4146: unary minus operator applied to unsigned type */ +/** Negate 32-bit value, ignore any overflows */ +#define NEG32_ovflw(a) (celt_mips+=2,(opus_val32)(0-(opus_uint32)(a))) + +static OPUS_INLINE short NEG16(int x) +{ + int res; + if (!VERIFY_SHORT(x)) + { + fprintf (stderr, "NEG16: input is not short: %d\n", (int)x); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = -x; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "NEG16: output is not short: %d\n", (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips++; + return res; +} +static OPUS_INLINE int NEG32(opus_int64 x) +{ + opus_int64 res; + if (!VERIFY_INT(x)) + { + fprintf (stderr, "NEG16: input is not int: %d\n", (int)x); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = -x; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "NEG16: output is not int: %d\n", (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=2; + return res; +} + +#define EXTRACT16(x) EXTRACT16_(x, __FILE__, __LINE__) +static OPUS_INLINE short EXTRACT16_(int x, char *file, int line) +{ + int res; + if (!VERIFY_SHORT(x)) + { + fprintf (stderr, "EXTRACT16: input is not short: %d in %s: line %d\n", x, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = x; + celt_mips++; + return res; +} + +#define EXTEND32(x) EXTEND32_(x, __FILE__, __LINE__) +static OPUS_INLINE int EXTEND32_(int x, char *file, int line) +{ + int res; + if (!VERIFY_SHORT(x)) + { + fprintf (stderr, "EXTEND32: input is not short: %d in %s: line %d\n", x, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = x; + celt_mips++; + return res; +} + +#define SHR16(a, shift) SHR16_(a, shift, __FILE__, __LINE__) +static OPUS_INLINE short SHR16_(int a, int shift, char *file, int line) +{ + int res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(shift)) + { + fprintf (stderr, "SHR16: inputs are not short: %d >> %d in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a>>shift; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "SHR16: output is not short: %d in %s: line %d\n", res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips++; + return res; +} +#define SHL16(a, shift) SHL16_(a, shift, __FILE__, __LINE__) +static OPUS_INLINE short SHL16_(int a, int shift, char *file, int line) +{ + opus_int32 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(shift)) + { + fprintf (stderr, "SHL16: inputs are not short: %d %d in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = (opus_int32)((opus_uint32)a<>shift; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "SHR32: output is not int: %d\n", (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=2; + return res; +} +#define SHL32(a, shift) SHL32_(a, shift, __FILE__, __LINE__) +static OPUS_INLINE int SHL32_(opus_int64 a, int shift, char *file, int line) +{ + opus_int64 res; + if (!VERIFY_INT(a) || !VERIFY_SHORT(shift)) + { + fprintf (stderr, "SHL32: inputs are not int: %lld %d in %s: line %d\n", (long long)a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = (opus_int64)((opus_uint64)a<>1))),shift)) +#define VSHR32(a, shift) (((shift)>0) ? SHR32(a, shift) : SHL32(a, -(shift))) + +#define ROUND16(x,a) (celt_mips--,EXTRACT16(PSHR32((x),(a)))) +#define SROUND16(x,a) (celt_mips--,EXTRACT16(SATURATE(PSHR32(x,a), 32767))); + +#define HALF16(x) (SHR16(x,1)) +#define HALF32(x) (SHR32(x,1)) + +#define ADD16(a, b) ADD16_(a, b, __FILE__, __LINE__) +static OPUS_INLINE short ADD16_(int a, int b, char *file, int line) +{ + int res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "ADD16: inputs are not short: %d %d in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a+b; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "ADD16: output is not short: %d+%d=%d in %s: line %d\n", a,b,res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips++; + return res; +} + +#define SUB16(a, b) SUB16_(a, b, __FILE__, __LINE__) +static OPUS_INLINE short SUB16_(int a, int b, char *file, int line) +{ + int res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "SUB16: inputs are not short: %d %d in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a-b; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "SUB16: output is not short: %d in %s: line %d\n", res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips++; + return res; +} + +#define ADD32(a, b) ADD32_(a, b, __FILE__, __LINE__) +static OPUS_INLINE int ADD32_(opus_int64 a, opus_int64 b, char *file, int line) +{ + opus_int64 res; + if (!VERIFY_INT(a) || !VERIFY_INT(b)) + { + fprintf (stderr, "ADD32: inputs are not int: %d %d in %s: line %d\n", (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a+b; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "ADD32: output is not int: %d in %s: line %d\n", (int)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=2; + return res; +} + +#define SUB32(a, b) SUB32_(a, b, __FILE__, __LINE__) +static OPUS_INLINE int SUB32_(opus_int64 a, opus_int64 b, char *file, int line) +{ + opus_int64 res; + if (!VERIFY_INT(a) || !VERIFY_INT(b)) + { + fprintf (stderr, "SUB32: inputs are not int: %d %d in %s: line %d\n", (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a-b; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "SUB32: output is not int: %d in %s: line %d\n", (int)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=2; + return res; +} + +#undef UADD32 +#define UADD32(a, b) UADD32_(a, b, __FILE__, __LINE__) +static OPUS_INLINE unsigned int UADD32_(opus_uint64 a, opus_uint64 b, char *file, int line) +{ + opus_uint64 res; + if (!VERIFY_UINT(a) || !VERIFY_UINT(b)) + { + fprintf (stderr, "UADD32: inputs are not uint32: %llu %llu in %s: line %d\n", (unsigned long long)a, (unsigned long long)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a+b; + if (!VERIFY_UINT(res)) + { + fprintf (stderr, "UADD32: output is not uint32: %llu in %s: line %d\n", (unsigned long long)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=2; + return res; +} + +#undef USUB32 +#define USUB32(a, b) USUB32_(a, b, __FILE__, __LINE__) +static OPUS_INLINE unsigned int USUB32_(opus_uint64 a, opus_uint64 b, char *file, int line) +{ + opus_uint64 res; + if (!VERIFY_UINT(a) || !VERIFY_UINT(b)) + { + fprintf (stderr, "USUB32: inputs are not uint32: %llu %llu in %s: line %d\n", (unsigned long long)a, (unsigned long long)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + if (a> 16; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT32_32_Q16: output is not int: %lld*%lld=%lld\n", (long long)a, (long long)b, (long long)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=5; + return res; +} + +#define MULT16_16(a, b) MULT16_16_(a, b, __FILE__, __LINE__) +static OPUS_INLINE int MULT16_16_(int a, int b, char *file, int line) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16: inputs are not short: %d %d in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT16_16: output is not int: %d in %s: line %d\n", (int)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips++; + return res; +} + +#define MAC16_16(c,a,b) (celt_mips-=2,ADD32((c),MULT16_16((a),(b)))) + +#define MULT16_32_QX(a, b, Q) MULT16_32_QX_(a, b, Q, __FILE__, __LINE__) +static OPUS_INLINE int MULT16_32_QX_(int a, opus_int64 b, int Q, char *file, int line) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_INT(b)) + { + fprintf (stderr, "MULT16_32_Q%d: inputs are not short+int: %d %d in %s: line %d\n", Q, (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + if (ABS32(b)>=((opus_int64)(1)<<(16+Q))) + { + fprintf (stderr, "MULT16_32_Q%d: second operand too large: %d %d in %s: line %d\n", Q, (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = (((opus_int64)a)*(opus_int64)b) >> Q; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT16_32_Q%d: output is not int: %d*%d=%d in %s: line %d\n", Q, (int)a, (int)b,(int)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + if (Q==15) + celt_mips+=3; + else + celt_mips+=4; + return res; +} + +#define MULT16_32_PX(a, b, Q) MULT16_32_PX_(a, b, Q, __FILE__, __LINE__) +static OPUS_INLINE int MULT16_32_PX_(int a, opus_int64 b, int Q, char *file, int line) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_INT(b)) + { + fprintf (stderr, "MULT16_32_P%d: inputs are not short+int: %d %d in %s: line %d\n\n", Q, (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + if (ABS32(b)>=((opus_int64)(1)<<(16+Q))) + { + fprintf (stderr, "MULT16_32_Q%d: second operand too large: %d %d in %s: line %d\n\n", Q, (int)a, (int)b,file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((((opus_int64)a)*(opus_int64)b) + (((opus_val32)(1)<>1))>> Q; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT16_32_P%d: output is not int: %d*%d=%d in %s: line %d\n\n", Q, (int)a, (int)b,(int)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + if (Q==15) + celt_mips+=4; + else + celt_mips+=5; + return res; +} + +#define MULT16_32_Q15(a,b) MULT16_32_QX(a,b,15) +#define MAC16_32_Q15(c,a,b) (celt_mips-=2,ADD32((c),MULT16_32_Q15((a),(b)))) +#define MAC16_32_Q16(c,a,b) (celt_mips-=2,ADD32((c),MULT16_32_Q16((a),(b)))) + +static OPUS_INLINE int SATURATE(int a, int b) +{ + if (a>b) + a=b; + if (a<-b) + a = -b; + celt_mips+=3; + return a; +} + +static OPUS_INLINE opus_int16 SATURATE16(opus_int32 a) +{ + celt_mips+=3; + if (a>32767) + return 32767; + else if (a<-32768) + return -32768; + else return a; +} + +static OPUS_INLINE int MULT16_16_Q11_32(int a, int b) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16_Q11: inputs are not short: %d %d\n", a, b); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + res >>= 11; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT16_16_Q11: output is not short: %d*%d=%d\n", (int)a, (int)b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=3; + return res; +} +static OPUS_INLINE short MULT16_16_Q13(int a, int b) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16_Q13: inputs are not short: %d %d\n", a, b); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + res >>= 13; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "MULT16_16_Q13: output is not short: %d*%d=%d\n", a, b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=3; + return res; +} +static OPUS_INLINE short MULT16_16_Q14(int a, int b) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16_Q14: inputs are not short: %d %d\n", a, b); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + res >>= 14; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "MULT16_16_Q14: output is not short: %d\n", (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=3; + return res; +} + +#define MULT16_16_Q15(a, b) MULT16_16_Q15_(a, b, __FILE__, __LINE__) +static OPUS_INLINE short MULT16_16_Q15_(int a, int b, char *file, int line) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16_Q15: inputs are not short: %d %d in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + res >>= 15; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "MULT16_16_Q15: output is not short: %d in %s: line %d\n", (int)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=1; + return res; +} + +static OPUS_INLINE short MULT16_16_P13(int a, int b) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16_P13: inputs are not short: %d %d\n", a, b); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + res += 4096; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT16_16_P13: overflow: %d*%d=%d\n", a, b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res >>= 13; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "MULT16_16_P13: output is not short: %d*%d=%d\n", a, b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=4; + return res; +} +static OPUS_INLINE short MULT16_16_P14(int a, int b) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16_P14: inputs are not short: %d %d\n", a, b); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + res += 8192; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT16_16_P14: overflow: %d*%d=%d\n", a, b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res >>= 14; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "MULT16_16_P14: output is not short: %d*%d=%d\n", a, b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=4; + return res; +} +static OPUS_INLINE short MULT16_16_P15(int a, int b) +{ + opus_int64 res; + if (!VERIFY_SHORT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "MULT16_16_P15: inputs are not short: %d %d\n", a, b); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = ((opus_int64)a)*b; + res += 16384; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "MULT16_16_P15: overflow: %d*%d=%d\n", a, b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res >>= 15; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "MULT16_16_P15: output is not short: %d*%d=%d\n", a, b, (int)res); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=2; + return res; +} + +#define DIV32_16(a, b) DIV32_16_(a, b, __FILE__, __LINE__) + +static OPUS_INLINE int DIV32_16_(opus_int64 a, opus_int64 b, char *file, int line) +{ + opus_int64 res; + if (b==0) + { + fprintf(stderr, "DIV32_16: divide by zero: %d/%d in %s: line %d\n", (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + return 0; + } + if (!VERIFY_INT(a) || !VERIFY_SHORT(b)) + { + fprintf (stderr, "DIV32_16: inputs are not int/short: %d %d in %s: line %d\n", (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a/b; + if (!VERIFY_SHORT(res)) + { + fprintf (stderr, "DIV32_16: output is not short: %d / %d = %d in %s: line %d\n", (int)a,(int)b,(int)res, file, line); + if (res>32767) + res = 32767; + if (res<-32768) + res = -32768; +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=35; + return res; +} + +#define DIV32(a, b) DIV32_(a, b, __FILE__, __LINE__) +static OPUS_INLINE int DIV32_(opus_int64 a, opus_int64 b, char *file, int line) +{ + opus_int64 res; + if (b==0) + { + fprintf(stderr, "DIV32: divide by zero: %d/%d in %s: line %d\n", (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + return 0; + } + + if (!VERIFY_INT(a) || !VERIFY_INT(b)) + { + fprintf (stderr, "DIV32: inputs are not int/short: %d %d in %s: line %d\n", (int)a, (int)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + res = a/b; + if (!VERIFY_INT(res)) + { + fprintf (stderr, "DIV32: output is not int: %d in %s: line %d\n", (int)res, file, line); +#ifdef FIXED_DEBUG_ASSERT + celt_assert(0); +#endif + } + celt_mips+=70; + return res; +} + +static OPUS_INLINE opus_val16 SIG2WORD16_generic(celt_sig x) +{ + x = PSHR32(x, SIG_SHIFT); + x = MAX32(x, -32768); + x = MIN32(x, 32767); + return EXTRACT16(x); +} +#define SIG2WORD16(x) (SIG2WORD16_generic(x)) + + +#undef PRINT_MIPS +#define PRINT_MIPS(file) do {fprintf (file, "total complexity = %llu MIPS\n", (unsigned long long)celt_mips);} while (0); + +#endif diff --git a/lib/opus/celt/fixed_generic.h b/lib/opus/celt/fixed_generic.h new file mode 100644 index 00000000..8f29d46b --- /dev/null +++ b/lib/opus/celt/fixed_generic.h @@ -0,0 +1,188 @@ +/* Copyright (C) 2007-2009 Xiph.Org Foundation + Copyright (C) 2003-2008 Jean-Marc Valin + Copyright (C) 2007-2008 CSIRO */ +/** + @file fixed_generic.h + @brief Generic fixed-point operations +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef FIXED_GENERIC_H +#define FIXED_GENERIC_H + +/** Multiply a 16-bit signed value by a 16-bit unsigned value. The result is a 32-bit signed value */ +#define MULT16_16SU(a,b) ((opus_val32)(opus_val16)(a)*(opus_val32)(opus_uint16)(b)) + +/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ +#if OPUS_FAST_INT64 +#define MULT16_32_Q16(a,b) ((opus_val32)SHR((opus_int64)((opus_val16)(a))*(b),16)) +#else +#define MULT16_32_Q16(a,b) ADD32(MULT16_16((a),SHR((b),16)), SHR(MULT16_16SU((a),((b)&0x0000ffff)),16)) +#endif + +/** 16x32 multiplication, followed by a 16-bit shift right (round-to-nearest). Results fits in 32 bits */ +#if OPUS_FAST_INT64 +#define MULT16_32_P16(a,b) ((opus_val32)PSHR((opus_int64)((opus_val16)(a))*(b),16)) +#else +#define MULT16_32_P16(a,b) ADD32(MULT16_16((a),SHR((b),16)), PSHR(MULT16_16SU((a),((b)&0x0000ffff)),16)) +#endif + +/** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */ +#if OPUS_FAST_INT64 +#define MULT16_32_Q15(a,b) ((opus_val32)SHR((opus_int64)((opus_val16)(a))*(b),15)) +#else +#define MULT16_32_Q15(a,b) ADD32(SHL(MULT16_16((a),SHR((b),16)),1), SHR(MULT16_16SU((a),((b)&0x0000ffff)),15)) +#endif + +/** 32x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ +#if OPUS_FAST_INT64 +#define MULT32_32_Q16(a,b) ((opus_val32)SHR((opus_int64)(a)*(opus_int64)(b),16)) +#else +#define MULT32_32_Q16(a,b) (ADD32(ADD32(ADD32((opus_val32)(SHR32(((opus_uint32)((a)&0x0000ffff)*(opus_uint32)((b)&0x0000ffff)),16)), MULT16_16SU(SHR32(a,16),((b)&0x0000ffff))), MULT16_16SU(SHR32(b,16),((a)&0x0000ffff))), SHL32(MULT16_16(SHR32(a,16),SHR32(b,16)),16))) +#endif + +/** 32x32 multiplication, followed by a 31-bit shift right. Results fits in 32 bits */ +#if OPUS_FAST_INT64 +#define MULT32_32_Q31(a,b) ((opus_val32)SHR((opus_int64)(a)*(opus_int64)(b),31)) +#else +#define MULT32_32_Q31(a,b) ADD32(ADD32(SHL(MULT16_16(SHR((a),16),SHR((b),16)),1), SHR(MULT16_16SU(SHR((a),16),((b)&0x0000ffff)),15)), SHR(MULT16_16SU(SHR((b),16),((a)&0x0000ffff)),15)) +#endif + +/** Compile-time conversion of float constant to 16-bit value */ +#define QCONST16(x,bits) ((opus_val16)(.5+(x)*(((opus_val32)1)<<(bits)))) + +/** Compile-time conversion of float constant to 32-bit value */ +#define QCONST32(x,bits) ((opus_val32)(.5+(x)*(((opus_val32)1)<<(bits)))) + +/** Negate a 16-bit value */ +#define NEG16(x) (-(x)) +/** Negate a 32-bit value */ +#define NEG32(x) (-(x)) + +/** Change a 32-bit value into a 16-bit value. The value is assumed to fit in 16-bit, otherwise the result is undefined */ +#define EXTRACT16(x) ((opus_val16)(x)) +/** Change a 16-bit value into a 32-bit value */ +#define EXTEND32(x) ((opus_val32)(x)) + +/** Arithmetic shift-right of a 16-bit value */ +#define SHR16(a,shift) ((a) >> (shift)) +/** Arithmetic shift-left of a 16-bit value */ +#define SHL16(a,shift) ((opus_int16)((opus_uint16)(a)<<(shift))) +/** Arithmetic shift-right of a 32-bit value */ +#define SHR32(a,shift) ((a) >> (shift)) +/** Arithmetic shift-left of a 32-bit value */ +#define SHL32(a,shift) ((opus_int32)((opus_uint32)(a)<<(shift))) + +/** 32-bit arithmetic shift right with rounding-to-nearest instead of rounding down */ +#define PSHR32(a,shift) (SHR32((a)+((EXTEND32(1)<<((shift))>>1)),shift)) +/** 32-bit arithmetic shift right where the argument can be negative */ +#define VSHR32(a, shift) (((shift)>0) ? SHR32(a, shift) : SHL32(a, -(shift))) + +/** "RAW" macros, should not be used outside of this header file */ +#define SHR(a,shift) ((a) >> (shift)) +#define SHL(a,shift) SHL32(a,shift) +#define PSHR(a,shift) (SHR((a)+((EXTEND32(1)<<((shift))>>1)),shift)) +#define SATURATE(x,a) (((x)>(a) ? (a) : (x)<-(a) ? -(a) : (x))) + +#define SATURATE16(x) (EXTRACT16((x)>32767 ? 32767 : (x)<-32768 ? -32768 : (x))) + +/** Shift by a and round-to-nearest 32-bit value. Result is a 16-bit value */ +#define ROUND16(x,a) (EXTRACT16(PSHR32((x),(a)))) +/** Shift by a and round-to-nearest 32-bit value. Result is a saturated 16-bit value */ +#define SROUND16(x,a) EXTRACT16(SATURATE(PSHR32(x,a), 32767)); + +/** Divide by two */ +#define HALF16(x) (SHR16(x,1)) +#define HALF32(x) (SHR32(x,1)) + +/** Add two 16-bit values */ +#define ADD16(a,b) ((opus_val16)((opus_val16)(a)+(opus_val16)(b))) +/** Subtract two 16-bit values */ +#define SUB16(a,b) ((opus_val16)(a)-(opus_val16)(b)) +/** Add two 32-bit values */ +#define ADD32(a,b) ((opus_val32)(a)+(opus_val32)(b)) +/** Subtract two 32-bit values */ +#define SUB32(a,b) ((opus_val32)(a)-(opus_val32)(b)) + +/** Add two 32-bit values, ignore any overflows */ +#define ADD32_ovflw(a,b) ((opus_val32)((opus_uint32)(a)+(opus_uint32)(b))) +/** Subtract two 32-bit values, ignore any overflows */ +#define SUB32_ovflw(a,b) ((opus_val32)((opus_uint32)(a)-(opus_uint32)(b))) +/* Avoid MSVC warning C4146: unary minus operator applied to unsigned type */ +/** Negate 32-bit value, ignore any overflows */ +#define NEG32_ovflw(a) ((opus_val32)(0-(opus_uint32)(a))) + +/** 16x16 multiplication where the result fits in 16 bits */ +#define MULT16_16_16(a,b) ((((opus_val16)(a))*((opus_val16)(b)))) + +/** 32x32 multiplication where the result fits in 32 bits */ +#define MULT32_32_32(a,b) ((((opus_val32)(a))*((opus_val32)(b)))) + +/* (opus_val32)(opus_val16) gives TI compiler a hint that it's 16x16->32 multiply */ +/** 16x16 multiplication where the result fits in 32 bits */ +#define MULT16_16(a,b) (((opus_val32)(opus_val16)(a))*((opus_val32)(opus_val16)(b))) + +/** 16x16 multiply-add where the result fits in 32 bits */ +#define MAC16_16(c,a,b) (ADD32((c),MULT16_16((a),(b)))) +/** 16x32 multiply, followed by a 15-bit shift right and 32-bit add. + b must fit in 31 bits. + Result fits in 32 bits. */ +#define MAC16_32_Q15(c,a,b) ADD32((c),ADD32(MULT16_16((a),SHR((b),15)), SHR(MULT16_16((a),((b)&0x00007fff)),15))) + +/** 16x32 multiplication, followed by a 16-bit shift right and 32-bit add. + Results fits in 32 bits */ +#define MAC16_32_Q16(c,a,b) ADD32((c),ADD32(MULT16_16((a),SHR((b),16)), SHR(MULT16_16SU((a),((b)&0x0000ffff)),16))) + +#define MULT16_16_Q11_32(a,b) (SHR(MULT16_16((a),(b)),11)) +#define MULT16_16_Q11(a,b) (SHR(MULT16_16((a),(b)),11)) +#define MULT16_16_Q13(a,b) (SHR(MULT16_16((a),(b)),13)) +#define MULT16_16_Q14(a,b) (SHR(MULT16_16((a),(b)),14)) +#define MULT16_16_Q15(a,b) (SHR(MULT16_16((a),(b)),15)) + +#define MULT16_16_P13(a,b) (SHR(ADD32(4096,MULT16_16((a),(b))),13)) +#define MULT16_16_P14(a,b) (SHR(ADD32(8192,MULT16_16((a),(b))),14)) +#define MULT16_16_P15(a,b) (SHR(ADD32(16384,MULT16_16((a),(b))),15)) + +/** Divide a 32-bit value by a 16-bit value. Result fits in 16 bits */ +#define DIV32_16(a,b) ((opus_val16)(((opus_val32)(a))/((opus_val16)(b)))) + +/** Divide a 32-bit value by a 32-bit value. Result fits in 32 bits */ +#define DIV32(a,b) (((opus_val32)(a))/((opus_val32)(b))) + +#if defined(MIPSr1_ASM) +#include "mips/fixed_generic_mipsr1.h" +#endif + +static OPUS_INLINE opus_val16 SIG2WORD16_generic(celt_sig x) +{ + x = PSHR32(x, SIG_SHIFT); + x = MAX32(x, -32768); + x = MIN32(x, 32767); + return EXTRACT16(x); +} +#define SIG2WORD16(x) (SIG2WORD16_generic(x)) + +#endif diff --git a/lib/opus/celt/float_cast.h b/lib/opus/celt/float_cast.h new file mode 100644 index 00000000..8915a5fd --- /dev/null +++ b/lib/opus/celt/float_cast.h @@ -0,0 +1,152 @@ +/* Copyright (C) 2001 Erik de Castro Lopo */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* Version 1.1 */ + +#ifndef FLOAT_CAST_H +#define FLOAT_CAST_H + + +#include "arch.h" + +/*============================================================================ +** On Intel Pentium processors (especially PIII and probably P4), converting +** from float to int is very slow. To meet the C specs, the code produced by +** most C compilers targeting Pentium needs to change the FPU rounding mode +** before the float to int conversion is performed. +** +** Changing the FPU rounding mode causes the FPU pipeline to be flushed. It +** is this flushing of the pipeline which is so slow. +** +** Fortunately the ISO C99 specifications define the functions lrint, lrintf, +** llrint and llrintf which fix this problem as a side effect. +** +** On Unix-like systems, the configure process should have detected the +** presence of these functions. If they weren't found we have to replace them +** here with a standard C cast. +*/ + +/* +** The C99 prototypes for lrint and lrintf are as follows: +** +** long int lrintf (float x) ; +** long int lrint (double x) ; +*/ + +/* The presence of the required functions are detected during the configure +** process and the values HAVE_LRINT and HAVE_LRINTF are set accordingly in +** the config.h file. +*/ + +/* With GCC, when SSE is available, the fastest conversion is cvtss2si. */ +#if defined(__GNUC__) && defined(__SSE__) + +#include +static OPUS_INLINE opus_int32 float2int(float x) {return _mm_cvt_ss2si(_mm_set_ss(x));} + +#elif (defined(_MSC_VER) && _MSC_VER >= 1400) && (defined(_M_X64) || (defined(_M_IX86_FP) && _M_IX86_FP >= 1)) + + #include + static OPUS_INLINE opus_int32 float2int(float value) + { + /* _mm_load_ss will generate same code as _mm_set_ss + ** in _MSC_VER >= 1914 /02 so keep __mm_load__ss + ** for backward compatibility. + */ + return _mm_cvtss_si32(_mm_load_ss(&value)); + } + +#elif (defined(_MSC_VER) && _MSC_VER >= 1400) && defined (_M_IX86) + + #include + + /* Win32 doesn't seem to have these functions. + ** Therefore implement OPUS_INLINE versions of these functions here. + */ + + static OPUS_INLINE opus_int32 + float2int (float flt) + { int intgr; + + _asm + { fld flt + fistp intgr + } ; + + return intgr ; + } + +#elif defined(HAVE_LRINTF) && defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L + +/* These defines enable functionality introduced with the 1999 ISO C +** standard. They must be defined before the inclusion of math.h to +** engage them. If optimisation is enabled, these functions will be +** inlined. With optimisation switched off, you have to link in the +** maths library using -lm. +*/ + +#define _ISOC9X_SOURCE 1 +#define _ISOC99_SOURCE 1 + +#define __USE_ISOC9X 1 +#define __USE_ISOC99 1 + +#include +#define float2int(x) lrintf(x) + +#elif defined(HAVE_LRINT) && defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L + +#define _ISOC9X_SOURCE 1 +#define _ISOC99_SOURCE 1 + +#define __USE_ISOC9X 1 +#define __USE_ISOC99 1 + +#include +#define float2int(x) lrint(x) + +#else + +#if (defined(__GNUC__) && defined(__STDC__) && __STDC__ && __STDC_VERSION__ >= 199901L) + /* supported by gcc in C99 mode, but not by all other compilers */ + #warning "Don't have the functions lrint() and lrintf ()." + #warning "Replacing these functions with a standard C cast." +#endif /* __STDC_VERSION__ >= 199901L */ + #include + #define float2int(flt) ((int)(floor(.5+flt))) +#endif + +#ifndef DISABLE_FLOAT_API +static OPUS_INLINE opus_int16 FLOAT2INT16(float x) +{ + x = x*CELT_SIG_SCALE; + x = MAX32(x, -32768); + x = MIN32(x, 32767); + return (opus_int16)float2int(x); +} +#endif /* DISABLE_FLOAT_API */ + +#endif /* FLOAT_CAST_H */ diff --git a/lib/opus/celt/kiss_fft.c b/lib/opus/celt/kiss_fft.c new file mode 100644 index 00000000..83775165 --- /dev/null +++ b/lib/opus/celt/kiss_fft.c @@ -0,0 +1,604 @@ +/*Copyright (c) 2003-2004, Mark Borgerding + Lots of modifications by Jean-Marc Valin + Copyright (c) 2005-2007, Xiph.Org Foundation + Copyright (c) 2008, Xiph.Org Foundation, CSIRO + + All rights reserved. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE.*/ + +/* This code is originally from Mark Borgerding's KISS-FFT but has been + heavily modified to better suit Opus */ + +#ifndef SKIP_CONFIG_H +# ifdef HAVE_CONFIG_H +# include "config.h" +# endif +#endif + +#include "_kiss_fft_guts.h" +#include "arch.h" +#include "os_support.h" +#include "mathops.h" +#include "stack_alloc.h" + +/* The guts header contains all the multiplication and addition macros that are defined for + complex numbers. It also delares the kf_ internal functions. +*/ + +static void kf_bfly2( + kiss_fft_cpx * Fout, + int m, + int N + ) +{ + kiss_fft_cpx * Fout2; + int i; + (void)m; +#ifdef CUSTOM_MODES + if (m==1) + { + celt_assert(m==1); + for (i=0;itwiddles; + /* m is guaranteed to be a multiple of 4. */ + for (j=0;jtwiddles[fstride*m]; +#endif + for (i=0;itwiddles; + /* For non-custom modes, m is guaranteed to be a multiple of 4. */ + k=m; + do { + + C_MUL(scratch[1],Fout[m] , *tw1); + C_MUL(scratch[2],Fout[m2] , *tw2); + + C_ADD(scratch[3],scratch[1],scratch[2]); + C_SUB(scratch[0],scratch[1],scratch[2]); + tw1 += fstride; + tw2 += fstride*2; + + Fout[m].r = SUB32_ovflw(Fout->r, HALF_OF(scratch[3].r)); + Fout[m].i = SUB32_ovflw(Fout->i, HALF_OF(scratch[3].i)); + + C_MULBYSCALAR( scratch[0] , epi3.i ); + + C_ADDTO(*Fout,scratch[3]); + + Fout[m2].r = ADD32_ovflw(Fout[m].r, scratch[0].i); + Fout[m2].i = SUB32_ovflw(Fout[m].i, scratch[0].r); + + Fout[m].r = SUB32_ovflw(Fout[m].r, scratch[0].i); + Fout[m].i = ADD32_ovflw(Fout[m].i, scratch[0].r); + + ++Fout; + } while(--k); + } +} + + +#ifndef OVERRIDE_kf_bfly5 +static void kf_bfly5( + kiss_fft_cpx * Fout, + const size_t fstride, + const kiss_fft_state *st, + int m, + int N, + int mm + ) +{ + kiss_fft_cpx *Fout0,*Fout1,*Fout2,*Fout3,*Fout4; + int i, u; + kiss_fft_cpx scratch[13]; + const kiss_twiddle_cpx *tw; + kiss_twiddle_cpx ya,yb; + kiss_fft_cpx * Fout_beg = Fout; + +#ifdef FIXED_POINT + ya.r = 10126; + ya.i = -31164; + yb.r = -26510; + yb.i = -19261; +#else + ya = st->twiddles[fstride*m]; + yb = st->twiddles[fstride*2*m]; +#endif + tw=st->twiddles; + + for (i=0;ir = ADD32_ovflw(Fout0->r, ADD32_ovflw(scratch[7].r, scratch[8].r)); + Fout0->i = ADD32_ovflw(Fout0->i, ADD32_ovflw(scratch[7].i, scratch[8].i)); + + scratch[5].r = ADD32_ovflw(scratch[0].r, ADD32_ovflw(S_MUL(scratch[7].r,ya.r), S_MUL(scratch[8].r,yb.r))); + scratch[5].i = ADD32_ovflw(scratch[0].i, ADD32_ovflw(S_MUL(scratch[7].i,ya.r), S_MUL(scratch[8].i,yb.r))); + + scratch[6].r = ADD32_ovflw(S_MUL(scratch[10].i,ya.i), S_MUL(scratch[9].i,yb.i)); + scratch[6].i = NEG32_ovflw(ADD32_ovflw(S_MUL(scratch[10].r,ya.i), S_MUL(scratch[9].r,yb.i))); + + C_SUB(*Fout1,scratch[5],scratch[6]); + C_ADD(*Fout4,scratch[5],scratch[6]); + + scratch[11].r = ADD32_ovflw(scratch[0].r, ADD32_ovflw(S_MUL(scratch[7].r,yb.r), S_MUL(scratch[8].r,ya.r))); + scratch[11].i = ADD32_ovflw(scratch[0].i, ADD32_ovflw(S_MUL(scratch[7].i,yb.r), S_MUL(scratch[8].i,ya.r))); + scratch[12].r = SUB32_ovflw(S_MUL(scratch[9].i,ya.i), S_MUL(scratch[10].i,yb.i)); + scratch[12].i = SUB32_ovflw(S_MUL(scratch[10].r,yb.i), S_MUL(scratch[9].r,ya.i)); + + C_ADD(*Fout2,scratch[11],scratch[12]); + C_SUB(*Fout3,scratch[11],scratch[12]); + + ++Fout0;++Fout1;++Fout2;++Fout3;++Fout4; + } + } +} +#endif /* OVERRIDE_kf_bfly5 */ + + +#endif + + +#ifdef CUSTOM_MODES + +static +void compute_bitrev_table( + int Fout, + opus_int16 *f, + const size_t fstride, + int in_stride, + opus_int16 * factors, + const kiss_fft_state *st + ) +{ + const int p=*factors++; /* the radix */ + const int m=*factors++; /* stage's fft length/p */ + + /*printf ("fft %d %d %d %d %d %d\n", p*m, m, p, s2, fstride*in_stride, N);*/ + if (m==1) + { + int j; + for (j=0;j32000 || (opus_int32)p*(opus_int32)p > n) + p = n; /* no more factors, skip to end */ + } + n /= p; +#ifdef RADIX_TWO_ONLY + if (p!=2 && p != 4) +#else + if (p>5) +#endif + { + return 0; + } + facbuf[2*stages] = p; + if (p==2 && stages > 1) + { + facbuf[2*stages] = 4; + facbuf[2] = 2; + } + stages++; + } while (n > 1); + n = nbak; + /* Reverse the order to get the radix 4 at the end, so we can use the + fast degenerate case. It turns out that reversing the order also + improves the noise behaviour. */ + for (i=0;i= memneeded) + st = (kiss_fft_state*)mem; + *lenmem = memneeded; + } + if (st) { + opus_int16 *bitrev; + kiss_twiddle_cpx *twiddles; + + st->nfft=nfft; +#ifdef FIXED_POINT + st->scale_shift = celt_ilog2(st->nfft); + if (st->nfft == 1<scale_shift) + st->scale = Q15ONE; + else + st->scale = (1073741824+st->nfft/2)/st->nfft>>(15-st->scale_shift); +#else + st->scale = 1.f/nfft; +#endif + if (base != NULL) + { + st->twiddles = base->twiddles; + st->shift = 0; + while (st->shift < 32 && nfft<shift != base->nfft) + st->shift++; + if (st->shift>=32) + goto fail; + } else { + st->twiddles = twiddles = (kiss_twiddle_cpx*)KISS_FFT_MALLOC(sizeof(kiss_twiddle_cpx)*nfft); + compute_twiddles(twiddles, nfft); + st->shift = -1; + } + if (!kf_factor(nfft,st->factors)) + { + goto fail; + } + + /* bitrev */ + st->bitrev = bitrev = (opus_int16*)KISS_FFT_MALLOC(sizeof(opus_int16)*nfft); + if (st->bitrev==NULL) + goto fail; + compute_bitrev_table(0, bitrev, 1,1, st->factors,st); + + /* Initialize architecture specific fft parameters */ + if (opus_fft_alloc_arch(st, arch)) + goto fail; + } + return st; +fail: + opus_fft_free(st, arch); + return NULL; +} + +kiss_fft_state *opus_fft_alloc(int nfft,void * mem,size_t * lenmem, int arch) +{ + return opus_fft_alloc_twiddles(nfft, mem, lenmem, NULL, arch); +} + +void opus_fft_free_arch_c(kiss_fft_state *st) { + (void)st; +} + +void opus_fft_free(const kiss_fft_state *cfg, int arch) +{ + if (cfg) + { + opus_fft_free_arch((kiss_fft_state *)cfg, arch); + opus_free((opus_int16*)cfg->bitrev); + if (cfg->shift < 0) + opus_free((kiss_twiddle_cpx*)cfg->twiddles); + opus_free((kiss_fft_state*)cfg); + } +} + +#endif /* CUSTOM_MODES */ + +void opus_fft_impl(const kiss_fft_state *st,kiss_fft_cpx *fout) +{ + int m2, m; + int p; + int L; + int fstride[MAXFACTORS]; + int i; + int shift; + + /* st->shift can be -1 */ + shift = st->shift>0 ? st->shift : 0; + + fstride[0] = 1; + L=0; + do { + p = st->factors[2*L]; + m = st->factors[2*L+1]; + fstride[L+1] = fstride[L]*p; + L++; + } while(m!=1); + m = st->factors[2*L-1]; + for (i=L-1;i>=0;i--) + { + if (i!=0) + m2 = st->factors[2*i-1]; + else + m2 = 1; + switch (st->factors[2*i]) + { + case 2: + kf_bfly2(fout, m, fstride[i]); + break; + case 4: + kf_bfly4(fout,fstride[i]<scale_shift-1; +#endif + scale = st->scale; + + celt_assert2 (fin != fout, "In-place FFT not supported"); + /* Bit-reverse the input */ + for (i=0;infft;i++) + { + kiss_fft_cpx x = fin[i]; + fout[st->bitrev[i]].r = SHR32(MULT16_32_Q16(scale, x.r), scale_shift); + fout[st->bitrev[i]].i = SHR32(MULT16_32_Q16(scale, x.i), scale_shift); + } + opus_fft_impl(st, fout); +} + + +void opus_ifft_c(const kiss_fft_state *st,const kiss_fft_cpx *fin,kiss_fft_cpx *fout) +{ + int i; + celt_assert2 (fin != fout, "In-place FFT not supported"); + /* Bit-reverse the input */ + for (i=0;infft;i++) + fout[st->bitrev[i]] = fin[i]; + for (i=0;infft;i++) + fout[i].i = -fout[i].i; + opus_fft_impl(st, fout); + for (i=0;infft;i++) + fout[i].i = -fout[i].i; +} diff --git a/lib/opus/celt/kiss_fft.h b/lib/opus/celt/kiss_fft.h new file mode 100644 index 00000000..267f72f9 --- /dev/null +++ b/lib/opus/celt/kiss_fft.h @@ -0,0 +1,210 @@ +/*Copyright (c) 2003-2004, Mark Borgerding + Lots of modifications by Jean-Marc Valin + Copyright (c) 2005-2007, Xiph.Org Foundation + Copyright (c) 2008, Xiph.Org Foundation, CSIRO + + All rights reserved. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are met: + + * Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + * Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE.*/ + +#ifndef KISS_FFT_H +#define KISS_FFT_H + +#include +#include +#include "arch.h" +#include "cpu_support.h" + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef USE_SIMD +# include +# define kiss_fft_scalar __m128 +#define KISS_FFT_MALLOC(nbytes) memalign(16,nbytes) +#else +#define KISS_FFT_MALLOC opus_alloc +#endif + +#ifdef FIXED_POINT +#include "arch.h" + +# define kiss_fft_scalar opus_int32 +# define kiss_twiddle_scalar opus_int16 + +/* Some 32-bit CPUs would load/store a kiss_twiddle_cpx with a single memory + * access, and could benefit from additional alignment. + */ +# define KISS_TWIDDLE_CPX_ALIGNMENT (sizeof(opus_int32)) + +#else +# ifndef kiss_fft_scalar +/* default is float */ +# define kiss_fft_scalar float +# define kiss_twiddle_scalar float +# define KF_SUFFIX _celt_single +# endif +#endif + +#if defined(__GNUC__) && defined(KISS_TWIDDLE_CPX_ALIGNMENT) +#define KISS_TWIDDLE_CPX_ALIGNED __attribute__((aligned(KISS_TWIDDLE_CPX_ALIGNMENT))) +#else +#define KISS_TWIDDLE_CPX_ALIGNED +#endif + +typedef struct { + kiss_fft_scalar r; + kiss_fft_scalar i; +}kiss_fft_cpx; + +typedef struct { + kiss_twiddle_scalar r; + kiss_twiddle_scalar i; +} KISS_TWIDDLE_CPX_ALIGNED kiss_twiddle_cpx; + +#define MAXFACTORS 8 +/* e.g. an fft of length 128 has 4 factors + as far as kissfft is concerned + 4*4*4*2 + */ + +typedef struct arch_fft_state{ + int is_supported; + void *priv; +} arch_fft_state; + +typedef struct kiss_fft_state{ + int nfft; + opus_val16 scale; +#ifdef FIXED_POINT + int scale_shift; +#endif + int shift; + opus_int16 factors[2*MAXFACTORS]; + const opus_int16 *bitrev; + const kiss_twiddle_cpx *twiddles; + arch_fft_state *arch_fft; +} kiss_fft_state; + +#if defined(HAVE_ARM_NE10) +#include "arm/fft_arm.h" +#endif + +/*typedef struct kiss_fft_state* kiss_fft_cfg;*/ + +/** + * opus_fft_alloc + * + * Initialize a FFT (or IFFT) algorithm's cfg/state buffer. + * + * typical usage: kiss_fft_cfg mycfg=opus_fft_alloc(1024,0,NULL,NULL); + * + * The return value from fft_alloc is a cfg buffer used internally + * by the fft routine or NULL. + * + * If lenmem is NULL, then opus_fft_alloc will allocate a cfg buffer using malloc. + * The returned value should be free()d when done to avoid memory leaks. + * + * The state can be placed in a user supplied buffer 'mem': + * If lenmem is not NULL and mem is not NULL and *lenmem is large enough, + * then the function places the cfg in mem and the size used in *lenmem + * and returns mem. + * + * If lenmem is not NULL and ( mem is NULL or *lenmem is not large enough), + * then the function returns NULL and places the minimum cfg + * buffer size in *lenmem. + * */ + +kiss_fft_state *opus_fft_alloc_twiddles(int nfft,void * mem,size_t * lenmem, const kiss_fft_state *base, int arch); + +kiss_fft_state *opus_fft_alloc(int nfft,void * mem,size_t * lenmem, int arch); + +/** + * opus_fft(cfg,in_out_buf) + * + * Perform an FFT on a complex input buffer. + * for a forward FFT, + * fin should be f[0] , f[1] , ... ,f[nfft-1] + * fout will be F[0] , F[1] , ... ,F[nfft-1] + * Note that each element is complex and can be accessed like + f[k].r and f[k].i + * */ +void opus_fft_c(const kiss_fft_state *cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout); +void opus_ifft_c(const kiss_fft_state *cfg,const kiss_fft_cpx *fin,kiss_fft_cpx *fout); + +void opus_fft_impl(const kiss_fft_state *st,kiss_fft_cpx *fout); +void opus_ifft_impl(const kiss_fft_state *st,kiss_fft_cpx *fout); + +void opus_fft_free(const kiss_fft_state *cfg, int arch); + + +void opus_fft_free_arch_c(kiss_fft_state *st); +int opus_fft_alloc_arch_c(kiss_fft_state *st); + +#if !defined(OVERRIDE_OPUS_FFT) +/* Is run-time CPU detection enabled on this platform? */ +#if defined(OPUS_HAVE_RTCD) && (defined(HAVE_ARM_NE10)) + +extern int (*const OPUS_FFT_ALLOC_ARCH_IMPL[OPUS_ARCHMASK+1])( + kiss_fft_state *st); + +#define opus_fft_alloc_arch(_st, arch) \ + ((*OPUS_FFT_ALLOC_ARCH_IMPL[(arch)&OPUS_ARCHMASK])(_st)) + +extern void (*const OPUS_FFT_FREE_ARCH_IMPL[OPUS_ARCHMASK+1])( + kiss_fft_state *st); +#define opus_fft_free_arch(_st, arch) \ + ((*OPUS_FFT_FREE_ARCH_IMPL[(arch)&OPUS_ARCHMASK])(_st)) + +extern void (*const OPUS_FFT[OPUS_ARCHMASK+1])(const kiss_fft_state *cfg, + const kiss_fft_cpx *fin, kiss_fft_cpx *fout); +#define opus_fft(_cfg, _fin, _fout, arch) \ + ((*OPUS_FFT[(arch)&OPUS_ARCHMASK])(_cfg, _fin, _fout)) + +extern void (*const OPUS_IFFT[OPUS_ARCHMASK+1])(const kiss_fft_state *cfg, + const kiss_fft_cpx *fin, kiss_fft_cpx *fout); +#define opus_ifft(_cfg, _fin, _fout, arch) \ + ((*OPUS_IFFT[(arch)&OPUS_ARCHMASK])(_cfg, _fin, _fout)) + +#else /* else for if defined(OPUS_HAVE_RTCD) && (defined(HAVE_ARM_NE10)) */ + +#define opus_fft_alloc_arch(_st, arch) \ + ((void)(arch), opus_fft_alloc_arch_c(_st)) + +#define opus_fft_free_arch(_st, arch) \ + ((void)(arch), opus_fft_free_arch_c(_st)) + +#define opus_fft(_cfg, _fin, _fout, arch) \ + ((void)(arch), opus_fft_c(_cfg, _fin, _fout)) + +#define opus_ifft(_cfg, _fin, _fout, arch) \ + ((void)(arch), opus_ifft_c(_cfg, _fin, _fout)) + +#endif /* end if defined(OPUS_HAVE_RTCD) && (defined(HAVE_ARM_NE10)) */ +#endif /* end if !defined(OVERRIDE_OPUS_FFT) */ + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/celt/laplace.c b/lib/opus/celt/laplace.c new file mode 100644 index 00000000..a7bca874 --- /dev/null +++ b/lib/opus/celt/laplace.c @@ -0,0 +1,134 @@ +/* Copyright (c) 2007 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "laplace.h" +#include "mathops.h" + +/* The minimum probability of an energy delta (out of 32768). */ +#define LAPLACE_LOG_MINP (0) +#define LAPLACE_MINP (1<>15; +} + +void ec_laplace_encode(ec_enc *enc, int *value, unsigned fs, int decay) +{ + unsigned fl; + int val = *value; + fl = 0; + if (val) + { + int s; + int i; + s = -(val<0); + val = (val+s)^s; + fl = fs; + fs = ec_laplace_get_freq1(fs, decay); + /* Search the decaying part of the PDF.*/ + for (i=1; fs > 0 && i < val; i++) + { + fs *= 2; + fl += fs+2*LAPLACE_MINP; + fs = (fs*(opus_int32)decay)>>15; + } + /* Everything beyond that has probability LAPLACE_MINP. */ + if (!fs) + { + int di; + int ndi_max; + ndi_max = (32768-fl+LAPLACE_MINP-1)>>LAPLACE_LOG_MINP; + ndi_max = (ndi_max-s)>>1; + di = IMIN(val - i, ndi_max - 1); + fl += (2*di+1+s)*LAPLACE_MINP; + fs = IMIN(LAPLACE_MINP, 32768-fl); + *value = (i+di+s)^s; + } + else + { + fs += LAPLACE_MINP; + fl += fs&~s; + } + celt_assert(fl+fs<=32768); + celt_assert(fs>0); + } + ec_encode_bin(enc, fl, fl+fs, 15); +} + +int ec_laplace_decode(ec_dec *dec, unsigned fs, int decay) +{ + int val=0; + unsigned fl; + unsigned fm; + fm = ec_decode_bin(dec, 15); + fl = 0; + if (fm >= fs) + { + val++; + fl = fs; + fs = ec_laplace_get_freq1(fs, decay)+LAPLACE_MINP; + /* Search the decaying part of the PDF.*/ + while(fs > LAPLACE_MINP && fm >= fl+2*fs) + { + fs *= 2; + fl += fs; + fs = ((fs-2*LAPLACE_MINP)*(opus_int32)decay)>>15; + fs += LAPLACE_MINP; + val++; + } + /* Everything beyond that has probability LAPLACE_MINP. */ + if (fs <= LAPLACE_MINP) + { + int di; + di = (fm-fl)>>(LAPLACE_LOG_MINP+1); + val += di; + fl += 2*di*LAPLACE_MINP; + } + if (fm < fl+fs) + val = -val; + else + fl += fs; + } + celt_assert(fl<32768); + celt_assert(fs>0); + celt_assert(fl<=fm); + celt_assert(fm>1; + b=1U<>=1; + bshift--; + } + while(bshift>=0); + return g; +} + +#ifdef FIXED_POINT + +opus_val32 frac_div32(opus_val32 a, opus_val32 b) +{ + opus_val16 rcp; + opus_val32 result, rem; + int shift = celt_ilog2(b)-29; + a = VSHR32(a,shift); + b = VSHR32(b,shift); + /* 16-bit reciprocal */ + rcp = ROUND16(celt_rcp(ROUND16(b,16)),3); + result = MULT16_32_Q15(rcp, a); + rem = PSHR32(a,2)-MULT32_32_Q31(result, b); + result = ADD32(result, SHL32(MULT16_32_Q15(rcp, rem),2)); + if (result >= 536870912) /* 2^29 */ + return 2147483647; /* 2^31 - 1 */ + else if (result <= -536870912) /* -2^29 */ + return -2147483647; /* -2^31 */ + else + return SHL32(result, 2); +} + +/** Reciprocal sqrt approximation in the range [0.25,1) (Q16 in, Q14 out) */ +opus_val16 celt_rsqrt_norm(opus_val32 x) +{ + opus_val16 n; + opus_val16 r; + opus_val16 r2; + opus_val16 y; + /* Range of n is [-16384,32767] ([-0.5,1) in Q15). */ + n = x-32768; + /* Get a rough initial guess for the root. + The optimal minimax quadratic approximation (using relative error) is + r = 1.437799046117536+n*(-0.823394375837328+n*0.4096419668459485). + Coefficients here, and the final result r, are Q14.*/ + r = ADD16(23557, MULT16_16_Q15(n, ADD16(-13490, MULT16_16_Q15(n, 6713)))); + /* We want y = x*r*r-1 in Q15, but x is 32-bit Q16 and r is Q14. + We can compute the result from n and r using Q15 multiplies with some + adjustment, carefully done to avoid overflow. + Range of y is [-1564,1594]. */ + r2 = MULT16_16_Q15(r, r); + y = SHL16(SUB16(ADD16(MULT16_16_Q15(r2, n), r2), 16384), 1); + /* Apply a 2nd-order Householder iteration: r += r*y*(y*0.375-0.5). + This yields the Q14 reciprocal square root of the Q16 x, with a maximum + relative error of 1.04956E-4, a (relative) RMSE of 2.80979E-5, and a + peak absolute error of 2.26591/16384. */ + return ADD16(r, MULT16_16_Q15(r, MULT16_16_Q15(y, + SUB16(MULT16_16_Q15(y, 12288), 16384)))); +} + +/** Sqrt approximation (QX input, QX/2 output) */ +opus_val32 celt_sqrt(opus_val32 x) +{ + int k; + opus_val16 n; + opus_val32 rt; + static const opus_val16 C[5] = {23175, 11561, -3011, 1699, -664}; + if (x==0) + return 0; + else if (x>=1073741824) + return 32767; + k = (celt_ilog2(x)>>1)-7; + x = VSHR32(x, 2*k); + n = x-32768; + rt = ADD16(C[0], MULT16_16_Q15(n, ADD16(C[1], MULT16_16_Q15(n, ADD16(C[2], + MULT16_16_Q15(n, ADD16(C[3], MULT16_16_Q15(n, (C[4]))))))))); + rt = VSHR32(rt,7-k); + return rt; +} + +#define L1 32767 +#define L2 -7651 +#define L3 8277 +#define L4 -626 + +static OPUS_INLINE opus_val16 _celt_cos_pi_2(opus_val16 x) +{ + opus_val16 x2; + + x2 = MULT16_16_P15(x,x); + return ADD16(1,MIN16(32766,ADD32(SUB16(L1,x2), MULT16_16_P15(x2, ADD32(L2, MULT16_16_P15(x2, ADD32(L3, MULT16_16_P15(L4, x2 + )))))))); +} + +#undef L1 +#undef L2 +#undef L3 +#undef L4 + +opus_val16 celt_cos_norm(opus_val32 x) +{ + x = x&0x0001ffff; + if (x>SHL32(EXTEND32(1), 16)) + x = SUB32(SHL32(EXTEND32(1), 17),x); + if (x&0x00007fff) + { + if (x0); + i = celt_ilog2(x); + /* n is Q15 with range [0,1). */ + n = VSHR32(x,i-15)-32768; + /* Start with a linear approximation: + r = 1.8823529411764706-0.9411764705882353*n. + The coefficients and the result are Q14 in the range [15420,30840].*/ + r = ADD16(30840, MULT16_16_Q15(-15420, n)); + /* Perform two Newton iterations: + r -= r*((r*n)-1.Q15) + = r*((r*n)+(r-1.Q15)). */ + r = SUB16(r, MULT16_16_Q15(r, + ADD16(MULT16_16_Q15(r, n), ADD16(r, -32768)))); + /* We subtract an extra 1 in the second iteration to avoid overflow; it also + neatly compensates for truncation error in the rest of the process. */ + r = SUB16(r, ADD16(1, MULT16_16_Q15(r, + ADD16(MULT16_16_Q15(r, n), ADD16(r, -32768))))); + /* r is now the Q15 solution to 2/(n+1), with a maximum relative error + of 7.05346E-5, a (relative) RMSE of 2.14418E-5, and a peak absolute + error of 1.24665/32768. */ + return VSHR32(EXTEND32(r),i-16); +} + +#endif diff --git a/lib/opus/celt/mathops.h b/lib/opus/celt/mathops.h new file mode 100644 index 00000000..478ac918 --- /dev/null +++ b/lib/opus/celt/mathops.h @@ -0,0 +1,290 @@ +/* Copyright (c) 2002-2008 Jean-Marc Valin + Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/** + @file mathops.h + @brief Various math functions +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef MATHOPS_H +#define MATHOPS_H + +#include "arch.h" +#include "entcode.h" +#include "os_support.h" + +#define PI 3.141592653f + +/* Multiplies two 16-bit fractional values. Bit-exactness of this macro is important */ +#define FRAC_MUL16(a,b) ((16384+((opus_int32)(opus_int16)(a)*(opus_int16)(b)))>>15) + +unsigned isqrt32(opus_uint32 _val); + +/* CELT doesn't need it for fixed-point, by analysis.c does. */ +#if !defined(FIXED_POINT) || defined(ANALYSIS_C) +#define cA 0.43157974f +#define cB 0.67848403f +#define cC 0.08595542f +#define cE ((float)PI/2) +static OPUS_INLINE float fast_atan2f(float y, float x) { + float x2, y2; + x2 = x*x; + y2 = y*y; + /* For very small values, we don't care about the answer, so + we can just return 0. */ + if (x2 + y2 < 1e-18f) + { + return 0; + } + if(x2>23)-127; + in.i -= (opus_uint32)integer<<23; + frac = in.f - 1.5f; + frac = -0.41445418f + frac*(0.95909232f + + frac*(-0.33951290f + frac*0.16541097f)); + return 1+integer+frac; +} + +/** Base-2 exponential approximation (2^x). */ +static OPUS_INLINE float celt_exp2(float x) +{ + int integer; + float frac; + union { + float f; + opus_uint32 i; + } res; + integer = (int)floor(x); + if (integer < -50) + return 0; + frac = x-integer; + /* K0 = 1, K1 = log(2), K2 = 3-4*log(2), K3 = 3*log(2) - 2 */ + res.f = 0.99992522f + frac * (0.69583354f + + frac * (0.22606716f + 0.078024523f*frac)); + res.i = (res.i + ((opus_uint32)integer<<23)) & 0x7fffffff; + return res.f; +} + +#else +#define celt_log2(x) ((float)(1.442695040888963387*log(x))) +#define celt_exp2(x) ((float)exp(0.6931471805599453094*(x))) +#endif + +#endif + +#ifdef FIXED_POINT + +#include "os_support.h" + +#ifndef OVERRIDE_CELT_ILOG2 +/** Integer log in base2. Undefined for zero and negative numbers */ +static OPUS_INLINE opus_int16 celt_ilog2(opus_int32 x) +{ + celt_sig_assert(x>0); + return EC_ILOG(x)-1; +} +#endif + + +/** Integer log in base2. Defined for zero, but not for negative numbers */ +static OPUS_INLINE opus_int16 celt_zlog2(opus_val32 x) +{ + return x <= 0 ? 0 : celt_ilog2(x); +} + +opus_val16 celt_rsqrt_norm(opus_val32 x); + +opus_val32 celt_sqrt(opus_val32 x); + +opus_val16 celt_cos_norm(opus_val32 x); + +/** Base-2 logarithm approximation (log2(x)). (Q14 input, Q10 output) */ +static OPUS_INLINE opus_val16 celt_log2(opus_val32 x) +{ + int i; + opus_val16 n, frac; + /* -0.41509302963303146, 0.9609890551383969, -0.31836011537636605, + 0.15530808010959576, -0.08556153059057618 */ + static const opus_val16 C[5] = {-6801+(1<<(13-DB_SHIFT)), 15746, -5217, 2545, -1401}; + if (x==0) + return -32767; + i = celt_ilog2(x); + n = VSHR32(x,i-15)-32768-16384; + frac = ADD16(C[0], MULT16_16_Q15(n, ADD16(C[1], MULT16_16_Q15(n, ADD16(C[2], MULT16_16_Q15(n, ADD16(C[3], MULT16_16_Q15(n, C[4])))))))); + return SHL16(i-13,DB_SHIFT)+SHR16(frac,14-DB_SHIFT); +} + +/* + K0 = 1 + K1 = log(2) + K2 = 3-4*log(2) + K3 = 3*log(2) - 2 +*/ +#define D0 16383 +#define D1 22804 +#define D2 14819 +#define D3 10204 + +static OPUS_INLINE opus_val32 celt_exp2_frac(opus_val16 x) +{ + opus_val16 frac; + frac = SHL16(x, 4); + return ADD16(D0, MULT16_16_Q15(frac, ADD16(D1, MULT16_16_Q15(frac, ADD16(D2 , MULT16_16_Q15(D3,frac)))))); +} +/** Base-2 exponential approximation (2^x). (Q10 input, Q16 output) */ +static OPUS_INLINE opus_val32 celt_exp2(opus_val16 x) +{ + int integer; + opus_val16 frac; + integer = SHR16(x,10); + if (integer>14) + return 0x7f000000; + else if (integer < -15) + return 0; + frac = celt_exp2_frac(x-SHL16(integer,10)); + return VSHR32(EXTEND32(frac), -integer-2); +} + +opus_val32 celt_rcp(opus_val32 x); + +#define celt_div(a,b) MULT32_32_Q31((opus_val32)(a),celt_rcp(b)) + +opus_val32 frac_div32(opus_val32 a, opus_val32 b); + +#define M1 32767 +#define M2 -21 +#define M3 -11943 +#define M4 4936 + +/* Atan approximation using a 4th order polynomial. Input is in Q15 format + and normalized by pi/4. Output is in Q15 format */ +static OPUS_INLINE opus_val16 celt_atan01(opus_val16 x) +{ + return MULT16_16_P15(x, ADD32(M1, MULT16_16_P15(x, ADD32(M2, MULT16_16_P15(x, ADD32(M3, MULT16_16_P15(M4, x))))))); +} + +#undef M1 +#undef M2 +#undef M3 +#undef M4 + +/* atan2() approximation valid for positive input values */ +static OPUS_INLINE opus_val16 celt_atan2p(opus_val16 y, opus_val16 x) +{ + if (y < x) + { + opus_val32 arg; + arg = celt_div(SHL32(EXTEND32(y),15),x); + if (arg >= 32767) + arg = 32767; + return SHR16(celt_atan01(EXTRACT16(arg)),1); + } else { + opus_val32 arg; + arg = celt_div(SHL32(EXTEND32(x),15),y); + if (arg >= 32767) + arg = 32767; + return 25736-SHR16(celt_atan01(EXTRACT16(arg)),1); + } +} + +#endif /* FIXED_POINT */ +#endif /* MATHOPS_H */ diff --git a/lib/opus/celt/mdct.c b/lib/opus/celt/mdct.c new file mode 100644 index 00000000..5c6dab5b --- /dev/null +++ b/lib/opus/celt/mdct.c @@ -0,0 +1,343 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2008 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* This is a simple MDCT implementation that uses a N/4 complex FFT + to do most of the work. It should be relatively straightforward to + plug in pretty much and FFT here. + + This replaces the Vorbis FFT (and uses the exact same API), which + was a bit too messy and that was ending up duplicating code + (might as well use the same FFT everywhere). + + The algorithm is similar to (and inspired from) Fabrice Bellard's + MDCT implementation in FFMPEG, but has differences in signs, ordering + and scaling in many places. +*/ + +#ifndef SKIP_CONFIG_H +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#endif + +#include "mdct.h" +#include "kiss_fft.h" +#include "_kiss_fft_guts.h" +#include +#include "os_support.h" +#include "mathops.h" +#include "stack_alloc.h" + +#if defined(MIPSr1_ASM) +#include "mips/mdct_mipsr1.h" +#endif + + +#ifdef CUSTOM_MODES + +int clt_mdct_init(mdct_lookup *l,int N, int maxshift, int arch) +{ + int i; + kiss_twiddle_scalar *trig; + int shift; + int N2=N>>1; + l->n = N; + l->maxshift = maxshift; + for (i=0;i<=maxshift;i++) + { + if (i==0) + l->kfft[i] = opus_fft_alloc(N>>2>>i, 0, 0, arch); + else + l->kfft[i] = opus_fft_alloc_twiddles(N>>2>>i, 0, 0, l->kfft[0], arch); +#ifndef ENABLE_TI_DSPLIB55 + if (l->kfft[i]==NULL) + return 0; +#endif + } + l->trig = trig = (kiss_twiddle_scalar*)opus_alloc((N-(N2>>maxshift))*sizeof(kiss_twiddle_scalar)); + if (l->trig==NULL) + return 0; + for (shift=0;shift<=maxshift;shift++) + { + /* We have enough points that sine isn't necessary */ +#if defined(FIXED_POINT) +#if 1 + for (i=0;i>= 1; + N >>= 1; + } + return 1; +} + +void clt_mdct_clear(mdct_lookup *l, int arch) +{ + int i; + for (i=0;i<=l->maxshift;i++) + opus_fft_free(l->kfft[i], arch); + opus_free((kiss_twiddle_scalar*)l->trig); +} + +#endif /* CUSTOM_MODES */ + +/* Forward MDCT trashes the input array */ +#ifndef OVERRIDE_clt_mdct_forward +void clt_mdct_forward_c(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, int overlap, int shift, int stride, int arch) +{ + int i; + int N, N2, N4; + VARDECL(kiss_fft_scalar, f); + VARDECL(kiss_fft_cpx, f2); + const kiss_fft_state *st = l->kfft[shift]; + const kiss_twiddle_scalar *trig; + opus_val16 scale; +#ifdef FIXED_POINT + /* Allows us to scale with MULT16_32_Q16(), which is faster than + MULT16_32_Q15() on ARM. */ + int scale_shift = st->scale_shift-1; +#endif + SAVE_STACK; + (void)arch; + scale = st->scale; + + N = l->n; + trig = l->trig; + for (i=0;i>= 1; + trig += N; + } + N2 = N>>1; + N4 = N>>2; + + ALLOC(f, N2, kiss_fft_scalar); + ALLOC(f2, N4, kiss_fft_cpx); + + /* Consider the input to be composed of four blocks: [a, b, c, d] */ + /* Window, shuffle, fold */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_scalar * OPUS_RESTRICT xp1 = in+(overlap>>1); + const kiss_fft_scalar * OPUS_RESTRICT xp2 = in+N2-1+(overlap>>1); + kiss_fft_scalar * OPUS_RESTRICT yp = f; + const opus_val16 * OPUS_RESTRICT wp1 = window+(overlap>>1); + const opus_val16 * OPUS_RESTRICT wp2 = window+(overlap>>1)-1; + for(i=0;i<((overlap+3)>>2);i++) + { + /* Real part arranged as -d-cR, Imag part arranged as -b+aR*/ + *yp++ = MULT16_32_Q15(*wp2, xp1[N2]) + MULT16_32_Q15(*wp1,*xp2); + *yp++ = MULT16_32_Q15(*wp1, *xp1) - MULT16_32_Q15(*wp2, xp2[-N2]); + xp1+=2; + xp2-=2; + wp1+=2; + wp2-=2; + } + wp1 = window; + wp2 = window+overlap-1; + for(;i>2);i++) + { + /* Real part arranged as a-bR, Imag part arranged as -c-dR */ + *yp++ = *xp2; + *yp++ = *xp1; + xp1+=2; + xp2-=2; + } + for(;ibitrev[i]] = yc; + } + } + + /* N/4 complex FFT, does not downscale anymore */ + opus_fft_impl(st, f2); + + /* Post-rotate */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_cpx * OPUS_RESTRICT fp = f2; + kiss_fft_scalar * OPUS_RESTRICT yp1 = out; + kiss_fft_scalar * OPUS_RESTRICT yp2 = out+stride*(N2-1); + const kiss_twiddle_scalar *t = &trig[0]; + /* Temp pointers to make it really clear to the compiler what we're doing */ + for(i=0;ii,t[N4+i]) - S_MUL(fp->r,t[i]); + yi = S_MUL(fp->r,t[N4+i]) + S_MUL(fp->i,t[i]); + *yp1 = yr; + *yp2 = yi; + fp++; + yp1 += 2*stride; + yp2 -= 2*stride; + } + } + RESTORE_STACK; +} +#endif /* OVERRIDE_clt_mdct_forward */ + +#ifndef OVERRIDE_clt_mdct_backward +void clt_mdct_backward_c(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 * OPUS_RESTRICT window, int overlap, int shift, int stride, int arch) +{ + int i; + int N, N2, N4; + const kiss_twiddle_scalar *trig; + (void) arch; + + N = l->n; + trig = l->trig; + for (i=0;i>= 1; + trig += N; + } + N2 = N>>1; + N4 = N>>2; + + /* Pre-rotate */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_scalar * OPUS_RESTRICT xp1 = in; + const kiss_fft_scalar * OPUS_RESTRICT xp2 = in+stride*(N2-1); + kiss_fft_scalar * OPUS_RESTRICT yp = out+(overlap>>1); + const kiss_twiddle_scalar * OPUS_RESTRICT t = &trig[0]; + const opus_int16 * OPUS_RESTRICT bitrev = l->kfft[shift]->bitrev; + for(i=0;ikfft[shift], (kiss_fft_cpx*)(out+(overlap>>1))); + + /* Post-rotate and de-shuffle from both ends of the buffer at once to make + it in-place. */ + { + kiss_fft_scalar * yp0 = out+(overlap>>1); + kiss_fft_scalar * yp1 = out+(overlap>>1)+N2-2; + const kiss_twiddle_scalar *t = &trig[0]; + /* Loop to (N4+1)>>1 to handle odd N4. When N4 is odd, the + middle pair will be computed twice. */ + for(i=0;i<(N4+1)>>1;i++) + { + kiss_fft_scalar re, im, yr, yi; + kiss_twiddle_scalar t0, t1; + /* We swap real and imag because we're using an FFT instead of an IFFT. */ + re = yp0[1]; + im = yp0[0]; + t0 = t[i]; + t1 = t[N4+i]; + /* We'd scale up by 2 here, but instead it's done when mixing the windows */ + yr = ADD32_ovflw(S_MUL(re,t0), S_MUL(im,t1)); + yi = SUB32_ovflw(S_MUL(re,t1), S_MUL(im,t0)); + /* We swap real and imag because we're using an FFT instead of an IFFT. */ + re = yp1[1]; + im = yp1[0]; + yp0[0] = yr; + yp1[1] = yi; + + t0 = t[(N4-i-1)]; + t1 = t[(N2-i-1)]; + /* We'd scale up by 2 here, but instead it's done when mixing the windows */ + yr = ADD32_ovflw(S_MUL(re,t0), S_MUL(im,t1)); + yi = SUB32_ovflw(S_MUL(re,t1), S_MUL(im,t0)); + yp1[0] = yr; + yp0[1] = yi; + yp0 += 2; + yp1 -= 2; + } + } + + /* Mirror on both sides for TDAC */ + { + kiss_fft_scalar * OPUS_RESTRICT xp1 = out+overlap-1; + kiss_fft_scalar * OPUS_RESTRICT yp1 = out; + const opus_val16 * OPUS_RESTRICT wp1 = window; + const opus_val16 * OPUS_RESTRICT wp2 = window+overlap-1; + + for(i = 0; i < overlap/2; i++) + { + kiss_fft_scalar x1, x2; + x1 = *xp1; + x2 = *yp1; + *yp1++ = SUB32_ovflw(MULT16_32_Q15(*wp2, x2), MULT16_32_Q15(*wp1, x1)); + *xp1-- = ADD32_ovflw(MULT16_32_Q15(*wp1, x2), MULT16_32_Q15(*wp2, x1)); + wp1++; + wp2--; + } + } +} +#endif /* OVERRIDE_clt_mdct_backward */ diff --git a/lib/opus/celt/mdct.h b/lib/opus/celt/mdct.h new file mode 100644 index 00000000..160ae4e0 --- /dev/null +++ b/lib/opus/celt/mdct.h @@ -0,0 +1,112 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2008 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* This is a simple MDCT implementation that uses a N/4 complex FFT + to do most of the work. It should be relatively straightforward to + plug in pretty much and FFT here. + + This replaces the Vorbis FFT (and uses the exact same API), which + was a bit too messy and that was ending up duplicating code + (might as well use the same FFT everywhere). + + The algorithm is similar to (and inspired from) Fabrice Bellard's + MDCT implementation in FFMPEG, but has differences in signs, ordering + and scaling in many places. +*/ + +#ifndef MDCT_H +#define MDCT_H + +#include "opus_defines.h" +#include "kiss_fft.h" +#include "arch.h" + +typedef struct { + int n; + int maxshift; + const kiss_fft_state *kfft[4]; + const kiss_twiddle_scalar * OPUS_RESTRICT trig; +} mdct_lookup; + +#if defined(HAVE_ARM_NE10) +#include "arm/mdct_arm.h" +#endif + + +int clt_mdct_init(mdct_lookup *l,int N, int maxshift, int arch); +void clt_mdct_clear(mdct_lookup *l, int arch); + +/** Compute a forward MDCT and scale by 4/N, trashes the input array */ +void clt_mdct_forward_c(const mdct_lookup *l, kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, int overlap, + int shift, int stride, int arch); + +/** Compute a backward MDCT (no scaling) and performs weighted overlap-add + (scales implicitly by 1/2) */ +void clt_mdct_backward_c(const mdct_lookup *l, kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 * OPUS_RESTRICT window, + int overlap, int shift, int stride, int arch); + +#if !defined(OVERRIDE_OPUS_MDCT) +/* Is run-time CPU detection enabled on this platform? */ +#if defined(OPUS_HAVE_RTCD) && defined(HAVE_ARM_NE10) + +extern void (*const CLT_MDCT_FORWARD_IMPL[OPUS_ARCHMASK+1])( + const mdct_lookup *l, kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, const opus_val16 *window, + int overlap, int shift, int stride, int arch); + +#define clt_mdct_forward(_l, _in, _out, _window, _overlap, _shift, _stride, _arch) \ + ((*CLT_MDCT_FORWARD_IMPL[(arch)&OPUS_ARCHMASK])(_l, _in, _out, \ + _window, _overlap, _shift, \ + _stride, _arch)) + +extern void (*const CLT_MDCT_BACKWARD_IMPL[OPUS_ARCHMASK+1])( + const mdct_lookup *l, kiss_fft_scalar *in, + kiss_fft_scalar * OPUS_RESTRICT out, const opus_val16 *window, + int overlap, int shift, int stride, int arch); + +#define clt_mdct_backward(_l, _in, _out, _window, _overlap, _shift, _stride, _arch) \ + (*CLT_MDCT_BACKWARD_IMPL[(arch)&OPUS_ARCHMASK])(_l, _in, _out, \ + _window, _overlap, _shift, \ + _stride, _arch) + +#else /* if defined(OPUS_HAVE_RTCD) && defined(HAVE_ARM_NE10) */ + +#define clt_mdct_forward(_l, _in, _out, _window, _overlap, _shift, _stride, _arch) \ + clt_mdct_forward_c(_l, _in, _out, _window, _overlap, _shift, _stride, _arch) + +#define clt_mdct_backward(_l, _in, _out, _window, _overlap, _shift, _stride, _arch) \ + clt_mdct_backward_c(_l, _in, _out, _window, _overlap, _shift, _stride, _arch) + +#endif /* end if defined(OPUS_HAVE_RTCD) && defined(HAVE_ARM_NE10) && !defined(FIXED_POINT) */ +#endif /* end if !defined(OVERRIDE_OPUS_MDCT) */ + +#endif diff --git a/lib/opus/celt/meson.build b/lib/opus/celt/meson.build new file mode 100644 index 00000000..ad95d949 --- /dev/null +++ b/lib/opus/celt/meson.build @@ -0,0 +1,69 @@ +celt_sources = sources['CELT_SOURCES'] + +celt_sse_sources = sources['CELT_SOURCES_SSE'] + +celt_sse2_sources = sources['CELT_SOURCES_SSE2'] + +celt_sse4_1_sources = sources['CELT_SOURCES_SSE4_1'] + +celt_neon_intr_sources = sources['CELT_SOURCES_ARM_NEON_INTR'] + +celt_static_libs = [] + +if host_cpu_family in ['x86', 'x86_64'] and opus_conf.has('OPUS_HAVE_RTCD') + celt_sources += sources['CELT_SOURCES_X86_RTCD'] +endif + +foreach intr_name : ['sse', 'sse2', 'sse4_1', 'neon_intr'] + have_intr = get_variable('have_' + intr_name) + if not have_intr + continue + endif + + intr_sources = get_variable('celt_@0@_sources'.format(intr_name)) + intr_args = get_variable('opus_@0@_args'.format(intr_name), []) + celt_static_libs += static_library('celt_' + intr_name, intr_sources, + c_args: intr_args, + include_directories: opus_includes, + install: false) +endforeach + +have_arm_intrinsics_or_asm = have_arm_ne10 +if (intrinsics_support.length() + asm_optimization.length() + inline_optimization.length()) > 0 + have_arm_intrinsics_or_asm = true +endif + +if host_cpu_family in ['arm', 'aarch64'] and have_arm_intrinsics_or_asm + if opus_conf.has('OPUS_HAVE_RTCD') + celt_sources += sources['CELT_SOURCES_ARM_RTCD'] + endif + if have_arm_ne10 + celt_sources += sources['CELT_SOURCES_ARM_NE10'] + endif + if opus_arm_external_asm + arm2gnu = [find_program('arm/arm2gnu.pl')] + arm2gnu_args + celt_sources_arm_asm = configure_file(input: 'arm/celt_pitch_xcorr_arm.s', + output: '@BASENAME@-gnu.S', + command: arm2gnu + ['@INPUT@'], + capture: true) + celt_arm_armopts_s = configure_file(input: 'arm/armopts.s.in', + output: 'arm/armopts.s', + configuration: opus_conf) + celt_static_libs += static_library('celt-armasm', + celt_arm_armopts_s, celt_sources_arm_asm, + install: false) + endif +endif + +celt_c_args = [] +if host_system == 'windows' + celt_c_args += ['-DDLL_EXPORT'] +endif + +celt_lib = static_library('opus-celt', + celt_sources, + c_args: celt_c_args, + include_directories: opus_includes, + link_whole: celt_static_libs, + dependencies: libm, + install: false) diff --git a/lib/opus/celt/mfrngcod.h b/lib/opus/celt/mfrngcod.h new file mode 100644 index 00000000..809152a5 --- /dev/null +++ b/lib/opus/celt/mfrngcod.h @@ -0,0 +1,48 @@ +/* Copyright (c) 2001-2008 Timothy B. Terriberry + Copyright (c) 2008-2009 Xiph.Org Foundation */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(_mfrngcode_H) +# define _mfrngcode_H (1) +# include "entcode.h" + +/*Constants used by the entropy encoder/decoder.*/ + +/*The number of bits to output at a time.*/ +# define EC_SYM_BITS (8) +/*The total number of bits in each of the state registers.*/ +# define EC_CODE_BITS (32) +/*The maximum symbol value.*/ +# define EC_SYM_MAX ((1U<>EC_SYM_BITS) +/*The number of bits available for the last, partial symbol in the code field.*/ +# define EC_CODE_EXTRA ((EC_CODE_BITS-2)%EC_SYM_BITS+1) +#endif diff --git a/lib/opus/celt/mips/celt_mipsr1.h b/lib/opus/celt/mips/celt_mipsr1.h new file mode 100644 index 00000000..c332fe04 --- /dev/null +++ b/lib/opus/celt/mips/celt_mipsr1.h @@ -0,0 +1,152 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2010 Xiph.Org Foundation + Copyright (c) 2008 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef __CELT_MIPSR1_H__ +#define __CELT_MIPSR1_H__ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#define CELT_C + +#include "os_support.h" +#include "mdct.h" +#include +#include "celt.h" +#include "pitch.h" +#include "bands.h" +#include "modes.h" +#include "entcode.h" +#include "quant_bands.h" +#include "rate.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "float_cast.h" +#include +#include "celt_lpc.h" +#include "vq.h" + +#define OVERRIDE_COMB_FILTER_CONST +#define OVERRIDE_comb_filter +void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N, + opus_val16 g0, opus_val16 g1, int tapset0, int tapset1, + const opus_val16 *window, int overlap, int arch) +{ + int i; + opus_val32 x0, x1, x2, x3, x4; + + (void)arch; + + /* printf ("%d %d %f %f\n", T0, T1, g0, g1); */ + opus_val16 g00, g01, g02, g10, g11, g12; + static const opus_val16 gains[3][3] = { + {QCONST16(0.3066406250f, 15), QCONST16(0.2170410156f, 15), QCONST16(0.1296386719f, 15)}, + {QCONST16(0.4638671875f, 15), QCONST16(0.2680664062f, 15), QCONST16(0.f, 15)}, + {QCONST16(0.7998046875f, 15), QCONST16(0.1000976562f, 15), QCONST16(0.f, 15)}}; + + if (g0==0 && g1==0) + { + /* OPT: Happens to work without the OPUS_MOVE(), but only because the current encoder already copies x to y */ + if (x!=y) + OPUS_MOVE(y, x, N); + return; + } + + g00 = MULT16_16_P15(g0, gains[tapset0][0]); + g01 = MULT16_16_P15(g0, gains[tapset0][1]); + g02 = MULT16_16_P15(g0, gains[tapset0][2]); + g10 = MULT16_16_P15(g1, gains[tapset1][0]); + g11 = MULT16_16_P15(g1, gains[tapset1][1]); + g12 = MULT16_16_P15(g1, gains[tapset1][2]); + x1 = x[-T1+1]; + x2 = x[-T1 ]; + x3 = x[-T1-1]; + x4 = x[-T1-2]; + /* If the filter didn't change, we don't need the overlap */ + if (g0==g1 && T0==T1 && tapset0==tapset1) + overlap=0; + + for (i=0;itwiddles[fstride*m]; + yb = st->twiddles[fstride*2*m]; +#endif + + tw=st->twiddles; + + for (i=0;ir += scratch[7].r + scratch[8].r; + Fout0->i += scratch[7].i + scratch[8].i; + scratch[5].r = scratch[0].r + S_MUL_ADD(scratch[7].r,ya.r,scratch[8].r,yb.r); + scratch[5].i = scratch[0].i + S_MUL_ADD(scratch[7].i,ya.r,scratch[8].i,yb.r); + + scratch[6].r = S_MUL_ADD(scratch[10].i,ya.i,scratch[9].i,yb.i); + scratch[6].i = -S_MUL_ADD(scratch[10].r,ya.i,scratch[9].r,yb.i); + + C_SUB(*Fout1,scratch[5],scratch[6]); + C_ADD(*Fout4,scratch[5],scratch[6]); + + scratch[11].r = scratch[0].r + S_MUL_ADD(scratch[7].r,yb.r,scratch[8].r,ya.r); + scratch[11].i = scratch[0].i + S_MUL_ADD(scratch[7].i,yb.r,scratch[8].i,ya.r); + + scratch[12].r = S_MUL_SUB(scratch[9].i,ya.i,scratch[10].i,yb.i); + scratch[12].i = S_MUL_SUB(scratch[10].r,yb.i,scratch[9].r,ya.i); + + C_ADD(*Fout2,scratch[11],scratch[12]); + C_SUB(*Fout3,scratch[11],scratch[12]); + + ++Fout0;++Fout1;++Fout2;++Fout3;++Fout4; + } + } +} + + +#endif /* KISS_FFT_MIPSR1_H */ diff --git a/lib/opus/celt/mips/mdct_mipsr1.h b/lib/opus/celt/mips/mdct_mipsr1.h new file mode 100644 index 00000000..2934dab7 --- /dev/null +++ b/lib/opus/celt/mips/mdct_mipsr1.h @@ -0,0 +1,288 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2008 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* This is a simple MDCT implementation that uses a N/4 complex FFT + to do most of the work. It should be relatively straightforward to + plug in pretty much and FFT here. + + This replaces the Vorbis FFT (and uses the exact same API), which + was a bit too messy and that was ending up duplicating code + (might as well use the same FFT everywhere). + + The algorithm is similar to (and inspired from) Fabrice Bellard's + MDCT implementation in FFMPEG, but has differences in signs, ordering + and scaling in many places. +*/ +#ifndef __MDCT_MIPSR1_H__ +#define __MDCT_MIPSR1_H__ + +#ifndef SKIP_CONFIG_H +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#endif + +#include "mdct.h" +#include "kiss_fft.h" +#include "_kiss_fft_guts.h" +#include +#include "os_support.h" +#include "mathops.h" +#include "stack_alloc.h" + +/* Forward MDCT trashes the input array */ +#define OVERRIDE_clt_mdct_forward +void clt_mdct_forward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 *window, int overlap, int shift, int stride, int arch) +{ + int i; + int N, N2, N4; + VARDECL(kiss_fft_scalar, f); + VARDECL(kiss_fft_cpx, f2); + const kiss_fft_state *st = l->kfft[shift]; + const kiss_twiddle_scalar *trig; + opus_val16 scale; +#ifdef FIXED_POINT + /* Allows us to scale with MULT16_32_Q16(), which is faster than + MULT16_32_Q15() on ARM. */ + int scale_shift = st->scale_shift-1; +#endif + + (void)arch; + + SAVE_STACK; + scale = st->scale; + + N = l->n; + trig = l->trig; + for (i=0;i>= 1; + trig += N; + } + N2 = N>>1; + N4 = N>>2; + + ALLOC(f, N2, kiss_fft_scalar); + ALLOC(f2, N4, kiss_fft_cpx); + + /* Consider the input to be composed of four blocks: [a, b, c, d] */ + /* Window, shuffle, fold */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_scalar * OPUS_RESTRICT xp1 = in+(overlap>>1); + const kiss_fft_scalar * OPUS_RESTRICT xp2 = in+N2-1+(overlap>>1); + kiss_fft_scalar * OPUS_RESTRICT yp = f; + const opus_val16 * OPUS_RESTRICT wp1 = window+(overlap>>1); + const opus_val16 * OPUS_RESTRICT wp2 = window+(overlap>>1)-1; + for(i=0;i<((overlap+3)>>2);i++) + { + /* Real part arranged as -d-cR, Imag part arranged as -b+aR*/ + *yp++ = S_MUL_ADD(*wp2, xp1[N2],*wp1,*xp2); + *yp++ = S_MUL_SUB(*wp1, *xp1,*wp2, xp2[-N2]); + xp1+=2; + xp2-=2; + wp1+=2; + wp2-=2; + } + wp1 = window; + wp2 = window+overlap-1; + for(;i>2);i++) + { + /* Real part arranged as a-bR, Imag part arranged as -c-dR */ + *yp++ = *xp2; + *yp++ = *xp1; + xp1+=2; + xp2-=2; + } + for(;ibitrev[i]] = yc; + } + } + + /* N/4 complex FFT, does not downscale anymore */ + opus_fft_impl(st, f2); + + /* Post-rotate */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_cpx * OPUS_RESTRICT fp = f2; + kiss_fft_scalar * OPUS_RESTRICT yp1 = out; + kiss_fft_scalar * OPUS_RESTRICT yp2 = out+stride*(N2-1); + const kiss_twiddle_scalar *t = &trig[0]; + /* Temp pointers to make it really clear to the compiler what we're doing */ + for(i=0;ii,t[N4+i] , fp->r,t[i]); + yi = S_MUL_ADD(fp->r,t[N4+i] ,fp->i,t[i]); + *yp1 = yr; + *yp2 = yi; + fp++; + yp1 += 2*stride; + yp2 -= 2*stride; + } + } + RESTORE_STACK; +} + +#define OVERRIDE_clt_mdct_backward +void clt_mdct_backward(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_scalar * OPUS_RESTRICT out, + const opus_val16 * OPUS_RESTRICT window, int overlap, int shift, int stride, int arch) +{ + int i; + int N, N2, N4; + const kiss_twiddle_scalar *trig; + + (void)arch; + + N = l->n; + trig = l->trig; + for (i=0;i>= 1; + trig += N; + } + N2 = N>>1; + N4 = N>>2; + + /* Pre-rotate */ + { + /* Temp pointers to make it really clear to the compiler what we're doing */ + const kiss_fft_scalar * OPUS_RESTRICT xp1 = in; + const kiss_fft_scalar * OPUS_RESTRICT xp2 = in+stride*(N2-1); + kiss_fft_scalar * OPUS_RESTRICT yp = out+(overlap>>1); + const kiss_twiddle_scalar * OPUS_RESTRICT t = &trig[0]; + const opus_int16 * OPUS_RESTRICT bitrev = l->kfft[shift]->bitrev; + for(i=0;ikfft[shift], (kiss_fft_cpx*)(out+(overlap>>1))); + + /* Post-rotate and de-shuffle from both ends of the buffer at once to make + it in-place. */ + { + kiss_fft_scalar * OPUS_RESTRICT yp0 = out+(overlap>>1); + kiss_fft_scalar * OPUS_RESTRICT yp1 = out+(overlap>>1)+N2-2; + const kiss_twiddle_scalar *t = &trig[0]; + /* Loop to (N4+1)>>1 to handle odd N4. When N4 is odd, the + middle pair will be computed twice. */ + for(i=0;i<(N4+1)>>1;i++) + { + kiss_fft_scalar re, im, yr, yi; + kiss_twiddle_scalar t0, t1; + /* We swap real and imag because we're using an FFT instead of an IFFT. */ + re = yp0[1]; + im = yp0[0]; + t0 = t[i]; + t1 = t[N4+i]; + /* We'd scale up by 2 here, but instead it's done when mixing the windows */ + yr = S_MUL_ADD(re,t0 , im,t1); + yi = S_MUL_SUB(re,t1 , im,t0); + /* We swap real and imag because we're using an FFT instead of an IFFT. */ + re = yp1[1]; + im = yp1[0]; + yp0[0] = yr; + yp1[1] = yi; + + t0 = t[(N4-i-1)]; + t1 = t[(N2-i-1)]; + /* We'd scale up by 2 here, but instead it's done when mixing the windows */ + yr = S_MUL_ADD(re,t0,im,t1); + yi = S_MUL_SUB(re,t1,im,t0); + yp1[0] = yr; + yp0[1] = yi; + yp0 += 2; + yp1 -= 2; + } + } + + /* Mirror on both sides for TDAC */ + { + kiss_fft_scalar * OPUS_RESTRICT xp1 = out+overlap-1; + kiss_fft_scalar * OPUS_RESTRICT yp1 = out; + const opus_val16 * OPUS_RESTRICT wp1 = window; + const opus_val16 * OPUS_RESTRICT wp2 = window+overlap-1; + + for(i = 0; i < overlap/2; i++) + { + kiss_fft_scalar x1, x2; + x1 = *xp1; + x2 = *yp1; + *yp1++ = MULT16_32_Q15(*wp2, x2) - MULT16_32_Q15(*wp1, x1); + *xp1-- = MULT16_32_Q15(*wp1, x2) + MULT16_32_Q15(*wp2, x1); + wp1++; + wp2--; + } + } +} +#endif /* __MDCT_MIPSR1_H__ */ diff --git a/lib/opus/celt/mips/pitch_mipsr1.h b/lib/opus/celt/mips/pitch_mipsr1.h new file mode 100644 index 00000000..a9500aff --- /dev/null +++ b/lib/opus/celt/mips/pitch_mipsr1.h @@ -0,0 +1,161 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/** + @file pitch.h + @brief Pitch analysis + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef PITCH_MIPSR1_H +#define PITCH_MIPSR1_H + +#define OVERRIDE_DUAL_INNER_PROD +static inline void dual_inner_prod(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02, + int N, opus_val32 *xy1, opus_val32 *xy2, int arch) +{ + int j; + opus_val32 xy01=0; + opus_val32 xy02=0; + + (void)arch; + + asm volatile("MULT $ac1, $0, $0"); + asm volatile("MULT $ac2, $0, $0"); + /* Compute the norm of X+Y and X-Y as |X|^2 + |Y|^2 +/- sum(xy) */ + for (j=0;j=0;i--) + { + celt_norm x1, x2; + x1 = Xptr[0]; + x2 = Xptr[stride]; + Xptr[stride] = EXTRACT16(PSHR32(MAC16_16(MULT16_16(c, x2), s, x1), 15)); + *Xptr-- = EXTRACT16(PSHR32(MAC16_16(MULT16_16(c, x1), ms, x2), 15)); + } +} + +#define OVERRIDE_renormalise_vector +void renormalise_vector(celt_norm *X, int N, opus_val16 gain, int arch) +{ + int i; +#ifdef FIXED_POINT + int k; +#endif + opus_val32 E = EPSILON; + opus_val16 g; + opus_val32 t; + celt_norm *xptr = X; + int X0, X1; + + (void)arch; + + asm volatile("mult $ac1, $0, $0"); + asm volatile("MTLO %0, $ac1" : :"r" (E)); + /*if(N %4) + printf("error");*/ + for (i=0;i>1; +#endif + t = VSHR32(E, 2*(k-7)); + g = MULT16_16_P15(celt_rsqrt_norm(t),gain); + + xptr = X; + for (i=0;i= Fs) + break; + + /* Find where the linear part ends (i.e. where the spacing is more than min_width */ + for (lin=0;lin= res) + break; + + low = (bark_freq[lin]+res/2)/res; + high = nBark-lin; + *nbEBands = low+high; + eBands = opus_alloc(sizeof(opus_int16)*(*nbEBands+2)); + + if (eBands==NULL) + return NULL; + + /* Linear spacing (min_width) */ + for (i=0;i0) + offset = eBands[low-1]*res - bark_freq[lin-1]; + /* Spacing follows critical bands */ + for (i=0;i frame_size) + eBands[*nbEBands] = frame_size; + for (i=1;i<*nbEBands-1;i++) + { + if (eBands[i+1]-eBands[i] < eBands[i]-eBands[i-1]) + { + eBands[i] -= (2*eBands[i]-eBands[i-1]-eBands[i+1])/2; + } + } + /* Remove any empty bands. */ + for (i=j=0;i<*nbEBands;i++) + if(eBands[i+1]>eBands[j]) + eBands[++j]=eBands[i+1]; + *nbEBands=j; + + for (i=1;i<*nbEBands;i++) + { + /* Every band must be smaller than the last band. */ + celt_assert(eBands[i]-eBands[i-1]<=eBands[*nbEBands]-eBands[*nbEBands-1]); + /* Each band must be no larger than twice the size of the previous one. */ + celt_assert(eBands[i+1]-eBands[i]<=2*(eBands[i]-eBands[i-1])); + } + + return eBands; +} + +static void compute_allocation_table(CELTMode *mode) +{ + int i, j; + unsigned char *allocVectors; + int maxBands = sizeof(eband5ms)/sizeof(eband5ms[0])-1; + + mode->nbAllocVectors = BITALLOC_SIZE; + allocVectors = opus_alloc(sizeof(unsigned char)*(BITALLOC_SIZE*mode->nbEBands)); + if (allocVectors==NULL) + { + mode->allocVectors = NULL; + return; + } + + /* Check for standard mode */ + if (mode->Fs == 400*(opus_int32)mode->shortMdctSize) + { + for (i=0;inbEBands;i++) + allocVectors[i] = band_allocation[i]; + mode->allocVectors = allocVectors; + return; + } + /* If not the standard mode, interpolate */ + /* Compute per-codec-band allocation from per-critical-band matrix */ + for (i=0;inbEBands;j++) + { + int k; + for (k=0;k mode->eBands[j]*(opus_int32)mode->Fs/mode->shortMdctSize) + break; + } + if (k>maxBands-1) + allocVectors[i*mode->nbEBands+j] = band_allocation[i*maxBands + maxBands-1]; + else { + opus_int32 a0, a1; + a1 = mode->eBands[j]*(opus_int32)mode->Fs/mode->shortMdctSize - 400*(opus_int32)eband5ms[k-1]; + a0 = 400*(opus_int32)eband5ms[k] - mode->eBands[j]*(opus_int32)mode->Fs/mode->shortMdctSize; + allocVectors[i*mode->nbEBands+j] = (a0*band_allocation[i*maxBands+k-1] + + a1*band_allocation[i*maxBands+k])/(a0+a1); + } + } + } + + /*printf ("\n"); + for (i=0;inbEBands;j++) + printf ("%d ", allocVectors[i*mode->nbEBands+j]); + printf ("\n"); + } + exit(0);*/ + + mode->allocVectors = allocVectors; +} + +#endif /* CUSTOM_MODES */ + +CELTMode *opus_custom_mode_create(opus_int32 Fs, int frame_size, int *error) +{ + int i; +#ifdef CUSTOM_MODES + CELTMode *mode=NULL; + int res; + opus_val16 *window; + opus_int16 *logN; + int LM; + int arch = opus_select_arch(); + ALLOC_STACK; +#if !defined(VAR_ARRAYS) && !defined(USE_ALLOCA) + if (global_stack==NULL) + goto failure; +#endif +#endif + +#ifndef CUSTOM_MODES_ONLY + for (i=0;iFs && + (frame_size<shortMdctSize*static_mode_list[i]->nbShortMdcts) + { + if (error) + *error = OPUS_OK; + return (CELTMode*)static_mode_list[i]; + } + } + } +#endif /* CUSTOM_MODES_ONLY */ + +#ifndef CUSTOM_MODES + if (error) + *error = OPUS_BAD_ARG; + return NULL; +#else + + /* The good thing here is that permutation of the arguments will automatically be invalid */ + + if (Fs < 8000 || Fs > 96000) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + if (frame_size < 40 || frame_size > 1024 || frame_size%2!=0) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + /* Frames of less than 1ms are not supported. */ + if ((opus_int32)frame_size*1000 < Fs) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + + if ((opus_int32)frame_size*75 >= Fs && (frame_size%16)==0) + { + LM = 3; + } else if ((opus_int32)frame_size*150 >= Fs && (frame_size%8)==0) + { + LM = 2; + } else if ((opus_int32)frame_size*300 >= Fs && (frame_size%4)==0) + { + LM = 1; + } else + { + LM = 0; + } + + /* Shorts longer than 3.3ms are not supported. */ + if ((opus_int32)(frame_size>>LM)*300 > Fs) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + + mode = opus_alloc(sizeof(CELTMode)); + if (mode==NULL) + goto failure; + mode->Fs = Fs; + + /* Pre/de-emphasis depends on sampling rate. The "standard" pre-emphasis + is defined as A(z) = 1 - 0.85*z^-1 at 48 kHz. Other rates should + approximate that. */ + if(Fs < 12000) /* 8 kHz */ + { + mode->preemph[0] = QCONST16(0.3500061035f, 15); + mode->preemph[1] = -QCONST16(0.1799926758f, 15); + mode->preemph[2] = QCONST16(0.2719968125f, SIG_SHIFT); /* exact 1/preemph[3] */ + mode->preemph[3] = QCONST16(3.6765136719f, 13); + } else if(Fs < 24000) /* 16 kHz */ + { + mode->preemph[0] = QCONST16(0.6000061035f, 15); + mode->preemph[1] = -QCONST16(0.1799926758f, 15); + mode->preemph[2] = QCONST16(0.4424998650f, SIG_SHIFT); /* exact 1/preemph[3] */ + mode->preemph[3] = QCONST16(2.2598876953f, 13); + } else if(Fs < 40000) /* 32 kHz */ + { + mode->preemph[0] = QCONST16(0.7799987793f, 15); + mode->preemph[1] = -QCONST16(0.1000061035f, 15); + mode->preemph[2] = QCONST16(0.7499771125f, SIG_SHIFT); /* exact 1/preemph[3] */ + mode->preemph[3] = QCONST16(1.3333740234f, 13); + } else /* 48 kHz */ + { + mode->preemph[0] = QCONST16(0.8500061035f, 15); + mode->preemph[1] = QCONST16(0.0f, 15); + mode->preemph[2] = QCONST16(1.f, SIG_SHIFT); + mode->preemph[3] = QCONST16(1.f, 13); + } + + mode->maxLM = LM; + mode->nbShortMdcts = 1<shortMdctSize = frame_size/mode->nbShortMdcts; + res = (mode->Fs+mode->shortMdctSize)/(2*mode->shortMdctSize); + + mode->eBands = compute_ebands(Fs, mode->shortMdctSize, res, &mode->nbEBands); + if (mode->eBands==NULL) + goto failure; +#if !defined(SMALL_FOOTPRINT) + /* Make sure we don't allocate a band larger than our PVQ table. + 208 should be enough, but let's be paranoid. */ + if ((mode->eBands[mode->nbEBands] - mode->eBands[mode->nbEBands-1])< + 208) { + goto failure; + } +#endif + + mode->effEBands = mode->nbEBands; + while (mode->eBands[mode->effEBands] > mode->shortMdctSize) + mode->effEBands--; + + /* Overlap must be divisible by 4 */ + mode->overlap = ((mode->shortMdctSize>>2)<<2); + + compute_allocation_table(mode); + if (mode->allocVectors==NULL) + goto failure; + + window = (opus_val16*)opus_alloc(mode->overlap*sizeof(opus_val16)); + if (window==NULL) + goto failure; + +#ifndef FIXED_POINT + for (i=0;ioverlap;i++) + window[i] = Q15ONE*sin(.5*M_PI* sin(.5*M_PI*(i+.5)/mode->overlap) * sin(.5*M_PI*(i+.5)/mode->overlap)); +#else + for (i=0;ioverlap;i++) + window[i] = MIN32(32767,floor(.5+32768.*sin(.5*M_PI* sin(.5*M_PI*(i+.5)/mode->overlap) * sin(.5*M_PI*(i+.5)/mode->overlap)))); +#endif + mode->window = window; + + logN = (opus_int16*)opus_alloc(mode->nbEBands*sizeof(opus_int16)); + if (logN==NULL) + goto failure; + + for (i=0;inbEBands;i++) + logN[i] = log2_frac(mode->eBands[i+1]-mode->eBands[i], BITRES); + mode->logN = logN; + + compute_pulse_cache(mode, mode->maxLM); + + if (clt_mdct_init(&mode->mdct, 2*mode->shortMdctSize*mode->nbShortMdcts, + mode->maxLM, arch) == 0) + goto failure; + + if (error) + *error = OPUS_OK; + + return mode; +failure: + if (error) + *error = OPUS_ALLOC_FAIL; + if (mode!=NULL) + opus_custom_mode_destroy(mode); + return NULL; +#endif /* !CUSTOM_MODES */ +} + +#ifdef CUSTOM_MODES +void opus_custom_mode_destroy(CELTMode *mode) +{ + int arch = opus_select_arch(); + + if (mode == NULL) + return; +#ifndef CUSTOM_MODES_ONLY + { + int i; + for (i=0;ieBands); + opus_free((unsigned char*)mode->allocVectors); + + opus_free((opus_val16*)mode->window); + opus_free((opus_int16*)mode->logN); + + opus_free((opus_int16*)mode->cache.index); + opus_free((unsigned char*)mode->cache.bits); + opus_free((unsigned char*)mode->cache.caps); + clt_mdct_clear(&mode->mdct, arch); + + opus_free((CELTMode *)mode); +} +#endif diff --git a/lib/opus/celt/modes.h b/lib/opus/celt/modes.h new file mode 100644 index 00000000..be813ccc --- /dev/null +++ b/lib/opus/celt/modes.h @@ -0,0 +1,75 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2008 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef MODES_H +#define MODES_H + +#include "opus_types.h" +#include "celt.h" +#include "arch.h" +#include "mdct.h" +#include "entenc.h" +#include "entdec.h" + +#define MAX_PERIOD 1024 + +typedef struct { + int size; + const opus_int16 *index; + const unsigned char *bits; + const unsigned char *caps; +} PulseCache; + +/** Mode definition (opaque) + @brief Mode definition + */ +struct OpusCustomMode { + opus_int32 Fs; + int overlap; + + int nbEBands; + int effEBands; + opus_val16 preemph[4]; + const opus_int16 *eBands; /**< Definition for each "pseudo-critical band" */ + + int maxLM; + int nbShortMdcts; + int shortMdctSize; + + int nbAllocVectors; /**< Number of lines in the matrix below */ + const unsigned char *allocVectors; /**< Number of bits in each band for several rates */ + const opus_int16 *logN; + + const opus_val16 *window; + mdct_lookup mdct; + PulseCache cache; +}; + + +#endif diff --git a/lib/opus/celt/opus_custom_demo.c b/lib/opus/celt/opus_custom_demo.c new file mode 100644 index 00000000..ae41c0de --- /dev/null +++ b/lib/opus/celt/opus_custom_demo.c @@ -0,0 +1,210 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_custom.h" +#include "arch.h" +#include +#include +#include +#include + +#define MAX_PACKET 1275 + +int main(int argc, char *argv[]) +{ + int err; + char *inFile, *outFile; + FILE *fin, *fout; + OpusCustomMode *mode=NULL; + OpusCustomEncoder *enc; + OpusCustomDecoder *dec; + int len; + opus_int32 frame_size, channels, rate; + int bytes_per_packet; + unsigned char data[MAX_PACKET]; + int complexity; +#if !(defined (FIXED_POINT) && !defined(CUSTOM_MODES)) && defined(RESYNTH) + int i; + double rmsd = 0; +#endif + int count = 0; + opus_int32 skip; + opus_int16 *in, *out; + if (argc != 9 && argc != 8 && argc != 7) + { + fprintf (stderr, "Usage: test_opus_custom " + " [ [packet loss rate]] " + " \n"); + return 1; + } + + rate = (opus_int32)atol(argv[1]); + channels = atoi(argv[2]); + frame_size = atoi(argv[3]); + mode = opus_custom_mode_create(rate, frame_size, NULL); + if (mode == NULL) + { + fprintf(stderr, "failed to create a mode\n"); + return 1; + } + + bytes_per_packet = atoi(argv[4]); + if (bytes_per_packet < 0 || bytes_per_packet > MAX_PACKET) + { + fprintf (stderr, "bytes per packet must be between 0 and %d\n", + MAX_PACKET); + return 1; + } + + inFile = argv[argc-2]; + fin = fopen(inFile, "rb"); + if (!fin) + { + fprintf (stderr, "Could not open input file %s\n", argv[argc-2]); + return 1; + } + outFile = argv[argc-1]; + fout = fopen(outFile, "wb+"); + if (!fout) + { + fprintf (stderr, "Could not open output file %s\n", argv[argc-1]); + fclose(fin); + return 1; + } + + enc = opus_custom_encoder_create(mode, channels, &err); + if (err != 0) + { + fprintf(stderr, "Failed to create the encoder: %s\n", opus_strerror(err)); + fclose(fin); + fclose(fout); + return 1; + } + dec = opus_custom_decoder_create(mode, channels, &err); + if (err != 0) + { + fprintf(stderr, "Failed to create the decoder: %s\n", opus_strerror(err)); + fclose(fin); + fclose(fout); + return 1; + } + opus_custom_decoder_ctl(dec, OPUS_GET_LOOKAHEAD(&skip)); + + if (argc>7) + { + complexity=atoi(argv[5]); + opus_custom_encoder_ctl(enc,OPUS_SET_COMPLEXITY(complexity)); + } + + in = (opus_int16*)malloc(frame_size*channels*sizeof(opus_int16)); + out = (opus_int16*)malloc(frame_size*channels*sizeof(opus_int16)); + + while (!feof(fin)) + { + int ret; + err = fread(in, sizeof(short), frame_size*channels, fin); + if (feof(fin)) + break; + len = opus_custom_encode(enc, in, frame_size, data, bytes_per_packet); + if (len <= 0) + fprintf (stderr, "opus_custom_encode() failed: %s\n", opus_strerror(len)); + + /* This is for simulating bit errors */ +#if 0 + int errors = 0; + int eid = 0; + /* This simulates random bit error */ + for (i=0;i 0) + { + rmsd = sqrt(rmsd/(1.0*frame_size*channels*count)); + fprintf (stderr, "Error: encoder doesn't match decoder\n"); + fprintf (stderr, "RMS mismatch is %f\n", rmsd); + return 1; + } else { + fprintf (stderr, "Encoder matches decoder!!\n"); + } +#endif + return 0; +} + diff --git a/lib/opus/celt/os_support.h b/lib/opus/celt/os_support.h new file mode 100644 index 00000000..009bf861 --- /dev/null +++ b/lib/opus/celt/os_support.h @@ -0,0 +1,91 @@ +/* Copyright (C) 2007 Jean-Marc Valin + + File: os_support.h + This is the (tiny) OS abstraction layer. Aside from math.h, this is the + only place where system headers are allowed. + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions are + met: + + 1. Redistributions of source code must retain the above copyright notice, + this list of conditions and the following disclaimer. + + 2. Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE + DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, + INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES + (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR + SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, + STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN + ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef OS_SUPPORT_H +#define OS_SUPPORT_H + +#ifdef CUSTOM_SUPPORT +# include "custom_support.h" +#endif + +#include "opus_types.h" +#include "opus_defines.h" + +#include +#include + +/** Opus wrapper for malloc(). To do your own dynamic allocation, all you need to do is replace this function and opus_free */ +#ifndef OVERRIDE_OPUS_ALLOC +static OPUS_INLINE void *opus_alloc (size_t size) +{ + return malloc(size); +} +#endif + +/** Same as celt_alloc(), except that the area is only needed inside a CELT call (might cause problem with wideband though) */ +#ifndef OVERRIDE_OPUS_ALLOC_SCRATCH +static OPUS_INLINE void *opus_alloc_scratch (size_t size) +{ + /* Scratch space doesn't need to be cleared */ + return opus_alloc(size); +} +#endif + +/** Opus wrapper for free(). To do your own dynamic allocation, all you need to do is replace this function and opus_alloc */ +#ifndef OVERRIDE_OPUS_FREE +static OPUS_INLINE void opus_free (void *ptr) +{ + free(ptr); +} +#endif + +/** Copy n elements from src to dst. The 0* term provides compile-time type checking */ +#ifndef OVERRIDE_OPUS_COPY +#define OPUS_COPY(dst, src, n) (memcpy((dst), (src), (n)*sizeof(*(dst)) + 0*((dst)-(src)) )) +#endif + +/** Copy n elements from src to dst, allowing overlapping regions. The 0* term + provides compile-time type checking */ +#ifndef OVERRIDE_OPUS_MOVE +#define OPUS_MOVE(dst, src, n) (memmove((dst), (src), (n)*sizeof(*(dst)) + 0*((dst)-(src)) )) +#endif + +/** Set n elements of dst to zero */ +#ifndef OVERRIDE_OPUS_CLEAR +#define OPUS_CLEAR(dst, n) (memset((dst), 0, (n)*sizeof(*(dst)))) +#endif + +/*#ifdef __GNUC__ +#pragma GCC poison printf sprintf +#pragma GCC poison malloc free realloc calloc +#endif*/ + +#endif /* OS_SUPPORT_H */ + diff --git a/lib/opus/celt/pitch.c b/lib/opus/celt/pitch.c new file mode 100644 index 00000000..7998db41 --- /dev/null +++ b/lib/opus/celt/pitch.c @@ -0,0 +1,546 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/** + @file pitch.c + @brief Pitch analysis + */ + +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "pitch.h" +#include "os_support.h" +#include "modes.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "celt_lpc.h" + +static void find_best_pitch(opus_val32 *xcorr, opus_val16 *y, int len, + int max_pitch, int *best_pitch +#ifdef FIXED_POINT + , int yshift, opus_val32 maxcorr +#endif + ) +{ + int i, j; + opus_val32 Syy=1; + opus_val16 best_num[2]; + opus_val32 best_den[2]; +#ifdef FIXED_POINT + int xshift; + + xshift = celt_ilog2(maxcorr)-14; +#endif + + best_num[0] = -1; + best_num[1] = -1; + best_den[0] = 0; + best_den[1] = 0; + best_pitch[0] = 0; + best_pitch[1] = 1; + for (j=0;j0) + { + opus_val16 num; + opus_val32 xcorr16; + xcorr16 = EXTRACT16(VSHR32(xcorr[i], xshift)); +#ifndef FIXED_POINT + /* Considering the range of xcorr16, this should avoid both underflows + and overflows (inf) when squaring xcorr16 */ + xcorr16 *= 1e-12f; +#endif + num = MULT16_16_Q15(xcorr16,xcorr16); + if (MULT16_32_Q15(num,best_den[1]) > MULT16_32_Q15(best_num[1],Syy)) + { + if (MULT16_32_Q15(num,best_den[0]) > MULT16_32_Q15(best_num[0],Syy)) + { + best_num[1] = best_num[0]; + best_den[1] = best_den[0]; + best_pitch[1] = best_pitch[0]; + best_num[0] = num; + best_den[0] = Syy; + best_pitch[0] = i; + } else { + best_num[1] = num; + best_den[1] = Syy; + best_pitch[1] = i; + } + } + } + Syy += SHR32(MULT16_16(y[i+len],y[i+len]),yshift) - SHR32(MULT16_16(y[i],y[i]),yshift); + Syy = MAX32(1, Syy); + } +} + +static void celt_fir5(opus_val16 *x, + const opus_val16 *num, + int N) +{ + int i; + opus_val16 num0, num1, num2, num3, num4; + opus_val32 mem0, mem1, mem2, mem3, mem4; + num0=num[0]; + num1=num[1]; + num2=num[2]; + num3=num[3]; + num4=num[4]; + mem0=0; + mem1=0; + mem2=0; + mem3=0; + mem4=0; + for (i=0;i>1;i++) + x_lp[i] = SHR32(x[0][(2*i-1)], shift+2) + SHR32(x[0][(2*i+1)], shift+2) + SHR32(x[0][2*i], shift+1); + x_lp[0] = SHR32(x[0][1], shift+2) + SHR32(x[0][0], shift+1); + if (C==2) + { + for (i=1;i>1;i++) + x_lp[i] += SHR32(x[1][(2*i-1)], shift+2) + SHR32(x[1][(2*i+1)], shift+2) + SHR32(x[1][2*i], shift+1); + x_lp[0] += SHR32(x[1][1], shift+2) + SHR32(x[1][0], shift+1); + } +#else + for (i=1;i>1;i++) + x_lp[i] = .25f*x[0][(2*i-1)] + .25f*x[0][(2*i+1)] + .5f*x[0][2*i]; + x_lp[0] = .25f*x[0][1] + .5f*x[0][0]; + if (C==2) + { + for (i=1;i>1;i++) + x_lp[i] += .25f*x[1][(2*i-1)] + .25f*x[1][(2*i+1)] + .5f*x[1][2*i]; + x_lp[0] += .25f*x[1][1] + .5f*x[1][0]; + } +#endif + _celt_autocorr(x_lp, ac, NULL, 0, + 4, len>>1, arch); + + /* Noise floor -40 dB */ +#ifdef FIXED_POINT + ac[0] += SHR32(ac[0],13); +#else + ac[0] *= 1.0001f; +#endif + /* Lag windowing */ + for (i=1;i<=4;i++) + { + /*ac[i] *= exp(-.5*(2*M_PI*.002*i)*(2*M_PI*.002*i));*/ +#ifdef FIXED_POINT + ac[i] -= MULT16_32_Q15(2*i*i, ac[i]); +#else + ac[i] -= ac[i]*(.008f*i)*(.008f*i); +#endif + } + + _celt_lpc(lpc, ac, 4); + for (i=0;i<4;i++) + { + tmp = MULT16_16_Q15(QCONST16(.9f,15), tmp); + lpc[i] = MULT16_16_Q15(lpc[i], tmp); + } + /* Add a zero */ + lpc2[0] = lpc[0] + QCONST16(.8f,SIG_SHIFT); + lpc2[1] = lpc[1] + MULT16_16_Q15(c1,lpc[0]); + lpc2[2] = lpc[2] + MULT16_16_Q15(c1,lpc[1]); + lpc2[3] = lpc[3] + MULT16_16_Q15(c1,lpc[2]); + lpc2[4] = MULT16_16_Q15(c1,lpc[3]); + celt_fir5(x_lp, lpc2, len>>1); +} + +/* Pure C implementation. */ +#ifdef FIXED_POINT +opus_val32 +#else +void +#endif +celt_pitch_xcorr_c(const opus_val16 *_x, const opus_val16 *_y, + opus_val32 *xcorr, int len, int max_pitch, int arch) +{ + +#if 0 /* This is a simple version of the pitch correlation that should work + well on DSPs like Blackfin and TI C5x/C6x */ + int i, j; +#ifdef FIXED_POINT + opus_val32 maxcorr=1; +#endif +#if !defined(OVERRIDE_PITCH_XCORR) + (void)arch; +#endif + for (i=0;i0); + celt_sig_assert(((size_t)_x&3)==0); + for (i=0;i0); + celt_assert(max_pitch>0); + lag = len+max_pitch; + + ALLOC(x_lp4, len>>2, opus_val16); + ALLOC(y_lp4, lag>>2, opus_val16); + ALLOC(xcorr, max_pitch>>1, opus_val32); + + /* Downsample by 2 again */ + for (j=0;j>2;j++) + x_lp4[j] = x_lp[2*j]; + for (j=0;j>2;j++) + y_lp4[j] = y[2*j]; + +#ifdef FIXED_POINT + xmax = celt_maxabs16(x_lp4, len>>2); + ymax = celt_maxabs16(y_lp4, lag>>2); + shift = celt_ilog2(MAX32(1, MAX32(xmax, ymax)))-11; + if (shift>0) + { + for (j=0;j>2;j++) + x_lp4[j] = SHR16(x_lp4[j], shift); + for (j=0;j>2;j++) + y_lp4[j] = SHR16(y_lp4[j], shift); + /* Use double the shift for a MAC */ + shift *= 2; + } else { + shift = 0; + } +#endif + + /* Coarse search with 4x decimation */ + +#ifdef FIXED_POINT + maxcorr = +#endif + celt_pitch_xcorr(x_lp4, y_lp4, xcorr, len>>2, max_pitch>>2, arch); + + find_best_pitch(xcorr, y_lp4, len>>2, max_pitch>>2, best_pitch +#ifdef FIXED_POINT + , 0, maxcorr +#endif + ); + + /* Finer search with 2x decimation */ +#ifdef FIXED_POINT + maxcorr=1; +#endif + for (i=0;i>1;i++) + { + opus_val32 sum; + xcorr[i] = 0; + if (abs(i-2*best_pitch[0])>2 && abs(i-2*best_pitch[1])>2) + continue; +#ifdef FIXED_POINT + sum = 0; + for (j=0;j>1;j++) + sum += SHR32(MULT16_16(x_lp[j],y[i+j]), shift); +#else + sum = celt_inner_prod(x_lp, y+i, len>>1, arch); +#endif + xcorr[i] = MAX32(-1, sum); +#ifdef FIXED_POINT + maxcorr = MAX32(maxcorr, sum); +#endif + } + find_best_pitch(xcorr, y, len>>1, max_pitch>>1, best_pitch +#ifdef FIXED_POINT + , shift+1, maxcorr +#endif + ); + + /* Refine by pseudo-interpolation */ + if (best_pitch[0]>0 && best_pitch[0]<(max_pitch>>1)-1) + { + opus_val32 a, b, c; + a = xcorr[best_pitch[0]-1]; + b = xcorr[best_pitch[0]]; + c = xcorr[best_pitch[0]+1]; + if ((c-a) > MULT16_32_Q15(QCONST16(.7f,15),b-a)) + offset = 1; + else if ((a-c) > MULT16_32_Q15(QCONST16(.7f,15),b-c)) + offset = -1; + else + offset = 0; + } else { + offset = 0; + } + *pitch = 2*best_pitch[0]-offset; + + RESTORE_STACK; +} + +#ifdef FIXED_POINT +static opus_val16 compute_pitch_gain(opus_val32 xy, opus_val32 xx, opus_val32 yy) +{ + opus_val32 x2y2; + int sx, sy, shift; + opus_val32 g; + opus_val16 den; + if (xy == 0 || xx == 0 || yy == 0) + return 0; + sx = celt_ilog2(xx)-14; + sy = celt_ilog2(yy)-14; + shift = sx + sy; + x2y2 = SHR32(MULT16_16(VSHR32(xx, sx), VSHR32(yy, sy)), 14); + if (shift & 1) { + if (x2y2 < 32768) + { + x2y2 <<= 1; + shift--; + } else { + x2y2 >>= 1; + shift++; + } + } + den = celt_rsqrt_norm(x2y2); + g = MULT16_32_Q15(den, xy); + g = VSHR32(g, (shift>>1)-1); + return EXTRACT16(MIN32(g, Q15ONE)); +} +#else +static opus_val16 compute_pitch_gain(opus_val32 xy, opus_val32 xx, opus_val32 yy) +{ + return xy/celt_sqrt(1+xx*yy); +} +#endif + +static const int second_check[16] = {0, 0, 3, 2, 3, 2, 5, 2, 3, 2, 3, 2, 5, 2, 3, 2}; +opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod, + int N, int *T0_, int prev_period, opus_val16 prev_gain, int arch) +{ + int k, i, T, T0; + opus_val16 g, g0; + opus_val16 pg; + opus_val32 xy,xx,yy,xy2; + opus_val32 xcorr[3]; + opus_val32 best_xy, best_yy; + int offset; + int minperiod0; + VARDECL(opus_val32, yy_lookup); + SAVE_STACK; + + minperiod0 = minperiod; + maxperiod /= 2; + minperiod /= 2; + *T0_ /= 2; + prev_period /= 2; + N /= 2; + x += maxperiod; + if (*T0_>=maxperiod) + *T0_=maxperiod-1; + + T = T0 = *T0_; + ALLOC(yy_lookup, maxperiod+1, opus_val32); + dual_inner_prod(x, x, x-T0, N, &xx, &xy, arch); + yy_lookup[0] = xx; + yy=xx; + for (i=1;i<=maxperiod;i++) + { + yy = yy+MULT16_16(x[-i],x[-i])-MULT16_16(x[N-i],x[N-i]); + yy_lookup[i] = MAX32(0, yy); + } + yy = yy_lookup[T0]; + best_xy = xy; + best_yy = yy; + g = g0 = compute_pitch_gain(xy, xx, yy); + /* Look for any pitch at T/k */ + for (k=2;k<=15;k++) + { + int T1, T1b; + opus_val16 g1; + opus_val16 cont=0; + opus_val16 thresh; + T1 = celt_udiv(2*T0+k, 2*k); + if (T1 < minperiod) + break; + /* Look for another strong correlation at T1b */ + if (k==2) + { + if (T1+T0>maxperiod) + T1b = T0; + else + T1b = T0+T1; + } else + { + T1b = celt_udiv(2*second_check[k]*T0+k, 2*k); + } + dual_inner_prod(x, &x[-T1], &x[-T1b], N, &xy, &xy2, arch); + xy = HALF32(xy + xy2); + yy = HALF32(yy_lookup[T1] + yy_lookup[T1b]); + g1 = compute_pitch_gain(xy, xx, yy); + if (abs(T1-prev_period)<=1) + cont = prev_gain; + else if (abs(T1-prev_period)<=2 && 5*k*k < T0) + cont = HALF16(prev_gain); + else + cont = 0; + thresh = MAX16(QCONST16(.3f,15), MULT16_16_Q15(QCONST16(.7f,15),g0)-cont); + /* Bias against very high pitch (very short period) to avoid false-positives + due to short-term correlation */ + if (T1<3*minperiod) + thresh = MAX16(QCONST16(.4f,15), MULT16_16_Q15(QCONST16(.85f,15),g0)-cont); + else if (T1<2*minperiod) + thresh = MAX16(QCONST16(.5f,15), MULT16_16_Q15(QCONST16(.9f,15),g0)-cont); + if (g1 > thresh) + { + best_xy = xy; + best_yy = yy; + T = T1; + g = g1; + } + } + best_xy = MAX32(0, best_xy); + if (best_yy <= best_xy) + pg = Q15ONE; + else + pg = SHR32(frac_div32(best_xy,best_yy+1),16); + + for (k=0;k<3;k++) + xcorr[k] = celt_inner_prod(x, x-(T+k-1), N, arch); + if ((xcorr[2]-xcorr[0]) > MULT16_32_Q15(QCONST16(.7f,15),xcorr[1]-xcorr[0])) + offset = 1; + else if ((xcorr[0]-xcorr[2]) > MULT16_32_Q15(QCONST16(.7f,15),xcorr[1]-xcorr[2])) + offset = -1; + else + offset = 0; + if (pg > g) + pg = g; + *T0_ = 2*T+offset; + + if (*T0_=3); + y_3=0; /* gcc doesn't realize that y_3 can't be used uninitialized */ + y_0=*y++; + y_1=*y++; + y_2=*y++; + for (j=0;j +#include "os_support.h" +#include "arch.h" +#include "mathops.h" +#include "stack_alloc.h" +#include "rate.h" + +#ifdef FIXED_POINT +/* Mean energy in each band quantized in Q4 */ +const signed char eMeans[25] = { + 103,100, 92, 85, 81, + 77, 72, 70, 78, 75, + 73, 71, 78, 74, 69, + 72, 70, 74, 76, 71, + 60, 60, 60, 60, 60 +}; +#else +/* Mean energy in each band quantized in Q4 and converted back to float */ +const opus_val16 eMeans[25] = { + 6.437500f, 6.250000f, 5.750000f, 5.312500f, 5.062500f, + 4.812500f, 4.500000f, 4.375000f, 4.875000f, 4.687500f, + 4.562500f, 4.437500f, 4.875000f, 4.625000f, 4.312500f, + 4.500000f, 4.375000f, 4.625000f, 4.750000f, 4.437500f, + 3.750000f, 3.750000f, 3.750000f, 3.750000f, 3.750000f +}; +#endif +/* prediction coefficients: 0.9, 0.8, 0.65, 0.5 */ +#ifdef FIXED_POINT +static const opus_val16 pred_coef[4] = {29440, 26112, 21248, 16384}; +static const opus_val16 beta_coef[4] = {30147, 22282, 12124, 6554}; +static const opus_val16 beta_intra = 4915; +#else +static const opus_val16 pred_coef[4] = {29440/32768., 26112/32768., 21248/32768., 16384/32768.}; +static const opus_val16 beta_coef[4] = {30147/32768., 22282/32768., 12124/32768., 6554/32768.}; +static const opus_val16 beta_intra = 4915/32768.; +#endif + +/*Parameters of the Laplace-like probability models used for the coarse energy. + There is one pair of parameters for each frame size, prediction type + (inter/intra), and band number. + The first number of each pair is the probability of 0, and the second is the + decay rate, both in Q8 precision.*/ +static const unsigned char e_prob_model[4][2][42] = { + /*120 sample frames.*/ + { + /*Inter*/ + { + 72, 127, 65, 129, 66, 128, 65, 128, 64, 128, 62, 128, 64, 128, + 64, 128, 92, 78, 92, 79, 92, 78, 90, 79, 116, 41, 115, 40, + 114, 40, 132, 26, 132, 26, 145, 17, 161, 12, 176, 10, 177, 11 + }, + /*Intra*/ + { + 24, 179, 48, 138, 54, 135, 54, 132, 53, 134, 56, 133, 55, 132, + 55, 132, 61, 114, 70, 96, 74, 88, 75, 88, 87, 74, 89, 66, + 91, 67, 100, 59, 108, 50, 120, 40, 122, 37, 97, 43, 78, 50 + } + }, + /*240 sample frames.*/ + { + /*Inter*/ + { + 83, 78, 84, 81, 88, 75, 86, 74, 87, 71, 90, 73, 93, 74, + 93, 74, 109, 40, 114, 36, 117, 34, 117, 34, 143, 17, 145, 18, + 146, 19, 162, 12, 165, 10, 178, 7, 189, 6, 190, 8, 177, 9 + }, + /*Intra*/ + { + 23, 178, 54, 115, 63, 102, 66, 98, 69, 99, 74, 89, 71, 91, + 73, 91, 78, 89, 86, 80, 92, 66, 93, 64, 102, 59, 103, 60, + 104, 60, 117, 52, 123, 44, 138, 35, 133, 31, 97, 38, 77, 45 + } + }, + /*480 sample frames.*/ + { + /*Inter*/ + { + 61, 90, 93, 60, 105, 42, 107, 41, 110, 45, 116, 38, 113, 38, + 112, 38, 124, 26, 132, 27, 136, 19, 140, 20, 155, 14, 159, 16, + 158, 18, 170, 13, 177, 10, 187, 8, 192, 6, 175, 9, 159, 10 + }, + /*Intra*/ + { + 21, 178, 59, 110, 71, 86, 75, 85, 84, 83, 91, 66, 88, 73, + 87, 72, 92, 75, 98, 72, 105, 58, 107, 54, 115, 52, 114, 55, + 112, 56, 129, 51, 132, 40, 150, 33, 140, 29, 98, 35, 77, 42 + } + }, + /*960 sample frames.*/ + { + /*Inter*/ + { + 42, 121, 96, 66, 108, 43, 111, 40, 117, 44, 123, 32, 120, 36, + 119, 33, 127, 33, 134, 34, 139, 21, 147, 23, 152, 20, 158, 25, + 154, 26, 166, 21, 173, 16, 184, 13, 184, 10, 150, 13, 139, 15 + }, + /*Intra*/ + { + 22, 178, 63, 114, 74, 82, 84, 83, 92, 82, 103, 62, 96, 72, + 96, 67, 101, 73, 107, 72, 113, 55, 118, 52, 125, 52, 118, 52, + 117, 55, 135, 49, 137, 39, 157, 32, 145, 29, 97, 33, 77, 40 + } + } +}; + +static const unsigned char small_energy_icdf[3]={2,1,0}; + +static opus_val32 loss_distortion(const opus_val16 *eBands, opus_val16 *oldEBands, int start, int end, int len, int C) +{ + int c, i; + opus_val32 dist = 0; + c=0; do { + for (i=start;inbEBands]; + oldE = MAX16(-QCONST16(9.f,DB_SHIFT), oldEBands[i+c*m->nbEBands]); +#ifdef FIXED_POINT + f = SHL32(EXTEND32(x),7) - PSHR32(MULT16_16(coef,oldE), 8) - prev[c]; + /* Rounding to nearest integer here is really important! */ + qi = (f+QCONST32(.5f,DB_SHIFT+7))>>(DB_SHIFT+7); + decay_bound = EXTRACT16(MAX32(-QCONST16(28.f,DB_SHIFT), + SUB32((opus_val32)oldEBands[i+c*m->nbEBands],max_decay))); +#else + f = x-coef*oldE-prev[c]; + /* Rounding to nearest integer here is really important! */ + qi = (int)floor(.5f+f); + decay_bound = MAX16(-QCONST16(28.f,DB_SHIFT), oldEBands[i+c*m->nbEBands]) - max_decay; +#endif + /* Prevent the energy from going down too quickly (e.g. for bands + that have just one bin) */ + if (qi < 0 && x < decay_bound) + { + qi += (int)SHR16(SUB16(decay_bound,x), DB_SHIFT); + if (qi > 0) + qi = 0; + } + qi0 = qi; + /* If we don't have enough bits to encode all the energy, just assume + something safe. */ + tell = ec_tell(enc); + bits_left = budget-tell-3*C*(end-i); + if (i!=start && bits_left < 30) + { + if (bits_left < 24) + qi = IMIN(1, qi); + if (bits_left < 16) + qi = IMAX(-1, qi); + } + if (lfe && i>=2) + qi = IMIN(qi, 0); + if (budget-tell >= 15) + { + int pi; + pi = 2*IMIN(i,20); + ec_laplace_encode(enc, &qi, + prob_model[pi]<<7, prob_model[pi+1]<<6); + } + else if(budget-tell >= 2) + { + qi = IMAX(-1, IMIN(qi, 1)); + ec_enc_icdf(enc, 2*qi^-(qi<0), small_energy_icdf, 2); + } + else if(budget-tell >= 1) + { + qi = IMIN(0, qi); + ec_enc_bit_logp(enc, -qi, 1); + } + else + qi = -1; + error[i+c*m->nbEBands] = PSHR32(f,7) - SHL16(qi,DB_SHIFT); + badness += abs(qi0-qi); + q = (opus_val32)SHL32(EXTEND32(qi),DB_SHIFT); + + tmp = PSHR32(MULT16_16(coef,oldE),8) + prev[c] + SHL32(q,7); +#ifdef FIXED_POINT + tmp = MAX32(-QCONST32(28.f, DB_SHIFT+7), tmp); +#endif + oldEBands[i+c*m->nbEBands] = PSHR32(tmp, 7); + prev[c] = prev[c] + SHL32(q,7) - MULT16_16(beta,PSHR32(q,8)); + } while (++c < C); + } + return lfe ? 0 : badness; +} + +void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, + const opus_val16 *eBands, opus_val16 *oldEBands, opus_uint32 budget, + opus_val16 *error, ec_enc *enc, int C, int LM, int nbAvailableBytes, + int force_intra, opus_val32 *delayedIntra, int two_pass, int loss_rate, int lfe) +{ + int intra; + opus_val16 max_decay; + VARDECL(opus_val16, oldEBands_intra); + VARDECL(opus_val16, error_intra); + ec_enc enc_start_state; + opus_uint32 tell; + int badness1=0; + opus_int32 intra_bias; + opus_val32 new_distortion; + SAVE_STACK; + + intra = force_intra || (!two_pass && *delayedIntra>2*C*(end-start) && nbAvailableBytes > (end-start)*C); + intra_bias = (opus_int32)((budget**delayedIntra*loss_rate)/(C*512)); + new_distortion = loss_distortion(eBands, oldEBands, start, effEnd, m->nbEBands, C); + + tell = ec_tell(enc); + if (tell+3 > budget) + two_pass = intra = 0; + + max_decay = QCONST16(16.f,DB_SHIFT); + if (end-start>10) + { +#ifdef FIXED_POINT + max_decay = MIN32(max_decay, SHL32(EXTEND32(nbAvailableBytes),DB_SHIFT-3)); +#else + max_decay = MIN32(max_decay, .125f*nbAvailableBytes); +#endif + } + if (lfe) + max_decay = QCONST16(3.f,DB_SHIFT); + enc_start_state = *enc; + + ALLOC(oldEBands_intra, C*m->nbEBands, opus_val16); + ALLOC(error_intra, C*m->nbEBands, opus_val16); + OPUS_COPY(oldEBands_intra, oldEBands, C*m->nbEBands); + + if (two_pass || intra) + { + badness1 = quant_coarse_energy_impl(m, start, end, eBands, oldEBands_intra, budget, + tell, e_prob_model[LM][1], error_intra, enc, C, LM, 1, max_decay, lfe); + } + + if (!intra) + { + unsigned char *intra_buf; + ec_enc enc_intra_state; + opus_int32 tell_intra; + opus_uint32 nstart_bytes; + opus_uint32 nintra_bytes; + opus_uint32 save_bytes; + int badness2; + VARDECL(unsigned char, intra_bits); + + tell_intra = ec_tell_frac(enc); + + enc_intra_state = *enc; + + nstart_bytes = ec_range_bytes(&enc_start_state); + nintra_bytes = ec_range_bytes(&enc_intra_state); + intra_buf = ec_get_buffer(&enc_intra_state) + nstart_bytes; + save_bytes = nintra_bytes-nstart_bytes; + if (save_bytes == 0) + save_bytes = ALLOC_NONE; + ALLOC(intra_bits, save_bytes, unsigned char); + /* Copy bits from intra bit-stream */ + OPUS_COPY(intra_bits, intra_buf, nintra_bytes - nstart_bytes); + + *enc = enc_start_state; + + badness2 = quant_coarse_energy_impl(m, start, end, eBands, oldEBands, budget, + tell, e_prob_model[LM][intra], error, enc, C, LM, 0, max_decay, lfe); + + if (two_pass && (badness1 < badness2 || (badness1 == badness2 && ((opus_int32)ec_tell_frac(enc))+intra_bias > tell_intra))) + { + *enc = enc_intra_state; + /* Copy intra bits to bit-stream */ + OPUS_COPY(intra_buf, intra_bits, nintra_bytes - nstart_bytes); + OPUS_COPY(oldEBands, oldEBands_intra, C*m->nbEBands); + OPUS_COPY(error, error_intra, C*m->nbEBands); + intra = 1; + } + } else { + OPUS_COPY(oldEBands, oldEBands_intra, C*m->nbEBands); + OPUS_COPY(error, error_intra, C*m->nbEBands); + } + + if (intra) + *delayedIntra = new_distortion; + else + *delayedIntra = ADD32(MULT16_32_Q15(MULT16_16_Q15(pred_coef[LM], pred_coef[LM]),*delayedIntra), + new_distortion); + + RESTORE_STACK; +} + +void quant_fine_energy(const CELTMode *m, int start, int end, opus_val16 *oldEBands, opus_val16 *error, int *fine_quant, ec_enc *enc, int C) +{ + int i, c; + + /* Encode finer resolution */ + for (i=start;inbEBands]+QCONST16(.5f,DB_SHIFT))>>(DB_SHIFT-fine_quant[i]); +#else + q2 = (int)floor((error[i+c*m->nbEBands]+.5f)*frac); +#endif + if (q2 > frac-1) + q2 = frac-1; + if (q2<0) + q2 = 0; + ec_enc_bits(enc, q2, fine_quant[i]); +#ifdef FIXED_POINT + offset = SUB16(SHR32(SHL32(EXTEND32(q2),DB_SHIFT)+QCONST16(.5f,DB_SHIFT),fine_quant[i]),QCONST16(.5f,DB_SHIFT)); +#else + offset = (q2+.5f)*(1<<(14-fine_quant[i]))*(1.f/16384) - .5f; +#endif + oldEBands[i+c*m->nbEBands] += offset; + error[i+c*m->nbEBands] -= offset; + /*printf ("%f ", error[i] - offset);*/ + } while (++c < C); + } +} + +void quant_energy_finalise(const CELTMode *m, int start, int end, opus_val16 *oldEBands, opus_val16 *error, int *fine_quant, int *fine_priority, int bits_left, ec_enc *enc, int C) +{ + int i, prio, c; + + /* Use up the remaining bits */ + for (prio=0;prio<2;prio++) + { + for (i=start;i=C ;i++) + { + if (fine_quant[i] >= MAX_FINE_BITS || fine_priority[i]!=prio) + continue; + c=0; + do { + int q2; + opus_val16 offset; + q2 = error[i+c*m->nbEBands]<0 ? 0 : 1; + ec_enc_bits(enc, q2, 1); +#ifdef FIXED_POINT + offset = SHR16(SHL16(q2,DB_SHIFT)-QCONST16(.5f,DB_SHIFT),fine_quant[i]+1); +#else + offset = (q2-.5f)*(1<<(14-fine_quant[i]-1))*(1.f/16384); +#endif + oldEBands[i+c*m->nbEBands] += offset; + error[i+c*m->nbEBands] -= offset; + bits_left--; + } while (++c < C); + } + } +} + +void unquant_coarse_energy(const CELTMode *m, int start, int end, opus_val16 *oldEBands, int intra, ec_dec *dec, int C, int LM) +{ + const unsigned char *prob_model = e_prob_model[LM][intra]; + int i, c; + opus_val32 prev[2] = {0, 0}; + opus_val16 coef; + opus_val16 beta; + opus_int32 budget; + opus_int32 tell; + + if (intra) + { + coef = 0; + beta = beta_intra; + } else { + beta = beta_coef[LM]; + coef = pred_coef[LM]; + } + + budget = dec->storage*8; + + /* Decode at a fixed coarse resolution */ + for (i=start;i=15) + { + int pi; + pi = 2*IMIN(i,20); + qi = ec_laplace_decode(dec, + prob_model[pi]<<7, prob_model[pi+1]<<6); + } + else if(budget-tell>=2) + { + qi = ec_dec_icdf(dec, small_energy_icdf, 2); + qi = (qi>>1)^-(qi&1); + } + else if(budget-tell>=1) + { + qi = -ec_dec_bit_logp(dec, 1); + } + else + qi = -1; + q = (opus_val32)SHL32(EXTEND32(qi),DB_SHIFT); + + oldEBands[i+c*m->nbEBands] = MAX16(-QCONST16(9.f,DB_SHIFT), oldEBands[i+c*m->nbEBands]); + tmp = PSHR32(MULT16_16(coef,oldEBands[i+c*m->nbEBands]),8) + prev[c] + SHL32(q,7); +#ifdef FIXED_POINT + tmp = MAX32(-QCONST32(28.f, DB_SHIFT+7), tmp); +#endif + oldEBands[i+c*m->nbEBands] = PSHR32(tmp, 7); + prev[c] = prev[c] + SHL32(q,7) - MULT16_16(beta,PSHR32(q,8)); + } while (++c < C); + } +} + +void unquant_fine_energy(const CELTMode *m, int start, int end, opus_val16 *oldEBands, int *fine_quant, ec_dec *dec, int C) +{ + int i, c; + /* Decode finer resolution */ + for (i=start;inbEBands] += offset; + } while (++c < C); + } +} + +void unquant_energy_finalise(const CELTMode *m, int start, int end, opus_val16 *oldEBands, int *fine_quant, int *fine_priority, int bits_left, ec_dec *dec, int C) +{ + int i, prio, c; + + /* Use up the remaining bits */ + for (prio=0;prio<2;prio++) + { + for (i=start;i=C ;i++) + { + if (fine_quant[i] >= MAX_FINE_BITS || fine_priority[i]!=prio) + continue; + c=0; + do { + int q2; + opus_val16 offset; + q2 = ec_dec_bits(dec, 1); +#ifdef FIXED_POINT + offset = SHR16(SHL16(q2,DB_SHIFT)-QCONST16(.5f,DB_SHIFT),fine_quant[i]+1); +#else + offset = (q2-.5f)*(1<<(14-fine_quant[i]-1))*(1.f/16384); +#endif + oldEBands[i+c*m->nbEBands] += offset; + bits_left--; + } while (++c < C); + } + } +} + +void amp2Log2(const CELTMode *m, int effEnd, int end, + celt_ener *bandE, opus_val16 *bandLogE, int C) +{ + int c, i; + c=0; + do { + for (i=0;inbEBands] = + celt_log2(bandE[i+c*m->nbEBands]) + - SHL16((opus_val16)eMeans[i],6); +#ifdef FIXED_POINT + /* Compensate for bandE[] being Q12 but celt_log2() taking a Q14 input. */ + bandLogE[i+c*m->nbEBands] += QCONST16(2.f, DB_SHIFT); +#endif + } + for (i=effEnd;inbEBands+i] = -QCONST16(14.f,DB_SHIFT); + } while (++c < C); +} diff --git a/lib/opus/celt/quant_bands.h b/lib/opus/celt/quant_bands.h new file mode 100644 index 00000000..0490bca4 --- /dev/null +++ b/lib/opus/celt/quant_bands.h @@ -0,0 +1,66 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef QUANT_BANDS +#define QUANT_BANDS + +#include "arch.h" +#include "modes.h" +#include "entenc.h" +#include "entdec.h" +#include "mathops.h" + +#ifdef FIXED_POINT +extern const signed char eMeans[25]; +#else +extern const opus_val16 eMeans[25]; +#endif + +void amp2Log2(const CELTMode *m, int effEnd, int end, + celt_ener *bandE, opus_val16 *bandLogE, int C); + +void log2Amp(const CELTMode *m, int start, int end, + celt_ener *eBands, const opus_val16 *oldEBands, int C); + +void quant_coarse_energy(const CELTMode *m, int start, int end, int effEnd, + const opus_val16 *eBands, opus_val16 *oldEBands, opus_uint32 budget, + opus_val16 *error, ec_enc *enc, int C, int LM, + int nbAvailableBytes, int force_intra, opus_val32 *delayedIntra, + int two_pass, int loss_rate, int lfe); + +void quant_fine_energy(const CELTMode *m, int start, int end, opus_val16 *oldEBands, opus_val16 *error, int *fine_quant, ec_enc *enc, int C); + +void quant_energy_finalise(const CELTMode *m, int start, int end, opus_val16 *oldEBands, opus_val16 *error, int *fine_quant, int *fine_priority, int bits_left, ec_enc *enc, int C); + +void unquant_coarse_energy(const CELTMode *m, int start, int end, opus_val16 *oldEBands, int intra, ec_dec *dec, int C, int LM); + +void unquant_fine_energy(const CELTMode *m, int start, int end, opus_val16 *oldEBands, int *fine_quant, ec_dec *dec, int C); + +void unquant_energy_finalise(const CELTMode *m, int start, int end, opus_val16 *oldEBands, int *fine_quant, int *fine_priority, int bits_left, ec_dec *dec, int C); + +#endif /* QUANT_BANDS */ diff --git a/lib/opus/celt/rate.c b/lib/opus/celt/rate.c new file mode 100644 index 00000000..7f7ad3fa --- /dev/null +++ b/lib/opus/celt/rate.c @@ -0,0 +1,646 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "modes.h" +#include "cwrs.h" +#include "arch.h" +#include "os_support.h" + +#include "entcode.h" +#include "rate.h" + +static const unsigned char LOG2_FRAC_TABLE[24]={ + 0, + 8,13, + 16,19,21,23, + 24,26,27,28,29,30,31,32, + 32,33,34,34,35,36,36,37,37 +}; + +#ifdef CUSTOM_MODES + +/*Determines if V(N,K) fits in a 32-bit unsigned integer. + N and K are themselves limited to 15 bits.*/ +static int fits_in32(int _n, int _k) +{ + static const opus_int16 maxN[15] = { + 32767, 32767, 32767, 1476, 283, 109, 60, 40, + 29, 24, 20, 18, 16, 14, 13}; + static const opus_int16 maxK[15] = { + 32767, 32767, 32767, 32767, 1172, 238, 95, 53, + 36, 27, 22, 18, 16, 15, 13}; + if (_n>=14) + { + if (_k>=14) + return 0; + else + return _n <= maxN[_k]; + } else { + return _k <= maxK[_n]; + } +} + +void compute_pulse_cache(CELTMode *m, int LM) +{ + int C; + int i; + int j; + int curr=0; + int nbEntries=0; + int entryN[100], entryK[100], entryI[100]; + const opus_int16 *eBands = m->eBands; + PulseCache *cache = &m->cache; + opus_int16 *cindex; + unsigned char *bits; + unsigned char *cap; + + cindex = (opus_int16 *)opus_alloc(sizeof(cache->index[0])*m->nbEBands*(LM+2)); + cache->index = cindex; + + /* Scan for all unique band sizes */ + for (i=0;i<=LM+1;i++) + { + for (j=0;jnbEBands;j++) + { + int k; + int N = (eBands[j+1]-eBands[j])<>1; + cindex[i*m->nbEBands+j] = -1; + /* Find other bands that have the same size */ + for (k=0;k<=i;k++) + { + int n; + for (n=0;nnbEBands && (k!=i || n>1) + { + cindex[i*m->nbEBands+j] = cindex[k*m->nbEBands+n]; + break; + } + } + } + if (cache->index[i*m->nbEBands+j] == -1 && N!=0) + { + int K; + entryN[nbEntries] = N; + K = 0; + while (fits_in32(N,get_pulses(K+1)) && KnbEBands+j] = curr; + entryI[nbEntries] = curr; + + curr += K+1; + nbEntries++; + } + } + } + bits = (unsigned char *)opus_alloc(sizeof(unsigned char)*curr); + cache->bits = bits; + cache->size = curr; + /* Compute the cache for all unique sizes */ + for (i=0;icaps = cap = (unsigned char *)opus_alloc(sizeof(cache->caps[0])*(LM+1)*2*m->nbEBands); + for (i=0;i<=LM;i++) + { + for (C=1;C<=2;C++) + { + for (j=0;jnbEBands;j++) + { + int N0; + int max_bits; + N0 = m->eBands[j+1]-m->eBands[j]; + /* N=1 bands only have a sign bit and fine bits. */ + if (N0<1 are even, including custom modes.*/ + if (N0 > 2) + { + N0>>=1; + LM0--; + } + /* N0=1 bands can't be split down to N<2. */ + else if (N0 <= 1) + { + LM0=IMIN(i,1); + N0<<=LM0; + } + /* Compute the cost for the lowest-level PVQ of a fully split + band. */ + pcache = bits + cindex[(LM0+1)*m->nbEBands+j]; + max_bits = pcache[pcache[0]]+1; + /* Add in the cost of coding regular splits. */ + N = N0; + for(k=0;klogN[j]+((LM0+k)<>1)-QTHETA_OFFSET; + /* The number of qtheta bits we'll allocate if the remainder + is to be max_bits. + The average measured cost for theta is 0.89701 times qb, + approximated here as 459/512. */ + num=459*(opus_int32)((2*N-1)*offset+max_bits); + den=((opus_int32)(2*N-1)<<9)-459; + qb = IMIN((num+(den>>1))/den, 57); + celt_assert(qb >= 0); + max_bits += qb; + N <<= 1; + } + /* Add in the cost of a stereo split, if necessary. */ + if (C==2) + { + max_bits <<= 1; + offset = ((m->logN[j]+(i<>1)-(N==2?QTHETA_OFFSET_TWOPHASE:QTHETA_OFFSET); + ndof = 2*N-1-(N==2); + /* The average measured cost for theta with the step PDF is + 0.95164 times qb, approximated here as 487/512. */ + num = (N==2?512:487)*(opus_int32)(max_bits+ndof*offset); + den = ((opus_int32)ndof<<9)-(N==2?512:487); + qb = IMIN((num+(den>>1))/den, (N==2?64:61)); + celt_assert(qb >= 0); + max_bits += qb; + } + /* Add the fine bits we'll use. */ + /* Compensate for the extra DoF in stereo */ + ndof = C*N + ((C==2 && N>2) ? 1 : 0); + /* Offset the number of fine bits by log2(N)/2 + FINE_OFFSET + compared to their "fair share" of total/N */ + offset = ((m->logN[j] + (i<>1)-FINE_OFFSET; + /* N=2 is the only point that doesn't match the curve */ + if (N==2) + offset += 1<>2; + /* The number of fine bits we'll allocate if the remainder is + to be max_bits. */ + num = max_bits+ndof*offset; + den = (ndof-1)<>1))/den, MAX_FINE_BITS); + celt_assert(qb >= 0); + max_bits += C*qb<eBands[j+1]-m->eBands[j])<= 0); + celt_assert(max_bits < 256); + *cap++ = (unsigned char)max_bits; + } + } + } +} + +#endif /* CUSTOM_MODES */ + +#define ALLOC_STEPS 6 + +static OPUS_INLINE int interp_bits2pulses(const CELTMode *m, int start, int end, int skip_start, + const int *bits1, const int *bits2, const int *thresh, const int *cap, opus_int32 total, opus_int32 *_balance, + int skip_rsv, int *intensity, int intensity_rsv, int *dual_stereo, int dual_stereo_rsv, int *bits, + int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev, int signalBandwidth) +{ + opus_int32 psum; + int lo, hi; + int i, j; + int logM; + int stereo; + int codedBands=-1; + int alloc_floor; + opus_int32 left, percoeff; + int done; + opus_int32 balance; + SAVE_STACK; + + alloc_floor = C<1; + + logM = LM<>1; + psum = 0; + done = 0; + for (j=end;j-->start;) + { + int tmp = bits1[j] + (mid*(opus_int32)bits2[j]>>ALLOC_STEPS); + if (tmp >= thresh[j] || done) + { + done = 1; + /* Don't allocate more than we can actually use */ + psum += IMIN(tmp, cap[j]); + } else { + if (tmp >= alloc_floor) + psum += alloc_floor; + } + } + if (psum > total) + hi = mid; + else + lo = mid; + } + psum = 0; + /*printf ("interp bisection gave %d\n", lo);*/ + done = 0; + for (j=end;j-->start;) + { + int tmp = bits1[j] + ((opus_int32)lo*bits2[j]>>ALLOC_STEPS); + if (tmp < thresh[j] && !done) + { + if (tmp >= alloc_floor) + tmp = alloc_floor; + else + tmp = 0; + } else + done = 1; + /* Don't allocate more than we can actually use */ + tmp = IMIN(tmp, cap[j]); + bits[j] = tmp; + psum += tmp; + } + + /* Decide which bands to skip, working backwards from the end. */ + for (codedBands=end;;codedBands--) + { + int band_width; + int band_bits; + int rem; + j = codedBands-1; + /* Never skip the first band, nor a band that has been boosted by + dynalloc. + In the first case, we'd be coding a bit to signal we're going to waste + all the other bits. + In the second case, we'd be coding a bit to redistribute all the bits + we just signaled should be cocentrated in this band. */ + if (j<=skip_start) + { + /* Give the bit we reserved to end skipping back. */ + total += skip_rsv; + break; + } + /*Figure out how many left-over bits we would be adding to this band. + This can include bits we've stolen back from higher, skipped bands.*/ + left = total-psum; + percoeff = celt_udiv(left, m->eBands[codedBands]-m->eBands[start]); + left -= (m->eBands[codedBands]-m->eBands[start])*percoeff; + rem = IMAX(left-(m->eBands[j]-m->eBands[start]),0); + band_width = m->eBands[codedBands]-m->eBands[j]; + band_bits = (int)(bits[j] + percoeff*band_width + rem); + /*Only code a skip decision if we're above the threshold for this band. + Otherwise it is force-skipped. + This ensures that we have enough bits to code the skip flag.*/ + if (band_bits >= IMAX(thresh[j], alloc_floor+(1< 17) + depth_threshold = j (depth_threshold*band_width<>4 && j<=signalBandwidth)) +#endif + { + ec_enc_bit_logp(ec, 1, 1); + break; + } + ec_enc_bit_logp(ec, 0, 1); + } else if (ec_dec_bit_logp(ec, 1)) { + break; + } + /*We used a bit to skip this band.*/ + psum += 1< 0) + intensity_rsv = LOG2_FRAC_TABLE[j-start]; + psum += intensity_rsv; + if (band_bits >= alloc_floor) + { + /*If we have enough for a fine energy bit per channel, use it.*/ + psum += alloc_floor; + bits[j] = alloc_floor; + } else { + /*Otherwise this band gets nothing at all.*/ + bits[j] = 0; + } + } + + celt_assert(codedBands > start); + /* Code the intensity and dual stereo parameters. */ + if (intensity_rsv > 0) + { + if (encode) + { + *intensity = IMIN(*intensity, codedBands); + ec_enc_uint(ec, *intensity-start, codedBands+1-start); + } + else + *intensity = start+ec_dec_uint(ec, codedBands+1-start); + } + else + *intensity = 0; + if (*intensity <= start) + { + total += dual_stereo_rsv; + dual_stereo_rsv = 0; + } + if (dual_stereo_rsv > 0) + { + if (encode) + ec_enc_bit_logp(ec, *dual_stereo, 1); + else + *dual_stereo = ec_dec_bit_logp(ec, 1); + } + else + *dual_stereo = 0; + + /* Allocate the remaining bits */ + left = total-psum; + percoeff = celt_udiv(left, m->eBands[codedBands]-m->eBands[start]); + left -= (m->eBands[codedBands]-m->eBands[start])*percoeff; + for (j=start;jeBands[j+1]-m->eBands[j])); + for (j=start;jeBands[j+1]-m->eBands[j]); + bits[j] += tmp; + left -= tmp; + } + /*for (j=0;j= 0); + N0 = m->eBands[j+1]-m->eBands[j]; + N=N0<1) + { + excess = MAX32(bit-cap[j],0); + bits[j] = bit-excess; + + /* Compensate for the extra DoF in stereo */ + den=(C*N+ ((C==2 && N>2 && !*dual_stereo && j<*intensity) ? 1 : 0)); + + NClogN = den*(m->logN[j] + logM); + + /* Offset for the number of fine bits by log2(N)/2 + FINE_OFFSET + compared to their "fair share" of total/N */ + offset = (NClogN>>1)-den*FINE_OFFSET; + + /* N=2 is the only point that doesn't match the curve */ + if (N==2) + offset += den<>2; + + /* Changing the offset for allocating the second and third + fine energy bit */ + if (bits[j] + offset < den*2<>2; + else if (bits[j] + offset < den*3<>3; + + /* Divide with rounding */ + ebits[j] = IMAX(0, (bits[j] + offset + (den<<(BITRES-1)))); + ebits[j] = celt_udiv(ebits[j], den)>>BITRES; + + /* Make sure not to bust */ + if (C*ebits[j] > (bits[j]>>BITRES)) + ebits[j] = bits[j] >> stereo >> BITRES; + + /* More than that is useless because that's about as far as PVQ can go */ + ebits[j] = IMIN(ebits[j], MAX_FINE_BITS); + + /* If we rounded down or capped this band, make it a candidate for the + final fine energy pass */ + fine_priority[j] = ebits[j]*(den<= bits[j]+offset; + + /* Remove the allocated fine bits; the rest are assigned to PVQ */ + bits[j] -= C*ebits[j]< 0) + { + int extra_fine; + int extra_bits; + extra_fine = IMIN(excess>>(stereo+BITRES),MAX_FINE_BITS-ebits[j]); + ebits[j] += extra_fine; + extra_bits = extra_fine*C<= excess-balance; + excess -= extra_bits; + } + balance = excess; + + celt_assert(bits[j] >= 0); + celt_assert(ebits[j] >= 0); + } + /* Save any remaining bits over the cap for the rebalancing in + quant_all_bands(). */ + *_balance = balance; + + /* The skipped bands use all their bits for fine energy. */ + for (;j> stereo >> BITRES; + celt_assert(C*ebits[j]<nbEBands; + skip_start = start; + /* Reserve a bit to signal the end of manually skipped bands. */ + skip_rsv = total >= 1<total) + intensity_rsv = 0; + else + { + total -= intensity_rsv; + dual_stereo_rsv = total>=1<eBands[j+1]-m->eBands[j])<>4); + /* Tilt of the allocation curve */ + trim_offset[j] = C*(m->eBands[j+1]-m->eBands[j])*(alloc_trim-5-LM)*(end-j-1) + *(1<<(LM+BITRES))>>6; + /* Giving less resolution to single-coefficient bands because they get + more benefit from having one coarse value per coefficient*/ + if ((m->eBands[j+1]-m->eBands[j])<nbAllocVectors - 1; + do + { + int done = 0; + int psum = 0; + int mid = (lo+hi) >> 1; + for (j=end;j-->start;) + { + int bitsj; + int N = m->eBands[j+1]-m->eBands[j]; + bitsj = C*N*m->allocVectors[mid*len+j]<>2; + if (bitsj > 0) + bitsj = IMAX(0, bitsj + trim_offset[j]); + bitsj += offsets[j]; + if (bitsj >= thresh[j] || done) + { + done = 1; + /* Don't allocate more than we can actually use */ + psum += IMIN(bitsj, cap[j]); + } else { + if (bitsj >= C< total) + hi = mid - 1; + else + lo = mid + 1; + /*printf ("lo = %d, hi = %d\n", lo, hi);*/ + } + while (lo <= hi); + hi = lo--; + /*printf ("interp between %d and %d\n", lo, hi);*/ + for (j=start;jeBands[j+1]-m->eBands[j]; + bits1j = C*N*m->allocVectors[lo*len+j]<>2; + bits2j = hi>=m->nbAllocVectors ? + cap[j] : C*N*m->allocVectors[hi*len+j]<>2; + if (bits1j > 0) + bits1j = IMAX(0, bits1j + trim_offset[j]); + if (bits2j > 0) + bits2j = IMAX(0, bits2j + trim_offset[j]); + if (lo > 0) + bits1j += offsets[j]; + bits2j += offsets[j]; + if (offsets[j]>0) + skip_start = j; + bits2j = IMAX(0,bits2j-bits1j); + bits1[j] = bits1j; + bits2[j] = bits2j; + } + codedBands = interp_bits2pulses(m, start, end, skip_start, bits1, bits2, thresh, cap, + total, balance, skip_rsv, intensity, intensity_rsv, dual_stereo, dual_stereo_rsv, + pulses, ebits, fine_priority, C, LM, ec, encode, prev, signalBandwidth); + RESTORE_STACK; + return codedBands; +} + diff --git a/lib/opus/celt/rate.h b/lib/opus/celt/rate.h new file mode 100644 index 00000000..fad5e412 --- /dev/null +++ b/lib/opus/celt/rate.h @@ -0,0 +1,101 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef RATE_H +#define RATE_H + +#define MAX_PSEUDO 40 +#define LOG_MAX_PSEUDO 6 + +#define CELT_MAX_PULSES 128 + +#define MAX_FINE_BITS 8 + +#define FINE_OFFSET 21 +#define QTHETA_OFFSET 4 +#define QTHETA_OFFSET_TWOPHASE 16 + +#include "cwrs.h" +#include "modes.h" + +void compute_pulse_cache(CELTMode *m, int LM); + +static OPUS_INLINE int get_pulses(int i) +{ + return i<8 ? i : (8 + (i&7)) << ((i>>3)-1); +} + +static OPUS_INLINE int bits2pulses(const CELTMode *m, int band, int LM, int bits) +{ + int i; + int lo, hi; + const unsigned char *cache; + + LM++; + cache = m->cache.bits + m->cache.index[LM*m->nbEBands+band]; + + lo = 0; + hi = cache[0]; + bits--; + for (i=0;i>1; + /* OPT: Make sure this is implemented with a conditional move */ + if ((int)cache[mid] >= bits) + hi = mid; + else + lo = mid; + } + if (bits- (lo == 0 ? -1 : (int)cache[lo]) <= (int)cache[hi]-bits) + return lo; + else + return hi; +} + +static OPUS_INLINE int pulses2bits(const CELTMode *m, int band, int LM, int pulses) +{ + const unsigned char *cache; + + LM++; + cache = m->cache.bits + m->cache.index[LM*m->nbEBands+band]; + return pulses == 0 ? 0 : cache[pulses]+1; +} + +/** Compute the pulse allocation, i.e. how many pulses will go in each + * band. + @param m mode + @param offsets Requested increase or decrease in the number of bits for + each band + @param total Number of bands + @param pulses Number of pulses per band (returned) + @return Total number of bits allocated +*/ +int clt_compute_allocation(const CELTMode *m, int start, int end, const int *offsets, const int *cap, int alloc_trim, int *intensity, int *dual_stereo, + opus_int32 total, opus_int32 *balance, int *pulses, int *ebits, int *fine_priority, int C, int LM, ec_ctx *ec, int encode, int prev, int signalBandwidth); + +#endif diff --git a/lib/opus/celt/stack_alloc.h b/lib/opus/celt/stack_alloc.h new file mode 100644 index 00000000..ae40e2a1 --- /dev/null +++ b/lib/opus/celt/stack_alloc.h @@ -0,0 +1,184 @@ +/* Copyright (C) 2002-2003 Jean-Marc Valin + Copyright (C) 2007-2009 Xiph.Org Foundation */ +/** + @file stack_alloc.h + @brief Temporary memory allocation on stack +*/ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef STACK_ALLOC_H +#define STACK_ALLOC_H + +#include "opus_types.h" +#include "opus_defines.h" + +#if (!defined (VAR_ARRAYS) && !defined (USE_ALLOCA) && !defined (NONTHREADSAFE_PSEUDOSTACK)) +#error "Opus requires one of VAR_ARRAYS, USE_ALLOCA, or NONTHREADSAFE_PSEUDOSTACK be defined to select the temporary allocation mode." +#endif + +#ifdef USE_ALLOCA +# ifdef _WIN32 +# include +# else +# ifdef HAVE_ALLOCA_H +# include +# else +# include +# endif +# endif +#endif + +/** + * @def ALIGN(stack, size) + * + * Aligns the stack to a 'size' boundary + * + * @param stack Stack + * @param size New size boundary + */ + +/** + * @def PUSH(stack, size, type) + * + * Allocates 'size' elements of type 'type' on the stack + * + * @param stack Stack + * @param size Number of elements + * @param type Type of element + */ + +/** + * @def VARDECL(var) + * + * Declare variable on stack + * + * @param var Variable to declare + */ + +/** + * @def ALLOC(var, size, type) + * + * Allocate 'size' elements of 'type' on stack + * + * @param var Name of variable to allocate + * @param size Number of elements + * @param type Type of element + */ + +#if defined(VAR_ARRAYS) + +#define VARDECL(type, var) +#define ALLOC(var, size, type) type var[size] +#define SAVE_STACK +#define RESTORE_STACK +#define ALLOC_STACK +/* C99 does not allow VLAs of size zero */ +#define ALLOC_NONE 1 + +#elif defined(USE_ALLOCA) + +#define VARDECL(type, var) type *var + +# ifdef _WIN32 +# define ALLOC(var, size, type) var = ((type*)_alloca(sizeof(type)*(size))) +# else +# define ALLOC(var, size, type) var = ((type*)alloca(sizeof(type)*(size))) +# endif + +#define SAVE_STACK +#define RESTORE_STACK +#define ALLOC_STACK +#define ALLOC_NONE 0 + +#else + +#ifdef CELT_C +char *scratch_ptr=0; +char *global_stack=0; +#else +extern char *global_stack; +extern char *scratch_ptr; +#endif /* CELT_C */ + +#ifdef ENABLE_VALGRIND + +#include + +#ifdef CELT_C +char *global_stack_top=0; +#else +extern char *global_stack_top; +#endif /* CELT_C */ + +#define ALIGN(stack, size) ((stack) += ((size) - (long)(stack)) & ((size) - 1)) +#define PUSH(stack, size, type) (VALGRIND_MAKE_MEM_NOACCESS(stack, global_stack_top-stack),ALIGN((stack),sizeof(type)/sizeof(char)),VALGRIND_MAKE_MEM_UNDEFINED(stack, ((size)*sizeof(type)/sizeof(char))),(stack)+=(2*(size)*sizeof(type)/sizeof(char)),(type*)((stack)-(2*(size)*sizeof(type)/sizeof(char)))) +#define RESTORE_STACK ((global_stack = _saved_stack),VALGRIND_MAKE_MEM_NOACCESS(global_stack, global_stack_top-global_stack)) +#define ALLOC_STACK char *_saved_stack; ((global_stack = (global_stack==0) ? ((global_stack_top=opus_alloc_scratch(GLOBAL_STACK_SIZE*2)+(GLOBAL_STACK_SIZE*2))-(GLOBAL_STACK_SIZE*2)) : global_stack),VALGRIND_MAKE_MEM_NOACCESS(global_stack, global_stack_top-global_stack)); _saved_stack = global_stack; + +#else + +#define ALIGN(stack, size) ((stack) += ((size) - (long)(stack)) & ((size) - 1)) +#define PUSH(stack, size, type) (ALIGN((stack),sizeof(type)/sizeof(char)),(stack)+=(size)*(sizeof(type)/sizeof(char)),(type*)((stack)-(size)*(sizeof(type)/sizeof(char)))) +#if 0 /* Set this to 1 to instrument pseudostack usage */ +#define RESTORE_STACK (printf("%ld %s:%d\n", global_stack-scratch_ptr, __FILE__, __LINE__),global_stack = _saved_stack) +#else +#define RESTORE_STACK (global_stack = _saved_stack) +#endif +#define ALLOC_STACK char *_saved_stack; (global_stack = (global_stack==0) ? (scratch_ptr=opus_alloc_scratch(GLOBAL_STACK_SIZE)) : global_stack); _saved_stack = global_stack; + +#endif /* ENABLE_VALGRIND */ + +#include "os_support.h" +#define VARDECL(type, var) type *var +#define ALLOC(var, size, type) var = PUSH(global_stack, size, type) +#define SAVE_STACK char *_saved_stack = global_stack; +#define ALLOC_NONE 0 + +#endif /* VAR_ARRAYS */ + + +#ifdef ENABLE_VALGRIND + +#include +#define OPUS_CHECK_ARRAY(ptr, len) VALGRIND_CHECK_MEM_IS_DEFINED(ptr, len*sizeof(*ptr)) +#define OPUS_CHECK_VALUE(value) VALGRIND_CHECK_VALUE_IS_DEFINED(value) +#define OPUS_CHECK_ARRAY_COND(ptr, len) VALGRIND_CHECK_MEM_IS_DEFINED(ptr, len*sizeof(*ptr)) +#define OPUS_CHECK_VALUE_COND(value) VALGRIND_CHECK_VALUE_IS_DEFINED(value) +#define OPUS_PRINT_INT(value) do {fprintf(stderr, #value " = %d at %s:%d\n", value, __FILE__, __LINE__);}while(0) +#define OPUS_FPRINTF fprintf + +#else + +static OPUS_INLINE int _opus_false(void) {return 0;} +#define OPUS_CHECK_ARRAY(ptr, len) _opus_false() +#define OPUS_CHECK_VALUE(value) _opus_false() +#define OPUS_PRINT_INT(value) do{}while(0) +#define OPUS_FPRINTF (void) + +#endif + + +#endif /* STACK_ALLOC_H */ diff --git a/lib/opus/celt/static_modes_fixed.h b/lib/opus/celt/static_modes_fixed.h new file mode 100644 index 00000000..8717d626 --- /dev/null +++ b/lib/opus/celt/static_modes_fixed.h @@ -0,0 +1,892 @@ +/* The contents of this file was automatically generated by dump_modes.c + with arguments: 48000 960 + It contains static definitions for some pre-defined modes. */ +#include "modes.h" +#include "rate.h" + +#ifdef HAVE_ARM_NE10 +#define OVERRIDE_FFT 1 +#include "static_modes_fixed_arm_ne10.h" +#endif + +#ifndef DEF_WINDOW120 +#define DEF_WINDOW120 +static const opus_val16 window120[120] = { +2, 20, 55, 108, 178, +266, 372, 494, 635, 792, +966, 1157, 1365, 1590, 1831, +2089, 2362, 2651, 2956, 3276, +3611, 3961, 4325, 4703, 5094, +5499, 5916, 6346, 6788, 7241, +7705, 8179, 8663, 9156, 9657, +10167, 10684, 11207, 11736, 12271, +12810, 13353, 13899, 14447, 14997, +15547, 16098, 16648, 17197, 17744, +18287, 18827, 19363, 19893, 20418, +20936, 21447, 21950, 22445, 22931, +23407, 23874, 24330, 24774, 25208, +25629, 26039, 26435, 26819, 27190, +27548, 27893, 28224, 28541, 28845, +29135, 29411, 29674, 29924, 30160, +30384, 30594, 30792, 30977, 31151, +31313, 31463, 31602, 31731, 31849, +31958, 32057, 32148, 32229, 32303, +32370, 32429, 32481, 32528, 32568, +32604, 32634, 32661, 32683, 32701, +32717, 32729, 32740, 32748, 32754, +32758, 32762, 32764, 32766, 32767, +32767, 32767, 32767, 32767, 32767, +}; +#endif + +#ifndef DEF_LOGN400 +#define DEF_LOGN400 +static const opus_int16 logN400[21] = { +0, 0, 0, 0, 0, 0, 0, 0, 8, 8, 8, 8, 16, 16, 16, 21, 21, 24, 29, 34, 36, }; +#endif + +#ifndef DEF_PULSE_CACHE50 +#define DEF_PULSE_CACHE50 +static const opus_int16 cache_index50[105] = { +-1, -1, -1, -1, -1, -1, -1, -1, 0, 0, 0, 0, 41, 41, 41, +82, 82, 123, 164, 200, 222, 0, 0, 0, 0, 0, 0, 0, 0, 41, +41, 41, 41, 123, 123, 123, 164, 164, 240, 266, 283, 295, 41, 41, 41, +41, 41, 41, 41, 41, 123, 123, 123, 123, 240, 240, 240, 266, 266, 305, +318, 328, 336, 123, 123, 123, 123, 123, 123, 123, 123, 240, 240, 240, 240, +305, 305, 305, 318, 318, 343, 351, 358, 364, 240, 240, 240, 240, 240, 240, +240, 240, 305, 305, 305, 305, 343, 343, 343, 351, 351, 370, 376, 382, 387, +}; +static const unsigned char cache_bits50[392] = { +40, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, +7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, +7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 40, 15, 23, 28, +31, 34, 36, 38, 39, 41, 42, 43, 44, 45, 46, 47, 47, 49, 50, +51, 52, 53, 54, 55, 55, 57, 58, 59, 60, 61, 62, 63, 63, 65, +66, 67, 68, 69, 70, 71, 71, 40, 20, 33, 41, 48, 53, 57, 61, +64, 66, 69, 71, 73, 75, 76, 78, 80, 82, 85, 87, 89, 91, 92, +94, 96, 98, 101, 103, 105, 107, 108, 110, 112, 114, 117, 119, 121, 123, +124, 126, 128, 40, 23, 39, 51, 60, 67, 73, 79, 83, 87, 91, 94, +97, 100, 102, 105, 107, 111, 115, 118, 121, 124, 126, 129, 131, 135, 139, +142, 145, 148, 150, 153, 155, 159, 163, 166, 169, 172, 174, 177, 179, 35, +28, 49, 65, 78, 89, 99, 107, 114, 120, 126, 132, 136, 141, 145, 149, +153, 159, 165, 171, 176, 180, 185, 189, 192, 199, 205, 211, 216, 220, 225, +229, 232, 239, 245, 251, 21, 33, 58, 79, 97, 112, 125, 137, 148, 157, +166, 174, 182, 189, 195, 201, 207, 217, 227, 235, 243, 251, 17, 35, 63, +86, 106, 123, 139, 152, 165, 177, 187, 197, 206, 214, 222, 230, 237, 250, +25, 31, 55, 75, 91, 105, 117, 128, 138, 146, 154, 161, 168, 174, 180, +185, 190, 200, 208, 215, 222, 229, 235, 240, 245, 255, 16, 36, 65, 89, +110, 128, 144, 159, 173, 185, 196, 207, 217, 226, 234, 242, 250, 11, 41, +74, 103, 128, 151, 172, 191, 209, 225, 241, 255, 9, 43, 79, 110, 138, +163, 186, 207, 227, 246, 12, 39, 71, 99, 123, 144, 164, 182, 198, 214, +228, 241, 253, 9, 44, 81, 113, 142, 168, 192, 214, 235, 255, 7, 49, +90, 127, 160, 191, 220, 247, 6, 51, 95, 134, 170, 203, 234, 7, 47, +87, 123, 155, 184, 212, 237, 6, 52, 97, 137, 174, 208, 240, 5, 57, +106, 151, 192, 231, 5, 59, 111, 158, 202, 243, 5, 55, 103, 147, 187, +224, 5, 60, 113, 161, 206, 248, 4, 65, 122, 175, 224, 4, 67, 127, +182, 234, }; +static const unsigned char cache_caps50[168] = { +224, 224, 224, 224, 224, 224, 224, 224, 160, 160, 160, 160, 185, 185, 185, +178, 178, 168, 134, 61, 37, 224, 224, 224, 224, 224, 224, 224, 224, 240, +240, 240, 240, 207, 207, 207, 198, 198, 183, 144, 66, 40, 160, 160, 160, +160, 160, 160, 160, 160, 185, 185, 185, 185, 193, 193, 193, 183, 183, 172, +138, 64, 38, 240, 240, 240, 240, 240, 240, 240, 240, 207, 207, 207, 207, +204, 204, 204, 193, 193, 180, 143, 66, 40, 185, 185, 185, 185, 185, 185, +185, 185, 193, 193, 193, 193, 193, 193, 193, 183, 183, 172, 138, 65, 39, +207, 207, 207, 207, 207, 207, 207, 207, 204, 204, 204, 204, 201, 201, 201, +188, 188, 176, 141, 66, 40, 193, 193, 193, 193, 193, 193, 193, 193, 193, +193, 193, 193, 194, 194, 194, 184, 184, 173, 139, 65, 39, 204, 204, 204, +204, 204, 204, 204, 204, 201, 201, 201, 201, 198, 198, 198, 187, 187, 175, +140, 66, 40, }; +#endif + +#ifndef FFT_TWIDDLES48000_960 +#define FFT_TWIDDLES48000_960 +static const kiss_twiddle_cpx fft_twiddles48000_960[480] = { +{32767, 0}, {32766, -429}, +{32757, -858}, {32743, -1287}, +{32724, -1715}, {32698, -2143}, +{32667, -2570}, {32631, -2998}, +{32588, -3425}, {32541, -3851}, +{32488, -4277}, {32429, -4701}, +{32364, -5125}, {32295, -5548}, +{32219, -5971}, {32138, -6393}, +{32051, -6813}, {31960, -7231}, +{31863, -7650}, {31760, -8067}, +{31652, -8481}, {31539, -8895}, +{31419, -9306}, {31294, -9716}, +{31165, -10126}, {31030, -10532}, +{30889, -10937}, {30743, -11340}, +{30592, -11741}, {30436, -12141}, +{30274, -12540}, {30107, -12935}, +{29936, -13328}, {29758, -13718}, +{29577, -14107}, {29390, -14493}, +{29197, -14875}, {29000, -15257}, +{28797, -15635}, {28590, -16010}, +{28379, -16384}, {28162, -16753}, +{27940, -17119}, {27714, -17484}, +{27482, -17845}, {27246, -18205}, +{27006, -18560}, {26760, -18911}, +{26510, -19260}, {26257, -19606}, +{25997, -19947}, {25734, -20286}, +{25466, -20621}, {25194, -20952}, +{24918, -21281}, {24637, -21605}, +{24353, -21926}, {24063, -22242}, +{23770, -22555}, {23473, -22865}, +{23171, -23171}, {22866, -23472}, +{22557, -23769}, {22244, -24063}, +{21927, -24352}, {21606, -24636}, +{21282, -24917}, {20954, -25194}, +{20622, -25465}, {20288, -25733}, +{19949, -25997}, {19607, -26255}, +{19261, -26509}, {18914, -26760}, +{18561, -27004}, {18205, -27246}, +{17846, -27481}, {17485, -27713}, +{17122, -27940}, {16755, -28162}, +{16385, -28378}, {16012, -28590}, +{15636, -28797}, {15258, -28999}, +{14878, -29197}, {14494, -29389}, +{14108, -29576}, {13720, -29757}, +{13329, -29934}, {12937, -30107}, +{12540, -30274}, {12142, -30435}, +{11744, -30592}, {11342, -30743}, +{10939, -30889}, {10534, -31030}, +{10127, -31164}, {9718, -31294}, +{9307, -31418}, {8895, -31537}, +{8482, -31652}, {8067, -31759}, +{7650, -31862}, {7233, -31960}, +{6815, -32051}, {6393, -32138}, +{5973, -32219}, {5549, -32294}, +{5127, -32364}, {4703, -32429}, +{4278, -32487}, {3852, -32541}, +{3426, -32588}, {2999, -32630}, +{2572, -32667}, {2144, -32698}, +{1716, -32724}, {1287, -32742}, +{860, -32757}, {430, -32766}, +{0, -32767}, {-429, -32766}, +{-858, -32757}, {-1287, -32743}, +{-1715, -32724}, {-2143, -32698}, +{-2570, -32667}, {-2998, -32631}, +{-3425, -32588}, {-3851, -32541}, +{-4277, -32488}, {-4701, -32429}, +{-5125, -32364}, {-5548, -32295}, +{-5971, -32219}, {-6393, -32138}, +{-6813, -32051}, {-7231, -31960}, +{-7650, -31863}, {-8067, -31760}, +{-8481, -31652}, {-8895, -31539}, +{-9306, -31419}, {-9716, -31294}, +{-10126, -31165}, {-10532, -31030}, +{-10937, -30889}, {-11340, -30743}, +{-11741, -30592}, {-12141, -30436}, +{-12540, -30274}, {-12935, -30107}, +{-13328, -29936}, {-13718, -29758}, +{-14107, -29577}, {-14493, -29390}, +{-14875, -29197}, {-15257, -29000}, +{-15635, -28797}, {-16010, -28590}, +{-16384, -28379}, {-16753, -28162}, +{-17119, -27940}, {-17484, -27714}, +{-17845, -27482}, {-18205, -27246}, +{-18560, -27006}, {-18911, -26760}, +{-19260, -26510}, {-19606, -26257}, +{-19947, -25997}, {-20286, -25734}, +{-20621, -25466}, {-20952, -25194}, +{-21281, -24918}, {-21605, -24637}, +{-21926, -24353}, {-22242, -24063}, +{-22555, -23770}, {-22865, -23473}, +{-23171, -23171}, {-23472, -22866}, +{-23769, -22557}, {-24063, -22244}, +{-24352, -21927}, {-24636, -21606}, +{-24917, -21282}, {-25194, -20954}, +{-25465, -20622}, {-25733, -20288}, +{-25997, -19949}, {-26255, -19607}, +{-26509, -19261}, {-26760, -18914}, +{-27004, -18561}, {-27246, -18205}, +{-27481, -17846}, {-27713, -17485}, +{-27940, -17122}, {-28162, -16755}, +{-28378, -16385}, {-28590, -16012}, +{-28797, -15636}, {-28999, -15258}, +{-29197, -14878}, {-29389, -14494}, +{-29576, -14108}, {-29757, -13720}, +{-29934, -13329}, {-30107, -12937}, +{-30274, -12540}, {-30435, -12142}, +{-30592, -11744}, {-30743, -11342}, +{-30889, -10939}, {-31030, -10534}, +{-31164, -10127}, {-31294, -9718}, +{-31418, -9307}, {-31537, -8895}, +{-31652, -8482}, {-31759, -8067}, +{-31862, -7650}, {-31960, -7233}, +{-32051, -6815}, {-32138, -6393}, +{-32219, -5973}, {-32294, -5549}, +{-32364, -5127}, {-32429, -4703}, +{-32487, -4278}, {-32541, -3852}, +{-32588, -3426}, {-32630, -2999}, +{-32667, -2572}, {-32698, -2144}, +{-32724, -1716}, {-32742, -1287}, +{-32757, -860}, {-32766, -430}, +{-32767, 0}, {-32766, 429}, +{-32757, 858}, {-32743, 1287}, +{-32724, 1715}, {-32698, 2143}, +{-32667, 2570}, {-32631, 2998}, +{-32588, 3425}, {-32541, 3851}, +{-32488, 4277}, {-32429, 4701}, +{-32364, 5125}, {-32295, 5548}, +{-32219, 5971}, {-32138, 6393}, +{-32051, 6813}, {-31960, 7231}, +{-31863, 7650}, {-31760, 8067}, +{-31652, 8481}, {-31539, 8895}, +{-31419, 9306}, {-31294, 9716}, +{-31165, 10126}, {-31030, 10532}, +{-30889, 10937}, {-30743, 11340}, +{-30592, 11741}, {-30436, 12141}, +{-30274, 12540}, {-30107, 12935}, +{-29936, 13328}, {-29758, 13718}, +{-29577, 14107}, {-29390, 14493}, +{-29197, 14875}, {-29000, 15257}, +{-28797, 15635}, {-28590, 16010}, +{-28379, 16384}, {-28162, 16753}, +{-27940, 17119}, {-27714, 17484}, +{-27482, 17845}, {-27246, 18205}, +{-27006, 18560}, {-26760, 18911}, +{-26510, 19260}, {-26257, 19606}, +{-25997, 19947}, {-25734, 20286}, +{-25466, 20621}, {-25194, 20952}, +{-24918, 21281}, {-24637, 21605}, +{-24353, 21926}, {-24063, 22242}, +{-23770, 22555}, {-23473, 22865}, +{-23171, 23171}, {-22866, 23472}, +{-22557, 23769}, {-22244, 24063}, +{-21927, 24352}, {-21606, 24636}, +{-21282, 24917}, {-20954, 25194}, +{-20622, 25465}, {-20288, 25733}, +{-19949, 25997}, {-19607, 26255}, +{-19261, 26509}, {-18914, 26760}, +{-18561, 27004}, {-18205, 27246}, +{-17846, 27481}, {-17485, 27713}, +{-17122, 27940}, {-16755, 28162}, +{-16385, 28378}, {-16012, 28590}, +{-15636, 28797}, {-15258, 28999}, +{-14878, 29197}, {-14494, 29389}, +{-14108, 29576}, {-13720, 29757}, +{-13329, 29934}, {-12937, 30107}, +{-12540, 30274}, {-12142, 30435}, +{-11744, 30592}, {-11342, 30743}, +{-10939, 30889}, {-10534, 31030}, +{-10127, 31164}, {-9718, 31294}, +{-9307, 31418}, {-8895, 31537}, +{-8482, 31652}, {-8067, 31759}, +{-7650, 31862}, {-7233, 31960}, +{-6815, 32051}, {-6393, 32138}, +{-5973, 32219}, {-5549, 32294}, +{-5127, 32364}, {-4703, 32429}, +{-4278, 32487}, {-3852, 32541}, +{-3426, 32588}, {-2999, 32630}, +{-2572, 32667}, {-2144, 32698}, +{-1716, 32724}, {-1287, 32742}, +{-860, 32757}, {-430, 32766}, +{0, 32767}, {429, 32766}, +{858, 32757}, {1287, 32743}, +{1715, 32724}, {2143, 32698}, +{2570, 32667}, {2998, 32631}, +{3425, 32588}, {3851, 32541}, +{4277, 32488}, {4701, 32429}, +{5125, 32364}, {5548, 32295}, +{5971, 32219}, {6393, 32138}, +{6813, 32051}, {7231, 31960}, +{7650, 31863}, {8067, 31760}, +{8481, 31652}, {8895, 31539}, +{9306, 31419}, {9716, 31294}, +{10126, 31165}, {10532, 31030}, +{10937, 30889}, {11340, 30743}, +{11741, 30592}, {12141, 30436}, +{12540, 30274}, {12935, 30107}, +{13328, 29936}, {13718, 29758}, +{14107, 29577}, {14493, 29390}, +{14875, 29197}, {15257, 29000}, +{15635, 28797}, {16010, 28590}, +{16384, 28379}, {16753, 28162}, +{17119, 27940}, {17484, 27714}, +{17845, 27482}, {18205, 27246}, +{18560, 27006}, {18911, 26760}, +{19260, 26510}, {19606, 26257}, +{19947, 25997}, {20286, 25734}, +{20621, 25466}, {20952, 25194}, +{21281, 24918}, {21605, 24637}, +{21926, 24353}, {22242, 24063}, +{22555, 23770}, {22865, 23473}, +{23171, 23171}, {23472, 22866}, +{23769, 22557}, {24063, 22244}, +{24352, 21927}, {24636, 21606}, +{24917, 21282}, {25194, 20954}, +{25465, 20622}, {25733, 20288}, +{25997, 19949}, {26255, 19607}, +{26509, 19261}, {26760, 18914}, +{27004, 18561}, {27246, 18205}, +{27481, 17846}, {27713, 17485}, +{27940, 17122}, {28162, 16755}, +{28378, 16385}, {28590, 16012}, +{28797, 15636}, {28999, 15258}, +{29197, 14878}, {29389, 14494}, +{29576, 14108}, {29757, 13720}, +{29934, 13329}, {30107, 12937}, +{30274, 12540}, {30435, 12142}, +{30592, 11744}, {30743, 11342}, +{30889, 10939}, {31030, 10534}, +{31164, 10127}, {31294, 9718}, +{31418, 9307}, {31537, 8895}, +{31652, 8482}, {31759, 8067}, +{31862, 7650}, {31960, 7233}, +{32051, 6815}, {32138, 6393}, +{32219, 5973}, {32294, 5549}, +{32364, 5127}, {32429, 4703}, +{32487, 4278}, {32541, 3852}, +{32588, 3426}, {32630, 2999}, +{32667, 2572}, {32698, 2144}, +{32724, 1716}, {32742, 1287}, +{32757, 860}, {32766, 430}, +}; +#ifndef FFT_BITREV480 +#define FFT_BITREV480 +static const opus_int16 fft_bitrev480[480] = { +0, 96, 192, 288, 384, 32, 128, 224, 320, 416, 64, 160, 256, 352, 448, +8, 104, 200, 296, 392, 40, 136, 232, 328, 424, 72, 168, 264, 360, 456, +16, 112, 208, 304, 400, 48, 144, 240, 336, 432, 80, 176, 272, 368, 464, +24, 120, 216, 312, 408, 56, 152, 248, 344, 440, 88, 184, 280, 376, 472, +4, 100, 196, 292, 388, 36, 132, 228, 324, 420, 68, 164, 260, 356, 452, +12, 108, 204, 300, 396, 44, 140, 236, 332, 428, 76, 172, 268, 364, 460, +20, 116, 212, 308, 404, 52, 148, 244, 340, 436, 84, 180, 276, 372, 468, +28, 124, 220, 316, 412, 60, 156, 252, 348, 444, 92, 188, 284, 380, 476, +1, 97, 193, 289, 385, 33, 129, 225, 321, 417, 65, 161, 257, 353, 449, +9, 105, 201, 297, 393, 41, 137, 233, 329, 425, 73, 169, 265, 361, 457, +17, 113, 209, 305, 401, 49, 145, 241, 337, 433, 81, 177, 273, 369, 465, +25, 121, 217, 313, 409, 57, 153, 249, 345, 441, 89, 185, 281, 377, 473, +5, 101, 197, 293, 389, 37, 133, 229, 325, 421, 69, 165, 261, 357, 453, +13, 109, 205, 301, 397, 45, 141, 237, 333, 429, 77, 173, 269, 365, 461, +21, 117, 213, 309, 405, 53, 149, 245, 341, 437, 85, 181, 277, 373, 469, +29, 125, 221, 317, 413, 61, 157, 253, 349, 445, 93, 189, 285, 381, 477, +2, 98, 194, 290, 386, 34, 130, 226, 322, 418, 66, 162, 258, 354, 450, +10, 106, 202, 298, 394, 42, 138, 234, 330, 426, 74, 170, 266, 362, 458, +18, 114, 210, 306, 402, 50, 146, 242, 338, 434, 82, 178, 274, 370, 466, +26, 122, 218, 314, 410, 58, 154, 250, 346, 442, 90, 186, 282, 378, 474, +6, 102, 198, 294, 390, 38, 134, 230, 326, 422, 70, 166, 262, 358, 454, +14, 110, 206, 302, 398, 46, 142, 238, 334, 430, 78, 174, 270, 366, 462, +22, 118, 214, 310, 406, 54, 150, 246, 342, 438, 86, 182, 278, 374, 470, +30, 126, 222, 318, 414, 62, 158, 254, 350, 446, 94, 190, 286, 382, 478, +3, 99, 195, 291, 387, 35, 131, 227, 323, 419, 67, 163, 259, 355, 451, +11, 107, 203, 299, 395, 43, 139, 235, 331, 427, 75, 171, 267, 363, 459, +19, 115, 211, 307, 403, 51, 147, 243, 339, 435, 83, 179, 275, 371, 467, +27, 123, 219, 315, 411, 59, 155, 251, 347, 443, 91, 187, 283, 379, 475, +7, 103, 199, 295, 391, 39, 135, 231, 327, 423, 71, 167, 263, 359, 455, +15, 111, 207, 303, 399, 47, 143, 239, 335, 431, 79, 175, 271, 367, 463, +23, 119, 215, 311, 407, 55, 151, 247, 343, 439, 87, 183, 279, 375, 471, +31, 127, 223, 319, 415, 63, 159, 255, 351, 447, 95, 191, 287, 383, 479, +}; +#endif + +#ifndef FFT_BITREV240 +#define FFT_BITREV240 +static const opus_int16 fft_bitrev240[240] = { +0, 48, 96, 144, 192, 16, 64, 112, 160, 208, 32, 80, 128, 176, 224, +4, 52, 100, 148, 196, 20, 68, 116, 164, 212, 36, 84, 132, 180, 228, +8, 56, 104, 152, 200, 24, 72, 120, 168, 216, 40, 88, 136, 184, 232, +12, 60, 108, 156, 204, 28, 76, 124, 172, 220, 44, 92, 140, 188, 236, +1, 49, 97, 145, 193, 17, 65, 113, 161, 209, 33, 81, 129, 177, 225, +5, 53, 101, 149, 197, 21, 69, 117, 165, 213, 37, 85, 133, 181, 229, +9, 57, 105, 153, 201, 25, 73, 121, 169, 217, 41, 89, 137, 185, 233, +13, 61, 109, 157, 205, 29, 77, 125, 173, 221, 45, 93, 141, 189, 237, +2, 50, 98, 146, 194, 18, 66, 114, 162, 210, 34, 82, 130, 178, 226, +6, 54, 102, 150, 198, 22, 70, 118, 166, 214, 38, 86, 134, 182, 230, +10, 58, 106, 154, 202, 26, 74, 122, 170, 218, 42, 90, 138, 186, 234, +14, 62, 110, 158, 206, 30, 78, 126, 174, 222, 46, 94, 142, 190, 238, +3, 51, 99, 147, 195, 19, 67, 115, 163, 211, 35, 83, 131, 179, 227, +7, 55, 103, 151, 199, 23, 71, 119, 167, 215, 39, 87, 135, 183, 231, +11, 59, 107, 155, 203, 27, 75, 123, 171, 219, 43, 91, 139, 187, 235, +15, 63, 111, 159, 207, 31, 79, 127, 175, 223, 47, 95, 143, 191, 239, +}; +#endif + +#ifndef FFT_BITREV120 +#define FFT_BITREV120 +static const opus_int16 fft_bitrev120[120] = { +0, 24, 48, 72, 96, 8, 32, 56, 80, 104, 16, 40, 64, 88, 112, +4, 28, 52, 76, 100, 12, 36, 60, 84, 108, 20, 44, 68, 92, 116, +1, 25, 49, 73, 97, 9, 33, 57, 81, 105, 17, 41, 65, 89, 113, +5, 29, 53, 77, 101, 13, 37, 61, 85, 109, 21, 45, 69, 93, 117, +2, 26, 50, 74, 98, 10, 34, 58, 82, 106, 18, 42, 66, 90, 114, +6, 30, 54, 78, 102, 14, 38, 62, 86, 110, 22, 46, 70, 94, 118, +3, 27, 51, 75, 99, 11, 35, 59, 83, 107, 19, 43, 67, 91, 115, +7, 31, 55, 79, 103, 15, 39, 63, 87, 111, 23, 47, 71, 95, 119, +}; +#endif + +#ifndef FFT_BITREV60 +#define FFT_BITREV60 +static const opus_int16 fft_bitrev60[60] = { +0, 12, 24, 36, 48, 4, 16, 28, 40, 52, 8, 20, 32, 44, 56, +1, 13, 25, 37, 49, 5, 17, 29, 41, 53, 9, 21, 33, 45, 57, +2, 14, 26, 38, 50, 6, 18, 30, 42, 54, 10, 22, 34, 46, 58, +3, 15, 27, 39, 51, 7, 19, 31, 43, 55, 11, 23, 35, 47, 59, +}; +#endif + +#ifndef FFT_STATE48000_960_0 +#define FFT_STATE48000_960_0 +static const kiss_fft_state fft_state48000_960_0 = { +480, /* nfft */ +17476, /* scale */ +8, /* scale_shift */ +-1, /* shift */ +{5, 96, 3, 32, 4, 8, 2, 4, 4, 1, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev480, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_480, +#else +NULL, +#endif +}; +#endif + +#ifndef FFT_STATE48000_960_1 +#define FFT_STATE48000_960_1 +static const kiss_fft_state fft_state48000_960_1 = { +240, /* nfft */ +17476, /* scale */ +7, /* scale_shift */ +1, /* shift */ +{5, 48, 3, 16, 4, 4, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev240, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_240, +#else +NULL, +#endif +}; +#endif + +#ifndef FFT_STATE48000_960_2 +#define FFT_STATE48000_960_2 +static const kiss_fft_state fft_state48000_960_2 = { +120, /* nfft */ +17476, /* scale */ +6, /* scale_shift */ +2, /* shift */ +{5, 24, 3, 8, 2, 4, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev120, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_120, +#else +NULL, +#endif +}; +#endif + +#ifndef FFT_STATE48000_960_3 +#define FFT_STATE48000_960_3 +static const kiss_fft_state fft_state48000_960_3 = { +60, /* nfft */ +17476, /* scale */ +5, /* scale_shift */ +3, /* shift */ +{5, 12, 3, 4, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev60, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_60, +#else +NULL, +#endif +}; +#endif + +#endif + +#ifndef MDCT_TWIDDLES960 +#define MDCT_TWIDDLES960 +static const opus_val16 mdct_twiddles960[1800] = { +32767, 32767, 32767, 32766, 32765, +32763, 32761, 32759, 32756, 32753, +32750, 32746, 32742, 32738, 32733, +32728, 32722, 32717, 32710, 32704, +32697, 32690, 32682, 32674, 32666, +32657, 32648, 32639, 32629, 32619, +32609, 32598, 32587, 32576, 32564, +32552, 32539, 32526, 32513, 32500, +32486, 32472, 32457, 32442, 32427, +32411, 32395, 32379, 32362, 32345, +32328, 32310, 32292, 32274, 32255, +32236, 32217, 32197, 32177, 32157, +32136, 32115, 32093, 32071, 32049, +32027, 32004, 31981, 31957, 31933, +31909, 31884, 31859, 31834, 31809, +31783, 31756, 31730, 31703, 31676, +31648, 31620, 31592, 31563, 31534, +31505, 31475, 31445, 31415, 31384, +31353, 31322, 31290, 31258, 31226, +31193, 31160, 31127, 31093, 31059, +31025, 30990, 30955, 30920, 30884, +30848, 30812, 30775, 30738, 30701, +30663, 30625, 30587, 30548, 30509, +30470, 30430, 30390, 30350, 30309, +30269, 30227, 30186, 30144, 30102, +30059, 30016, 29973, 29930, 29886, +29842, 29797, 29752, 29707, 29662, +29616, 29570, 29524, 29477, 29430, +29383, 29335, 29287, 29239, 29190, +29142, 29092, 29043, 28993, 28943, +28892, 28842, 28791, 28739, 28688, +28636, 28583, 28531, 28478, 28425, +28371, 28317, 28263, 28209, 28154, +28099, 28044, 27988, 27932, 27876, +27820, 27763, 27706, 27648, 27591, +27533, 27474, 27416, 27357, 27298, +27238, 27178, 27118, 27058, 26997, +26936, 26875, 26814, 26752, 26690, +26628, 26565, 26502, 26439, 26375, +26312, 26247, 26183, 26119, 26054, +25988, 25923, 25857, 25791, 25725, +25658, 25592, 25524, 25457, 25389, +25322, 25253, 25185, 25116, 25047, +24978, 24908, 24838, 24768, 24698, +24627, 24557, 24485, 24414, 24342, +24270, 24198, 24126, 24053, 23980, +23907, 23834, 23760, 23686, 23612, +23537, 23462, 23387, 23312, 23237, +23161, 23085, 23009, 22932, 22856, +22779, 22701, 22624, 22546, 22468, +22390, 22312, 22233, 22154, 22075, +21996, 21916, 21836, 21756, 21676, +21595, 21515, 21434, 21352, 21271, +21189, 21107, 21025, 20943, 20860, +20777, 20694, 20611, 20528, 20444, +20360, 20276, 20192, 20107, 20022, +19937, 19852, 19767, 19681, 19595, +19509, 19423, 19336, 19250, 19163, +19076, 18988, 18901, 18813, 18725, +18637, 18549, 18460, 18372, 18283, +18194, 18104, 18015, 17925, 17835, +17745, 17655, 17565, 17474, 17383, +17292, 17201, 17110, 17018, 16927, +16835, 16743, 16650, 16558, 16465, +16372, 16279, 16186, 16093, 15999, +15906, 15812, 15718, 15624, 15529, +15435, 15340, 15245, 15150, 15055, +14960, 14864, 14769, 14673, 14577, +14481, 14385, 14288, 14192, 14095, +13998, 13901, 13804, 13706, 13609, +13511, 13414, 13316, 13218, 13119, +13021, 12923, 12824, 12725, 12626, +12527, 12428, 12329, 12230, 12130, +12030, 11930, 11831, 11730, 11630, +11530, 11430, 11329, 11228, 11128, +11027, 10926, 10824, 10723, 10622, +10520, 10419, 10317, 10215, 10113, +10011, 9909, 9807, 9704, 9602, +9499, 9397, 9294, 9191, 9088, +8985, 8882, 8778, 8675, 8572, +8468, 8364, 8261, 8157, 8053, +7949, 7845, 7741, 7637, 7532, +7428, 7323, 7219, 7114, 7009, +6905, 6800, 6695, 6590, 6485, +6380, 6274, 6169, 6064, 5958, +5853, 5747, 5642, 5536, 5430, +5325, 5219, 5113, 5007, 4901, +4795, 4689, 4583, 4476, 4370, +4264, 4157, 4051, 3945, 3838, +3732, 3625, 3518, 3412, 3305, +3198, 3092, 2985, 2878, 2771, +2664, 2558, 2451, 2344, 2237, +2130, 2023, 1916, 1809, 1702, +1594, 1487, 1380, 1273, 1166, +1059, 952, 844, 737, 630, +523, 416, 308, 201, 94, +-13, -121, -228, -335, -442, +-550, -657, -764, -871, -978, +-1086, -1193, -1300, -1407, -1514, +-1621, -1728, -1835, -1942, -2049, +-2157, -2263, -2370, -2477, -2584, +-2691, -2798, -2905, -3012, -3118, +-3225, -3332, -3439, -3545, -3652, +-3758, -3865, -3971, -4078, -4184, +-4290, -4397, -4503, -4609, -4715, +-4821, -4927, -5033, -5139, -5245, +-5351, -5457, -5562, -5668, -5774, +-5879, -5985, -6090, -6195, -6301, +-6406, -6511, -6616, -6721, -6826, +-6931, -7036, -7140, -7245, -7349, +-7454, -7558, -7663, -7767, -7871, +-7975, -8079, -8183, -8287, -8390, +-8494, -8597, -8701, -8804, -8907, +-9011, -9114, -9217, -9319, -9422, +-9525, -9627, -9730, -9832, -9934, +-10037, -10139, -10241, -10342, -10444, +-10546, -10647, -10748, -10850, -10951, +-11052, -11153, -11253, -11354, -11455, +-11555, -11655, -11756, -11856, -11955, +-12055, -12155, -12254, -12354, -12453, +-12552, -12651, -12750, -12849, -12947, +-13046, -13144, -13242, -13340, -13438, +-13536, -13633, -13731, -13828, -13925, +-14022, -14119, -14216, -14312, -14409, +-14505, -14601, -14697, -14793, -14888, +-14984, -15079, -15174, -15269, -15364, +-15459, -15553, -15647, -15741, -15835, +-15929, -16023, -16116, -16210, -16303, +-16396, -16488, -16581, -16673, -16766, +-16858, -16949, -17041, -17133, -17224, +-17315, -17406, -17497, -17587, -17678, +-17768, -17858, -17948, -18037, -18127, +-18216, -18305, -18394, -18483, -18571, +-18659, -18747, -18835, -18923, -19010, +-19098, -19185, -19271, -19358, -19444, +-19531, -19617, -19702, -19788, -19873, +-19959, -20043, -20128, -20213, -20297, +-20381, -20465, -20549, -20632, -20715, +-20798, -20881, -20963, -21046, -21128, +-21210, -21291, -21373, -21454, -21535, +-21616, -21696, -21776, -21856, -21936, +-22016, -22095, -22174, -22253, -22331, +-22410, -22488, -22566, -22643, -22721, +-22798, -22875, -22951, -23028, -23104, +-23180, -23256, -23331, -23406, -23481, +-23556, -23630, -23704, -23778, -23852, +-23925, -23998, -24071, -24144, -24216, +-24288, -24360, -24432, -24503, -24574, +-24645, -24716, -24786, -24856, -24926, +-24995, -25064, -25133, -25202, -25270, +-25339, -25406, -25474, -25541, -25608, +-25675, -25742, -25808, -25874, -25939, +-26005, -26070, -26135, -26199, -26264, +-26327, -26391, -26455, -26518, -26581, +-26643, -26705, -26767, -26829, -26891, +-26952, -27013, -27073, -27133, -27193, +-27253, -27312, -27372, -27430, -27489, +-27547, -27605, -27663, -27720, -27777, +-27834, -27890, -27946, -28002, -28058, +-28113, -28168, -28223, -28277, -28331, +-28385, -28438, -28491, -28544, -28596, +-28649, -28701, -28752, -28803, -28854, +-28905, -28955, -29006, -29055, -29105, +-29154, -29203, -29251, -29299, -29347, +-29395, -29442, -29489, -29535, -29582, +-29628, -29673, -29719, -29764, -29808, +-29853, -29897, -29941, -29984, -30027, +-30070, -30112, -30154, -30196, -30238, +-30279, -30320, -30360, -30400, -30440, +-30480, -30519, -30558, -30596, -30635, +-30672, -30710, -30747, -30784, -30821, +-30857, -30893, -30929, -30964, -30999, +-31033, -31068, -31102, -31135, -31168, +-31201, -31234, -31266, -31298, -31330, +-31361, -31392, -31422, -31453, -31483, +-31512, -31541, -31570, -31599, -31627, +-31655, -31682, -31710, -31737, -31763, +-31789, -31815, -31841, -31866, -31891, +-31915, -31939, -31963, -31986, -32010, +-32032, -32055, -32077, -32099, -32120, +-32141, -32162, -32182, -32202, -32222, +-32241, -32260, -32279, -32297, -32315, +-32333, -32350, -32367, -32383, -32399, +-32415, -32431, -32446, -32461, -32475, +-32489, -32503, -32517, -32530, -32542, +-32555, -32567, -32579, -32590, -32601, +-32612, -32622, -32632, -32641, -32651, +-32659, -32668, -32676, -32684, -32692, +-32699, -32706, -32712, -32718, -32724, +-32729, -32734, -32739, -32743, -32747, +-32751, -32754, -32757, -32760, -32762, +-32764, -32765, -32767, -32767, -32767, +32767, 32767, 32765, 32761, 32756, +32750, 32742, 32732, 32722, 32710, +32696, 32681, 32665, 32647, 32628, +32608, 32586, 32562, 32538, 32512, +32484, 32455, 32425, 32393, 32360, +32326, 32290, 32253, 32214, 32174, +32133, 32090, 32046, 32001, 31954, +31906, 31856, 31805, 31753, 31700, +31645, 31588, 31530, 31471, 31411, +31349, 31286, 31222, 31156, 31089, +31020, 30951, 30880, 30807, 30733, +30658, 30582, 30504, 30425, 30345, +30263, 30181, 30096, 30011, 29924, +29836, 29747, 29656, 29564, 29471, +29377, 29281, 29184, 29086, 28987, +28886, 28784, 28681, 28577, 28471, +28365, 28257, 28147, 28037, 27925, +27812, 27698, 27583, 27467, 27349, +27231, 27111, 26990, 26868, 26744, +26620, 26494, 26367, 26239, 26110, +25980, 25849, 25717, 25583, 25449, +25313, 25176, 25038, 24900, 24760, +24619, 24477, 24333, 24189, 24044, +23898, 23751, 23602, 23453, 23303, +23152, 22999, 22846, 22692, 22537, +22380, 22223, 22065, 21906, 21746, +21585, 21423, 21261, 21097, 20933, +20767, 20601, 20434, 20265, 20096, +19927, 19756, 19584, 19412, 19239, +19065, 18890, 18714, 18538, 18361, +18183, 18004, 17824, 17644, 17463, +17281, 17098, 16915, 16731, 16546, +16361, 16175, 15988, 15800, 15612, +15423, 15234, 15043, 14852, 14661, +14469, 14276, 14083, 13889, 13694, +13499, 13303, 13107, 12910, 12713, +12515, 12317, 12118, 11918, 11718, +11517, 11316, 11115, 10913, 10710, +10508, 10304, 10100, 9896, 9691, +9486, 9281, 9075, 8869, 8662, +8455, 8248, 8040, 7832, 7623, +7415, 7206, 6996, 6787, 6577, +6366, 6156, 5945, 5734, 5523, +5311, 5100, 4888, 4675, 4463, +4251, 4038, 3825, 3612, 3399, +3185, 2972, 2758, 2544, 2330, +2116, 1902, 1688, 1474, 1260, +1045, 831, 617, 402, 188, +-27, -241, -456, -670, -885, +-1099, -1313, -1528, -1742, -1956, +-2170, -2384, -2598, -2811, -3025, +-3239, -3452, -3665, -3878, -4091, +-4304, -4516, -4728, -4941, -5153, +-5364, -5576, -5787, -5998, -6209, +-6419, -6629, -6839, -7049, -7258, +-7467, -7676, -7884, -8092, -8300, +-8507, -8714, -8920, -9127, -9332, +-9538, -9743, -9947, -10151, -10355, +-10558, -10761, -10963, -11165, -11367, +-11568, -11768, -11968, -12167, -12366, +-12565, -12762, -12960, -13156, -13352, +-13548, -13743, -13937, -14131, -14324, +-14517, -14709, -14900, -15091, -15281, +-15470, -15659, -15847, -16035, -16221, +-16407, -16593, -16777, -16961, -17144, +-17326, -17508, -17689, -17869, -18049, +-18227, -18405, -18582, -18758, -18934, +-19108, -19282, -19455, -19627, -19799, +-19969, -20139, -20308, -20475, -20642, +-20809, -20974, -21138, -21301, -21464, +-21626, -21786, -21946, -22105, -22263, +-22420, -22575, -22730, -22884, -23037, +-23189, -23340, -23490, -23640, -23788, +-23935, -24080, -24225, -24369, -24512, +-24654, -24795, -24934, -25073, -25211, +-25347, -25482, -25617, -25750, -25882, +-26013, -26143, -26272, -26399, -26526, +-26651, -26775, -26898, -27020, -27141, +-27260, -27379, -27496, -27612, -27727, +-27841, -27953, -28065, -28175, -28284, +-28391, -28498, -28603, -28707, -28810, +-28911, -29012, -29111, -29209, -29305, +-29401, -29495, -29587, -29679, -29769, +-29858, -29946, -30032, -30118, -30201, +-30284, -30365, -30445, -30524, -30601, +-30677, -30752, -30825, -30897, -30968, +-31038, -31106, -31172, -31238, -31302, +-31365, -31426, -31486, -31545, -31602, +-31658, -31713, -31766, -31818, -31869, +-31918, -31966, -32012, -32058, -32101, +-32144, -32185, -32224, -32262, -32299, +-32335, -32369, -32401, -32433, -32463, +-32491, -32518, -32544, -32568, -32591, +-32613, -32633, -32652, -32669, -32685, +-32700, -32713, -32724, -32735, -32744, +-32751, -32757, -32762, -32766, -32767, +32767, 32764, 32755, 32741, 32720, +32694, 32663, 32626, 32583, 32535, +32481, 32421, 32356, 32286, 32209, +32128, 32041, 31948, 31850, 31747, +31638, 31523, 31403, 31278, 31148, +31012, 30871, 30724, 30572, 30415, +30253, 30086, 29913, 29736, 29553, +29365, 29172, 28974, 28771, 28564, +28351, 28134, 27911, 27684, 27452, +27216, 26975, 26729, 26478, 26223, +25964, 25700, 25432, 25159, 24882, +24601, 24315, 24026, 23732, 23434, +23133, 22827, 22517, 22204, 21886, +21565, 21240, 20912, 20580, 20244, +19905, 19563, 19217, 18868, 18516, +18160, 17802, 17440, 17075, 16708, +16338, 15964, 15588, 15210, 14829, +14445, 14059, 13670, 13279, 12886, +12490, 12093, 11693, 11291, 10888, +10482, 10075, 9666, 9255, 8843, +8429, 8014, 7597, 7180, 6760, +6340, 5919, 5496, 5073, 4649, +4224, 3798, 3372, 2945, 2517, +2090, 1661, 1233, 804, 375, +-54, -483, -911, -1340, -1768, +-2197, -2624, -3052, -3479, -3905, +-4330, -4755, -5179, -5602, -6024, +-6445, -6865, -7284, -7702, -8118, +-8533, -8946, -9358, -9768, -10177, +-10584, -10989, -11392, -11793, -12192, +-12589, -12984, -13377, -13767, -14155, +-14541, -14924, -15305, -15683, -16058, +-16430, -16800, -17167, -17531, -17892, +-18249, -18604, -18956, -19304, -19649, +-19990, -20329, -20663, -20994, -21322, +-21646, -21966, -22282, -22595, -22904, +-23208, -23509, -23806, -24099, -24387, +-24672, -24952, -25228, -25499, -25766, +-26029, -26288, -26541, -26791, -27035, +-27275, -27511, -27741, -27967, -28188, +-28405, -28616, -28823, -29024, -29221, +-29412, -29599, -29780, -29957, -30128, +-30294, -30455, -30611, -30761, -30906, +-31046, -31181, -31310, -31434, -31552, +-31665, -31773, -31875, -31972, -32063, +-32149, -32229, -32304, -32373, -32437, +-32495, -32547, -32594, -32635, -32671, +-32701, -32726, -32745, -32758, -32766, +32767, 32754, 32717, 32658, 32577, +32473, 32348, 32200, 32029, 31837, +31624, 31388, 31131, 30853, 30553, +30232, 29891, 29530, 29148, 28746, +28324, 27883, 27423, 26944, 26447, +25931, 25398, 24847, 24279, 23695, +23095, 22478, 21846, 21199, 20538, +19863, 19174, 18472, 17757, 17030, +16291, 15541, 14781, 14010, 13230, +12441, 11643, 10837, 10024, 9204, +8377, 7545, 6708, 5866, 5020, +4171, 3319, 2464, 1608, 751, +-107, -965, -1822, -2678, -3532, +-4383, -5232, -6077, -6918, -7754, +-8585, -9409, -10228, -11039, -11843, +-12639, -13426, -14204, -14972, -15730, +-16477, -17213, -17937, -18648, -19347, +-20033, -20705, -21363, -22006, -22634, +-23246, -23843, -24423, -24986, -25533, +-26062, -26573, -27066, -27540, -27995, +-28431, -28848, -29245, -29622, -29979, +-30315, -30630, -30924, -31197, -31449, +-31679, -31887, -32074, -32239, -32381, +-32501, -32600, -32675, -32729, -32759, +}; +#endif + +static const CELTMode mode48000_960_120 = { +48000, /* Fs */ +120, /* overlap */ +21, /* nbEBands */ +21, /* effEBands */ +{27853, 0, 4096, 8192, }, /* preemph */ +eband5ms, /* eBands */ +3, /* maxLM */ +8, /* nbShortMdcts */ +120, /* shortMdctSize */ +11, /* nbAllocVectors */ +band_allocation, /* allocVectors */ +logN400, /* logN */ +window120, /* window */ +{1920, 3, {&fft_state48000_960_0, &fft_state48000_960_1, &fft_state48000_960_2, &fft_state48000_960_3, }, mdct_twiddles960}, /* mdct */ +{392, cache_index50, cache_bits50, cache_caps50}, /* cache */ +}; + +/* List of all the available modes */ +#define TOTAL_MODES 1 +static const CELTMode * const static_mode_list[TOTAL_MODES] = { +&mode48000_960_120, +}; diff --git a/lib/opus/celt/static_modes_fixed_arm_ne10.h b/lib/opus/celt/static_modes_fixed_arm_ne10.h new file mode 100644 index 00000000..76230921 --- /dev/null +++ b/lib/opus/celt/static_modes_fixed_arm_ne10.h @@ -0,0 +1,388 @@ +/* The contents of this file was automatically generated by + * dump_mode_arm_ne10.c with arguments: 48000 960 + * It contains static definitions for some pre-defined modes. */ +#include + +#ifndef NE10_FFT_PARAMS48000_960 +#define NE10_FFT_PARAMS48000_960 +static const ne10_int32_t ne10_factors_480[64] = { +4, 40, 4, 30, 2, 15, 5, 3, 3, 1, 1, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_int32_t ne10_factors_240[64] = { +3, 20, 4, 15, 5, 3, 3, 1, 1, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_int32_t ne10_factors_120[64] = { +3, 10, 2, 15, 5, 3, 3, 1, 1, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_int32_t ne10_factors_60[64] = { +2, 5, 5, 3, 3, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_fft_cpx_int32_t ne10_twiddles_480[480] = { +{0,0}, {2147483647,0}, {2147483647,0}, +{2147483647,0}, {1961823921,-873460313}, {1436946998,-1595891394}, +{2147483647,0}, {1436946998,-1595891394}, {-224473265,-2135719496}, +{2147483647,0}, {663608871,-2042378339}, {-1737350854,-1262259096}, +{2147483647,0}, {-224473265,-2135719496}, {-2100555935,446487152}, +{2147483647,0}, {2100555974,-446486968}, {1961823921,-873460313}, +{1737350743,-1262259248}, {1436946998,-1595891394}, {1073741769,-1859775424}, +{663608871,-2042378339}, {224473078,-2135719516}, {-224473265,-2135719496}, +{-663609049,-2042378281}, {-1073741932,-1859775330}, {-1436947137,-1595891268}, +{-1737350854,-1262259096}, {-1961823997,-873460141}, {-2100556013,-446486785}, +{2147483647,0}, {2144540595,-112390613}, {2135719506,-224473172}, +{2121044558,-335940465}, {2100555974,-446486968}, {2074309912,-555809682}, +{2042378310,-663608960}, {2004848691,-769589332}, {1961823921,-873460313}, +{1913421927,-974937199}, {1859775377,-1073741851}, {1801031311,-1169603450}, +{1737350743,-1262259248}, {1668908218,-1351455280}, {1595891331,-1436947067}, +{1518500216,-1518500282}, {1436946998,-1595891394}, {1351455207,-1668908277}, +{1262259172,-1737350799}, {1169603371,-1801031362}, {1073741769,-1859775424}, +{974937230,-1913421912}, {873460227,-1961823959}, {769589125,-2004848771}, +{663608871,-2042378339}, {555809715,-2074309903}, {446486876,-2100555994}, +{335940246,-2121044593}, {224473078,-2135719516}, {112390647,-2144540593}, +{2147483647,0}, {2135719506,-224473172}, {2100555974,-446486968}, +{2042378310,-663608960}, {1961823921,-873460313}, {1859775377,-1073741851}, +{1737350743,-1262259248}, {1595891331,-1436947067}, {1436946998,-1595891394}, +{1262259172,-1737350799}, {1073741769,-1859775424}, {873460227,-1961823959}, +{663608871,-2042378339}, {446486876,-2100555994}, {224473078,-2135719516}, +{-94,-2147483647}, {-224473265,-2135719496}, {-446487060,-2100555955}, +{-663609049,-2042378281}, {-873460398,-1961823883}, {-1073741932,-1859775330}, +{-1262259116,-1737350839}, {-1436947137,-1595891268}, {-1595891628,-1436946738}, +{-1737350854,-1262259096}, {-1859775343,-1073741910}, {-1961823997,-873460141}, +{-2042378447,-663608538}, {-2100556013,-446486785}, {-2135719499,-224473240}, +{2147483647,0}, {2121044558,-335940465}, {2042378310,-663608960}, +{1913421927,-974937199}, {1737350743,-1262259248}, {1518500216,-1518500282}, +{1262259172,-1737350799}, {974937230,-1913421912}, {663608871,-2042378339}, +{335940246,-2121044593}, {-94,-2147483647}, {-335940431,-2121044564}, +{-663609049,-2042378281}, {-974937397,-1913421827}, {-1262259116,-1737350839}, +{-1518500258,-1518500240}, {-1737350854,-1262259096}, {-1913422071,-974936918}, +{-2042378447,-663608538}, {-2121044568,-335940406}, {-2147483647,188}, +{-2121044509,335940777}, {-2042378331,663608895}, {-1913421900,974937252}, +{-1737350633,1262259400}, {-1518499993,1518500506}, {-1262258813,1737351059}, +{-974936606,1913422229}, {-663609179,2042378239}, {-335940566,2121044542}, +{2147483647,0}, {2147299667,-28109693}, {2146747758,-56214570}, +{2145828015,-84309815}, {2144540595,-112390613}, {2142885719,-140452154}, +{2140863671,-168489630}, {2138474797,-196498235}, {2135719506,-224473172}, +{2132598271,-252409646}, {2129111626,-280302871}, {2125260168,-308148068}, +{2121044558,-335940465}, {2116465518,-363675300}, {2111523833,-391347822}, +{2106220349,-418953288}, {2100555974,-446486968}, {2094531681,-473944146}, +{2088148500,-501320115}, {2081407525,-528610186}, {2074309912,-555809682}, +{2066856885,-582913912}, {2059049696,-609918325}, {2050889698,-636818231}, +{2042378310,-663608960}, {2033516972,-690285983}, {2024307180,-716844791}, +{2014750533,-743280770}, {2004848691,-769589332}, {1994603329,-795766029}, +{1984016179,-821806435}, {1973089077,-847706028}, {1961823921,-873460313}, +{1950222618,-899064934}, {1938287127,-924515564}, {1926019520,-949807783}, +{1913421927,-974937199}, {1900496481,-999899565}, {1887245364,-1024690661}, +{1873670877,-1049306180}, {1859775377,-1073741851}, {1845561215,-1097993541}, +{1831030826,-1122057097}, {1816186632,-1145928502}, {1801031311,-1169603450}, +{1785567394,-1193077993}, {1769797456,-1216348214}, {1753724345,-1239409914}, +{1737350743,-1262259248}, {1720679456,-1284892300}, {1703713340,-1307305194}, +{1686455222,-1329494189}, {1668908218,-1351455280}, {1651075255,-1373184807}, +{1632959307,-1394679144}, {1614563642,-1415934412}, {1595891331,-1436947067}, +{1576945572,-1457713510}, {1557729613,-1478230181}, {1538246655,-1498493658}, +{1518500216,-1518500282}, {1498493590,-1538246721}, {1478230113,-1557729677}, +{1457713441,-1576945636}, {1436946998,-1595891394}, {1415934341,-1614563704}, +{1394679073,-1632959368}, {1373184735,-1651075315}, {1351455207,-1668908277}, +{1329494115,-1686455280}, {1307305120,-1703713397}, {1284892225,-1720679512}, +{1262259172,-1737350799}, {1239409837,-1753724400}, {1216348136,-1769797510}, +{1193077915,-1785567446}, {1169603371,-1801031362}, {1145928423,-1816186682}, +{1122057017,-1831030875}, {1097993571,-1845561197}, {1073741769,-1859775424}, +{1049305987,-1873670985}, {1024690635,-1887245378}, {999899482,-1900496524}, +{974937230,-1913421912}, {949807699,-1926019561}, {924515422,-1938287195}, +{899064965,-1950222603}, {873460227,-1961823959}, {847705824,-1973089164}, +{821806407,-1984016190}, {795765941,-1994603364}, {769589125,-2004848771}, +{743280682,-2014750566}, {716844642,-2024307233}, {690286016,-2033516961}, +{663608871,-2042378339}, {636818019,-2050889764}, {609918296,-2059049705}, +{582913822,-2066856911}, {555809715,-2074309903}, {528610126,-2081407540}, +{501319962,-2088148536}, {473944148,-2094531680}, {446486876,-2100555994}, +{418953102,-2106220386}, {391347792,-2111523838}, {363675176,-2116465540}, +{335940246,-2121044593}, {308148006,-2125260177}, {280302715,-2129111646}, +{252409648,-2132598271}, {224473078,-2135719516}, {196498046,-2138474814}, +{168489600,-2140863674}, {140452029,-2142885728}, {112390647,-2144540593}, +{84309753,-2145828017}, {56214412,-2146747762}, {28109695,-2147299667}, +{2147483647,0}, {2146747758,-56214570}, {2144540595,-112390613}, +{2140863671,-168489630}, {2135719506,-224473172}, {2129111626,-280302871}, +{2121044558,-335940465}, {2111523833,-391347822}, {2100555974,-446486968}, +{2088148500,-501320115}, {2074309912,-555809682}, {2059049696,-609918325}, +{2042378310,-663608960}, {2024307180,-716844791}, {2004848691,-769589332}, +{1984016179,-821806435}, {1961823921,-873460313}, {1938287127,-924515564}, +{1913421927,-974937199}, {1887245364,-1024690661}, {1859775377,-1073741851}, +{1831030826,-1122057097}, {1801031311,-1169603450}, {1769797456,-1216348214}, +{1737350743,-1262259248}, {1703713340,-1307305194}, {1668908218,-1351455280}, +{1632959307,-1394679144}, {1595891331,-1436947067}, {1557729613,-1478230181}, +{1518500216,-1518500282}, {1478230113,-1557729677}, {1436946998,-1595891394}, +{1394679073,-1632959368}, {1351455207,-1668908277}, {1307305120,-1703713397}, +{1262259172,-1737350799}, {1216348136,-1769797510}, {1169603371,-1801031362}, +{1122057017,-1831030875}, {1073741769,-1859775424}, {1024690635,-1887245378}, +{974937230,-1913421912}, {924515422,-1938287195}, {873460227,-1961823959}, +{821806407,-1984016190}, {769589125,-2004848771}, {716844642,-2024307233}, +{663608871,-2042378339}, {609918296,-2059049705}, {555809715,-2074309903}, +{501319962,-2088148536}, {446486876,-2100555994}, {391347792,-2111523838}, +{335940246,-2121044593}, {280302715,-2129111646}, {224473078,-2135719516}, +{168489600,-2140863674}, {112390647,-2144540593}, {56214412,-2146747762}, +{-94,-2147483647}, {-56214600,-2146747757}, {-112390835,-2144540584}, +{-168489787,-2140863659}, {-224473265,-2135719496}, {-280302901,-2129111622}, +{-335940431,-2121044564}, {-391347977,-2111523804}, {-446487060,-2100555955}, +{-501320144,-2088148493}, {-555809896,-2074309855}, {-609918476,-2059049651}, +{-663609049,-2042378281}, {-716844819,-2024307170}, {-769589300,-2004848703}, +{-821806581,-1984016118}, {-873460398,-1961823883}, {-924515591,-1938287114}, +{-974937397,-1913421827}, {-1024690575,-1887245411}, {-1073741932,-1859775330}, +{-1122057395,-1831030643}, {-1169603421,-1801031330}, {-1216348291,-1769797403}, +{-1262259116,-1737350839}, {-1307305268,-1703713283}, {-1351455453,-1668908078}, +{-1394679021,-1632959413}, {-1436947137,-1595891268}, {-1478230435,-1557729372}, +{-1518500258,-1518500240}, {-1557729742,-1478230045}, {-1595891628,-1436946738}, +{-1632959429,-1394679001}, {-1668908417,-1351455035}, {-1703713298,-1307305248}, +{-1737350854,-1262259096}, {-1769797708,-1216347848}, {-1801031344,-1169603400}, +{-1831030924,-1122056937}, {-1859775343,-1073741910}, {-1887245423,-1024690552}, +{-1913422071,-974936918}, {-1938287125,-924515568}, {-1961823997,-873460141}, +{-1984016324,-821806084}, {-2004848713,-769589276}, {-2024307264,-716844553}, +{-2042378447,-663608538}, {-2059049731,-609918206}, {-2074309994,-555809377}, +{-2088148499,-501320119}, {-2100556013,-446486785}, {-2111523902,-391347448}, +{-2121044568,-335940406}, {-2129111659,-280302621}, {-2135719499,-224473240}, +{-2140863681,-168489506}, {-2144540612,-112390298}, {-2146747758,-56214574}, +{2147483647,0}, {2145828015,-84309815}, {2140863671,-168489630}, +{2132598271,-252409646}, {2121044558,-335940465}, {2106220349,-418953288}, +{2088148500,-501320115}, {2066856885,-582913912}, {2042378310,-663608960}, +{2014750533,-743280770}, {1984016179,-821806435}, {1950222618,-899064934}, +{1913421927,-974937199}, {1873670877,-1049306180}, {1831030826,-1122057097}, +{1785567394,-1193077993}, {1737350743,-1262259248}, {1686455222,-1329494189}, +{1632959307,-1394679144}, {1576945572,-1457713510}, {1518500216,-1518500282}, +{1457713441,-1576945636}, {1394679073,-1632959368}, {1329494115,-1686455280}, +{1262259172,-1737350799}, {1193077915,-1785567446}, {1122057017,-1831030875}, +{1049305987,-1873670985}, {974937230,-1913421912}, {899064965,-1950222603}, +{821806407,-1984016190}, {743280682,-2014750566}, {663608871,-2042378339}, +{582913822,-2066856911}, {501319962,-2088148536}, {418953102,-2106220386}, +{335940246,-2121044593}, {252409648,-2132598271}, {168489600,-2140863674}, +{84309753,-2145828017}, {-94,-2147483647}, {-84309940,-2145828010}, +{-168489787,-2140863659}, {-252409834,-2132598249}, {-335940431,-2121044564}, +{-418953286,-2106220349}, {-501320144,-2088148493}, {-582914003,-2066856860}, +{-663609049,-2042378281}, {-743280858,-2014750501}, {-821806581,-1984016118}, +{-899065136,-1950222525}, {-974937397,-1913421827}, {-1049306374,-1873670768}, +{-1122057395,-1831030643}, {-1193078284,-1785567199}, {-1262259116,-1737350839}, +{-1329494061,-1686455323}, {-1394679021,-1632959413}, {-1457713485,-1576945595}, +{-1518500258,-1518500240}, {-1576945613,-1457713466}, {-1632959429,-1394679001}, +{-1686455338,-1329494041}, {-1737350854,-1262259096}, {-1785567498,-1193077837}, +{-1831030924,-1122056937}, {-1873671031,-1049305905}, {-1913422071,-974936918}, +{-1950222750,-899064648}, {-1984016324,-821806084}, {-2014750687,-743280354}, +{-2042378447,-663608538}, {-2066856867,-582913978}, {-2088148499,-501320119}, +{-2106220354,-418953261}, {-2121044568,-335940406}, {-2132598282,-252409555}, +{-2140863681,-168489506}, {-2145828021,-84309659}, {-2147483647,188}, +{-2145828006,84310034}, {-2140863651,168489881}, {-2132598237,252409928}, +{-2121044509,335940777}, {-2106220281,418953629}, {-2088148411,501320484}, +{-2066856765,582914339}, {-2042378331,663608895}, {-2014750557,743280706}, +{-1984016181,821806431}, {-1950222593,899064989}, {-1913421900,974937252}, +{-1873670848,1049306232}, {-1831030728,1122057257}, {-1785567289,1193078149}, +{-1737350633,1262259400}, {-1686455106,1329494336}, {-1632959185,1394679287}, +{-1576945358,1457713742}, {-1518499993,1518500506}, {-1457713209,1576945850}, +{-1394678735,1632959656}, {-1329493766,1686455555}, {-1262258813,1737351059}, +{-1193077546,1785567692}, {-1122056638,1831031107}, {-1049305599,1873671202}, +{-974936606,1913422229}, {-899064330,1950222896}, {-821805761,1984016458}, +{-743280025,2014750808}, {-663609179,2042378239}, {-582914134,2066856823}, +{-501320277,2088148461}, {-418953420,2106220322}, {-335940566,2121044542}, +{-252409716,2132598263}, {-168489668,2140863668}, {-84309821,2145828015}, +}; +static const ne10_fft_cpx_int32_t ne10_twiddles_240[240] = { +{0,0}, {2147483647,0}, {2147483647,0}, +{2147483647,0}, {1961823921,-873460313}, {1436946998,-1595891394}, +{2147483647,0}, {1436946998,-1595891394}, {-224473265,-2135719496}, +{2147483647,0}, {663608871,-2042378339}, {-1737350854,-1262259096}, +{2147483647,0}, {-224473265,-2135719496}, {-2100555935,446487152}, +{2147483647,0}, {2135719506,-224473172}, {2100555974,-446486968}, +{2042378310,-663608960}, {1961823921,-873460313}, {1859775377,-1073741851}, +{1737350743,-1262259248}, {1595891331,-1436947067}, {1436946998,-1595891394}, +{1262259172,-1737350799}, {1073741769,-1859775424}, {873460227,-1961823959}, +{663608871,-2042378339}, {446486876,-2100555994}, {224473078,-2135719516}, +{2147483647,0}, {2100555974,-446486968}, {1961823921,-873460313}, +{1737350743,-1262259248}, {1436946998,-1595891394}, {1073741769,-1859775424}, +{663608871,-2042378339}, {224473078,-2135719516}, {-224473265,-2135719496}, +{-663609049,-2042378281}, {-1073741932,-1859775330}, {-1436947137,-1595891268}, +{-1737350854,-1262259096}, {-1961823997,-873460141}, {-2100556013,-446486785}, +{2147483647,0}, {2042378310,-663608960}, {1737350743,-1262259248}, +{1262259172,-1737350799}, {663608871,-2042378339}, {-94,-2147483647}, +{-663609049,-2042378281}, {-1262259116,-1737350839}, {-1737350854,-1262259096}, +{-2042378447,-663608538}, {-2147483647,188}, {-2042378331,663608895}, +{-1737350633,1262259400}, {-1262258813,1737351059}, {-663609179,2042378239}, +{2147483647,0}, {2146747758,-56214570}, {2144540595,-112390613}, +{2140863671,-168489630}, {2135719506,-224473172}, {2129111626,-280302871}, +{2121044558,-335940465}, {2111523833,-391347822}, {2100555974,-446486968}, +{2088148500,-501320115}, {2074309912,-555809682}, {2059049696,-609918325}, +{2042378310,-663608960}, {2024307180,-716844791}, {2004848691,-769589332}, +{1984016179,-821806435}, {1961823921,-873460313}, {1938287127,-924515564}, +{1913421927,-974937199}, {1887245364,-1024690661}, {1859775377,-1073741851}, +{1831030826,-1122057097}, {1801031311,-1169603450}, {1769797456,-1216348214}, +{1737350743,-1262259248}, {1703713340,-1307305194}, {1668908218,-1351455280}, +{1632959307,-1394679144}, {1595891331,-1436947067}, {1557729613,-1478230181}, +{1518500216,-1518500282}, {1478230113,-1557729677}, {1436946998,-1595891394}, +{1394679073,-1632959368}, {1351455207,-1668908277}, {1307305120,-1703713397}, +{1262259172,-1737350799}, {1216348136,-1769797510}, {1169603371,-1801031362}, +{1122057017,-1831030875}, {1073741769,-1859775424}, {1024690635,-1887245378}, +{974937230,-1913421912}, {924515422,-1938287195}, {873460227,-1961823959}, +{821806407,-1984016190}, {769589125,-2004848771}, {716844642,-2024307233}, +{663608871,-2042378339}, {609918296,-2059049705}, {555809715,-2074309903}, +{501319962,-2088148536}, {446486876,-2100555994}, {391347792,-2111523838}, +{335940246,-2121044593}, {280302715,-2129111646}, {224473078,-2135719516}, +{168489600,-2140863674}, {112390647,-2144540593}, {56214412,-2146747762}, +{2147483647,0}, {2144540595,-112390613}, {2135719506,-224473172}, +{2121044558,-335940465}, {2100555974,-446486968}, {2074309912,-555809682}, +{2042378310,-663608960}, {2004848691,-769589332}, {1961823921,-873460313}, +{1913421927,-974937199}, {1859775377,-1073741851}, {1801031311,-1169603450}, +{1737350743,-1262259248}, {1668908218,-1351455280}, {1595891331,-1436947067}, +{1518500216,-1518500282}, {1436946998,-1595891394}, {1351455207,-1668908277}, +{1262259172,-1737350799}, {1169603371,-1801031362}, {1073741769,-1859775424}, +{974937230,-1913421912}, {873460227,-1961823959}, {769589125,-2004848771}, +{663608871,-2042378339}, {555809715,-2074309903}, {446486876,-2100555994}, +{335940246,-2121044593}, {224473078,-2135719516}, {112390647,-2144540593}, +{-94,-2147483647}, {-112390835,-2144540584}, {-224473265,-2135719496}, +{-335940431,-2121044564}, {-446487060,-2100555955}, {-555809896,-2074309855}, +{-663609049,-2042378281}, {-769589300,-2004848703}, {-873460398,-1961823883}, +{-974937397,-1913421827}, {-1073741932,-1859775330}, {-1169603421,-1801031330}, +{-1262259116,-1737350839}, {-1351455453,-1668908078}, {-1436947137,-1595891268}, +{-1518500258,-1518500240}, {-1595891628,-1436946738}, {-1668908417,-1351455035}, +{-1737350854,-1262259096}, {-1801031344,-1169603400}, {-1859775343,-1073741910}, +{-1913422071,-974936918}, {-1961823997,-873460141}, {-2004848713,-769589276}, +{-2042378447,-663608538}, {-2074309994,-555809377}, {-2100556013,-446486785}, +{-2121044568,-335940406}, {-2135719499,-224473240}, {-2144540612,-112390298}, +{2147483647,0}, {2140863671,-168489630}, {2121044558,-335940465}, +{2088148500,-501320115}, {2042378310,-663608960}, {1984016179,-821806435}, +{1913421927,-974937199}, {1831030826,-1122057097}, {1737350743,-1262259248}, +{1632959307,-1394679144}, {1518500216,-1518500282}, {1394679073,-1632959368}, +{1262259172,-1737350799}, {1122057017,-1831030875}, {974937230,-1913421912}, +{821806407,-1984016190}, {663608871,-2042378339}, {501319962,-2088148536}, +{335940246,-2121044593}, {168489600,-2140863674}, {-94,-2147483647}, +{-168489787,-2140863659}, {-335940431,-2121044564}, {-501320144,-2088148493}, +{-663609049,-2042378281}, {-821806581,-1984016118}, {-974937397,-1913421827}, +{-1122057395,-1831030643}, {-1262259116,-1737350839}, {-1394679021,-1632959413}, +{-1518500258,-1518500240}, {-1632959429,-1394679001}, {-1737350854,-1262259096}, +{-1831030924,-1122056937}, {-1913422071,-974936918}, {-1984016324,-821806084}, +{-2042378447,-663608538}, {-2088148499,-501320119}, {-2121044568,-335940406}, +{-2140863681,-168489506}, {-2147483647,188}, {-2140863651,168489881}, +{-2121044509,335940777}, {-2088148411,501320484}, {-2042378331,663608895}, +{-1984016181,821806431}, {-1913421900,974937252}, {-1831030728,1122057257}, +{-1737350633,1262259400}, {-1632959185,1394679287}, {-1518499993,1518500506}, +{-1394678735,1632959656}, {-1262258813,1737351059}, {-1122056638,1831031107}, +{-974936606,1913422229}, {-821805761,1984016458}, {-663609179,2042378239}, +{-501320277,2088148461}, {-335940566,2121044542}, {-168489668,2140863668}, +}; +static const ne10_fft_cpx_int32_t ne10_twiddles_120[120] = { +{0,0}, {2147483647,0}, {2147483647,0}, +{2147483647,0}, {1961823921,-873460313}, {1436946998,-1595891394}, +{2147483647,0}, {1436946998,-1595891394}, {-224473265,-2135719496}, +{2147483647,0}, {663608871,-2042378339}, {-1737350854,-1262259096}, +{2147483647,0}, {-224473265,-2135719496}, {-2100555935,446487152}, +{2147483647,0}, {2100555974,-446486968}, {1961823921,-873460313}, +{1737350743,-1262259248}, {1436946998,-1595891394}, {1073741769,-1859775424}, +{663608871,-2042378339}, {224473078,-2135719516}, {-224473265,-2135719496}, +{-663609049,-2042378281}, {-1073741932,-1859775330}, {-1436947137,-1595891268}, +{-1737350854,-1262259096}, {-1961823997,-873460141}, {-2100556013,-446486785}, +{2147483647,0}, {2144540595,-112390613}, {2135719506,-224473172}, +{2121044558,-335940465}, {2100555974,-446486968}, {2074309912,-555809682}, +{2042378310,-663608960}, {2004848691,-769589332}, {1961823921,-873460313}, +{1913421927,-974937199}, {1859775377,-1073741851}, {1801031311,-1169603450}, +{1737350743,-1262259248}, {1668908218,-1351455280}, {1595891331,-1436947067}, +{1518500216,-1518500282}, {1436946998,-1595891394}, {1351455207,-1668908277}, +{1262259172,-1737350799}, {1169603371,-1801031362}, {1073741769,-1859775424}, +{974937230,-1913421912}, {873460227,-1961823959}, {769589125,-2004848771}, +{663608871,-2042378339}, {555809715,-2074309903}, {446486876,-2100555994}, +{335940246,-2121044593}, {224473078,-2135719516}, {112390647,-2144540593}, +{2147483647,0}, {2135719506,-224473172}, {2100555974,-446486968}, +{2042378310,-663608960}, {1961823921,-873460313}, {1859775377,-1073741851}, +{1737350743,-1262259248}, {1595891331,-1436947067}, {1436946998,-1595891394}, +{1262259172,-1737350799}, {1073741769,-1859775424}, {873460227,-1961823959}, +{663608871,-2042378339}, {446486876,-2100555994}, {224473078,-2135719516}, +{-94,-2147483647}, {-224473265,-2135719496}, {-446487060,-2100555955}, +{-663609049,-2042378281}, {-873460398,-1961823883}, {-1073741932,-1859775330}, +{-1262259116,-1737350839}, {-1436947137,-1595891268}, {-1595891628,-1436946738}, +{-1737350854,-1262259096}, {-1859775343,-1073741910}, {-1961823997,-873460141}, +{-2042378447,-663608538}, {-2100556013,-446486785}, {-2135719499,-224473240}, +{2147483647,0}, {2121044558,-335940465}, {2042378310,-663608960}, +{1913421927,-974937199}, {1737350743,-1262259248}, {1518500216,-1518500282}, +{1262259172,-1737350799}, {974937230,-1913421912}, {663608871,-2042378339}, +{335940246,-2121044593}, {-94,-2147483647}, {-335940431,-2121044564}, +{-663609049,-2042378281}, {-974937397,-1913421827}, {-1262259116,-1737350839}, +{-1518500258,-1518500240}, {-1737350854,-1262259096}, {-1913422071,-974936918}, +{-2042378447,-663608538}, {-2121044568,-335940406}, {-2147483647,188}, +{-2121044509,335940777}, {-2042378331,663608895}, {-1913421900,974937252}, +{-1737350633,1262259400}, {-1518499993,1518500506}, {-1262258813,1737351059}, +{-974936606,1913422229}, {-663609179,2042378239}, {-335940566,2121044542}, +}; +static const ne10_fft_cpx_int32_t ne10_twiddles_60[60] = { +{0,0}, {2147483647,0}, {2147483647,0}, +{2147483647,0}, {1961823921,-873460313}, {1436946998,-1595891394}, +{2147483647,0}, {1436946998,-1595891394}, {-224473265,-2135719496}, +{2147483647,0}, {663608871,-2042378339}, {-1737350854,-1262259096}, +{2147483647,0}, {-224473265,-2135719496}, {-2100555935,446487152}, +{2147483647,0}, {2135719506,-224473172}, {2100555974,-446486968}, +{2042378310,-663608960}, {1961823921,-873460313}, {1859775377,-1073741851}, +{1737350743,-1262259248}, {1595891331,-1436947067}, {1436946998,-1595891394}, +{1262259172,-1737350799}, {1073741769,-1859775424}, {873460227,-1961823959}, +{663608871,-2042378339}, {446486876,-2100555994}, {224473078,-2135719516}, +{2147483647,0}, {2100555974,-446486968}, {1961823921,-873460313}, +{1737350743,-1262259248}, {1436946998,-1595891394}, {1073741769,-1859775424}, +{663608871,-2042378339}, {224473078,-2135719516}, {-224473265,-2135719496}, +{-663609049,-2042378281}, {-1073741932,-1859775330}, {-1436947137,-1595891268}, +{-1737350854,-1262259096}, {-1961823997,-873460141}, {-2100556013,-446486785}, +{2147483647,0}, {2042378310,-663608960}, {1737350743,-1262259248}, +{1262259172,-1737350799}, {663608871,-2042378339}, {-94,-2147483647}, +{-663609049,-2042378281}, {-1262259116,-1737350839}, {-1737350854,-1262259096}, +{-2042378447,-663608538}, {-2147483647,188}, {-2042378331,663608895}, +{-1737350633,1262259400}, {-1262258813,1737351059}, {-663609179,2042378239}, +}; +static const ne10_fft_state_int32_t ne10_fft_state_int32_t_480 = { +120, +(ne10_int32_t *)ne10_factors_480, +(ne10_fft_cpx_int32_t *)ne10_twiddles_480, +NULL, +(ne10_fft_cpx_int32_t *)&ne10_twiddles_480[120], +}; +static const arch_fft_state cfg_arch_480 = { +1, +(void *)&ne10_fft_state_int32_t_480, +}; + +static const ne10_fft_state_int32_t ne10_fft_state_int32_t_240 = { +60, +(ne10_int32_t *)ne10_factors_240, +(ne10_fft_cpx_int32_t *)ne10_twiddles_240, +NULL, +(ne10_fft_cpx_int32_t *)&ne10_twiddles_240[60], +}; +static const arch_fft_state cfg_arch_240 = { +1, +(void *)&ne10_fft_state_int32_t_240, +}; + +static const ne10_fft_state_int32_t ne10_fft_state_int32_t_120 = { +30, +(ne10_int32_t *)ne10_factors_120, +(ne10_fft_cpx_int32_t *)ne10_twiddles_120, +NULL, +(ne10_fft_cpx_int32_t *)&ne10_twiddles_120[30], +}; +static const arch_fft_state cfg_arch_120 = { +1, +(void *)&ne10_fft_state_int32_t_120, +}; + +static const ne10_fft_state_int32_t ne10_fft_state_int32_t_60 = { +15, +(ne10_int32_t *)ne10_factors_60, +(ne10_fft_cpx_int32_t *)ne10_twiddles_60, +NULL, +(ne10_fft_cpx_int32_t *)&ne10_twiddles_60[15], +}; +static const arch_fft_state cfg_arch_60 = { +1, +(void *)&ne10_fft_state_int32_t_60, +}; + +#endif /* end NE10_FFT_PARAMS48000_960 */ diff --git a/lib/opus/celt/static_modes_float.h b/lib/opus/celt/static_modes_float.h new file mode 100644 index 00000000..e102a383 --- /dev/null +++ b/lib/opus/celt/static_modes_float.h @@ -0,0 +1,888 @@ +/* The contents of this file was automatically generated by dump_modes.c + with arguments: 48000 960 + It contains static definitions for some pre-defined modes. */ +#include "modes.h" +#include "rate.h" + +#ifdef HAVE_ARM_NE10 +#define OVERRIDE_FFT 1 +#include "static_modes_float_arm_ne10.h" +#endif + +#ifndef DEF_WINDOW120 +#define DEF_WINDOW120 +static const opus_val16 window120[120] = { +6.7286966e-05f, 0.00060551348f, 0.0016815970f, 0.0032947962f, 0.0054439943f, +0.0081276923f, 0.011344001f, 0.015090633f, 0.019364886f, 0.024163635f, +0.029483315f, 0.035319905f, 0.041668911f, 0.048525347f, 0.055883718f, +0.063737999f, 0.072081616f, 0.080907428f, 0.090207705f, 0.099974111f, +0.11019769f, 0.12086883f, 0.13197729f, 0.14351214f, 0.15546177f, +0.16781389f, 0.18055550f, 0.19367290f, 0.20715171f, 0.22097682f, +0.23513243f, 0.24960208f, 0.26436860f, 0.27941419f, 0.29472040f, +0.31026818f, 0.32603788f, 0.34200931f, 0.35816177f, 0.37447407f, +0.39092462f, 0.40749142f, 0.42415215f, 0.44088423f, 0.45766484f, +0.47447104f, 0.49127978f, 0.50806798f, 0.52481261f, 0.54149077f, +0.55807973f, 0.57455701f, 0.59090049f, 0.60708841f, 0.62309951f, +0.63891306f, 0.65450896f, 0.66986776f, 0.68497077f, 0.69980010f, +0.71433873f, 0.72857055f, 0.74248043f, 0.75605424f, 0.76927895f, +0.78214257f, 0.79463430f, 0.80674445f, 0.81846456f, 0.82978733f, +0.84070669f, 0.85121779f, 0.86131698f, 0.87100183f, 0.88027111f, +0.88912479f, 0.89756398f, 0.90559094f, 0.91320904f, 0.92042270f, +0.92723738f, 0.93365955f, 0.93969656f, 0.94535671f, 0.95064907f, +0.95558353f, 0.96017067f, 0.96442171f, 0.96834849f, 0.97196334f, +0.97527906f, 0.97830883f, 0.98106616f, 0.98356480f, 0.98581869f, +0.98784191f, 0.98964856f, 0.99125274f, 0.99266849f, 0.99390969f, +0.99499004f, 0.99592297f, 0.99672162f, 0.99739874f, 0.99796667f, +0.99843728f, 0.99882195f, 0.99913147f, 0.99937606f, 0.99956527f, +0.99970802f, 0.99981248f, 0.99988613f, 0.99993565f, 0.99996697f, +0.99998518f, 0.99999457f, 0.99999859f, 0.99999982f, 1.0000000f, +}; +#endif + +#ifndef DEF_LOGN400 +#define DEF_LOGN400 +static const opus_int16 logN400[21] = { +0, 0, 0, 0, 0, 0, 0, 0, 8, 8, 8, 8, 16, 16, 16, 21, 21, 24, 29, 34, 36, }; +#endif + +#ifndef DEF_PULSE_CACHE50 +#define DEF_PULSE_CACHE50 +static const opus_int16 cache_index50[105] = { +-1, -1, -1, -1, -1, -1, -1, -1, 0, 0, 0, 0, 41, 41, 41, +82, 82, 123, 164, 200, 222, 0, 0, 0, 0, 0, 0, 0, 0, 41, +41, 41, 41, 123, 123, 123, 164, 164, 240, 266, 283, 295, 41, 41, 41, +41, 41, 41, 41, 41, 123, 123, 123, 123, 240, 240, 240, 266, 266, 305, +318, 328, 336, 123, 123, 123, 123, 123, 123, 123, 123, 240, 240, 240, 240, +305, 305, 305, 318, 318, 343, 351, 358, 364, 240, 240, 240, 240, 240, 240, +240, 240, 305, 305, 305, 305, 343, 343, 343, 351, 351, 370, 376, 382, 387, +}; +static const unsigned char cache_bits50[392] = { +40, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, +7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, +7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 40, 15, 23, 28, +31, 34, 36, 38, 39, 41, 42, 43, 44, 45, 46, 47, 47, 49, 50, +51, 52, 53, 54, 55, 55, 57, 58, 59, 60, 61, 62, 63, 63, 65, +66, 67, 68, 69, 70, 71, 71, 40, 20, 33, 41, 48, 53, 57, 61, +64, 66, 69, 71, 73, 75, 76, 78, 80, 82, 85, 87, 89, 91, 92, +94, 96, 98, 101, 103, 105, 107, 108, 110, 112, 114, 117, 119, 121, 123, +124, 126, 128, 40, 23, 39, 51, 60, 67, 73, 79, 83, 87, 91, 94, +97, 100, 102, 105, 107, 111, 115, 118, 121, 124, 126, 129, 131, 135, 139, +142, 145, 148, 150, 153, 155, 159, 163, 166, 169, 172, 174, 177, 179, 35, +28, 49, 65, 78, 89, 99, 107, 114, 120, 126, 132, 136, 141, 145, 149, +153, 159, 165, 171, 176, 180, 185, 189, 192, 199, 205, 211, 216, 220, 225, +229, 232, 239, 245, 251, 21, 33, 58, 79, 97, 112, 125, 137, 148, 157, +166, 174, 182, 189, 195, 201, 207, 217, 227, 235, 243, 251, 17, 35, 63, +86, 106, 123, 139, 152, 165, 177, 187, 197, 206, 214, 222, 230, 237, 250, +25, 31, 55, 75, 91, 105, 117, 128, 138, 146, 154, 161, 168, 174, 180, +185, 190, 200, 208, 215, 222, 229, 235, 240, 245, 255, 16, 36, 65, 89, +110, 128, 144, 159, 173, 185, 196, 207, 217, 226, 234, 242, 250, 11, 41, +74, 103, 128, 151, 172, 191, 209, 225, 241, 255, 9, 43, 79, 110, 138, +163, 186, 207, 227, 246, 12, 39, 71, 99, 123, 144, 164, 182, 198, 214, +228, 241, 253, 9, 44, 81, 113, 142, 168, 192, 214, 235, 255, 7, 49, +90, 127, 160, 191, 220, 247, 6, 51, 95, 134, 170, 203, 234, 7, 47, +87, 123, 155, 184, 212, 237, 6, 52, 97, 137, 174, 208, 240, 5, 57, +106, 151, 192, 231, 5, 59, 111, 158, 202, 243, 5, 55, 103, 147, 187, +224, 5, 60, 113, 161, 206, 248, 4, 65, 122, 175, 224, 4, 67, 127, +182, 234, }; +static const unsigned char cache_caps50[168] = { +224, 224, 224, 224, 224, 224, 224, 224, 160, 160, 160, 160, 185, 185, 185, +178, 178, 168, 134, 61, 37, 224, 224, 224, 224, 224, 224, 224, 224, 240, +240, 240, 240, 207, 207, 207, 198, 198, 183, 144, 66, 40, 160, 160, 160, +160, 160, 160, 160, 160, 185, 185, 185, 185, 193, 193, 193, 183, 183, 172, +138, 64, 38, 240, 240, 240, 240, 240, 240, 240, 240, 207, 207, 207, 207, +204, 204, 204, 193, 193, 180, 143, 66, 40, 185, 185, 185, 185, 185, 185, +185, 185, 193, 193, 193, 193, 193, 193, 193, 183, 183, 172, 138, 65, 39, +207, 207, 207, 207, 207, 207, 207, 207, 204, 204, 204, 204, 201, 201, 201, +188, 188, 176, 141, 66, 40, 193, 193, 193, 193, 193, 193, 193, 193, 193, +193, 193, 193, 194, 194, 194, 184, 184, 173, 139, 65, 39, 204, 204, 204, +204, 204, 204, 204, 204, 201, 201, 201, 201, 198, 198, 198, 187, 187, 175, +140, 66, 40, }; +#endif + +#ifndef FFT_TWIDDLES48000_960 +#define FFT_TWIDDLES48000_960 +static const kiss_twiddle_cpx fft_twiddles48000_960[480] = { +{1.0000000f, -0.0000000f}, {0.99991433f, -0.013089596f}, +{0.99965732f, -0.026176948f}, {0.99922904f, -0.039259816f}, +{0.99862953f, -0.052335956f}, {0.99785892f, -0.065403129f}, +{0.99691733f, -0.078459096f}, {0.99580493f, -0.091501619f}, +{0.99452190f, -0.10452846f}, {0.99306846f, -0.11753740f}, +{0.99144486f, -0.13052619f}, {0.98965139f, -0.14349262f}, +{0.98768834f, -0.15643447f}, {0.98555606f, -0.16934950f}, +{0.98325491f, -0.18223553f}, {0.98078528f, -0.19509032f}, +{0.97814760f, -0.20791169f}, {0.97534232f, -0.22069744f}, +{0.97236992f, -0.23344536f}, {0.96923091f, -0.24615329f}, +{0.96592583f, -0.25881905f}, {0.96245524f, -0.27144045f}, +{0.95881973f, -0.28401534f}, {0.95501994f, -0.29654157f}, +{0.95105652f, -0.30901699f}, {0.94693013f, -0.32143947f}, +{0.94264149f, -0.33380686f}, {0.93819134f, -0.34611706f}, +{0.93358043f, -0.35836795f}, {0.92880955f, -0.37055744f}, +{0.92387953f, -0.38268343f}, {0.91879121f, -0.39474386f}, +{0.91354546f, -0.40673664f}, {0.90814317f, -0.41865974f}, +{0.90258528f, -0.43051110f}, {0.89687274f, -0.44228869f}, +{0.89100652f, -0.45399050f}, {0.88498764f, -0.46561452f}, +{0.87881711f, -0.47715876f}, {0.87249601f, -0.48862124f}, +{0.86602540f, -0.50000000f}, {0.85940641f, -0.51129309f}, +{0.85264016f, -0.52249856f}, {0.84572782f, -0.53361452f}, +{0.83867057f, -0.54463904f}, {0.83146961f, -0.55557023f}, +{0.82412619f, -0.56640624f}, {0.81664156f, -0.57714519f}, +{0.80901699f, -0.58778525f}, {0.80125381f, -0.59832460f}, +{0.79335334f, -0.60876143f}, {0.78531693f, -0.61909395f}, +{0.77714596f, -0.62932039f}, {0.76884183f, -0.63943900f}, +{0.76040597f, -0.64944805f}, {0.75183981f, -0.65934582f}, +{0.74314483f, -0.66913061f}, {0.73432251f, -0.67880075f}, +{0.72537437f, -0.68835458f}, {0.71630194f, -0.69779046f}, +{0.70710678f, -0.70710678f}, {0.69779046f, -0.71630194f}, +{0.68835458f, -0.72537437f}, {0.67880075f, -0.73432251f}, +{0.66913061f, -0.74314483f}, {0.65934582f, -0.75183981f}, +{0.64944805f, -0.76040597f}, {0.63943900f, -0.76884183f}, +{0.62932039f, -0.77714596f}, {0.61909395f, -0.78531693f}, +{0.60876143f, -0.79335334f}, {0.59832460f, -0.80125381f}, +{0.58778525f, -0.80901699f}, {0.57714519f, -0.81664156f}, +{0.56640624f, -0.82412619f}, {0.55557023f, -0.83146961f}, +{0.54463904f, -0.83867057f}, {0.53361452f, -0.84572782f}, +{0.52249856f, -0.85264016f}, {0.51129309f, -0.85940641f}, +{0.50000000f, -0.86602540f}, {0.48862124f, -0.87249601f}, +{0.47715876f, -0.87881711f}, {0.46561452f, -0.88498764f}, +{0.45399050f, -0.89100652f}, {0.44228869f, -0.89687274f}, +{0.43051110f, -0.90258528f}, {0.41865974f, -0.90814317f}, +{0.40673664f, -0.91354546f}, {0.39474386f, -0.91879121f}, +{0.38268343f, -0.92387953f}, {0.37055744f, -0.92880955f}, +{0.35836795f, -0.93358043f}, {0.34611706f, -0.93819134f}, +{0.33380686f, -0.94264149f}, {0.32143947f, -0.94693013f}, +{0.30901699f, -0.95105652f}, {0.29654157f, -0.95501994f}, +{0.28401534f, -0.95881973f}, {0.27144045f, -0.96245524f}, +{0.25881905f, -0.96592583f}, {0.24615329f, -0.96923091f}, +{0.23344536f, -0.97236992f}, {0.22069744f, -0.97534232f}, +{0.20791169f, -0.97814760f}, {0.19509032f, -0.98078528f}, +{0.18223553f, -0.98325491f}, {0.16934950f, -0.98555606f}, +{0.15643447f, -0.98768834f}, {0.14349262f, -0.98965139f}, +{0.13052619f, -0.99144486f}, {0.11753740f, -0.99306846f}, +{0.10452846f, -0.99452190f}, {0.091501619f, -0.99580493f}, +{0.078459096f, -0.99691733f}, {0.065403129f, -0.99785892f}, +{0.052335956f, -0.99862953f}, {0.039259816f, -0.99922904f}, +{0.026176948f, -0.99965732f}, {0.013089596f, -0.99991433f}, +{6.1230318e-17f, -1.0000000f}, {-0.013089596f, -0.99991433f}, +{-0.026176948f, -0.99965732f}, {-0.039259816f, -0.99922904f}, +{-0.052335956f, -0.99862953f}, {-0.065403129f, -0.99785892f}, +{-0.078459096f, -0.99691733f}, {-0.091501619f, -0.99580493f}, +{-0.10452846f, -0.99452190f}, {-0.11753740f, -0.99306846f}, +{-0.13052619f, -0.99144486f}, {-0.14349262f, -0.98965139f}, +{-0.15643447f, -0.98768834f}, {-0.16934950f, -0.98555606f}, +{-0.18223553f, -0.98325491f}, {-0.19509032f, -0.98078528f}, +{-0.20791169f, -0.97814760f}, {-0.22069744f, -0.97534232f}, +{-0.23344536f, -0.97236992f}, {-0.24615329f, -0.96923091f}, +{-0.25881905f, -0.96592583f}, {-0.27144045f, -0.96245524f}, +{-0.28401534f, -0.95881973f}, {-0.29654157f, -0.95501994f}, +{-0.30901699f, -0.95105652f}, {-0.32143947f, -0.94693013f}, +{-0.33380686f, -0.94264149f}, {-0.34611706f, -0.93819134f}, +{-0.35836795f, -0.93358043f}, {-0.37055744f, -0.92880955f}, +{-0.38268343f, -0.92387953f}, {-0.39474386f, -0.91879121f}, +{-0.40673664f, -0.91354546f}, {-0.41865974f, -0.90814317f}, +{-0.43051110f, -0.90258528f}, {-0.44228869f, -0.89687274f}, +{-0.45399050f, -0.89100652f}, {-0.46561452f, -0.88498764f}, +{-0.47715876f, -0.87881711f}, {-0.48862124f, -0.87249601f}, +{-0.50000000f, -0.86602540f}, {-0.51129309f, -0.85940641f}, +{-0.52249856f, -0.85264016f}, {-0.53361452f, -0.84572782f}, +{-0.54463904f, -0.83867057f}, {-0.55557023f, -0.83146961f}, +{-0.56640624f, -0.82412619f}, {-0.57714519f, -0.81664156f}, +{-0.58778525f, -0.80901699f}, {-0.59832460f, -0.80125381f}, +{-0.60876143f, -0.79335334f}, {-0.61909395f, -0.78531693f}, +{-0.62932039f, -0.77714596f}, {-0.63943900f, -0.76884183f}, +{-0.64944805f, -0.76040597f}, {-0.65934582f, -0.75183981f}, +{-0.66913061f, -0.74314483f}, {-0.67880075f, -0.73432251f}, +{-0.68835458f, -0.72537437f}, {-0.69779046f, -0.71630194f}, +{-0.70710678f, -0.70710678f}, {-0.71630194f, -0.69779046f}, +{-0.72537437f, -0.68835458f}, {-0.73432251f, -0.67880075f}, +{-0.74314483f, -0.66913061f}, {-0.75183981f, -0.65934582f}, +{-0.76040597f, -0.64944805f}, {-0.76884183f, -0.63943900f}, +{-0.77714596f, -0.62932039f}, {-0.78531693f, -0.61909395f}, +{-0.79335334f, -0.60876143f}, {-0.80125381f, -0.59832460f}, +{-0.80901699f, -0.58778525f}, {-0.81664156f, -0.57714519f}, +{-0.82412619f, -0.56640624f}, {-0.83146961f, -0.55557023f}, +{-0.83867057f, -0.54463904f}, {-0.84572782f, -0.53361452f}, +{-0.85264016f, -0.52249856f}, {-0.85940641f, -0.51129309f}, +{-0.86602540f, -0.50000000f}, {-0.87249601f, -0.48862124f}, +{-0.87881711f, -0.47715876f}, {-0.88498764f, -0.46561452f}, +{-0.89100652f, -0.45399050f}, {-0.89687274f, -0.44228869f}, +{-0.90258528f, -0.43051110f}, {-0.90814317f, -0.41865974f}, +{-0.91354546f, -0.40673664f}, {-0.91879121f, -0.39474386f}, +{-0.92387953f, -0.38268343f}, {-0.92880955f, -0.37055744f}, +{-0.93358043f, -0.35836795f}, {-0.93819134f, -0.34611706f}, +{-0.94264149f, -0.33380686f}, {-0.94693013f, -0.32143947f}, +{-0.95105652f, -0.30901699f}, {-0.95501994f, -0.29654157f}, +{-0.95881973f, -0.28401534f}, {-0.96245524f, -0.27144045f}, +{-0.96592583f, -0.25881905f}, {-0.96923091f, -0.24615329f}, +{-0.97236992f, -0.23344536f}, {-0.97534232f, -0.22069744f}, +{-0.97814760f, -0.20791169f}, {-0.98078528f, -0.19509032f}, +{-0.98325491f, -0.18223553f}, {-0.98555606f, -0.16934950f}, +{-0.98768834f, -0.15643447f}, {-0.98965139f, -0.14349262f}, +{-0.99144486f, -0.13052619f}, {-0.99306846f, -0.11753740f}, +{-0.99452190f, -0.10452846f}, {-0.99580493f, -0.091501619f}, +{-0.99691733f, -0.078459096f}, {-0.99785892f, -0.065403129f}, +{-0.99862953f, -0.052335956f}, {-0.99922904f, -0.039259816f}, +{-0.99965732f, -0.026176948f}, {-0.99991433f, -0.013089596f}, +{-1.0000000f, -1.2246064e-16f}, {-0.99991433f, 0.013089596f}, +{-0.99965732f, 0.026176948f}, {-0.99922904f, 0.039259816f}, +{-0.99862953f, 0.052335956f}, {-0.99785892f, 0.065403129f}, +{-0.99691733f, 0.078459096f}, {-0.99580493f, 0.091501619f}, +{-0.99452190f, 0.10452846f}, {-0.99306846f, 0.11753740f}, +{-0.99144486f, 0.13052619f}, {-0.98965139f, 0.14349262f}, +{-0.98768834f, 0.15643447f}, {-0.98555606f, 0.16934950f}, +{-0.98325491f, 0.18223553f}, {-0.98078528f, 0.19509032f}, +{-0.97814760f, 0.20791169f}, {-0.97534232f, 0.22069744f}, +{-0.97236992f, 0.23344536f}, {-0.96923091f, 0.24615329f}, +{-0.96592583f, 0.25881905f}, {-0.96245524f, 0.27144045f}, +{-0.95881973f, 0.28401534f}, {-0.95501994f, 0.29654157f}, +{-0.95105652f, 0.30901699f}, {-0.94693013f, 0.32143947f}, +{-0.94264149f, 0.33380686f}, {-0.93819134f, 0.34611706f}, +{-0.93358043f, 0.35836795f}, {-0.92880955f, 0.37055744f}, +{-0.92387953f, 0.38268343f}, {-0.91879121f, 0.39474386f}, +{-0.91354546f, 0.40673664f}, {-0.90814317f, 0.41865974f}, +{-0.90258528f, 0.43051110f}, {-0.89687274f, 0.44228869f}, +{-0.89100652f, 0.45399050f}, {-0.88498764f, 0.46561452f}, +{-0.87881711f, 0.47715876f}, {-0.87249601f, 0.48862124f}, +{-0.86602540f, 0.50000000f}, {-0.85940641f, 0.51129309f}, +{-0.85264016f, 0.52249856f}, {-0.84572782f, 0.53361452f}, +{-0.83867057f, 0.54463904f}, {-0.83146961f, 0.55557023f}, +{-0.82412619f, 0.56640624f}, {-0.81664156f, 0.57714519f}, +{-0.80901699f, 0.58778525f}, {-0.80125381f, 0.59832460f}, +{-0.79335334f, 0.60876143f}, {-0.78531693f, 0.61909395f}, +{-0.77714596f, 0.62932039f}, {-0.76884183f, 0.63943900f}, +{-0.76040597f, 0.64944805f}, {-0.75183981f, 0.65934582f}, +{-0.74314483f, 0.66913061f}, {-0.73432251f, 0.67880075f}, +{-0.72537437f, 0.68835458f}, {-0.71630194f, 0.69779046f}, +{-0.70710678f, 0.70710678f}, {-0.69779046f, 0.71630194f}, +{-0.68835458f, 0.72537437f}, {-0.67880075f, 0.73432251f}, +{-0.66913061f, 0.74314483f}, {-0.65934582f, 0.75183981f}, +{-0.64944805f, 0.76040597f}, {-0.63943900f, 0.76884183f}, +{-0.62932039f, 0.77714596f}, {-0.61909395f, 0.78531693f}, +{-0.60876143f, 0.79335334f}, {-0.59832460f, 0.80125381f}, +{-0.58778525f, 0.80901699f}, {-0.57714519f, 0.81664156f}, +{-0.56640624f, 0.82412619f}, {-0.55557023f, 0.83146961f}, +{-0.54463904f, 0.83867057f}, {-0.53361452f, 0.84572782f}, +{-0.52249856f, 0.85264016f}, {-0.51129309f, 0.85940641f}, +{-0.50000000f, 0.86602540f}, {-0.48862124f, 0.87249601f}, +{-0.47715876f, 0.87881711f}, {-0.46561452f, 0.88498764f}, +{-0.45399050f, 0.89100652f}, {-0.44228869f, 0.89687274f}, +{-0.43051110f, 0.90258528f}, {-0.41865974f, 0.90814317f}, +{-0.40673664f, 0.91354546f}, {-0.39474386f, 0.91879121f}, +{-0.38268343f, 0.92387953f}, {-0.37055744f, 0.92880955f}, +{-0.35836795f, 0.93358043f}, {-0.34611706f, 0.93819134f}, +{-0.33380686f, 0.94264149f}, {-0.32143947f, 0.94693013f}, +{-0.30901699f, 0.95105652f}, {-0.29654157f, 0.95501994f}, +{-0.28401534f, 0.95881973f}, {-0.27144045f, 0.96245524f}, +{-0.25881905f, 0.96592583f}, {-0.24615329f, 0.96923091f}, +{-0.23344536f, 0.97236992f}, {-0.22069744f, 0.97534232f}, +{-0.20791169f, 0.97814760f}, {-0.19509032f, 0.98078528f}, +{-0.18223553f, 0.98325491f}, {-0.16934950f, 0.98555606f}, +{-0.15643447f, 0.98768834f}, {-0.14349262f, 0.98965139f}, +{-0.13052619f, 0.99144486f}, {-0.11753740f, 0.99306846f}, +{-0.10452846f, 0.99452190f}, {-0.091501619f, 0.99580493f}, +{-0.078459096f, 0.99691733f}, {-0.065403129f, 0.99785892f}, +{-0.052335956f, 0.99862953f}, {-0.039259816f, 0.99922904f}, +{-0.026176948f, 0.99965732f}, {-0.013089596f, 0.99991433f}, +{-1.8369095e-16f, 1.0000000f}, {0.013089596f, 0.99991433f}, +{0.026176948f, 0.99965732f}, {0.039259816f, 0.99922904f}, +{0.052335956f, 0.99862953f}, {0.065403129f, 0.99785892f}, +{0.078459096f, 0.99691733f}, {0.091501619f, 0.99580493f}, +{0.10452846f, 0.99452190f}, {0.11753740f, 0.99306846f}, +{0.13052619f, 0.99144486f}, {0.14349262f, 0.98965139f}, +{0.15643447f, 0.98768834f}, {0.16934950f, 0.98555606f}, +{0.18223553f, 0.98325491f}, {0.19509032f, 0.98078528f}, +{0.20791169f, 0.97814760f}, {0.22069744f, 0.97534232f}, +{0.23344536f, 0.97236992f}, {0.24615329f, 0.96923091f}, +{0.25881905f, 0.96592583f}, {0.27144045f, 0.96245524f}, +{0.28401534f, 0.95881973f}, {0.29654157f, 0.95501994f}, +{0.30901699f, 0.95105652f}, {0.32143947f, 0.94693013f}, +{0.33380686f, 0.94264149f}, {0.34611706f, 0.93819134f}, +{0.35836795f, 0.93358043f}, {0.37055744f, 0.92880955f}, +{0.38268343f, 0.92387953f}, {0.39474386f, 0.91879121f}, +{0.40673664f, 0.91354546f}, {0.41865974f, 0.90814317f}, +{0.43051110f, 0.90258528f}, {0.44228869f, 0.89687274f}, +{0.45399050f, 0.89100652f}, {0.46561452f, 0.88498764f}, +{0.47715876f, 0.87881711f}, {0.48862124f, 0.87249601f}, +{0.50000000f, 0.86602540f}, {0.51129309f, 0.85940641f}, +{0.52249856f, 0.85264016f}, {0.53361452f, 0.84572782f}, +{0.54463904f, 0.83867057f}, {0.55557023f, 0.83146961f}, +{0.56640624f, 0.82412619f}, {0.57714519f, 0.81664156f}, +{0.58778525f, 0.80901699f}, {0.59832460f, 0.80125381f}, +{0.60876143f, 0.79335334f}, {0.61909395f, 0.78531693f}, +{0.62932039f, 0.77714596f}, {0.63943900f, 0.76884183f}, +{0.64944805f, 0.76040597f}, {0.65934582f, 0.75183981f}, +{0.66913061f, 0.74314483f}, {0.67880075f, 0.73432251f}, +{0.68835458f, 0.72537437f}, {0.69779046f, 0.71630194f}, +{0.70710678f, 0.70710678f}, {0.71630194f, 0.69779046f}, +{0.72537437f, 0.68835458f}, {0.73432251f, 0.67880075f}, +{0.74314483f, 0.66913061f}, {0.75183981f, 0.65934582f}, +{0.76040597f, 0.64944805f}, {0.76884183f, 0.63943900f}, +{0.77714596f, 0.62932039f}, {0.78531693f, 0.61909395f}, +{0.79335334f, 0.60876143f}, {0.80125381f, 0.59832460f}, +{0.80901699f, 0.58778525f}, {0.81664156f, 0.57714519f}, +{0.82412619f, 0.56640624f}, {0.83146961f, 0.55557023f}, +{0.83867057f, 0.54463904f}, {0.84572782f, 0.53361452f}, +{0.85264016f, 0.52249856f}, {0.85940641f, 0.51129309f}, +{0.86602540f, 0.50000000f}, {0.87249601f, 0.48862124f}, +{0.87881711f, 0.47715876f}, {0.88498764f, 0.46561452f}, +{0.89100652f, 0.45399050f}, {0.89687274f, 0.44228869f}, +{0.90258528f, 0.43051110f}, {0.90814317f, 0.41865974f}, +{0.91354546f, 0.40673664f}, {0.91879121f, 0.39474386f}, +{0.92387953f, 0.38268343f}, {0.92880955f, 0.37055744f}, +{0.93358043f, 0.35836795f}, {0.93819134f, 0.34611706f}, +{0.94264149f, 0.33380686f}, {0.94693013f, 0.32143947f}, +{0.95105652f, 0.30901699f}, {0.95501994f, 0.29654157f}, +{0.95881973f, 0.28401534f}, {0.96245524f, 0.27144045f}, +{0.96592583f, 0.25881905f}, {0.96923091f, 0.24615329f}, +{0.97236992f, 0.23344536f}, {0.97534232f, 0.22069744f}, +{0.97814760f, 0.20791169f}, {0.98078528f, 0.19509032f}, +{0.98325491f, 0.18223553f}, {0.98555606f, 0.16934950f}, +{0.98768834f, 0.15643447f}, {0.98965139f, 0.14349262f}, +{0.99144486f, 0.13052619f}, {0.99306846f, 0.11753740f}, +{0.99452190f, 0.10452846f}, {0.99580493f, 0.091501619f}, +{0.99691733f, 0.078459096f}, {0.99785892f, 0.065403129f}, +{0.99862953f, 0.052335956f}, {0.99922904f, 0.039259816f}, +{0.99965732f, 0.026176948f}, {0.99991433f, 0.013089596f}, +}; +#ifndef FFT_BITREV480 +#define FFT_BITREV480 +static const opus_int16 fft_bitrev480[480] = { +0, 96, 192, 288, 384, 32, 128, 224, 320, 416, 64, 160, 256, 352, 448, +8, 104, 200, 296, 392, 40, 136, 232, 328, 424, 72, 168, 264, 360, 456, +16, 112, 208, 304, 400, 48, 144, 240, 336, 432, 80, 176, 272, 368, 464, +24, 120, 216, 312, 408, 56, 152, 248, 344, 440, 88, 184, 280, 376, 472, +4, 100, 196, 292, 388, 36, 132, 228, 324, 420, 68, 164, 260, 356, 452, +12, 108, 204, 300, 396, 44, 140, 236, 332, 428, 76, 172, 268, 364, 460, +20, 116, 212, 308, 404, 52, 148, 244, 340, 436, 84, 180, 276, 372, 468, +28, 124, 220, 316, 412, 60, 156, 252, 348, 444, 92, 188, 284, 380, 476, +1, 97, 193, 289, 385, 33, 129, 225, 321, 417, 65, 161, 257, 353, 449, +9, 105, 201, 297, 393, 41, 137, 233, 329, 425, 73, 169, 265, 361, 457, +17, 113, 209, 305, 401, 49, 145, 241, 337, 433, 81, 177, 273, 369, 465, +25, 121, 217, 313, 409, 57, 153, 249, 345, 441, 89, 185, 281, 377, 473, +5, 101, 197, 293, 389, 37, 133, 229, 325, 421, 69, 165, 261, 357, 453, +13, 109, 205, 301, 397, 45, 141, 237, 333, 429, 77, 173, 269, 365, 461, +21, 117, 213, 309, 405, 53, 149, 245, 341, 437, 85, 181, 277, 373, 469, +29, 125, 221, 317, 413, 61, 157, 253, 349, 445, 93, 189, 285, 381, 477, +2, 98, 194, 290, 386, 34, 130, 226, 322, 418, 66, 162, 258, 354, 450, +10, 106, 202, 298, 394, 42, 138, 234, 330, 426, 74, 170, 266, 362, 458, +18, 114, 210, 306, 402, 50, 146, 242, 338, 434, 82, 178, 274, 370, 466, +26, 122, 218, 314, 410, 58, 154, 250, 346, 442, 90, 186, 282, 378, 474, +6, 102, 198, 294, 390, 38, 134, 230, 326, 422, 70, 166, 262, 358, 454, +14, 110, 206, 302, 398, 46, 142, 238, 334, 430, 78, 174, 270, 366, 462, +22, 118, 214, 310, 406, 54, 150, 246, 342, 438, 86, 182, 278, 374, 470, +30, 126, 222, 318, 414, 62, 158, 254, 350, 446, 94, 190, 286, 382, 478, +3, 99, 195, 291, 387, 35, 131, 227, 323, 419, 67, 163, 259, 355, 451, +11, 107, 203, 299, 395, 43, 139, 235, 331, 427, 75, 171, 267, 363, 459, +19, 115, 211, 307, 403, 51, 147, 243, 339, 435, 83, 179, 275, 371, 467, +27, 123, 219, 315, 411, 59, 155, 251, 347, 443, 91, 187, 283, 379, 475, +7, 103, 199, 295, 391, 39, 135, 231, 327, 423, 71, 167, 263, 359, 455, +15, 111, 207, 303, 399, 47, 143, 239, 335, 431, 79, 175, 271, 367, 463, +23, 119, 215, 311, 407, 55, 151, 247, 343, 439, 87, 183, 279, 375, 471, +31, 127, 223, 319, 415, 63, 159, 255, 351, 447, 95, 191, 287, 383, 479, +}; +#endif + +#ifndef FFT_BITREV240 +#define FFT_BITREV240 +static const opus_int16 fft_bitrev240[240] = { +0, 48, 96, 144, 192, 16, 64, 112, 160, 208, 32, 80, 128, 176, 224, +4, 52, 100, 148, 196, 20, 68, 116, 164, 212, 36, 84, 132, 180, 228, +8, 56, 104, 152, 200, 24, 72, 120, 168, 216, 40, 88, 136, 184, 232, +12, 60, 108, 156, 204, 28, 76, 124, 172, 220, 44, 92, 140, 188, 236, +1, 49, 97, 145, 193, 17, 65, 113, 161, 209, 33, 81, 129, 177, 225, +5, 53, 101, 149, 197, 21, 69, 117, 165, 213, 37, 85, 133, 181, 229, +9, 57, 105, 153, 201, 25, 73, 121, 169, 217, 41, 89, 137, 185, 233, +13, 61, 109, 157, 205, 29, 77, 125, 173, 221, 45, 93, 141, 189, 237, +2, 50, 98, 146, 194, 18, 66, 114, 162, 210, 34, 82, 130, 178, 226, +6, 54, 102, 150, 198, 22, 70, 118, 166, 214, 38, 86, 134, 182, 230, +10, 58, 106, 154, 202, 26, 74, 122, 170, 218, 42, 90, 138, 186, 234, +14, 62, 110, 158, 206, 30, 78, 126, 174, 222, 46, 94, 142, 190, 238, +3, 51, 99, 147, 195, 19, 67, 115, 163, 211, 35, 83, 131, 179, 227, +7, 55, 103, 151, 199, 23, 71, 119, 167, 215, 39, 87, 135, 183, 231, +11, 59, 107, 155, 203, 27, 75, 123, 171, 219, 43, 91, 139, 187, 235, +15, 63, 111, 159, 207, 31, 79, 127, 175, 223, 47, 95, 143, 191, 239, +}; +#endif + +#ifndef FFT_BITREV120 +#define FFT_BITREV120 +static const opus_int16 fft_bitrev120[120] = { +0, 24, 48, 72, 96, 8, 32, 56, 80, 104, 16, 40, 64, 88, 112, +4, 28, 52, 76, 100, 12, 36, 60, 84, 108, 20, 44, 68, 92, 116, +1, 25, 49, 73, 97, 9, 33, 57, 81, 105, 17, 41, 65, 89, 113, +5, 29, 53, 77, 101, 13, 37, 61, 85, 109, 21, 45, 69, 93, 117, +2, 26, 50, 74, 98, 10, 34, 58, 82, 106, 18, 42, 66, 90, 114, +6, 30, 54, 78, 102, 14, 38, 62, 86, 110, 22, 46, 70, 94, 118, +3, 27, 51, 75, 99, 11, 35, 59, 83, 107, 19, 43, 67, 91, 115, +7, 31, 55, 79, 103, 15, 39, 63, 87, 111, 23, 47, 71, 95, 119, +}; +#endif + +#ifndef FFT_BITREV60 +#define FFT_BITREV60 +static const opus_int16 fft_bitrev60[60] = { +0, 12, 24, 36, 48, 4, 16, 28, 40, 52, 8, 20, 32, 44, 56, +1, 13, 25, 37, 49, 5, 17, 29, 41, 53, 9, 21, 33, 45, 57, +2, 14, 26, 38, 50, 6, 18, 30, 42, 54, 10, 22, 34, 46, 58, +3, 15, 27, 39, 51, 7, 19, 31, 43, 55, 11, 23, 35, 47, 59, +}; +#endif + +#ifndef FFT_STATE48000_960_0 +#define FFT_STATE48000_960_0 +static const kiss_fft_state fft_state48000_960_0 = { +480, /* nfft */ +0.002083333f, /* scale */ +-1, /* shift */ +{5, 96, 3, 32, 4, 8, 2, 4, 4, 1, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev480, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_480, +#else +NULL, +#endif +}; +#endif + +#ifndef FFT_STATE48000_960_1 +#define FFT_STATE48000_960_1 +static const kiss_fft_state fft_state48000_960_1 = { +240, /* nfft */ +0.004166667f, /* scale */ +1, /* shift */ +{5, 48, 3, 16, 4, 4, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev240, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_240, +#else +NULL, +#endif +}; +#endif + +#ifndef FFT_STATE48000_960_2 +#define FFT_STATE48000_960_2 +static const kiss_fft_state fft_state48000_960_2 = { +120, /* nfft */ +0.008333333f, /* scale */ +2, /* shift */ +{5, 24, 3, 8, 2, 4, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev120, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_120, +#else +NULL, +#endif +}; +#endif + +#ifndef FFT_STATE48000_960_3 +#define FFT_STATE48000_960_3 +static const kiss_fft_state fft_state48000_960_3 = { +60, /* nfft */ +0.016666667f, /* scale */ +3, /* shift */ +{5, 12, 3, 4, 4, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, }, /* factors */ +fft_bitrev60, /* bitrev */ +fft_twiddles48000_960, /* bitrev */ +#ifdef OVERRIDE_FFT +(arch_fft_state *)&cfg_arch_60, +#else +NULL, +#endif +}; +#endif + +#endif + +#ifndef MDCT_TWIDDLES960 +#define MDCT_TWIDDLES960 +static const opus_val16 mdct_twiddles960[1800] = { +0.99999994f, 0.99999321f, 0.99997580f, 0.99994773f, 0.99990886f, +0.99985933f, 0.99979913f, 0.99972820f, 0.99964654f, 0.99955416f, +0.99945110f, 0.99933738f, 0.99921292f, 0.99907774f, 0.99893188f, +0.99877530f, 0.99860805f, 0.99843007f, 0.99824142f, 0.99804211f, +0.99783206f, 0.99761140f, 0.99737996f, 0.99713790f, 0.99688518f, +0.99662173f, 0.99634761f, 0.99606287f, 0.99576741f, 0.99546129f, +0.99514455f, 0.99481714f, 0.99447906f, 0.99413031f, 0.99377096f, +0.99340093f, 0.99302030f, 0.99262899f, 0.99222708f, 0.99181455f, +0.99139136f, 0.99095762f, 0.99051321f, 0.99005818f, 0.98959261f, +0.98911643f, 0.98862964f, 0.98813224f, 0.98762429f, 0.98710573f, +0.98657662f, 0.98603696f, 0.98548669f, 0.98492593f, 0.98435456f, +0.98377270f, 0.98318028f, 0.98257732f, 0.98196387f, 0.98133987f, +0.98070538f, 0.98006040f, 0.97940493f, 0.97873890f, 0.97806245f, +0.97737551f, 0.97667813f, 0.97597027f, 0.97525197f, 0.97452319f, +0.97378403f, 0.97303438f, 0.97227436f, 0.97150391f, 0.97072303f, +0.96993178f, 0.96913016f, 0.96831810f, 0.96749574f, 0.96666300f, +0.96581990f, 0.96496642f, 0.96410263f, 0.96322852f, 0.96234411f, +0.96144938f, 0.96054435f, 0.95962906f, 0.95870346f, 0.95776761f, +0.95682150f, 0.95586514f, 0.95489854f, 0.95392174f, 0.95293468f, +0.95193744f, 0.95093000f, 0.94991243f, 0.94888461f, 0.94784665f, +0.94679856f, 0.94574034f, 0.94467193f, 0.94359344f, 0.94250488f, +0.94140619f, 0.94029742f, 0.93917859f, 0.93804967f, 0.93691075f, +0.93576175f, 0.93460274f, 0.93343377f, 0.93225473f, 0.93106574f, +0.92986679f, 0.92865789f, 0.92743903f, 0.92621022f, 0.92497152f, +0.92372292f, 0.92246443f, 0.92119598f, 0.91991776f, 0.91862965f, +0.91733170f, 0.91602397f, 0.91470635f, 0.91337901f, 0.91204184f, +0.91069490f, 0.90933824f, 0.90797186f, 0.90659571f, 0.90520984f, +0.90381432f, 0.90240908f, 0.90099424f, 0.89956969f, 0.89813554f, +0.89669174f, 0.89523834f, 0.89377540f, 0.89230281f, 0.89082074f, +0.88932908f, 0.88782793f, 0.88631725f, 0.88479710f, 0.88326746f, +0.88172835f, 0.88017982f, 0.87862182f, 0.87705445f, 0.87547767f, +0.87389153f, 0.87229604f, 0.87069118f, 0.86907703f, 0.86745358f, +0.86582077f, 0.86417878f, 0.86252749f, 0.86086690f, 0.85919720f, +0.85751826f, 0.85583007f, 0.85413277f, 0.85242635f, 0.85071075f, +0.84898609f, 0.84725231f, 0.84550947f, 0.84375757f, 0.84199661f, +0.84022665f, 0.83844769f, 0.83665979f, 0.83486289f, 0.83305705f, +0.83124226f, 0.82941860f, 0.82758605f, 0.82574469f, 0.82389444f, +0.82203537f, 0.82016748f, 0.81829083f, 0.81640542f, 0.81451124f, +0.81260836f, 0.81069672f, 0.80877650f, 0.80684757f, 0.80490994f, +0.80296379f, 0.80100900f, 0.79904562f, 0.79707366f, 0.79509324f, +0.79310423f, 0.79110676f, 0.78910083f, 0.78708643f, 0.78506362f, +0.78303236f, 0.78099275f, 0.77894479f, 0.77688843f, 0.77482378f, +0.77275085f, 0.77066964f, 0.76858020f, 0.76648247f, 0.76437658f, +0.76226246f, 0.76014024f, 0.75800985f, 0.75587130f, 0.75372469f, +0.75157005f, 0.74940729f, 0.74723655f, 0.74505776f, 0.74287105f, +0.74067634f, 0.73847371f, 0.73626316f, 0.73404479f, 0.73181850f, +0.72958434f, 0.72734243f, 0.72509271f, 0.72283524f, 0.72057003f, +0.71829706f, 0.71601641f, 0.71372813f, 0.71143216f, 0.70912862f, +0.70681745f, 0.70449871f, 0.70217246f, 0.69983864f, 0.69749737f, +0.69514859f, 0.69279242f, 0.69042879f, 0.68805778f, 0.68567938f, +0.68329364f, 0.68090063f, 0.67850029f, 0.67609268f, 0.67367786f, +0.67125577f, 0.66882652f, 0.66639012f, 0.66394657f, 0.66149592f, +0.65903819f, 0.65657341f, 0.65410155f, 0.65162271f, 0.64913690f, +0.64664418f, 0.64414448f, 0.64163786f, 0.63912445f, 0.63660413f, +0.63407701f, 0.63154310f, 0.62900239f, 0.62645501f, 0.62390089f, +0.62134010f, 0.61877263f, 0.61619854f, 0.61361790f, 0.61103064f, +0.60843682f, 0.60583651f, 0.60322970f, 0.60061646f, 0.59799677f, +0.59537065f, 0.59273821f, 0.59009939f, 0.58745426f, 0.58480281f, +0.58214509f, 0.57948118f, 0.57681108f, 0.57413477f, 0.57145232f, +0.56876373f, 0.56606907f, 0.56336832f, 0.56066155f, 0.55794877f, +0.55523002f, 0.55250537f, 0.54977477f, 0.54703826f, 0.54429591f, +0.54154772f, 0.53879374f, 0.53603399f, 0.53326851f, 0.53049731f, +0.52772039f, 0.52493787f, 0.52214974f, 0.51935595f, 0.51655668f, +0.51375180f, 0.51094145f, 0.50812566f, 0.50530440f, 0.50247771f, +0.49964568f, 0.49680826f, 0.49396557f, 0.49111754f, 0.48826426f, +0.48540577f, 0.48254207f, 0.47967321f, 0.47679919f, 0.47392011f, +0.47103590f, 0.46814668f, 0.46525243f, 0.46235323f, 0.45944905f, +0.45653993f, 0.45362595f, 0.45070711f, 0.44778344f, 0.44485497f, +0.44192174f, 0.43898380f, 0.43604112f, 0.43309379f, 0.43014181f, +0.42718524f, 0.42422408f, 0.42125839f, 0.41828820f, 0.41531351f, +0.41233435f, 0.40935081f, 0.40636289f, 0.40337059f, 0.40037400f, +0.39737311f, 0.39436796f, 0.39135858f, 0.38834500f, 0.38532731f, +0.38230544f, 0.37927949f, 0.37624949f, 0.37321547f, 0.37017745f, +0.36713544f, 0.36408952f, 0.36103970f, 0.35798600f, 0.35492846f, +0.35186714f, 0.34880206f, 0.34573323f, 0.34266070f, 0.33958447f, +0.33650464f, 0.33342120f, 0.33033419f, 0.32724363f, 0.32414958f, +0.32105204f, 0.31795108f, 0.31484672f, 0.31173897f, 0.30862790f, +0.30551350f, 0.30239585f, 0.29927495f, 0.29615086f, 0.29302359f, +0.28989318f, 0.28675964f, 0.28362307f, 0.28048345f, 0.27734083f, +0.27419522f, 0.27104670f, 0.26789525f, 0.26474094f, 0.26158381f, +0.25842386f, 0.25526115f, 0.25209570f, 0.24892756f, 0.24575676f, +0.24258332f, 0.23940729f, 0.23622867f, 0.23304754f, 0.22986393f, +0.22667783f, 0.22348931f, 0.22029841f, 0.21710514f, 0.21390954f, +0.21071166f, 0.20751151f, 0.20430915f, 0.20110460f, 0.19789790f, +0.19468907f, 0.19147816f, 0.18826519f, 0.18505022f, 0.18183327f, +0.17861435f, 0.17539354f, 0.17217083f, 0.16894630f, 0.16571994f, +0.16249183f, 0.15926196f, 0.15603039f, 0.15279715f, 0.14956227f, +0.14632578f, 0.14308774f, 0.13984816f, 0.13660708f, 0.13336454f, +0.13012058f, 0.12687522f, 0.12362850f, 0.12038045f, 0.11713112f, +0.11388054f, 0.11062872f, 0.10737573f, 0.10412160f, 0.10086634f, +0.097609997f, 0.094352618f, 0.091094226f, 0.087834857f, 0.084574550f, +0.081313334f, 0.078051247f, 0.074788325f, 0.071524605f, 0.068260118f, +0.064994894f, 0.061728980f, 0.058462404f, 0.055195201f, 0.051927410f, +0.048659060f, 0.045390189f, 0.042120833f, 0.038851023f, 0.035580799f, +0.032310195f, 0.029039243f, 0.025767982f, 0.022496443f, 0.019224664f, +0.015952680f, 0.012680525f, 0.0094082337f, 0.0061358409f, 0.0028633832f, +-0.00040910527f, -0.0036815894f, -0.0069540343f, -0.010226404f, -0.013498665f, +-0.016770782f, -0.020042717f, -0.023314439f, -0.026585912f, -0.029857099f, +-0.033127967f, -0.036398482f, -0.039668605f, -0.042938303f, -0.046207540f, +-0.049476285f, -0.052744497f, -0.056012146f, -0.059279196f, -0.062545612f, +-0.065811358f, -0.069076397f, -0.072340697f, -0.075604223f, -0.078866936f, +-0.082128808f, -0.085389800f, -0.088649876f, -0.091909006f, -0.095167145f, +-0.098424271f, -0.10168034f, -0.10493532f, -0.10818918f, -0.11144188f, +-0.11469338f, -0.11794366f, -0.12119267f, -0.12444039f, -0.12768677f, +-0.13093179f, -0.13417540f, -0.13741758f, -0.14065829f, -0.14389749f, +-0.14713514f, -0.15037122f, -0.15360570f, -0.15683852f, -0.16006967f, +-0.16329910f, -0.16652679f, -0.16975269f, -0.17297678f, -0.17619900f, +-0.17941935f, -0.18263777f, -0.18585424f, -0.18906870f, -0.19228116f, +-0.19549155f, -0.19869985f, -0.20190603f, -0.20511003f, -0.20831184f, +-0.21151142f, -0.21470875f, -0.21790376f, -0.22109644f, -0.22428675f, +-0.22747467f, -0.23066014f, -0.23384315f, -0.23702365f, -0.24020162f, +-0.24337701f, -0.24654980f, -0.24971995f, -0.25288740f, -0.25605217f, +-0.25921419f, -0.26237345f, -0.26552987f, -0.26868346f, -0.27183419f, +-0.27498198f, -0.27812684f, -0.28126872f, -0.28440759f, -0.28754342f, +-0.29067615f, -0.29380578f, -0.29693225f, -0.30005556f, -0.30317566f, +-0.30629250f, -0.30940607f, -0.31251630f, -0.31562322f, -0.31872672f, +-0.32182685f, -0.32492352f, -0.32801670f, -0.33110636f, -0.33419248f, +-0.33727503f, -0.34035397f, -0.34342924f, -0.34650084f, -0.34956875f, +-0.35263291f, -0.35569328f, -0.35874987f, -0.36180258f, -0.36485144f, +-0.36789638f, -0.37093741f, -0.37397444f, -0.37700745f, -0.38003644f, +-0.38306138f, -0.38608220f, -0.38909888f, -0.39211139f, -0.39511973f, +-0.39812380f, -0.40112361f, -0.40411916f, -0.40711036f, -0.41009718f, +-0.41307965f, -0.41605768f, -0.41903123f, -0.42200032f, -0.42496487f, +-0.42792490f, -0.43088034f, -0.43383113f, -0.43677729f, -0.43971881f, +-0.44265559f, -0.44558764f, -0.44851488f, -0.45143735f, -0.45435500f, +-0.45726776f, -0.46017563f, -0.46307856f, -0.46597654f, -0.46886954f, +-0.47175750f, -0.47464043f, -0.47751826f, -0.48039100f, -0.48325855f, +-0.48612097f, -0.48897815f, -0.49183011f, -0.49467680f, -0.49751821f, +-0.50035429f, -0.50318497f, -0.50601029f, -0.50883019f, -0.51164466f, +-0.51445359f, -0.51725709f, -0.52005500f, -0.52284735f, -0.52563411f, +-0.52841520f, -0.53119069f, -0.53396046f, -0.53672451f, -0.53948283f, +-0.54223537f, -0.54498214f, -0.54772300f, -0.55045801f, -0.55318713f, +-0.55591035f, -0.55862761f, -0.56133890f, -0.56404412f, -0.56674337f, +-0.56943649f, -0.57212353f, -0.57480448f, -0.57747924f, -0.58014780f, +-0.58281022f, -0.58546633f, -0.58811617f, -0.59075975f, -0.59339696f, +-0.59602785f, -0.59865236f, -0.60127044f, -0.60388207f, -0.60648727f, +-0.60908598f, -0.61167812f, -0.61426371f, -0.61684275f, -0.61941516f, +-0.62198097f, -0.62454009f, -0.62709254f, -0.62963831f, -0.63217729f, +-0.63470948f, -0.63723493f, -0.63975352f, -0.64226526f, -0.64477009f, +-0.64726806f, -0.64975911f, -0.65224314f, -0.65472025f, -0.65719032f, +-0.65965337f, -0.66210932f, -0.66455823f, -0.66700000f, -0.66943461f, +-0.67186207f, -0.67428231f, -0.67669535f, -0.67910111f, -0.68149966f, +-0.68389088f, -0.68627477f, -0.68865126f, -0.69102043f, -0.69338220f, +-0.69573659f, -0.69808346f, -0.70042288f, -0.70275480f, -0.70507920f, +-0.70739603f, -0.70970529f, -0.71200693f, -0.71430099f, -0.71658736f, +-0.71886611f, -0.72113711f, -0.72340041f, -0.72565591f, -0.72790372f, +-0.73014367f, -0.73237586f, -0.73460019f, -0.73681659f, -0.73902518f, +-0.74122584f, -0.74341851f, -0.74560326f, -0.74778003f, -0.74994880f, +-0.75210953f, -0.75426215f, -0.75640678f, -0.75854325f, -0.76067162f, +-0.76279181f, -0.76490390f, -0.76700771f, -0.76910341f, -0.77119076f, +-0.77326995f, -0.77534080f, -0.77740335f, -0.77945763f, -0.78150350f, +-0.78354102f, -0.78557014f, -0.78759086f, -0.78960317f, -0.79160696f, +-0.79360235f, -0.79558921f, -0.79756755f, -0.79953730f, -0.80149853f, +-0.80345118f, -0.80539525f, -0.80733067f, -0.80925739f, -0.81117553f, +-0.81308490f, -0.81498563f, -0.81687760f, -0.81876087f, -0.82063532f, +-0.82250100f, -0.82435787f, -0.82620591f, -0.82804507f, -0.82987541f, +-0.83169687f, -0.83350939f, -0.83531296f, -0.83710766f, -0.83889335f, +-0.84067005f, -0.84243774f, -0.84419644f, -0.84594607f, -0.84768665f, +-0.84941816f, -0.85114056f, -0.85285389f, -0.85455805f, -0.85625303f, +-0.85793889f, -0.85961550f, -0.86128294f, -0.86294121f, -0.86459017f, +-0.86622989f, -0.86786032f, -0.86948150f, -0.87109333f, -0.87269586f, +-0.87428904f, -0.87587279f, -0.87744725f, -0.87901229f, -0.88056785f, +-0.88211405f, -0.88365078f, -0.88517809f, -0.88669586f, -0.88820416f, +-0.88970292f, -0.89119220f, -0.89267188f, -0.89414203f, -0.89560264f, +-0.89705360f, -0.89849502f, -0.89992678f, -0.90134889f, -0.90276134f, +-0.90416414f, -0.90555727f, -0.90694070f, -0.90831441f, -0.90967834f, +-0.91103262f, -0.91237706f, -0.91371179f, -0.91503674f, -0.91635185f, +-0.91765714f, -0.91895264f, -0.92023826f, -0.92151409f, -0.92277998f, +-0.92403603f, -0.92528218f, -0.92651838f, -0.92774469f, -0.92896110f, +-0.93016750f, -0.93136400f, -0.93255049f, -0.93372697f, -0.93489349f, +-0.93604994f, -0.93719643f, -0.93833286f, -0.93945926f, -0.94057560f, +-0.94168180f, -0.94277799f, -0.94386405f, -0.94494003f, -0.94600588f, +-0.94706154f, -0.94810712f, -0.94914252f, -0.95016778f, -0.95118284f, +-0.95218778f, -0.95318246f, -0.95416695f, -0.95514119f, -0.95610523f, +-0.95705903f, -0.95800257f, -0.95893586f, -0.95985889f, -0.96077162f, +-0.96167403f, -0.96256620f, -0.96344805f, -0.96431959f, -0.96518075f, +-0.96603161f, -0.96687216f, -0.96770233f, -0.96852213f, -0.96933156f, +-0.97013056f, -0.97091925f, -0.97169751f, -0.97246534f, -0.97322279f, +-0.97396982f, -0.97470641f, -0.97543252f, -0.97614825f, -0.97685349f, +-0.97754824f, -0.97823256f, -0.97890645f, -0.97956979f, -0.98022264f, +-0.98086500f, -0.98149687f, -0.98211825f, -0.98272908f, -0.98332942f, +-0.98391914f, -0.98449844f, -0.98506713f, -0.98562527f, -0.98617285f, +-0.98670989f, -0.98723638f, -0.98775226f, -0.98825759f, -0.98875231f, +-0.98923647f, -0.98971003f, -0.99017298f, -0.99062532f, -0.99106705f, +-0.99149817f, -0.99191868f, -0.99232858f, -0.99272782f, -0.99311644f, +-0.99349445f, -0.99386179f, -0.99421853f, -0.99456459f, -0.99489999f, +-0.99522477f, -0.99553883f, -0.99584228f, -0.99613506f, -0.99641716f, +-0.99668860f, -0.99694937f, -0.99719942f, -0.99743885f, -0.99766755f, +-0.99788558f, -0.99809295f, -0.99828959f, -0.99847561f, -0.99865085f, +-0.99881548f, -0.99896932f, -0.99911255f, -0.99924499f, -0.99936682f, +-0.99947786f, -0.99957830f, -0.99966794f, -0.99974692f, -0.99981517f, +-0.99987274f, -0.99991959f, -0.99995571f, -0.99998116f, -0.99999589f, +0.99999964f, 0.99997288f, 0.99990326f, 0.99979085f, 0.99963558f, +0.99943751f, 0.99919659f, 0.99891287f, 0.99858636f, 0.99821711f, +0.99780506f, 0.99735034f, 0.99685282f, 0.99631262f, 0.99572974f, +0.99510419f, 0.99443603f, 0.99372530f, 0.99297196f, 0.99217612f, +0.99133772f, 0.99045694f, 0.98953366f, 0.98856801f, 0.98756003f, +0.98650974f, 0.98541719f, 0.98428243f, 0.98310548f, 0.98188645f, +0.98062533f, 0.97932225f, 0.97797716f, 0.97659022f, 0.97516143f, +0.97369087f, 0.97217858f, 0.97062469f, 0.96902919f, 0.96739221f, +0.96571374f, 0.96399397f, 0.96223283f, 0.96043050f, 0.95858705f, +0.95670253f, 0.95477700f, 0.95281059f, 0.95080340f, 0.94875544f, +0.94666684f, 0.94453770f, 0.94236809f, 0.94015813f, 0.93790787f, +0.93561745f, 0.93328691f, 0.93091643f, 0.92850608f, 0.92605597f, +0.92356616f, 0.92103678f, 0.91846794f, 0.91585976f, 0.91321236f, +0.91052586f, 0.90780038f, 0.90503591f, 0.90223277f, 0.89939094f, +0.89651060f, 0.89359182f, 0.89063478f, 0.88763964f, 0.88460642f, +0.88153529f, 0.87842643f, 0.87527996f, 0.87209594f, 0.86887461f, +0.86561602f, 0.86232042f, 0.85898781f, 0.85561842f, 0.85221243f, +0.84876984f, 0.84529096f, 0.84177583f, 0.83822471f, 0.83463764f, +0.83101481f, 0.82735640f, 0.82366252f, 0.81993335f, 0.81616908f, +0.81236988f, 0.80853581f, 0.80466717f, 0.80076402f, 0.79682660f, +0.79285502f, 0.78884947f, 0.78481019f, 0.78073722f, 0.77663082f, +0.77249116f, 0.76831841f, 0.76411277f, 0.75987434f, 0.75560343f, +0.75130010f, 0.74696463f, 0.74259710f, 0.73819780f, 0.73376691f, +0.72930455f, 0.72481096f, 0.72028631f, 0.71573079f, 0.71114463f, +0.70652801f, 0.70188117f, 0.69720417f, 0.69249737f, 0.68776089f, +0.68299496f, 0.67819971f, 0.67337549f, 0.66852236f, 0.66364062f, +0.65873051f, 0.65379208f, 0.64882571f, 0.64383155f, 0.63880974f, +0.63376063f, 0.62868434f, 0.62358117f, 0.61845124f, 0.61329484f, +0.60811216f, 0.60290343f, 0.59766883f, 0.59240872f, 0.58712316f, +0.58181250f, 0.57647687f, 0.57111657f, 0.56573176f, 0.56032276f, +0.55488980f, 0.54943299f, 0.54395270f, 0.53844911f, 0.53292239f, +0.52737290f, 0.52180082f, 0.51620632f, 0.51058978f, 0.50495136f, +0.49929130f, 0.49360985f, 0.48790723f, 0.48218375f, 0.47643960f, +0.47067502f, 0.46489030f, 0.45908567f, 0.45326138f, 0.44741765f, +0.44155475f, 0.43567297f, 0.42977250f, 0.42385364f, 0.41791660f, +0.41196167f, 0.40598908f, 0.39999911f, 0.39399201f, 0.38796803f, +0.38192743f, 0.37587047f, 0.36979741f, 0.36370850f, 0.35760403f, +0.35148421f, 0.34534934f, 0.33919969f, 0.33303553f, 0.32685706f, +0.32066461f, 0.31445843f, 0.30823877f, 0.30200592f, 0.29576012f, +0.28950164f, 0.28323078f, 0.27694780f, 0.27065292f, 0.26434645f, +0.25802869f, 0.25169984f, 0.24536023f, 0.23901010f, 0.23264973f, +0.22627939f, 0.21989937f, 0.21350993f, 0.20711134f, 0.20070387f, +0.19428782f, 0.18786344f, 0.18143101f, 0.17499080f, 0.16854310f, +0.16208819f, 0.15562633f, 0.14915779f, 0.14268288f, 0.13620184f, +0.12971498f, 0.12322257f, 0.11672486f, 0.11022217f, 0.10371475f, +0.097202882f, 0.090686858f, 0.084166944f, 0.077643424f, 0.071116582f, +0.064586692f, 0.058054037f, 0.051518895f, 0.044981543f, 0.038442269f, +0.031901345f, 0.025359053f, 0.018815678f, 0.012271495f, 0.0057267868f, +-0.00081816671f, -0.0073630852f, -0.013907688f, -0.020451695f, -0.026994826f, +-0.033536803f, -0.040077340f, -0.046616159f, -0.053152986f, -0.059687532f, +-0.066219524f, -0.072748676f, -0.079274714f, -0.085797355f, -0.092316322f, +-0.098831341f, -0.10534211f, -0.11184838f, -0.11834986f, -0.12484626f, +-0.13133731f, -0.13782275f, -0.14430228f, -0.15077563f, -0.15724251f, +-0.16370267f, -0.17015581f, -0.17660165f, -0.18303993f, -0.18947038f, +-0.19589271f, -0.20230664f, -0.20871192f, -0.21510825f, -0.22149536f, +-0.22787298f, -0.23424086f, -0.24059868f, -0.24694622f, -0.25328314f, +-0.25960925f, -0.26592422f, -0.27222782f, -0.27851975f, -0.28479972f, +-0.29106751f, -0.29732284f, -0.30356544f, -0.30979502f, -0.31601134f, +-0.32221413f, -0.32840309f, -0.33457801f, -0.34073856f, -0.34688455f, +-0.35301566f, -0.35913166f, -0.36523229f, -0.37131724f, -0.37738630f, +-0.38343921f, -0.38947567f, -0.39549544f, -0.40149832f, -0.40748394f, +-0.41345215f, -0.41940263f, -0.42533514f, -0.43124944f, -0.43714526f, +-0.44302234f, -0.44888046f, -0.45471936f, -0.46053877f, -0.46633846f, +-0.47211814f, -0.47787762f, -0.48361665f, -0.48933494f, -0.49503228f, +-0.50070840f, -0.50636309f, -0.51199609f, -0.51760709f, -0.52319598f, +-0.52876246f, -0.53430629f, -0.53982723f, -0.54532504f, -0.55079949f, +-0.55625033f, -0.56167740f, -0.56708032f, -0.57245898f, -0.57781315f, +-0.58314258f, -0.58844697f, -0.59372622f, -0.59897995f, -0.60420811f, +-0.60941035f, -0.61458647f, -0.61973625f, -0.62485951f, -0.62995601f, +-0.63502556f, -0.64006782f, -0.64508271f, -0.65007001f, -0.65502942f, +-0.65996075f, -0.66486382f, -0.66973841f, -0.67458433f, -0.67940134f, +-0.68418926f, -0.68894786f, -0.69367695f, -0.69837630f, -0.70304573f, +-0.70768511f, -0.71229410f, -0.71687263f, -0.72142041f, -0.72593731f, +-0.73042315f, -0.73487765f, -0.73930067f, -0.74369204f, -0.74805158f, +-0.75237900f, -0.75667429f, -0.76093709f, -0.76516730f, -0.76936477f, +-0.77352923f, -0.77766061f, -0.78175867f, -0.78582323f, -0.78985411f, +-0.79385114f, -0.79781419f, -0.80174309f, -0.80563760f, -0.80949765f, +-0.81332302f, -0.81711352f, -0.82086903f, -0.82458937f, -0.82827437f, +-0.83192390f, -0.83553779f, -0.83911592f, -0.84265804f, -0.84616417f, +-0.84963393f, -0.85306740f, -0.85646427f, -0.85982448f, -0.86314780f, +-0.86643422f, -0.86968350f, -0.87289548f, -0.87607014f, -0.87920725f, +-0.88230664f, -0.88536829f, -0.88839203f, -0.89137769f, -0.89432514f, +-0.89723432f, -0.90010506f, -0.90293723f, -0.90573072f, -0.90848541f, +-0.91120118f, -0.91387796f, -0.91651553f, -0.91911387f, -0.92167282f, +-0.92419231f, -0.92667222f, -0.92911243f, -0.93151283f, -0.93387336f, +-0.93619382f, -0.93847424f, -0.94071442f, -0.94291431f, -0.94507378f, +-0.94719279f, -0.94927126f, -0.95130903f, -0.95330608f, -0.95526224f, +-0.95717752f, -0.95905179f, -0.96088499f, -0.96267700f, -0.96442777f, +-0.96613729f, -0.96780539f, -0.96943200f, -0.97101706f, -0.97256058f, +-0.97406244f, -0.97552258f, -0.97694093f, -0.97831738f, -0.97965199f, +-0.98094457f, -0.98219514f, -0.98340368f, -0.98457009f, -0.98569429f, +-0.98677629f, -0.98781598f, -0.98881340f, -0.98976845f, -0.99068111f, +-0.99155134f, -0.99237907f, -0.99316430f, -0.99390697f, -0.99460709f, +-0.99526459f, -0.99587947f, -0.99645168f, -0.99698120f, -0.99746799f, +-0.99791211f, -0.99831343f, -0.99867201f, -0.99898779f, -0.99926084f, +-0.99949104f, -0.99967843f, -0.99982297f, -0.99992472f, -0.99998361f, +0.99999869f, 0.99989158f, 0.99961317f, 0.99916345f, 0.99854255f, +0.99775058f, 0.99678761f, 0.99565387f, 0.99434954f, 0.99287480f, +0.99122995f, 0.98941529f, 0.98743105f, 0.98527765f, 0.98295540f, +0.98046476f, 0.97780609f, 0.97497988f, 0.97198665f, 0.96882683f, +0.96550101f, 0.96200979f, 0.95835376f, 0.95453346f, 0.95054960f, +0.94640291f, 0.94209403f, 0.93762374f, 0.93299282f, 0.92820197f, +0.92325211f, 0.91814411f, 0.91287869f, 0.90745693f, 0.90187967f, +0.89614785f, 0.89026248f, 0.88422459f, 0.87803519f, 0.87169534f, +0.86520612f, 0.85856867f, 0.85178405f, 0.84485358f, 0.83777827f, +0.83055943f, 0.82319832f, 0.81569612f, 0.80805415f, 0.80027372f, +0.79235619f, 0.78430289f, 0.77611518f, 0.76779449f, 0.75934225f, +0.75075996f, 0.74204898f, 0.73321080f, 0.72424710f, 0.71515924f, +0.70594883f, 0.69661748f, 0.68716675f, 0.67759830f, 0.66791373f, +0.65811473f, 0.64820296f, 0.63818014f, 0.62804794f, 0.61780810f, +0.60746247f, 0.59701276f, 0.58646071f, 0.57580817f, 0.56505698f, +0.55420899f, 0.54326600f, 0.53222996f, 0.52110273f, 0.50988621f, +0.49858227f, 0.48719296f, 0.47572014f, 0.46416581f, 0.45253196f, +0.44082057f, 0.42903364f, 0.41717321f, 0.40524128f, 0.39323992f, +0.38117120f, 0.36903715f, 0.35683987f, 0.34458145f, 0.33226398f, +0.31988961f, 0.30746040f, 0.29497850f, 0.28244606f, 0.26986524f, +0.25723818f, 0.24456702f, 0.23185398f, 0.21910121f, 0.20631088f, +0.19348522f, 0.18062639f, 0.16773662f, 0.15481812f, 0.14187308f, +0.12890373f, 0.11591230f, 0.10290100f, 0.089872077f, 0.076827750f, +0.063770257f, 0.050701842f, 0.037624735f, 0.024541186f, 0.011453429f, +-0.0016362892f, -0.014725727f, -0.027812643f, -0.040894791f, -0.053969935f, +-0.067035832f, -0.080090240f, -0.093130924f, -0.10615565f, -0.11916219f, +-0.13214831f, -0.14511178f, -0.15805040f, -0.17096193f, -0.18384418f, +-0.19669491f, -0.20951195f, -0.22229309f, -0.23503613f, -0.24773891f, +-0.26039925f, -0.27301496f, -0.28558388f, -0.29810387f, -0.31057280f, +-0.32298848f, -0.33534884f, -0.34765175f, -0.35989508f, -0.37207675f, +-0.38419467f, -0.39624676f, -0.40823093f, -0.42014518f, -0.43198743f, +-0.44375566f, -0.45544785f, -0.46706200f, -0.47859612f, -0.49004826f, +-0.50141639f, -0.51269865f, -0.52389306f, -0.53499764f, -0.54601061f, +-0.55693001f, -0.56775403f, -0.57848072f, -0.58910829f, -0.59963489f, +-0.61005878f, -0.62037814f, -0.63059121f, -0.64069623f, -0.65069145f, +-0.66057515f, -0.67034572f, -0.68000144f, -0.68954057f, -0.69896162f, +-0.70826286f, -0.71744281f, -0.72649974f, -0.73543227f, -0.74423873f, +-0.75291771f, -0.76146764f, -0.76988715f, -0.77817470f, -0.78632891f, +-0.79434842f, -0.80223179f, -0.80997771f, -0.81758487f, -0.82505190f, +-0.83237761f, -0.83956063f, -0.84659988f, -0.85349399f, -0.86024189f, +-0.86684239f, -0.87329435f, -0.87959671f, -0.88574833f, -0.89174819f, +-0.89759529f, -0.90328854f, -0.90882701f, -0.91420978f, -0.91943592f, +-0.92450452f, -0.92941469f, -0.93416560f, -0.93875647f, -0.94318646f, +-0.94745487f, -0.95156091f, -0.95550388f, -0.95928317f, -0.96289814f, +-0.96634805f, -0.96963239f, -0.97275060f, -0.97570217f, -0.97848648f, +-0.98110318f, -0.98355180f, -0.98583186f, -0.98794299f, -0.98988485f, +-0.99165714f, -0.99325943f, -0.99469161f, -0.99595332f, -0.99704438f, +-0.99796462f, -0.99871385f, -0.99929196f, -0.99969882f, -0.99993443f, +0.99999464f, 0.99956632f, 0.99845290f, 0.99665523f, 0.99417448f, +0.99101239f, 0.98717111f, 0.98265326f, 0.97746199f, 0.97160077f, +0.96507365f, 0.95788515f, 0.95004016f, 0.94154406f, 0.93240267f, +0.92262226f, 0.91220951f, 0.90117162f, 0.88951606f, 0.87725091f, +0.86438453f, 0.85092574f, 0.83688372f, 0.82226819f, 0.80708915f, +0.79135692f, 0.77508235f, 0.75827658f, 0.74095112f, 0.72311783f, +0.70478898f, 0.68597710f, 0.66669506f, 0.64695615f, 0.62677377f, +0.60616189f, 0.58513457f, 0.56370622f, 0.54189157f, 0.51970547f, +0.49716324f, 0.47428027f, 0.45107225f, 0.42755505f, 0.40374488f, +0.37965798f, 0.35531086f, 0.33072025f, 0.30590299f, 0.28087607f, +0.25565663f, 0.23026201f, 0.20470956f, 0.17901683f, 0.15320139f, +0.12728097f, 0.10127331f, 0.075196236f, 0.049067631f, 0.022905400f, +-0.0032725304f, -0.029448219f, -0.055603724f, -0.081721120f, -0.10778251f, +-0.13377003f, -0.15966587f, -0.18545228f, -0.21111161f, -0.23662624f, +-0.26197869f, -0.28715160f, -0.31212771f, -0.33688989f, -0.36142120f, +-0.38570482f, -0.40972409f, -0.43346253f, -0.45690393f, -0.48003218f, +-0.50283146f, -0.52528608f, -0.54738069f, -0.56910020f, -0.59042966f, +-0.61135447f, -0.63186026f, -0.65193301f, -0.67155898f, -0.69072473f, +-0.70941705f, -0.72762316f, -0.74533063f, -0.76252723f, -0.77920127f, +-0.79534131f, -0.81093621f, -0.82597536f, -0.84044844f, -0.85434550f, +-0.86765707f, -0.88037395f, -0.89248747f, -0.90398932f, -0.91487163f, +-0.92512697f, -0.93474823f, -0.94372886f, -0.95206273f, -0.95974404f, +-0.96676767f, -0.97312868f, -0.97882277f, -0.98384601f, -0.98819500f, +-0.99186671f, -0.99485862f, -0.99716878f, -0.99879545f, -0.99973762f, +}; +#endif + +static const CELTMode mode48000_960_120 = { +48000, /* Fs */ +120, /* overlap */ +21, /* nbEBands */ +21, /* effEBands */ +{0.85000610f, 0.0000000f, 1.0000000f, 1.0000000f, }, /* preemph */ +eband5ms, /* eBands */ +3, /* maxLM */ +8, /* nbShortMdcts */ +120, /* shortMdctSize */ +11, /* nbAllocVectors */ +band_allocation, /* allocVectors */ +logN400, /* logN */ +window120, /* window */ +{1920, 3, {&fft_state48000_960_0, &fft_state48000_960_1, &fft_state48000_960_2, &fft_state48000_960_3, }, mdct_twiddles960}, /* mdct */ +{392, cache_index50, cache_bits50, cache_caps50}, /* cache */ +}; + +/* List of all the available modes */ +#define TOTAL_MODES 1 +static const CELTMode * const static_mode_list[TOTAL_MODES] = { +&mode48000_960_120, +}; diff --git a/lib/opus/celt/static_modes_float_arm_ne10.h b/lib/opus/celt/static_modes_float_arm_ne10.h new file mode 100644 index 00000000..66e1abb1 --- /dev/null +++ b/lib/opus/celt/static_modes_float_arm_ne10.h @@ -0,0 +1,404 @@ +/* The contents of this file was automatically generated by + * dump_mode_arm_ne10.c with arguments: 48000 960 + * It contains static definitions for some pre-defined modes. */ +#include + +#ifndef NE10_FFT_PARAMS48000_960 +#define NE10_FFT_PARAMS48000_960 +static const ne10_int32_t ne10_factors_480[64] = { +4, 40, 4, 30, 2, 15, 5, 3, 3, 1, 1, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_int32_t ne10_factors_240[64] = { +3, 20, 4, 15, 5, 3, 3, 1, 1, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_int32_t ne10_factors_120[64] = { +3, 10, 2, 15, 5, 3, 3, 1, 1, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_int32_t ne10_factors_60[64] = { +2, 5, 5, 3, 3, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, +0, 0, 0, 0, }; +static const ne10_fft_cpx_float32_t ne10_twiddles_480[480] = { +{1.0000000f,0.0000000f}, {1.0000000f,-0.0000000f}, {1.0000000f,-0.0000000f}, +{1.0000000f,-0.0000000f}, {0.91354543f,-0.40673664f}, {0.66913056f,-0.74314487f}, +{1.0000000f,-0.0000000f}, {0.66913056f,-0.74314487f}, {-0.10452851f,-0.99452192f}, +{1.0000000f,-0.0000000f}, {0.30901697f,-0.95105654f}, {-0.80901700f,-0.58778518f}, +{1.0000000f,-0.0000000f}, {-0.10452851f,-0.99452192f}, {-0.97814757f,0.20791179f}, +{1.0000000f,-0.0000000f}, {0.97814763f,-0.20791170f}, {0.91354543f,-0.40673664f}, +{0.80901700f,-0.58778524f}, {0.66913056f,-0.74314487f}, {0.49999997f,-0.86602545f}, +{0.30901697f,-0.95105654f}, {0.10452842f,-0.99452192f}, {-0.10452851f,-0.99452192f}, +{-0.30901703f,-0.95105648f}, {-0.50000006f,-0.86602533f}, {-0.66913068f,-0.74314475f}, +{-0.80901700f,-0.58778518f}, {-0.91354549f,-0.40673658f}, {-0.97814763f,-0.20791161f}, +{1.0000000f,-0.0000000f}, {0.99862951f,-0.052335959f}, {0.99452192f,-0.10452846f}, +{0.98768836f,-0.15643448f}, {0.97814763f,-0.20791170f}, {0.96592581f,-0.25881904f}, +{0.95105648f,-0.30901700f}, {0.93358040f,-0.35836795f}, {0.91354543f,-0.40673664f}, +{0.89100653f,-0.45399052f}, {0.86602545f,-0.50000000f}, {0.83867055f,-0.54463905f}, +{0.80901700f,-0.58778524f}, {0.77714598f,-0.62932038f}, {0.74314475f,-0.66913062f}, +{0.70710677f,-0.70710683f}, {0.66913056f,-0.74314487f}, {0.62932038f,-0.77714598f}, +{0.58778524f,-0.80901700f}, {0.54463899f,-0.83867055f}, {0.49999997f,-0.86602545f}, +{0.45399052f,-0.89100653f}, {0.40673661f,-0.91354549f}, {0.35836786f,-0.93358046f}, +{0.30901697f,-0.95105654f}, {0.25881907f,-0.96592581f}, {0.20791166f,-0.97814763f}, +{0.15643437f,-0.98768836f}, {0.10452842f,-0.99452192f}, {0.052335974f,-0.99862951f}, +{1.0000000f,-0.0000000f}, {0.99452192f,-0.10452846f}, {0.97814763f,-0.20791170f}, +{0.95105648f,-0.30901700f}, {0.91354543f,-0.40673664f}, {0.86602545f,-0.50000000f}, +{0.80901700f,-0.58778524f}, {0.74314475f,-0.66913062f}, {0.66913056f,-0.74314487f}, +{0.58778524f,-0.80901700f}, {0.49999997f,-0.86602545f}, {0.40673661f,-0.91354549f}, +{0.30901697f,-0.95105654f}, {0.20791166f,-0.97814763f}, {0.10452842f,-0.99452192f}, +{-4.3711388e-08f,-1.0000000f}, {-0.10452851f,-0.99452192f}, {-0.20791174f,-0.97814757f}, +{-0.30901703f,-0.95105648f}, {-0.40673670f,-0.91354543f}, {-0.50000006f,-0.86602533f}, +{-0.58778518f,-0.80901700f}, {-0.66913068f,-0.74314475f}, {-0.74314493f,-0.66913044f}, +{-0.80901700f,-0.58778518f}, {-0.86602539f,-0.50000006f}, {-0.91354549f,-0.40673658f}, +{-0.95105654f,-0.30901679f}, {-0.97814763f,-0.20791161f}, {-0.99452192f,-0.10452849f}, +{1.0000000f,-0.0000000f}, {0.98768836f,-0.15643448f}, {0.95105648f,-0.30901700f}, +{0.89100653f,-0.45399052f}, {0.80901700f,-0.58778524f}, {0.70710677f,-0.70710683f}, +{0.58778524f,-0.80901700f}, {0.45399052f,-0.89100653f}, {0.30901697f,-0.95105654f}, +{0.15643437f,-0.98768836f}, {-4.3711388e-08f,-1.0000000f}, {-0.15643445f,-0.98768836f}, +{-0.30901703f,-0.95105648f}, {-0.45399061f,-0.89100647f}, {-0.58778518f,-0.80901700f}, +{-0.70710677f,-0.70710677f}, {-0.80901700f,-0.58778518f}, {-0.89100659f,-0.45399037f}, +{-0.95105654f,-0.30901679f}, {-0.98768836f,-0.15643445f}, {-1.0000000f,8.7422777e-08f}, +{-0.98768830f,0.15643461f}, {-0.95105654f,0.30901697f}, {-0.89100653f,0.45399055f}, +{-0.80901694f,0.58778536f}, {-0.70710665f,0.70710689f}, {-0.58778507f,0.80901712f}, +{-0.45399022f,0.89100665f}, {-0.30901709f,0.95105648f}, {-0.15643452f,0.98768830f}, +{1.0000000f,-0.0000000f}, {0.99991435f,-0.013089596f}, {0.99965733f,-0.026176950f}, +{0.99922901f,-0.039259817f}, {0.99862951f,-0.052335959f}, {0.99785894f,-0.065403134f}, +{0.99691731f,-0.078459099f}, {0.99580491f,-0.091501623f}, {0.99452192f,-0.10452846f}, +{0.99306846f,-0.11753740f}, {0.99144489f,-0.13052620f}, {0.98965138f,-0.14349262f}, +{0.98768836f,-0.15643448f}, {0.98555607f,-0.16934951f}, {0.98325491f,-0.18223552f}, +{0.98078525f,-0.19509032f}, {0.97814763f,-0.20791170f}, {0.97534233f,-0.22069745f}, +{0.97236991f,-0.23344538f}, {0.96923089f,-0.24615330f}, {0.96592581f,-0.25881904f}, +{0.96245521f,-0.27144045f}, {0.95881975f,-0.28401536f}, {0.95501995f,-0.29654160f}, +{0.95105648f,-0.30901700f}, {0.94693011f,-0.32143945f}, {0.94264150f,-0.33380687f}, +{0.93819129f,-0.34611708f}, {0.93358040f,-0.35836795f}, {0.92880952f,-0.37055743f}, +{0.92387956f,-0.38268346f}, {0.91879117f,-0.39474389f}, {0.91354543f,-0.40673664f}, +{0.90814316f,-0.41865975f}, {0.90258527f,-0.43051112f}, {0.89687270f,-0.44228873f}, +{0.89100653f,-0.45399052f}, {0.88498765f,-0.46561453f}, {0.87881708f,-0.47715878f}, +{0.87249601f,-0.48862126f}, {0.86602545f,-0.50000000f}, {0.85940641f,-0.51129311f}, +{0.85264015f,-0.52249855f}, {0.84572786f,-0.53361452f}, {0.83867055f,-0.54463905f}, +{0.83146960f,-0.55557024f}, {0.82412618f,-0.56640625f}, {0.81664151f,-0.57714522f}, +{0.80901700f,-0.58778524f}, {0.80125380f,-0.59832460f}, {0.79335332f,-0.60876143f}, +{0.78531694f,-0.61909395f}, {0.77714598f,-0.62932038f}, {0.76884180f,-0.63943899f}, +{0.76040596f,-0.64944810f}, {0.75183982f,-0.65934587f}, {0.74314475f,-0.66913062f}, +{0.73432249f,-0.67880076f}, {0.72537434f,-0.68835455f}, {0.71630192f,-0.69779050f}, +{0.70710677f,-0.70710683f}, {0.69779044f,-0.71630198f}, {0.68835455f,-0.72537440f}, +{0.67880070f,-0.73432255f}, {0.66913056f,-0.74314487f}, {0.65934581f,-0.75183982f}, +{0.64944804f,-0.76040596f}, {0.63943899f,-0.76884186f}, {0.62932038f,-0.77714598f}, +{0.61909395f,-0.78531694f}, {0.60876137f,-0.79335338f}, {0.59832460f,-0.80125386f}, +{0.58778524f,-0.80901700f}, {0.57714516f,-0.81664151f}, {0.56640625f,-0.82412618f}, +{0.55557019f,-0.83146960f}, {0.54463899f,-0.83867055f}, {0.53361452f,-0.84572786f}, +{0.52249849f,-0.85264015f}, {0.51129311f,-0.85940641f}, {0.49999997f,-0.86602545f}, +{0.48862118f,-0.87249601f}, {0.47715876f,-0.87881708f}, {0.46561447f,-0.88498765f}, +{0.45399052f,-0.89100653f}, {0.44228867f,-0.89687276f}, {0.43051103f,-0.90258533f}, +{0.41865975f,-0.90814316f}, {0.40673661f,-0.91354549f}, {0.39474380f,-0.91879129f}, +{0.38268343f,-0.92387956f}, {0.37055740f,-0.92880958f}, {0.35836786f,-0.93358046f}, +{0.34611705f,-0.93819135f}, {0.33380681f,-0.94264150f}, {0.32143947f,-0.94693011f}, +{0.30901697f,-0.95105654f}, {0.29654151f,-0.95501995f}, {0.28401533f,-0.95881975f}, +{0.27144039f,-0.96245527f}, {0.25881907f,-0.96592581f}, {0.24615327f,-0.96923089f}, +{0.23344530f,-0.97236991f}, {0.22069745f,-0.97534233f}, {0.20791166f,-0.97814763f}, +{0.19509023f,-0.98078531f}, {0.18223552f,-0.98325491f}, {0.16934945f,-0.98555607f}, +{0.15643437f,-0.98768836f}, {0.14349259f,-0.98965138f}, {0.13052613f,-0.99144489f}, +{0.11753740f,-0.99306846f}, {0.10452842f,-0.99452192f}, {0.091501534f,-0.99580491f}, +{0.078459084f,-0.99691731f}, {0.065403074f,-0.99785894f}, {0.052335974f,-0.99862951f}, +{0.039259788f,-0.99922901f}, {0.026176875f,-0.99965733f}, {0.013089597f,-0.99991435f}, +{1.0000000f,-0.0000000f}, {0.99965733f,-0.026176950f}, {0.99862951f,-0.052335959f}, +{0.99691731f,-0.078459099f}, {0.99452192f,-0.10452846f}, {0.99144489f,-0.13052620f}, +{0.98768836f,-0.15643448f}, {0.98325491f,-0.18223552f}, {0.97814763f,-0.20791170f}, +{0.97236991f,-0.23344538f}, {0.96592581f,-0.25881904f}, {0.95881975f,-0.28401536f}, +{0.95105648f,-0.30901700f}, {0.94264150f,-0.33380687f}, {0.93358040f,-0.35836795f}, +{0.92387956f,-0.38268346f}, {0.91354543f,-0.40673664f}, {0.90258527f,-0.43051112f}, +{0.89100653f,-0.45399052f}, {0.87881708f,-0.47715878f}, {0.86602545f,-0.50000000f}, +{0.85264015f,-0.52249855f}, {0.83867055f,-0.54463905f}, {0.82412618f,-0.56640625f}, +{0.80901700f,-0.58778524f}, {0.79335332f,-0.60876143f}, {0.77714598f,-0.62932038f}, +{0.76040596f,-0.64944810f}, {0.74314475f,-0.66913062f}, {0.72537434f,-0.68835455f}, +{0.70710677f,-0.70710683f}, {0.68835455f,-0.72537440f}, {0.66913056f,-0.74314487f}, +{0.64944804f,-0.76040596f}, {0.62932038f,-0.77714598f}, {0.60876137f,-0.79335338f}, +{0.58778524f,-0.80901700f}, {0.56640625f,-0.82412618f}, {0.54463899f,-0.83867055f}, +{0.52249849f,-0.85264015f}, {0.49999997f,-0.86602545f}, {0.47715876f,-0.87881708f}, +{0.45399052f,-0.89100653f}, {0.43051103f,-0.90258533f}, {0.40673661f,-0.91354549f}, +{0.38268343f,-0.92387956f}, {0.35836786f,-0.93358046f}, {0.33380681f,-0.94264150f}, +{0.30901697f,-0.95105654f}, {0.28401533f,-0.95881975f}, {0.25881907f,-0.96592581f}, +{0.23344530f,-0.97236991f}, {0.20791166f,-0.97814763f}, {0.18223552f,-0.98325491f}, +{0.15643437f,-0.98768836f}, {0.13052613f,-0.99144489f}, {0.10452842f,-0.99452192f}, +{0.078459084f,-0.99691731f}, {0.052335974f,-0.99862951f}, {0.026176875f,-0.99965733f}, +{-4.3711388e-08f,-1.0000000f}, {-0.026176963f,-0.99965733f}, {-0.052336060f,-0.99862951f}, +{-0.078459173f,-0.99691731f}, {-0.10452851f,-0.99452192f}, {-0.13052621f,-0.99144489f}, +{-0.15643445f,-0.98768836f}, {-0.18223560f,-0.98325491f}, {-0.20791174f,-0.97814757f}, +{-0.23344538f,-0.97236991f}, {-0.25881916f,-0.96592581f}, {-0.28401542f,-0.95881969f}, +{-0.30901703f,-0.95105648f}, {-0.33380687f,-0.94264150f}, {-0.35836795f,-0.93358040f}, +{-0.38268352f,-0.92387950f}, {-0.40673670f,-0.91354543f}, {-0.43051112f,-0.90258527f}, +{-0.45399061f,-0.89100647f}, {-0.47715873f,-0.87881708f}, {-0.50000006f,-0.86602533f}, +{-0.52249867f,-0.85264009f}, {-0.54463905f,-0.83867055f}, {-0.56640631f,-0.82412612f}, +{-0.58778518f,-0.80901700f}, {-0.60876143f,-0.79335332f}, {-0.62932050f,-0.77714586f}, +{-0.64944804f,-0.76040596f}, {-0.66913068f,-0.74314475f}, {-0.68835467f,-0.72537428f}, +{-0.70710677f,-0.70710677f}, {-0.72537446f,-0.68835449f}, {-0.74314493f,-0.66913044f}, +{-0.76040596f,-0.64944804f}, {-0.77714604f,-0.62932026f}, {-0.79335332f,-0.60876143f}, +{-0.80901700f,-0.58778518f}, {-0.82412624f,-0.56640613f}, {-0.83867055f,-0.54463899f}, +{-0.85264021f,-0.52249849f}, {-0.86602539f,-0.50000006f}, {-0.87881714f,-0.47715873f}, +{-0.89100659f,-0.45399037f}, {-0.90258527f,-0.43051112f}, {-0.91354549f,-0.40673658f}, +{-0.92387956f,-0.38268328f}, {-0.93358040f,-0.35836792f}, {-0.94264150f,-0.33380675f}, +{-0.95105654f,-0.30901679f}, {-0.95881975f,-0.28401530f}, {-0.96592587f,-0.25881892f}, +{-0.97236991f,-0.23344538f}, {-0.97814763f,-0.20791161f}, {-0.98325491f,-0.18223536f}, +{-0.98768836f,-0.15643445f}, {-0.99144489f,-0.13052608f}, {-0.99452192f,-0.10452849f}, +{-0.99691737f,-0.078459039f}, {-0.99862957f,-0.052335810f}, {-0.99965733f,-0.026176952f}, +{1.0000000f,-0.0000000f}, {0.99922901f,-0.039259817f}, {0.99691731f,-0.078459099f}, +{0.99306846f,-0.11753740f}, {0.98768836f,-0.15643448f}, {0.98078525f,-0.19509032f}, +{0.97236991f,-0.23344538f}, {0.96245521f,-0.27144045f}, {0.95105648f,-0.30901700f}, +{0.93819129f,-0.34611708f}, {0.92387956f,-0.38268346f}, {0.90814316f,-0.41865975f}, +{0.89100653f,-0.45399052f}, {0.87249601f,-0.48862126f}, {0.85264015f,-0.52249855f}, +{0.83146960f,-0.55557024f}, {0.80901700f,-0.58778524f}, {0.78531694f,-0.61909395f}, +{0.76040596f,-0.64944810f}, {0.73432249f,-0.67880076f}, {0.70710677f,-0.70710683f}, +{0.67880070f,-0.73432255f}, {0.64944804f,-0.76040596f}, {0.61909395f,-0.78531694f}, +{0.58778524f,-0.80901700f}, {0.55557019f,-0.83146960f}, {0.52249849f,-0.85264015f}, +{0.48862118f,-0.87249601f}, {0.45399052f,-0.89100653f}, {0.41865975f,-0.90814316f}, +{0.38268343f,-0.92387956f}, {0.34611705f,-0.93819135f}, {0.30901697f,-0.95105654f}, +{0.27144039f,-0.96245527f}, {0.23344530f,-0.97236991f}, {0.19509023f,-0.98078531f}, +{0.15643437f,-0.98768836f}, {0.11753740f,-0.99306846f}, {0.078459084f,-0.99691731f}, +{0.039259788f,-0.99922901f}, {-4.3711388e-08f,-1.0000000f}, {-0.039259877f,-0.99922901f}, +{-0.078459173f,-0.99691731f}, {-0.11753749f,-0.99306846f}, {-0.15643445f,-0.98768836f}, +{-0.19509032f,-0.98078525f}, {-0.23344538f,-0.97236991f}, {-0.27144048f,-0.96245521f}, +{-0.30901703f,-0.95105648f}, {-0.34611711f,-0.93819129f}, {-0.38268352f,-0.92387950f}, +{-0.41865984f,-0.90814310f}, {-0.45399061f,-0.89100647f}, {-0.48862135f,-0.87249595f}, +{-0.52249867f,-0.85264009f}, {-0.55557036f,-0.83146954f}, {-0.58778518f,-0.80901700f}, +{-0.61909389f,-0.78531694f}, {-0.64944804f,-0.76040596f}, {-0.67880076f,-0.73432249f}, +{-0.70710677f,-0.70710677f}, {-0.73432249f,-0.67880070f}, {-0.76040596f,-0.64944804f}, +{-0.78531694f,-0.61909389f}, {-0.80901700f,-0.58778518f}, {-0.83146966f,-0.55557019f}, +{-0.85264021f,-0.52249849f}, {-0.87249607f,-0.48862115f}, {-0.89100659f,-0.45399037f}, +{-0.90814322f,-0.41865960f}, {-0.92387956f,-0.38268328f}, {-0.93819135f,-0.34611690f}, +{-0.95105654f,-0.30901679f}, {-0.96245521f,-0.27144048f}, {-0.97236991f,-0.23344538f}, +{-0.98078531f,-0.19509031f}, {-0.98768836f,-0.15643445f}, {-0.99306846f,-0.11753736f}, +{-0.99691737f,-0.078459039f}, {-0.99922901f,-0.039259743f}, {-1.0000000f,8.7422777e-08f}, +{-0.99922901f,0.039259918f}, {-0.99691731f,0.078459218f}, {-0.99306846f,0.11753753f}, +{-0.98768830f,0.15643461f}, {-0.98078525f,0.19509049f}, {-0.97236985f,0.23344554f}, +{-0.96245515f,0.27144065f}, {-0.95105654f,0.30901697f}, {-0.93819135f,0.34611705f}, +{-0.92387956f,0.38268346f}, {-0.90814316f,0.41865975f}, {-0.89100653f,0.45399055f}, +{-0.87249601f,0.48862129f}, {-0.85264015f,0.52249861f}, {-0.83146960f,0.55557030f}, +{-0.80901694f,0.58778536f}, {-0.78531688f,0.61909401f}, {-0.76040590f,0.64944816f}, +{-0.73432243f,0.67880082f}, {-0.70710665f,0.70710689f}, {-0.67880058f,0.73432261f}, +{-0.64944792f,0.76040608f}, {-0.61909378f,0.78531706f}, {-0.58778507f,0.80901712f}, +{-0.55557001f,0.83146977f}, {-0.52249837f,0.85264033f}, {-0.48862100f,0.87249613f}, +{-0.45399022f,0.89100665f}, {-0.41865945f,0.90814328f}, {-0.38268313f,0.92387968f}, +{-0.34611672f,0.93819147f}, {-0.30901709f,0.95105648f}, {-0.27144054f,0.96245521f}, +{-0.23344545f,0.97236991f}, {-0.19509038f,0.98078525f}, {-0.15643452f,0.98768830f}, +{-0.11753743f,0.99306846f}, {-0.078459114f,0.99691731f}, {-0.039259821f,0.99922901f}, +}; +static const ne10_fft_cpx_float32_t ne10_twiddles_240[240] = { +{1.0000000f,0.0000000f}, {1.0000000f,-0.0000000f}, {1.0000000f,-0.0000000f}, +{1.0000000f,-0.0000000f}, {0.91354543f,-0.40673664f}, {0.66913056f,-0.74314487f}, +{1.0000000f,-0.0000000f}, {0.66913056f,-0.74314487f}, {-0.10452851f,-0.99452192f}, +{1.0000000f,-0.0000000f}, {0.30901697f,-0.95105654f}, {-0.80901700f,-0.58778518f}, +{1.0000000f,-0.0000000f}, {-0.10452851f,-0.99452192f}, {-0.97814757f,0.20791179f}, +{1.0000000f,-0.0000000f}, {0.99452192f,-0.10452846f}, {0.97814763f,-0.20791170f}, +{0.95105648f,-0.30901700f}, {0.91354543f,-0.40673664f}, {0.86602545f,-0.50000000f}, +{0.80901700f,-0.58778524f}, {0.74314475f,-0.66913062f}, {0.66913056f,-0.74314487f}, +{0.58778524f,-0.80901700f}, {0.49999997f,-0.86602545f}, {0.40673661f,-0.91354549f}, +{0.30901697f,-0.95105654f}, {0.20791166f,-0.97814763f}, {0.10452842f,-0.99452192f}, +{1.0000000f,-0.0000000f}, {0.97814763f,-0.20791170f}, {0.91354543f,-0.40673664f}, +{0.80901700f,-0.58778524f}, {0.66913056f,-0.74314487f}, {0.49999997f,-0.86602545f}, +{0.30901697f,-0.95105654f}, {0.10452842f,-0.99452192f}, {-0.10452851f,-0.99452192f}, +{-0.30901703f,-0.95105648f}, {-0.50000006f,-0.86602533f}, {-0.66913068f,-0.74314475f}, +{-0.80901700f,-0.58778518f}, {-0.91354549f,-0.40673658f}, {-0.97814763f,-0.20791161f}, +{1.0000000f,-0.0000000f}, {0.95105648f,-0.30901700f}, {0.80901700f,-0.58778524f}, +{0.58778524f,-0.80901700f}, {0.30901697f,-0.95105654f}, {-4.3711388e-08f,-1.0000000f}, +{-0.30901703f,-0.95105648f}, {-0.58778518f,-0.80901700f}, {-0.80901700f,-0.58778518f}, +{-0.95105654f,-0.30901679f}, {-1.0000000f,8.7422777e-08f}, {-0.95105654f,0.30901697f}, +{-0.80901694f,0.58778536f}, {-0.58778507f,0.80901712f}, {-0.30901709f,0.95105648f}, +{1.0000000f,-0.0000000f}, {0.99965733f,-0.026176950f}, {0.99862951f,-0.052335959f}, +{0.99691731f,-0.078459099f}, {0.99452192f,-0.10452846f}, {0.99144489f,-0.13052620f}, +{0.98768836f,-0.15643448f}, {0.98325491f,-0.18223552f}, {0.97814763f,-0.20791170f}, +{0.97236991f,-0.23344538f}, {0.96592581f,-0.25881904f}, {0.95881975f,-0.28401536f}, +{0.95105648f,-0.30901700f}, {0.94264150f,-0.33380687f}, {0.93358040f,-0.35836795f}, +{0.92387956f,-0.38268346f}, {0.91354543f,-0.40673664f}, {0.90258527f,-0.43051112f}, +{0.89100653f,-0.45399052f}, {0.87881708f,-0.47715878f}, {0.86602545f,-0.50000000f}, +{0.85264015f,-0.52249855f}, {0.83867055f,-0.54463905f}, {0.82412618f,-0.56640625f}, +{0.80901700f,-0.58778524f}, {0.79335332f,-0.60876143f}, {0.77714598f,-0.62932038f}, +{0.76040596f,-0.64944810f}, {0.74314475f,-0.66913062f}, {0.72537434f,-0.68835455f}, +{0.70710677f,-0.70710683f}, {0.68835455f,-0.72537440f}, {0.66913056f,-0.74314487f}, +{0.64944804f,-0.76040596f}, {0.62932038f,-0.77714598f}, {0.60876137f,-0.79335338f}, +{0.58778524f,-0.80901700f}, {0.56640625f,-0.82412618f}, {0.54463899f,-0.83867055f}, +{0.52249849f,-0.85264015f}, {0.49999997f,-0.86602545f}, {0.47715876f,-0.87881708f}, +{0.45399052f,-0.89100653f}, {0.43051103f,-0.90258533f}, {0.40673661f,-0.91354549f}, +{0.38268343f,-0.92387956f}, {0.35836786f,-0.93358046f}, {0.33380681f,-0.94264150f}, +{0.30901697f,-0.95105654f}, {0.28401533f,-0.95881975f}, {0.25881907f,-0.96592581f}, +{0.23344530f,-0.97236991f}, {0.20791166f,-0.97814763f}, {0.18223552f,-0.98325491f}, +{0.15643437f,-0.98768836f}, {0.13052613f,-0.99144489f}, {0.10452842f,-0.99452192f}, +{0.078459084f,-0.99691731f}, {0.052335974f,-0.99862951f}, {0.026176875f,-0.99965733f}, +{1.0000000f,-0.0000000f}, {0.99862951f,-0.052335959f}, {0.99452192f,-0.10452846f}, +{0.98768836f,-0.15643448f}, {0.97814763f,-0.20791170f}, {0.96592581f,-0.25881904f}, +{0.95105648f,-0.30901700f}, {0.93358040f,-0.35836795f}, {0.91354543f,-0.40673664f}, +{0.89100653f,-0.45399052f}, {0.86602545f,-0.50000000f}, {0.83867055f,-0.54463905f}, +{0.80901700f,-0.58778524f}, {0.77714598f,-0.62932038f}, {0.74314475f,-0.66913062f}, +{0.70710677f,-0.70710683f}, {0.66913056f,-0.74314487f}, {0.62932038f,-0.77714598f}, +{0.58778524f,-0.80901700f}, {0.54463899f,-0.83867055f}, {0.49999997f,-0.86602545f}, +{0.45399052f,-0.89100653f}, {0.40673661f,-0.91354549f}, {0.35836786f,-0.93358046f}, +{0.30901697f,-0.95105654f}, {0.25881907f,-0.96592581f}, {0.20791166f,-0.97814763f}, +{0.15643437f,-0.98768836f}, {0.10452842f,-0.99452192f}, {0.052335974f,-0.99862951f}, +{-4.3711388e-08f,-1.0000000f}, {-0.052336060f,-0.99862951f}, {-0.10452851f,-0.99452192f}, +{-0.15643445f,-0.98768836f}, {-0.20791174f,-0.97814757f}, {-0.25881916f,-0.96592581f}, +{-0.30901703f,-0.95105648f}, {-0.35836795f,-0.93358040f}, {-0.40673670f,-0.91354543f}, +{-0.45399061f,-0.89100647f}, {-0.50000006f,-0.86602533f}, {-0.54463905f,-0.83867055f}, +{-0.58778518f,-0.80901700f}, {-0.62932050f,-0.77714586f}, {-0.66913068f,-0.74314475f}, +{-0.70710677f,-0.70710677f}, {-0.74314493f,-0.66913044f}, {-0.77714604f,-0.62932026f}, +{-0.80901700f,-0.58778518f}, {-0.83867055f,-0.54463899f}, {-0.86602539f,-0.50000006f}, +{-0.89100659f,-0.45399037f}, {-0.91354549f,-0.40673658f}, {-0.93358040f,-0.35836792f}, +{-0.95105654f,-0.30901679f}, {-0.96592587f,-0.25881892f}, {-0.97814763f,-0.20791161f}, +{-0.98768836f,-0.15643445f}, {-0.99452192f,-0.10452849f}, {-0.99862957f,-0.052335810f}, +{1.0000000f,-0.0000000f}, {0.99691731f,-0.078459099f}, {0.98768836f,-0.15643448f}, +{0.97236991f,-0.23344538f}, {0.95105648f,-0.30901700f}, {0.92387956f,-0.38268346f}, +{0.89100653f,-0.45399052f}, {0.85264015f,-0.52249855f}, {0.80901700f,-0.58778524f}, +{0.76040596f,-0.64944810f}, {0.70710677f,-0.70710683f}, {0.64944804f,-0.76040596f}, +{0.58778524f,-0.80901700f}, {0.52249849f,-0.85264015f}, {0.45399052f,-0.89100653f}, +{0.38268343f,-0.92387956f}, {0.30901697f,-0.95105654f}, {0.23344530f,-0.97236991f}, +{0.15643437f,-0.98768836f}, {0.078459084f,-0.99691731f}, {-4.3711388e-08f,-1.0000000f}, +{-0.078459173f,-0.99691731f}, {-0.15643445f,-0.98768836f}, {-0.23344538f,-0.97236991f}, +{-0.30901703f,-0.95105648f}, {-0.38268352f,-0.92387950f}, {-0.45399061f,-0.89100647f}, +{-0.52249867f,-0.85264009f}, {-0.58778518f,-0.80901700f}, {-0.64944804f,-0.76040596f}, +{-0.70710677f,-0.70710677f}, {-0.76040596f,-0.64944804f}, {-0.80901700f,-0.58778518f}, +{-0.85264021f,-0.52249849f}, {-0.89100659f,-0.45399037f}, {-0.92387956f,-0.38268328f}, +{-0.95105654f,-0.30901679f}, {-0.97236991f,-0.23344538f}, {-0.98768836f,-0.15643445f}, +{-0.99691737f,-0.078459039f}, {-1.0000000f,8.7422777e-08f}, {-0.99691731f,0.078459218f}, +{-0.98768830f,0.15643461f}, {-0.97236985f,0.23344554f}, {-0.95105654f,0.30901697f}, +{-0.92387956f,0.38268346f}, {-0.89100653f,0.45399055f}, {-0.85264015f,0.52249861f}, +{-0.80901694f,0.58778536f}, {-0.76040590f,0.64944816f}, {-0.70710665f,0.70710689f}, +{-0.64944792f,0.76040608f}, {-0.58778507f,0.80901712f}, {-0.52249837f,0.85264033f}, +{-0.45399022f,0.89100665f}, {-0.38268313f,0.92387968f}, {-0.30901709f,0.95105648f}, +{-0.23344545f,0.97236991f}, {-0.15643452f,0.98768830f}, {-0.078459114f,0.99691731f}, +}; +static const ne10_fft_cpx_float32_t ne10_twiddles_120[120] = { +{1.0000000f,0.0000000f}, {1.0000000f,-0.0000000f}, {1.0000000f,-0.0000000f}, +{1.0000000f,-0.0000000f}, {0.91354543f,-0.40673664f}, {0.66913056f,-0.74314487f}, +{1.0000000f,-0.0000000f}, {0.66913056f,-0.74314487f}, {-0.10452851f,-0.99452192f}, +{1.0000000f,-0.0000000f}, {0.30901697f,-0.95105654f}, {-0.80901700f,-0.58778518f}, +{1.0000000f,-0.0000000f}, {-0.10452851f,-0.99452192f}, {-0.97814757f,0.20791179f}, +{1.0000000f,-0.0000000f}, {0.97814763f,-0.20791170f}, {0.91354543f,-0.40673664f}, +{0.80901700f,-0.58778524f}, {0.66913056f,-0.74314487f}, {0.49999997f,-0.86602545f}, +{0.30901697f,-0.95105654f}, {0.10452842f,-0.99452192f}, {-0.10452851f,-0.99452192f}, +{-0.30901703f,-0.95105648f}, {-0.50000006f,-0.86602533f}, {-0.66913068f,-0.74314475f}, +{-0.80901700f,-0.58778518f}, {-0.91354549f,-0.40673658f}, {-0.97814763f,-0.20791161f}, +{1.0000000f,-0.0000000f}, {0.99862951f,-0.052335959f}, {0.99452192f,-0.10452846f}, +{0.98768836f,-0.15643448f}, {0.97814763f,-0.20791170f}, {0.96592581f,-0.25881904f}, +{0.95105648f,-0.30901700f}, {0.93358040f,-0.35836795f}, {0.91354543f,-0.40673664f}, +{0.89100653f,-0.45399052f}, {0.86602545f,-0.50000000f}, {0.83867055f,-0.54463905f}, +{0.80901700f,-0.58778524f}, {0.77714598f,-0.62932038f}, {0.74314475f,-0.66913062f}, +{0.70710677f,-0.70710683f}, {0.66913056f,-0.74314487f}, {0.62932038f,-0.77714598f}, +{0.58778524f,-0.80901700f}, {0.54463899f,-0.83867055f}, {0.49999997f,-0.86602545f}, +{0.45399052f,-0.89100653f}, {0.40673661f,-0.91354549f}, {0.35836786f,-0.93358046f}, +{0.30901697f,-0.95105654f}, {0.25881907f,-0.96592581f}, {0.20791166f,-0.97814763f}, +{0.15643437f,-0.98768836f}, {0.10452842f,-0.99452192f}, {0.052335974f,-0.99862951f}, +{1.0000000f,-0.0000000f}, {0.99452192f,-0.10452846f}, {0.97814763f,-0.20791170f}, +{0.95105648f,-0.30901700f}, {0.91354543f,-0.40673664f}, {0.86602545f,-0.50000000f}, +{0.80901700f,-0.58778524f}, {0.74314475f,-0.66913062f}, {0.66913056f,-0.74314487f}, +{0.58778524f,-0.80901700f}, {0.49999997f,-0.86602545f}, {0.40673661f,-0.91354549f}, +{0.30901697f,-0.95105654f}, {0.20791166f,-0.97814763f}, {0.10452842f,-0.99452192f}, +{-4.3711388e-08f,-1.0000000f}, {-0.10452851f,-0.99452192f}, {-0.20791174f,-0.97814757f}, +{-0.30901703f,-0.95105648f}, {-0.40673670f,-0.91354543f}, {-0.50000006f,-0.86602533f}, +{-0.58778518f,-0.80901700f}, {-0.66913068f,-0.74314475f}, {-0.74314493f,-0.66913044f}, +{-0.80901700f,-0.58778518f}, {-0.86602539f,-0.50000006f}, {-0.91354549f,-0.40673658f}, +{-0.95105654f,-0.30901679f}, {-0.97814763f,-0.20791161f}, {-0.99452192f,-0.10452849f}, +{1.0000000f,-0.0000000f}, {0.98768836f,-0.15643448f}, {0.95105648f,-0.30901700f}, +{0.89100653f,-0.45399052f}, {0.80901700f,-0.58778524f}, {0.70710677f,-0.70710683f}, +{0.58778524f,-0.80901700f}, {0.45399052f,-0.89100653f}, {0.30901697f,-0.95105654f}, +{0.15643437f,-0.98768836f}, {-4.3711388e-08f,-1.0000000f}, {-0.15643445f,-0.98768836f}, +{-0.30901703f,-0.95105648f}, {-0.45399061f,-0.89100647f}, {-0.58778518f,-0.80901700f}, +{-0.70710677f,-0.70710677f}, {-0.80901700f,-0.58778518f}, {-0.89100659f,-0.45399037f}, +{-0.95105654f,-0.30901679f}, {-0.98768836f,-0.15643445f}, {-1.0000000f,8.7422777e-08f}, +{-0.98768830f,0.15643461f}, {-0.95105654f,0.30901697f}, {-0.89100653f,0.45399055f}, +{-0.80901694f,0.58778536f}, {-0.70710665f,0.70710689f}, {-0.58778507f,0.80901712f}, +{-0.45399022f,0.89100665f}, {-0.30901709f,0.95105648f}, {-0.15643452f,0.98768830f}, +}; +static const ne10_fft_cpx_float32_t ne10_twiddles_60[60] = { +{1.0000000f,0.0000000f}, {1.0000000f,-0.0000000f}, {1.0000000f,-0.0000000f}, +{1.0000000f,-0.0000000f}, {0.91354543f,-0.40673664f}, {0.66913056f,-0.74314487f}, +{1.0000000f,-0.0000000f}, {0.66913056f,-0.74314487f}, {-0.10452851f,-0.99452192f}, +{1.0000000f,-0.0000000f}, {0.30901697f,-0.95105654f}, {-0.80901700f,-0.58778518f}, +{1.0000000f,-0.0000000f}, {-0.10452851f,-0.99452192f}, {-0.97814757f,0.20791179f}, +{1.0000000f,-0.0000000f}, {0.99452192f,-0.10452846f}, {0.97814763f,-0.20791170f}, +{0.95105648f,-0.30901700f}, {0.91354543f,-0.40673664f}, {0.86602545f,-0.50000000f}, +{0.80901700f,-0.58778524f}, {0.74314475f,-0.66913062f}, {0.66913056f,-0.74314487f}, +{0.58778524f,-0.80901700f}, {0.49999997f,-0.86602545f}, {0.40673661f,-0.91354549f}, +{0.30901697f,-0.95105654f}, {0.20791166f,-0.97814763f}, {0.10452842f,-0.99452192f}, +{1.0000000f,-0.0000000f}, {0.97814763f,-0.20791170f}, {0.91354543f,-0.40673664f}, +{0.80901700f,-0.58778524f}, {0.66913056f,-0.74314487f}, {0.49999997f,-0.86602545f}, +{0.30901697f,-0.95105654f}, {0.10452842f,-0.99452192f}, {-0.10452851f,-0.99452192f}, +{-0.30901703f,-0.95105648f}, {-0.50000006f,-0.86602533f}, {-0.66913068f,-0.74314475f}, +{-0.80901700f,-0.58778518f}, {-0.91354549f,-0.40673658f}, {-0.97814763f,-0.20791161f}, +{1.0000000f,-0.0000000f}, {0.95105648f,-0.30901700f}, {0.80901700f,-0.58778524f}, +{0.58778524f,-0.80901700f}, {0.30901697f,-0.95105654f}, {-4.3711388e-08f,-1.0000000f}, +{-0.30901703f,-0.95105648f}, {-0.58778518f,-0.80901700f}, {-0.80901700f,-0.58778518f}, +{-0.95105654f,-0.30901679f}, {-1.0000000f,8.7422777e-08f}, {-0.95105654f,0.30901697f}, +{-0.80901694f,0.58778536f}, {-0.58778507f,0.80901712f}, {-0.30901709f,0.95105648f}, +}; +static const ne10_fft_state_float32_t ne10_fft_state_float32_t_480 = { +120, +(ne10_int32_t *)ne10_factors_480, +(ne10_fft_cpx_float32_t *)ne10_twiddles_480, +NULL, +(ne10_fft_cpx_float32_t *)&ne10_twiddles_480[120], +/* is_forward_scaled = true */ +(ne10_int32_t) 1, +/* is_backward_scaled = false */ +(ne10_int32_t) 0, +}; +static const arch_fft_state cfg_arch_480 = { +1, +(void *)&ne10_fft_state_float32_t_480, +}; + +static const ne10_fft_state_float32_t ne10_fft_state_float32_t_240 = { +60, +(ne10_int32_t *)ne10_factors_240, +(ne10_fft_cpx_float32_t *)ne10_twiddles_240, +NULL, +(ne10_fft_cpx_float32_t *)&ne10_twiddles_240[60], +/* is_forward_scaled = true */ +(ne10_int32_t) 1, +/* is_backward_scaled = false */ +(ne10_int32_t) 0, +}; +static const arch_fft_state cfg_arch_240 = { +1, +(void *)&ne10_fft_state_float32_t_240, +}; + +static const ne10_fft_state_float32_t ne10_fft_state_float32_t_120 = { +30, +(ne10_int32_t *)ne10_factors_120, +(ne10_fft_cpx_float32_t *)ne10_twiddles_120, +NULL, +(ne10_fft_cpx_float32_t *)&ne10_twiddles_120[30], +/* is_forward_scaled = true */ +(ne10_int32_t) 1, +/* is_backward_scaled = false */ +(ne10_int32_t) 0, +}; +static const arch_fft_state cfg_arch_120 = { +1, +(void *)&ne10_fft_state_float32_t_120, +}; + +static const ne10_fft_state_float32_t ne10_fft_state_float32_t_60 = { +15, +(ne10_int32_t *)ne10_factors_60, +(ne10_fft_cpx_float32_t *)ne10_twiddles_60, +NULL, +(ne10_fft_cpx_float32_t *)&ne10_twiddles_60[15], +/* is_forward_scaled = true */ +(ne10_int32_t) 1, +/* is_backward_scaled = false */ +(ne10_int32_t) 0, +}; +static const arch_fft_state cfg_arch_60 = { +1, +(void *)&ne10_fft_state_float32_t_60, +}; + +#endif /* end NE10_FFT_PARAMS48000_960 */ diff --git a/lib/opus/celt/tests/meson.build b/lib/opus/celt/tests/meson.build new file mode 100644 index 00000000..0e6d2e62 --- /dev/null +++ b/lib/opus/celt/tests/meson.build @@ -0,0 +1,19 @@ +tests = [ + 'test_unit_types', + 'test_unit_mathops', + 'test_unit_entropy', + 'test_unit_laplace', + 'test_unit_dft', + 'test_unit_mdct', + 'test_unit_rotation', + 'test_unit_cwrs32', +] + +foreach test_name : tests + exe = executable(test_name, '@0@.c'.format(test_name), + include_directories : opus_includes, + link_with : [celt_lib, celt_static_libs], + dependencies : libm, + install : false) + test(test_name, exe) +endforeach diff --git a/lib/opus/celt/tests/test_unit_cwrs32.c b/lib/opus/celt/tests/test_unit_cwrs32.c new file mode 100644 index 00000000..36dd8af5 --- /dev/null +++ b/lib/opus/celt/tests/test_unit_cwrs32.c @@ -0,0 +1,161 @@ +/* Copyright (c) 2008-2011 Xiph.Org Foundation, Mozilla Corporation, + Gregory Maxwell + Written by Jean-Marc Valin, Gregory Maxwell, and Timothy B. Terriberry */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include + +#ifndef CUSTOM_MODES +#define CUSTOM_MODES +#else +#define TEST_CUSTOM_MODES +#endif + +#define CELT_C +#include "stack_alloc.h" +#include "entenc.c" +#include "entdec.c" +#include "entcode.c" +#include "cwrs.c" +#include "mathops.c" +#include "rate.h" + +#define NMAX (240) +#define KMAX (128) + +#ifdef TEST_CUSTOM_MODES + +#define NDIMS (44) +static const int pn[NDIMS]={ + 2, 3, 4, 5, 6, 7, 8, 9, 10, + 11, 12, 13, 14, 15, 16, 18, 20, 22, + 24, 26, 28, 30, 32, 36, 40, 44, 48, + 52, 56, 60, 64, 72, 80, 88, 96, 104, + 112, 120, 128, 144, 160, 176, 192, 208 +}; +static const int pkmax[NDIMS]={ + 128, 128, 128, 128, 88, 52, 36, 26, 22, + 18, 16, 15, 13, 12, 12, 11, 10, 9, + 9, 8, 8, 7, 7, 7, 7, 6, 6, + 6, 6, 6, 5, 5, 5, 5, 5, 5, + 4, 4, 4, 4, 4, 4, 4, 4 +}; + +#else /* TEST_CUSTOM_MODES */ + +#define NDIMS (22) +static const int pn[NDIMS]={ + 2, 3, 4, 6, 8, 9, 11, 12, 16, + 18, 22, 24, 32, 36, 44, 48, 64, 72, + 88, 96, 144, 176 +}; +static const int pkmax[NDIMS]={ + 128, 128, 128, 88, 36, 26, 18, 16, 12, + 11, 9, 9, 7, 7, 6, 6, 5, 5, + 5, 5, 4, 4 +}; + +#endif + +int main(void){ + int t; + int n; + ALLOC_STACK; + for(t=0;tpkmax[t])break; + printf("Testing CWRS with N=%i, K=%i...\n",n,k); +#if defined(SMALL_FOOTPRINT) + nc=ncwrs_urow(n,k,uu); +#else + nc=CELT_PVQ_V(n,k); +#endif + inc=nc/20000; + if(inc<1)inc=1; + for(i=0;i");*/ +#if defined(SMALL_FOOTPRINT) + ii=icwrs(n,k,&v,y,u); +#else + ii=icwrs(n,y); + v=CELT_PVQ_V(n,k); +#endif + if(ii!=i){ + fprintf(stderr,"Combination-index mismatch (%lu!=%lu).\n", + (long)ii,(long)i); + return 1; + } + if(v!=nc){ + fprintf(stderr,"Combination count mismatch (%lu!=%lu).\n", + (long)v,(long)nc); + return 2; + } + /*printf(" %6u\n",i);*/ + } + /*printf("\n");*/ + } + } + return 0; +} diff --git a/lib/opus/celt/tests/test_unit_dft.c b/lib/opus/celt/tests/test_unit_dft.c new file mode 100644 index 00000000..ae9a7b56 --- /dev/null +++ b/lib/opus/celt/tests/test_unit_dft.c @@ -0,0 +1,180 @@ +/* Copyright (c) 2008 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include + +#include "stack_alloc.h" +#include "kiss_fft.h" +#include "mathops.h" +#include "modes.h" + +#ifndef M_PI +#define M_PI 3.141592653 +#endif + +int ret = 0; + +void check(kiss_fft_cpx * in,kiss_fft_cpx * out,int nfft,int isinverse) +{ + int bin,k; + double errpow=0,sigpow=0, snr; + + for (bin=0;binmdct.kfft[id]; +#endif + + in = (kiss_fft_cpx*)malloc(buflen); + out = (kiss_fft_cpx*)malloc(buflen); + + for (k=0;k1) { + int k; + for (k=1;k +#include +#include +#include +#define CELT_C +#include "entcode.h" +#include "entenc.h" +#include "entdec.h" +#include + +#include "entenc.c" +#include "entdec.c" +#include "entcode.c" + +#ifndef M_LOG2E +# define M_LOG2E 1.4426950408889634074 +#endif +#define DATA_SIZE 10000000 +#define DATA_SIZE2 10000 + +int main(int _argc,char **_argv){ + ec_enc enc; + ec_dec dec; + long nbits; + long nbits2; + double entropy; + int ft; + int ftb; + int sz; + int i; + int ret; + unsigned int sym; + unsigned int seed; + unsigned char *ptr; + const char *env_seed; + ret=0; + entropy=0; + if (_argc > 2) { + fprintf(stderr, "Usage: %s []\n", _argv[0]); + return 1; + } + env_seed = getenv("SEED"); + if (_argc > 1) + seed = atoi(_argv[1]); + else if (env_seed) + seed = atoi(env_seed); + else + seed = time(NULL); + /*Testing encoding of raw bit values.*/ + ptr = (unsigned char *)malloc(DATA_SIZE); + ec_enc_init(&enc,ptr, DATA_SIZE); + for(ft=2;ft<1024;ft++){ + for(i=0;i>(rand()%11U))+1U)+10; + sz=rand()/((RAND_MAX>>(rand()%9U))+1U); + data=(unsigned *)malloc(sz*sizeof(*data)); + tell=(unsigned *)malloc((sz+1)*sizeof(*tell)); + ec_enc_init(&enc,ptr,DATA_SIZE2); + zeros = rand()%13==0; + tell[0]=ec_tell_frac(&enc); + for(j=0;j>(rand()%9U))+1U); + logp1=(unsigned *)malloc(sz*sizeof(*logp1)); + data=(unsigned *)malloc(sz*sizeof(*data)); + tell=(unsigned *)malloc((sz+1)*sizeof(*tell)); + enc_method=(unsigned *)malloc(sz*sizeof(*enc_method)); + ec_enc_init(&enc,ptr,DATA_SIZE2); + tell[0]=ec_tell_frac(&enc); + for(j=0;j>1)+1); + logp1[j]=(rand()%15)+1; + enc_method[j]=rand()/((RAND_MAX>>2)+1); + switch(enc_method[j]){ + case 0:{ + ec_encode(&enc,data[j]?(1<>2)+1); + switch(dec_method){ + case 0:{ + fs=ec_decode(&dec,1<=(1<=(1< +#include +#define CELT_C +#include "laplace.h" +#include "stack_alloc.h" + +#include "entenc.c" +#include "entdec.c" +#include "entcode.c" +#include "laplace.c" + +#define DATA_SIZE 40000 + +int ec_laplace_get_start_freq(int decay) +{ + opus_uint32 ft = 32768 - LAPLACE_MINP*(2*LAPLACE_NMIN+1); + int fs = (ft*(16384-decay))/(16384+decay); + return fs+LAPLACE_MINP; +} + +int main(void) +{ + int i; + int ret = 0; + ec_enc enc; + ec_dec dec; + unsigned char *ptr; + int val[10000], decay[10000]; + ALLOC_STACK; + ptr = (unsigned char *)malloc(DATA_SIZE); + ec_enc_init(&enc,ptr,DATA_SIZE); + + val[0] = 3; decay[0] = 6000; + val[1] = 0; decay[1] = 5800; + val[2] = -1; decay[2] = 5600; + for (i=3;i<10000;i++) + { + val[i] = rand()%15-7; + decay[i] = rand()%11000+5000; + } + for (i=0;i<10000;i++) + ec_laplace_encode(&enc, &val[i], + ec_laplace_get_start_freq(decay[i]), decay[i]); + + ec_enc_done(&enc); + + ec_dec_init(&dec,ec_get_buffer(&enc),ec_range_bytes(&enc)); + + for (i=0;i<10000;i++) + { + int d = ec_laplace_decode(&dec, + ec_laplace_get_start_freq(decay[i]), decay[i]); + if (d != val[i]) + { + fprintf (stderr, "Got %d instead of %d\n", d, val[i]); + ret = 1; + } + } + + free(ptr); + return ret; +} diff --git a/lib/opus/celt/tests/test_unit_mathops.c b/lib/opus/celt/tests/test_unit_mathops.c new file mode 100644 index 00000000..0615448d --- /dev/null +++ b/lib/opus/celt/tests/test_unit_mathops.c @@ -0,0 +1,266 @@ +/* Copyright (c) 2008-2011 Xiph.Org Foundation, Mozilla Corporation, + Gregory Maxwell + Written by Jean-Marc Valin, Gregory Maxwell, and Timothy B. Terriberry */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#ifndef CUSTOM_MODES +#define CUSTOM_MODES +#endif + +#include +#include +#include "mathops.h" +#include "bands.h" + +#ifdef FIXED_POINT +#define WORD "%d" +#else +#define WORD "%f" +#endif + +int ret = 0; + +void testdiv(void) +{ + opus_int32 i; + for (i=1;i<=327670;i++) + { + double prod; + opus_val32 val; + val = celt_rcp(i); +#ifdef FIXED_POINT + prod = (1./32768./65526.)*val*i; +#else + prod = val*i; +#endif + if (fabs(prod-1) > .00025) + { + fprintf (stderr, "div failed: 1/%d="WORD" (product = %f)\n", i, val, prod); + ret = 1; + } + } +} + +void testsqrt(void) +{ + opus_int32 i; + for (i=1;i<=1000000000;i++) + { + double ratio; + opus_val16 val; + val = celt_sqrt(i); + ratio = val/sqrt(i); + if (fabs(ratio - 1) > .0005 && fabs(val-sqrt(i)) > 2) + { + fprintf (stderr, "sqrt failed: sqrt(%d)="WORD" (ratio = %f)\n", i, val, ratio); + ret = 1; + } + i+= i>>10; + } +} + +void testbitexactcos(void) +{ + int i; + opus_int32 min_d,max_d,last,chk; + chk=max_d=0; + last=min_d=32767; + for(i=64;i<=16320;i++) + { + opus_int32 d; + opus_int32 q=bitexact_cos(i); + chk ^= q*i; + d = last - q; + if (d>max_d)max_d=d; + if (dmax_d)max_d=d; + if (d0.0009) + { + fprintf (stderr, "celt_log2 failed: fabs((1.442695040888963387*log(x))-celt_log2(x))>0.001 (x = %f, error = %f)\n", x,error); + ret = 1; + } + } +} + +void testexp2(void) +{ + float x; + for (x=-11.0;x<24.0;x+=0.0007f) + { + float error = fabs(x-(1.442695040888963387*log(celt_exp2(x)))); + if (error>0.0002) + { + fprintf (stderr, "celt_exp2 failed: fabs(x-(1.442695040888963387*log(celt_exp2(x))))>0.0005 (x = %f, error = %f)\n", x,error); + ret = 1; + } + } +} + +void testexp2log2(void) +{ + float x; + for (x=-11.0;x<24.0;x+=0.0007f) + { + float error = fabs(x-(celt_log2(celt_exp2(x)))); + if (error>0.001) + { + fprintf (stderr, "celt_log2/celt_exp2 failed: fabs(x-(celt_log2(celt_exp2(x))))>0.001 (x = %f, error = %f)\n", x,error); + ret = 1; + } + } +} +#else +void testlog2(void) +{ + opus_val32 x; + for (x=8;x<1073741824;x+=(x>>3)) + { + float error = fabs((1.442695040888963387*log(x/16384.0))-celt_log2(x)/1024.0); + if (error>0.003) + { + fprintf (stderr, "celt_log2 failed: x = %ld, error = %f\n", (long)x,error); + ret = 1; + } + } +} + +void testexp2(void) +{ + opus_val16 x; + for (x=-32768;x<15360;x++) + { + float error1 = fabs(x/1024.0-(1.442695040888963387*log(celt_exp2(x)/65536.0))); + float error2 = fabs(exp(0.6931471805599453094*x/1024.0)-celt_exp2(x)/65536.0); + if (error1>0.0002&&error2>0.00004) + { + fprintf (stderr, "celt_exp2 failed: x = "WORD", error1 = %f, error2 = %f\n", x,error1,error2); + ret = 1; + } + } +} + +void testexp2log2(void) +{ + opus_val32 x; + for (x=8;x<65536;x+=(x>>3)) + { + float error = fabs(x-0.25*celt_exp2(celt_log2(x)))/16384; + if (error>0.004) + { + fprintf (stderr, "celt_log2/celt_exp2 failed: fabs(x-(celt_exp2(celt_log2(x))))>0.001 (x = %ld, error = %f)\n", (long)x,error); + ret = 1; + } + } +} + +void testilog2(void) +{ + opus_val32 x; + for (x=1;x<=268435455;x+=127) + { + opus_val32 lg; + opus_val32 y; + + lg = celt_ilog2(x); + if (lg<0 || lg>=31) + { + printf("celt_ilog2 failed: 0<=celt_ilog2(x)<31 (x = %d, celt_ilog2(x) = %d)\n",x,lg); + ret = 1; + } + y = 1<>1)>=y) + { + printf("celt_ilog2 failed: 2**celt_ilog2(x)<=x<2**(celt_ilog2(x)+1) (x = %d, 2**celt_ilog2(x) = %d)\n",x,y); + ret = 1; + } + } +} +#endif + +int main(void) +{ + testbitexactcos(); + testbitexactlog2tan(); + testdiv(); + testsqrt(); + testlog2(); + testexp2(); + testexp2log2(); +#ifdef FIXED_POINT + testilog2(); +#endif + return ret; +} diff --git a/lib/opus/celt/tests/test_unit_mdct.c b/lib/opus/celt/tests/test_unit_mdct.c new file mode 100644 index 00000000..844c5b48 --- /dev/null +++ b/lib/opus/celt/tests/test_unit_mdct.c @@ -0,0 +1,228 @@ +/* Copyright (c) 2008-2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include + +#include "mdct.h" +#include "stack_alloc.h" +#include "kiss_fft.h" +#include "mdct.h" +#include "modes.h" + +#ifndef M_PI +#define M_PI 3.141592653 +#endif + +int ret = 0; +void check(kiss_fft_scalar * in,kiss_fft_scalar * out,int nfft,int isinverse) +{ + int bin,k; + double errpow=0,sigpow=0; + double snr; + for (bin=0;binmdct; +#endif + + in = (kiss_fft_scalar*)malloc(buflen); + in_copy = (kiss_fft_scalar*)malloc(buflen); + out = (kiss_fft_scalar*)malloc(buflen); + window = (opus_val16*)malloc(sizeof(opus_val16)*nfft/2); + + for (k=0;k1) { + int k; + for (k=1;k +#include +#include "vq.h" +#include "bands.h" +#include "stack_alloc.h" +#include + + +#define MAX_SIZE 100 + +int ret=0; +void test_rotation(int N, int K) +{ + int i; + double err = 0, ener = 0, snr, snr0; + opus_val16 x0[MAX_SIZE]; + opus_val16 x1[MAX_SIZE]; + for (i=0;i 20) + { + fprintf(stderr, "FAIL!\n"); + ret = 1; + } +} + +int main(void) +{ + ALLOC_STACK; + test_rotation(15, 3); + test_rotation(23, 5); + test_rotation(50, 3); + test_rotation(80, 1); + return ret; +} diff --git a/lib/opus/celt/tests/test_unit_types.c b/lib/opus/celt/tests/test_unit_types.c new file mode 100644 index 00000000..67a0fb8e --- /dev/null +++ b/lib/opus/celt/tests/test_unit_types.c @@ -0,0 +1,50 @@ +/* Copyright (c) 2008-2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_types.h" +#include + +int main(void) +{ + opus_int16 i = 1; + i <<= 14; + if (i>>14 != 1) + { + fprintf(stderr, "opus_int16 isn't 16 bits\n"); + return 1; + } + if (sizeof(opus_int16)*2 != sizeof(opus_int32)) + { + fprintf(stderr, "16*2 != 32\n"); + return 1; + } + return 0; +} diff --git a/lib/opus/celt/vq.c b/lib/opus/celt/vq.c new file mode 100644 index 00000000..8011e225 --- /dev/null +++ b/lib/opus/celt/vq.c @@ -0,0 +1,442 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "mathops.h" +#include "cwrs.h" +#include "vq.h" +#include "arch.h" +#include "os_support.h" +#include "bands.h" +#include "rate.h" +#include "pitch.h" + +#if defined(MIPSr1_ASM) +#include "mips/vq_mipsr1.h" +#endif + +#ifndef OVERRIDE_vq_exp_rotation1 +static void exp_rotation1(celt_norm *X, int len, int stride, opus_val16 c, opus_val16 s) +{ + int i; + opus_val16 ms; + celt_norm *Xptr; + Xptr = X; + ms = NEG16(s); + for (i=0;i=0;i--) + { + celt_norm x1, x2; + x1 = Xptr[0]; + x2 = Xptr[stride]; + Xptr[stride] = EXTRACT16(PSHR32(MAC16_16(MULT16_16(c, x2), s, x1), 15)); + *Xptr-- = EXTRACT16(PSHR32(MAC16_16(MULT16_16(c, x1), ms, x2), 15)); + } +} +#endif /* OVERRIDE_vq_exp_rotation1 */ + +void exp_rotation(celt_norm *X, int len, int dir, int stride, int K, int spread) +{ + static const int SPREAD_FACTOR[3]={15,10,5}; + int i; + opus_val16 c, s; + opus_val16 gain, theta; + int stride2=0; + int factor; + + if (2*K>=len || spread==SPREAD_NONE) + return; + factor = SPREAD_FACTOR[spread-1]; + + gain = celt_div((opus_val32)MULT16_16(Q15_ONE,len),(opus_val32)(len+factor*K)); + theta = HALF16(MULT16_16_Q15(gain,gain)); + + c = celt_cos_norm(EXTEND32(theta)); + s = celt_cos_norm(EXTEND32(SUB16(Q15ONE,theta))); /* sin(theta) */ + + if (len>=8*stride) + { + stride2 = 1; + /* This is just a simple (equivalent) way of computing sqrt(len/stride) with rounding. + It's basically incrementing long as (stride2+0.5)^2 < len/stride. */ + while ((stride2*stride2+stride2)*stride + (stride>>2) < len) + stride2++; + } + /*NOTE: As a minor optimization, we could be passing around log2(B), not B, for both this and for + extract_collapse_mask().*/ + len = celt_udiv(len, stride); + for (i=0;i>1; +#endif + t = VSHR32(Ryy, 2*(k-7)); + g = MULT16_16_P15(celt_rsqrt_norm(t),gain); + + i=0; + do + X[i] = EXTRACT16(PSHR32(MULT16_16(g, iy[i]), k+1)); + while (++i < N); +} + +static unsigned extract_collapse_mask(int *iy, int N, int B) +{ + unsigned collapse_mask; + int N0; + int i; + if (B<=1) + return 1; + /*NOTE: As a minor optimization, we could be passing around log2(B), not B, for both this and for + exp_rotation().*/ + N0 = celt_udiv(N, B); + collapse_mask = 0; + i=0; do { + int j; + unsigned tmp=0; + j=0; do { + tmp |= iy[i*N0+j]; + } while (++j (N>>1)) + { + opus_val16 rcp; + j=0; do { + sum += X[j]; + } while (++j EPSILON && sum < 64)) +#endif + { + X[0] = QCONST16(1.f,14); + j=1; do + X[j]=0; + while (++j=0); + + /* This should never happen, but just in case it does (e.g. on silence) + we fill the first bin with pulses. */ +#ifdef FIXED_POINT_DEBUG + celt_sig_assert(pulsesLeft<=N+3); +#endif + if (pulsesLeft > N+3) + { + opus_val16 tmp = (opus_val16)pulsesLeft; + yy = MAC16_16(yy, tmp, tmp); + yy = MAC16_16(yy, tmp, y[0]); + iy[0] += pulsesLeft; + pulsesLeft=0; + } + + for (i=0;i= best_num/best_den, but that way + we can do it without any division */ + /* OPT: It's not clear whether a cmov is faster than a branch here + since the condition is more often false than true and using + a cmov introduces data dependencies across iterations. The optimal + choice may be architecture-dependent. */ + if (opus_unlikely(MULT16_16(best_den, Rxy) > MULT16_16(Ryy, best_num))) + { + best_den = Ryy; + best_num = Rxy; + best_id = j; + } + } while (++j0, "alg_quant() needs at least one pulse"); + celt_assert2(N>1, "alg_quant() needs at least two dimensions"); + + /* Covers vectorization by up to 4. */ + ALLOC(iy, N+3, int); + + exp_rotation(X, N, 1, B, K, spread); + + yy = op_pvq_search(X, iy, K, N, arch); + + encode_pulses(iy, N, K, enc); + + if (resynth) + { + normalise_residual(iy, X, N, yy, gain); + exp_rotation(X, N, -1, B, K, spread); + } + + collapse_mask = extract_collapse_mask(iy, N, B); + RESTORE_STACK; + return collapse_mask; +} + +/** Decode pulse vector and combine the result with the pitch vector to produce + the final normalised signal in the current band. */ +unsigned alg_unquant(celt_norm *X, int N, int K, int spread, int B, + ec_dec *dec, opus_val16 gain) +{ + opus_val32 Ryy; + unsigned collapse_mask; + VARDECL(int, iy); + SAVE_STACK; + + celt_assert2(K>0, "alg_unquant() needs at least one pulse"); + celt_assert2(N>1, "alg_unquant() needs at least two dimensions"); + ALLOC(iy, N, int); + Ryy = decode_pulses(iy, N, K, dec); + normalise_residual(iy, X, N, Ryy, gain); + exp_rotation(X, N, -1, B, K, spread); + collapse_mask = extract_collapse_mask(iy, N, B); + RESTORE_STACK; + return collapse_mask; +} + +#ifndef OVERRIDE_renormalise_vector +void renormalise_vector(celt_norm *X, int N, opus_val16 gain, int arch) +{ + int i; +#ifdef FIXED_POINT + int k; +#endif + opus_val32 E; + opus_val16 g; + opus_val32 t; + celt_norm *xptr; + E = EPSILON + celt_inner_prod(X, X, N, arch); +#ifdef FIXED_POINT + k = celt_ilog2(E)>>1; +#endif + t = VSHR32(E, 2*(k-7)); + g = MULT16_16_P15(celt_rsqrt_norm(t),gain); + + xptr = X; + for (i=0;i +#include +#include +#include "celt_lpc.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "pitch.h" +#include "x86cpu.h" + +#if defined(FIXED_POINT) + +void celt_fir_sse4_1(const opus_val16 *x, + const opus_val16 *num, + opus_val16 *y, + int N, + int ord, + int arch) +{ + int i,j; + VARDECL(opus_val16, rnum); + + __m128i vecNoA; + opus_int32 noA ; + SAVE_STACK; + + ALLOC(rnum, ord, opus_val16); + for(i=0;i> 1; + vecNoA = _mm_set_epi32(noA, noA, noA, noA); + + for (i=0;i +#include "arch.h" + +void xcorr_kernel_sse(const opus_val16 *x, const opus_val16 *y, opus_val32 sum[4], int len) +{ + int j; + __m128 xsum1, xsum2; + xsum1 = _mm_loadu_ps(sum); + xsum2 = _mm_setzero_ps(); + + for (j = 0; j < len-3; j += 4) + { + __m128 x0 = _mm_loadu_ps(x+j); + __m128 yj = _mm_loadu_ps(y+j); + __m128 y3 = _mm_loadu_ps(y+j+3); + + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0x00),yj)); + xsum2 = _mm_add_ps(xsum2,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0x55), + _mm_shuffle_ps(yj,y3,0x49))); + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0xaa), + _mm_shuffle_ps(yj,y3,0x9e))); + xsum2 = _mm_add_ps(xsum2,_mm_mul_ps(_mm_shuffle_ps(x0,x0,0xff),y3)); + } + if (j < len) + { + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_load1_ps(x+j),_mm_loadu_ps(y+j))); + if (++j < len) + { + xsum2 = _mm_add_ps(xsum2,_mm_mul_ps(_mm_load1_ps(x+j),_mm_loadu_ps(y+j))); + if (++j < len) + { + xsum1 = _mm_add_ps(xsum1,_mm_mul_ps(_mm_load1_ps(x+j),_mm_loadu_ps(y+j))); + } + } + } + _mm_storeu_ps(sum,_mm_add_ps(xsum1,xsum2)); +} + + +void dual_inner_prod_sse(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02, + int N, opus_val32 *xy1, opus_val32 *xy2) +{ + int i; + __m128 xsum1, xsum2; + xsum1 = _mm_setzero_ps(); + xsum2 = _mm_setzero_ps(); + for (i=0;i +#include + +#include "macros.h" +#include "celt_lpc.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "pitch.h" + +#if defined(OPUS_X86_MAY_HAVE_SSE2) && defined(FIXED_POINT) +opus_val32 celt_inner_prod_sse2(const opus_val16 *x, const opus_val16 *y, + int N) +{ + opus_int i, dataSize16; + opus_int32 sum; + + __m128i inVec1_76543210, inVec1_FEDCBA98, acc1; + __m128i inVec2_76543210, inVec2_FEDCBA98, acc2; + + sum = 0; + dataSize16 = N & ~15; + + acc1 = _mm_setzero_si128(); + acc2 = _mm_setzero_si128(); + + for (i=0;i= 8) + { + inVec1_76543210 = _mm_loadu_si128((__m128i *)(&x[i + 0])); + inVec2_76543210 = _mm_loadu_si128((__m128i *)(&y[i + 0])); + + inVec1_76543210 = _mm_madd_epi16(inVec1_76543210, inVec2_76543210); + + acc1 = _mm_add_epi32(acc1, inVec1_76543210); + i += 8; + } + + acc1 = _mm_add_epi32(acc1, _mm_unpackhi_epi64( acc1, acc1)); + acc1 = _mm_add_epi32(acc1, _mm_shufflelo_epi16( acc1, 0x0E)); + sum += _mm_cvtsi128_si32(acc1); + + for (;i +#include + +#include "macros.h" +#include "celt_lpc.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "pitch.h" + +#if defined(OPUS_X86_MAY_HAVE_SSE4_1) && defined(FIXED_POINT) +#include +#include "x86cpu.h" + +opus_val32 celt_inner_prod_sse4_1(const opus_val16 *x, const opus_val16 *y, + int N) +{ + opus_int i, dataSize16; + opus_int32 sum; + __m128i inVec1_76543210, inVec1_FEDCBA98, acc1; + __m128i inVec2_76543210, inVec2_FEDCBA98, acc2; + __m128i inVec1_3210, inVec2_3210; + + sum = 0; + dataSize16 = N & ~15; + + acc1 = _mm_setzero_si128(); + acc2 = _mm_setzero_si128(); + + for (i=0;i= 8) + { + inVec1_76543210 = _mm_loadu_si128((__m128i *)(&x[i + 0])); + inVec2_76543210 = _mm_loadu_si128((__m128i *)(&y[i + 0])); + + inVec1_76543210 = _mm_madd_epi16(inVec1_76543210, inVec2_76543210); + + acc1 = _mm_add_epi32(acc1, inVec1_76543210); + i += 8; + } + + if (N - i >= 4) + { + inVec1_3210 = OP_CVTEPI16_EPI32_M64(&x[i + 0]); + inVec2_3210 = OP_CVTEPI16_EPI32_M64(&y[i + 0]); + + inVec1_3210 = _mm_mullo_epi32(inVec1_3210, inVec2_3210); + + acc1 = _mm_add_epi32(acc1, inVec1_3210); + i += 4; + } + + acc1 = _mm_add_epi32(acc1, _mm_unpackhi_epi64(acc1, acc1)); + acc1 = _mm_add_epi32(acc1, _mm_shufflelo_epi16(acc1, 0x0E)); + + sum += _mm_cvtsi128_si32(acc1); + + for (;i= 3); + + sum0 = _mm_setzero_si128(); + sum1 = _mm_setzero_si128(); + sum2 = _mm_setzero_si128(); + sum3 = _mm_setzero_si128(); + + for (j=0;j<(len-7);j+=8) + { + vecX = _mm_loadu_si128((__m128i *)(&x[j + 0])); + vecY0 = _mm_loadu_si128((__m128i *)(&y[j + 0])); + vecY1 = _mm_loadu_si128((__m128i *)(&y[j + 1])); + vecY2 = _mm_loadu_si128((__m128i *)(&y[j + 2])); + vecY3 = _mm_loadu_si128((__m128i *)(&y[j + 3])); + + sum0 = _mm_add_epi32(sum0, _mm_madd_epi16(vecX, vecY0)); + sum1 = _mm_add_epi32(sum1, _mm_madd_epi16(vecX, vecY1)); + sum2 = _mm_add_epi32(sum2, _mm_madd_epi16(vecX, vecY2)); + sum3 = _mm_add_epi32(sum3, _mm_madd_epi16(vecX, vecY3)); + } + + sum0 = _mm_add_epi32(sum0, _mm_unpackhi_epi64( sum0, sum0)); + sum0 = _mm_add_epi32(sum0, _mm_shufflelo_epi16( sum0, 0x0E)); + + sum1 = _mm_add_epi32(sum1, _mm_unpackhi_epi64( sum1, sum1)); + sum1 = _mm_add_epi32(sum1, _mm_shufflelo_epi16( sum1, 0x0E)); + + sum2 = _mm_add_epi32(sum2, _mm_unpackhi_epi64( sum2, sum2)); + sum2 = _mm_add_epi32(sum2, _mm_shufflelo_epi16( sum2, 0x0E)); + + sum3 = _mm_add_epi32(sum3, _mm_unpackhi_epi64( sum3, sum3)); + sum3 = _mm_add_epi32(sum3, _mm_shufflelo_epi16( sum3, 0x0E)); + + vecSum = _mm_unpacklo_epi64(_mm_unpacklo_epi32(sum0, sum1), + _mm_unpacklo_epi32(sum2, sum3)); + + for (;j<(len-3);j+=4) + { + vecX = OP_CVTEPI16_EPI32_M64(&x[j + 0]); + vecX0 = _mm_shuffle_epi32(vecX, 0x00); + vecX1 = _mm_shuffle_epi32(vecX, 0x55); + vecX2 = _mm_shuffle_epi32(vecX, 0xaa); + vecX3 = _mm_shuffle_epi32(vecX, 0xff); + + vecY0 = OP_CVTEPI16_EPI32_M64(&y[j + 0]); + vecY1 = OP_CVTEPI16_EPI32_M64(&y[j + 1]); + vecY2 = OP_CVTEPI16_EPI32_M64(&y[j + 2]); + vecY3 = OP_CVTEPI16_EPI32_M64(&y[j + 3]); + + sum0 = _mm_mullo_epi32(vecX0, vecY0); + sum1 = _mm_mullo_epi32(vecX1, vecY1); + sum2 = _mm_mullo_epi32(vecX2, vecY2); + sum3 = _mm_mullo_epi32(vecX3, vecY3); + + sum0 = _mm_add_epi32(sum0, sum1); + sum2 = _mm_add_epi32(sum2, sum3); + vecSum = _mm_add_epi32(vecSum, sum0); + vecSum = _mm_add_epi32(vecSum, sum2); + } + + vecX = OP_CVTEPI16_EPI32_M64(&x[len - 4]); + if (len - j == 3) + { + vecX0 = _mm_shuffle_epi32(vecX, 0x55); + vecX1 = _mm_shuffle_epi32(vecX, 0xaa); + vecX2 = _mm_shuffle_epi32(vecX, 0xff); + + vecY0 = OP_CVTEPI16_EPI32_M64(&y[j + 0]); + vecY1 = OP_CVTEPI16_EPI32_M64(&y[j + 1]); + vecY2 = OP_CVTEPI16_EPI32_M64(&y[j + 2]); + + sum0 = _mm_mullo_epi32(vecX0, vecY0); + sum1 = _mm_mullo_epi32(vecX1, vecY1); + sum2 = _mm_mullo_epi32(vecX2, vecY2); + + vecSum = _mm_add_epi32(vecSum, sum0); + vecSum = _mm_add_epi32(vecSum, sum1); + vecSum = _mm_add_epi32(vecSum, sum2); + } + else if (len - j == 2) + { + vecX0 = _mm_shuffle_epi32(vecX, 0xaa); + vecX1 = _mm_shuffle_epi32(vecX, 0xff); + + vecY0 = OP_CVTEPI16_EPI32_M64(&y[j + 0]); + vecY1 = OP_CVTEPI16_EPI32_M64(&y[j + 1]); + + sum0 = _mm_mullo_epi32(vecX0, vecY0); + sum1 = _mm_mullo_epi32(vecX1, vecY1); + + vecSum = _mm_add_epi32(vecSum, sum0); + vecSum = _mm_add_epi32(vecSum, sum1); + } + else if (len - j == 1) + { + vecX0 = _mm_shuffle_epi32(vecX, 0xff); + + vecY0 = OP_CVTEPI16_EPI32_M64(&y[j + 0]); + + sum0 = _mm_mullo_epi32(vecX0, vecY0); + + vecSum = _mm_add_epi32(vecSum, sum0); + } + + initSum = _mm_loadu_si128((__m128i *)(&sum[0])); + initSum = _mm_add_epi32(initSum, vecSum); + _mm_storeu_si128((__m128i *)sum, initSum); + +#ifdef OPUS_CHECK_ASM + celt_assert(!memcmp(sum_c, sum, sizeof(sum_c))); +#endif +} +#endif diff --git a/lib/opus/celt/x86/vq_sse.h b/lib/opus/celt/x86/vq_sse.h new file mode 100644 index 00000000..b4efe8f2 --- /dev/null +++ b/lib/opus/celt/x86/vq_sse.h @@ -0,0 +1,50 @@ +/* Copyright (c) 2016 Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef VQ_SSE_H +#define VQ_SSE_H + +#if defined(OPUS_X86_MAY_HAVE_SSE2) && !defined(FIXED_POINT) +#define OVERRIDE_OP_PVQ_SEARCH + +opus_val16 op_pvq_search_sse2(celt_norm *_X, int *iy, int K, int N, int arch); + +#if defined(OPUS_X86_PRESUME_SSE2) +#define op_pvq_search(x, iy, K, N, arch) \ + (op_pvq_search_sse2(x, iy, K, N, arch)) + +#else + +extern opus_val16 (*const OP_PVQ_SEARCH_IMPL[OPUS_ARCHMASK + 1])( + celt_norm *_X, int *iy, int K, int N, int arch); + +# define op_pvq_search(X, iy, K, N, arch) \ + ((*OP_PVQ_SEARCH_IMPL[(arch) & OPUS_ARCHMASK])(X, iy, K, N, arch)) + +#endif +#endif + +#endif diff --git a/lib/opus/celt/x86/vq_sse2.c b/lib/opus/celt/x86/vq_sse2.c new file mode 100644 index 00000000..77504286 --- /dev/null +++ b/lib/opus/celt/x86/vq_sse2.c @@ -0,0 +1,217 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2007-2016 Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include "celt_lpc.h" +#include "stack_alloc.h" +#include "mathops.h" +#include "vq.h" +#include "x86cpu.h" + + +#ifndef FIXED_POINT + +opus_val16 op_pvq_search_sse2(celt_norm *_X, int *iy, int K, int N, int arch) +{ + int i, j; + int pulsesLeft; + float xy, yy; + VARDECL(celt_norm, y); + VARDECL(celt_norm, X); + VARDECL(float, signy); + __m128 signmask; + __m128 sums; + __m128i fours; + SAVE_STACK; + + (void)arch; + /* All bits set to zero, except for the sign bit. */ + signmask = _mm_set_ps1(-0.f); + fours = _mm_set_epi32(4, 4, 4, 4); + ALLOC(y, N+3, celt_norm); + ALLOC(X, N+3, celt_norm); + ALLOC(signy, N+3, float); + + OPUS_COPY(X, _X, N); + X[N] = X[N+1] = X[N+2] = 0; + sums = _mm_setzero_ps(); + for (j=0;j (N>>1)) + { + __m128i pulses_sum; + __m128 yy4, xy4; + __m128 rcp4; + opus_val32 sum = _mm_cvtss_f32(sums); + /* If X is too small, just replace it with a pulse at 0 */ + /* Prevents infinities and NaNs from causing too many pulses + to be allocated. 64 is an approximation of infinity here. */ + if (!(sum > EPSILON && sum < 64)) + { + X[0] = QCONST16(1.f,14); + j=1; do + X[j]=0; + while (++j=0); + + /* This should never happen, but just in case it does (e.g. on silence) + we fill the first bin with pulses. */ + if (pulsesLeft > N+3) + { + opus_val16 tmp = (opus_val16)pulsesLeft; + yy = MAC16_16(yy, tmp, tmp); + yy = MAC16_16(yy, tmp, y[0]); + iy[0] += pulsesLeft; + pulsesLeft=0; + } + + for (i=0;i +static _inline void cpuid(unsigned int CPUInfo[4], unsigned int InfoType) +{ + __cpuid((int*)CPUInfo, InfoType); +} + +#else + +#if defined(CPU_INFO_BY_C) +#include +#endif + +static void cpuid(unsigned int CPUInfo[4], unsigned int InfoType) +{ +#if defined(CPU_INFO_BY_ASM) +#if defined(__i386__) && defined(__PIC__) +/* %ebx is PIC register in 32-bit, so mustn't clobber it. */ + __asm__ __volatile__ ( + "xchg %%ebx, %1\n" + "cpuid\n" + "xchg %%ebx, %1\n": + "=a" (CPUInfo[0]), + "=r" (CPUInfo[1]), + "=c" (CPUInfo[2]), + "=d" (CPUInfo[3]) : + /* We clear ECX to avoid a valgrind false-positive prior to v3.17.0. */ + "0" (InfoType), "2" (0) + ); +#else + __asm__ __volatile__ ( + "cpuid": + "=a" (CPUInfo[0]), + "=b" (CPUInfo[1]), + "=c" (CPUInfo[2]), + "=d" (CPUInfo[3]) : + /* We clear ECX to avoid a valgrind false-positive prior to v3.17.0. */ + "0" (InfoType), "2" (0) + ); +#endif +#elif defined(CPU_INFO_BY_C) + /* We use __get_cpuid_count to clear ECX to avoid a valgrind false-positive + prior to v3.17.0.*/ + if (!__get_cpuid_count(InfoType, 0, &(CPUInfo[0]), &(CPUInfo[1]), &(CPUInfo[2]), &(CPUInfo[3]))) { + /* Our function cannot fail, but __get_cpuid{_count} can. + Returning all zeroes will effectively disable all SIMD, which is + what we want on CPUs that don't support CPUID. */ + CPUInfo[3] = CPUInfo[2] = CPUInfo[1] = CPUInfo[0] = 0; + } +#else +# error "Configured to use x86 RTCD, but no CPU detection method available. " \ + "Reconfigure with --disable-rtcd (or send patches)." +#endif +} + +#endif + +typedef struct CPU_Feature{ + /* SIMD: 128-bit */ + int HW_SSE; + int HW_SSE2; + int HW_SSE41; + /* SIMD: 256-bit */ + int HW_AVX; +} CPU_Feature; + +static void opus_cpu_feature_check(CPU_Feature *cpu_feature) +{ + unsigned int info[4]; + unsigned int nIds = 0; + + cpuid(info, 0); + nIds = info[0]; + + if (nIds >= 1){ + cpuid(info, 1); + cpu_feature->HW_SSE = (info[3] & (1 << 25)) != 0; + cpu_feature->HW_SSE2 = (info[3] & (1 << 26)) != 0; + cpu_feature->HW_SSE41 = (info[2] & (1 << 19)) != 0; + cpu_feature->HW_AVX = (info[2] & (1 << 28)) != 0; + } + else { + cpu_feature->HW_SSE = 0; + cpu_feature->HW_SSE2 = 0; + cpu_feature->HW_SSE41 = 0; + cpu_feature->HW_AVX = 0; + } +} + +static int opus_select_arch_impl(void) +{ + CPU_Feature cpu_feature; + int arch; + + opus_cpu_feature_check(&cpu_feature); + + arch = 0; + if (!cpu_feature.HW_SSE) + { + return arch; + } + arch++; + + if (!cpu_feature.HW_SSE2) + { + return arch; + } + arch++; + + if (!cpu_feature.HW_SSE41) + { + return arch; + } + arch++; + + if (!cpu_feature.HW_AVX) + { + return arch; + } + arch++; + + return arch; +} + +int opus_select_arch(void) { + int arch = opus_select_arch_impl(); +#ifdef FUZZING + /* Randomly downgrade the architecture. */ + arch = rand()%(arch+1); +#endif + return arch; +} + +#endif diff --git a/lib/opus/celt/x86/x86cpu.h b/lib/opus/celt/x86/x86cpu.h new file mode 100644 index 00000000..04e80489 --- /dev/null +++ b/lib/opus/celt/x86/x86cpu.h @@ -0,0 +1,73 @@ +/* Copyright (c) 2014, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if !defined(X86CPU_H) +# define X86CPU_H + +# if defined(OPUS_X86_MAY_HAVE_SSE) +# define MAY_HAVE_SSE(name) name ## _sse +# else +# define MAY_HAVE_SSE(name) name ## _c +# endif + +# if defined(OPUS_X86_MAY_HAVE_SSE2) +# define MAY_HAVE_SSE2(name) name ## _sse2 +# else +# define MAY_HAVE_SSE2(name) name ## _c +# endif + +# if defined(OPUS_X86_MAY_HAVE_SSE4_1) +# define MAY_HAVE_SSE4_1(name) name ## _sse4_1 +# else +# define MAY_HAVE_SSE4_1(name) name ## _c +# endif + +# if defined(OPUS_X86_MAY_HAVE_AVX) +# define MAY_HAVE_AVX(name) name ## _avx +# else +# define MAY_HAVE_AVX(name) name ## _c +# endif + +# if defined(OPUS_HAVE_RTCD) +int opus_select_arch(void); +# endif + +/*MOVD should not impose any alignment restrictions, but the C standard does, + and UBSan will report errors if we actually make unaligned accesses. + Use this to work around those restrictions (which should hopefully all get + optimized to a single MOVD instruction).*/ +#define OP_LOADU_EPI32(x) \ + (int)((*(unsigned char *)(x) | *((unsigned char *)(x) + 1) << 8U |\ + *((unsigned char *)(x) + 2) << 16U | (opus_uint32)*((unsigned char *)(x) + 3) << 24U)) + +#define OP_CVTEPI8_EPI32_M32(x) \ + (_mm_cvtepi8_epi32(_mm_cvtsi32_si128(OP_LOADU_EPI32(x)))) + +#define OP_CVTEPI16_EPI32_M64(x) \ + (_mm_cvtepi16_epi32(_mm_loadl_epi64((__m128i *)(x)))) + +#endif diff --git a/lib/opus/celt_headers.mk b/lib/opus/celt_headers.mk new file mode 100644 index 00000000..706185da --- /dev/null +++ b/lib/opus/celt_headers.mk @@ -0,0 +1,53 @@ +CELT_HEAD = \ +celt/arch.h \ +celt/bands.h \ +celt/celt.h \ +celt/cpu_support.h \ +include/opus_types.h \ +include/opus_defines.h \ +include/opus_custom.h \ +celt/cwrs.h \ +celt/ecintrin.h \ +celt/entcode.h \ +celt/entdec.h \ +celt/entenc.h \ +celt/fixed_debug.h \ +celt/fixed_generic.h \ +celt/float_cast.h \ +celt/_kiss_fft_guts.h \ +celt/kiss_fft.h \ +celt/laplace.h \ +celt/mathops.h \ +celt/mdct.h \ +celt/mfrngcod.h \ +celt/modes.h \ +celt/os_support.h \ +celt/pitch.h \ +celt/celt_lpc.h \ +celt/x86/celt_lpc_sse.h \ +celt/quant_bands.h \ +celt/rate.h \ +celt/stack_alloc.h \ +celt/vq.h \ +celt/static_modes_float.h \ +celt/static_modes_fixed.h \ +celt/static_modes_float_arm_ne10.h \ +celt/static_modes_fixed_arm_ne10.h \ +celt/arm/armcpu.h \ +celt/arm/fixed_armv4.h \ +celt/arm/fixed_armv5e.h \ +celt/arm/fixed_arm64.h \ +celt/arm/kiss_fft_armv4.h \ +celt/arm/kiss_fft_armv5e.h \ +celt/arm/pitch_arm.h \ +celt/arm/fft_arm.h \ +celt/arm/mdct_arm.h \ +celt/mips/celt_mipsr1.h \ +celt/mips/fixed_generic_mipsr1.h \ +celt/mips/kiss_fft_mipsr1.h \ +celt/mips/mdct_mipsr1.h \ +celt/mips/pitch_mipsr1.h \ +celt/mips/vq_mipsr1.h \ +celt/x86/pitch_sse.h \ +celt/x86/vq_sse.h \ +celt/x86/x86cpu.h diff --git a/lib/opus/celt_sources.mk b/lib/opus/celt_sources.mk new file mode 100644 index 00000000..d6b6765b --- /dev/null +++ b/lib/opus/celt_sources.mk @@ -0,0 +1,52 @@ +CELT_SOURCES = \ +celt/bands.c \ +celt/celt.c \ +celt/celt_encoder.c \ +celt/celt_decoder.c \ +celt/cwrs.c \ +celt/entcode.c \ +celt/entdec.c \ +celt/entenc.c \ +celt/kiss_fft.c \ +celt/laplace.c \ +celt/mathops.c \ +celt/mdct.c \ +celt/modes.c \ +celt/pitch.c \ +celt/celt_lpc.c \ +celt/quant_bands.c \ +celt/rate.c \ +celt/vq.c + +CELT_SOURCES_X86_RTCD = \ +celt/x86/x86cpu.c \ +celt/x86/x86_celt_map.c + +CELT_SOURCES_SSE = \ +celt/x86/pitch_sse.c + +CELT_SOURCES_SSE2 = \ +celt/x86/pitch_sse2.c \ +celt/x86/vq_sse2.c + +CELT_SOURCES_SSE4_1 = \ +celt/x86/celt_lpc_sse4_1.c \ +celt/x86/pitch_sse4_1.c + +CELT_SOURCES_ARM_RTCD = \ +celt/arm/armcpu.c \ +celt/arm/arm_celt_map.c + +CELT_SOURCES_ARM_ASM = \ +celt/arm/celt_pitch_xcorr_arm.s + +CELT_AM_SOURCES_ARM_ASM = \ +celt/arm/armopts.s.in + +CELT_SOURCES_ARM_NEON_INTR = \ +celt/arm/celt_neon_intr.c \ +celt/arm/pitch_neon_intr.c + +CELT_SOURCES_ARM_NE10 = \ +celt/arm/celt_fft_ne10.c \ +celt/arm/celt_mdct_ne10.c diff --git a/lib/opus/cmake/CFeatureCheck.cmake b/lib/opus/cmake/CFeatureCheck.cmake new file mode 100644 index 00000000..08828f58 --- /dev/null +++ b/lib/opus/cmake/CFeatureCheck.cmake @@ -0,0 +1,39 @@ +# - Compile and run code to check for C features +# +# This functions compiles a source file under the `cmake` folder +# and adds the corresponding `HAVE_[FILENAME]` flag to the CMake +# environment +# +# c_feature_check( []) +# +# - Example +# +# include(CFeatureCheck) +# c_feature_check(VLA) + +if(__c_feature_check) + return() +endif() +set(__c_feature_check INCLUDED) + +function(c_feature_check FILE) + string(TOLOWER ${FILE} FILE) + string(TOUPPER ${FILE} VAR) + string(TOUPPER "${VAR}_SUPPORTED" FEATURE) + if (DEFINED ${VAR}_SUPPORTED) + set(${VAR}_SUPPORTED 1 PARENT_SCOPE) + return() + endif() + + if (NOT DEFINED COMPILE_${FEATURE}) + message(STATUS "Performing Test ${FEATURE}") + try_compile(COMPILE_${FEATURE} ${PROJECT_BINARY_DIR} ${PROJECT_SOURCE_DIR}/cmake/${FILE}.c) + endif() + + if(COMPILE_${FEATURE}) + message(STATUS "Performing Test ${FEATURE} -- success") + set(${VAR}_SUPPORTED 1 PARENT_SCOPE) + else() + message(STATUS "Performing Test ${FEATURE} -- failed to compile") + endif() +endfunction() diff --git a/lib/opus/cmake/OpusBuildtype.cmake b/lib/opus/cmake/OpusBuildtype.cmake new file mode 100644 index 00000000..557cc89b --- /dev/null +++ b/lib/opus/cmake/OpusBuildtype.cmake @@ -0,0 +1,27 @@ +# Set a default build type if none was specified +if(__opus_buildtype) + return() +endif() +set(__opus_buildtype INCLUDED) + +if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES) + if(CMAKE_C_FLAGS) + message(STATUS "CMAKE_C_FLAGS: " ${CMAKE_C_FLAGS}) + else() + set(default_build_type "Release") + message( + STATUS + "Setting build type to '${default_build_type}' as none was specified and no CFLAGS was exported." + ) + set(CMAKE_BUILD_TYPE "${default_build_type}" + CACHE STRING "Choose the type of build." + FORCE) + # Set the possible values of build type for cmake-gui + set_property(CACHE CMAKE_BUILD_TYPE + PROPERTY STRINGS + "Debug" + "Release" + "MinSizeRel" + "RelWithDebInfo") + endif() +endif() diff --git a/lib/opus/cmake/OpusConfig.cmake b/lib/opus/cmake/OpusConfig.cmake new file mode 100644 index 00000000..5cb899d7 --- /dev/null +++ b/lib/opus/cmake/OpusConfig.cmake @@ -0,0 +1,117 @@ +if(__opus_config) + return() +endif() +set(__opus_config INCLUDED) + +include(OpusFunctions) + +configure_file(cmake/config.h.cmake.in config.h @ONLY) +add_definitions(-DHAVE_CONFIG_H) + +set_property(GLOBAL PROPERTY USE_FOLDERS ON) + +if(MSVC) + # For compilers that have no notion of a C standard level, + # such as Microsoft Visual C++ before VS 16.7, + # this property has no effect. + set(CMAKE_C_STANDARD 11) +else() + set(CMAKE_C_STANDARD 99) +endif() + +if(MSVC) + add_definitions(-D_CRT_SECURE_NO_WARNINGS) +endif() + +include(CFeatureCheck) +c_feature_check(VLA) + +include(CheckIncludeFile) +check_include_file(alloca.h HAVE_ALLOCA_H) + +include(CheckSymbolExists) +if(HAVE_ALLOCA_H) + add_definitions(-DHAVE_ALLOCA_H) + check_symbol_exists(alloca "alloca.h" USE_ALLOCA_SUPPORTED) +else() + check_symbol_exists(alloca "stdlib.h;malloc.h" USE_ALLOCA_SUPPORTED) +endif() + +include(CMakePushCheckState) +cmake_push_check_state(RESET) +include(CheckLibraryExists) +check_library_exists(m floor "" HAVE_LIBM) +if(HAVE_LIBM) + list(APPEND OPUS_REQUIRED_LIBRARIES m) + set(CMAKE_REQUIRED_LIBRARIES m) +endif() + +check_symbol_exists(lrintf "math.h" HAVE_LRINTF) +check_symbol_exists(lrint "math.h" HAVE_LRINT) +cmake_pop_check_state() + +if(CMAKE_SYSTEM_PROCESSOR MATCHES "(i[0-9]86|x86|X86|amd64|AMD64|x86_64)") + if(CMAKE_SIZEOF_VOID_P EQUAL 8) + set(OPUS_CPU_X64 1) + else() + set(OPUS_CPU_X86 1) + endif() +elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "(arm|aarch64)") + set(OPUS_CPU_ARM 1) +endif() + +if(NOT OPUS_DISABLE_INTRINSICS) + opus_supports_cpu_detection(RUNTIME_CPU_CAPABILITY_DETECTION) +endif() + +if(OPUS_CPU_X86 OR OPUS_CPU_X64 AND NOT OPUS_DISABLE_INTRINSICS) + opus_detect_sse(COMPILER_SUPPORT_SIMD) +elseif(OPUS_CPU_ARM AND NOT OPUS_DISABLE_INTRINSICS) + opus_detect_neon(COMPILER_SUPPORT_NEON) + if(COMPILER_SUPPORT_NEON) + option(OPUS_USE_NEON "Option to enable NEON" ON) + option(OPUS_MAY_HAVE_NEON "Does runtime check for neon support" ON) + option(OPUS_PRESUME_NEON "Assume target CPU has NEON support" OFF) + if(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64") + set(OPUS_PRESUME_NEON ON) + elseif(CMAKE_SYSTEM_NAME MATCHES "iOS") + set(OPUS_PRESUME_NEON ON) + endif() + endif() +endif() + +if(MSVC) + check_flag(FAST_MATH /fp:fast) + check_flag(STACK_PROTECTOR /GS) + check_flag(STACK_PROTECTOR_DISABLED /GS-) +else() + check_flag(FAST_MATH -ffast-math) + check_flag(STACK_PROTECTOR -fstack-protector-strong) + check_flag(HIDDEN_VISIBILITY -fvisibility=hidden) + set(FORTIFY_SOURCE_SUPPORTED 1) +endif() + +if(MINGW) + # For MINGW we need to link ssp lib for security features such as + # stack protector and fortify_sources + check_library_exists(ssp __stack_chk_fail "" HAVE_LIBSSP) + if(NOT HAVE_LIBSSP) + message(WARNING "Could not find libssp in MinGW, disabling STACK_PROTECTOR and FORTIFY_SOURCE") + set(STACK_PROTECTOR_SUPPORTED 0) + set(FORTIFY_SOURCE_SUPPORTED 0) + endif() +endif() + +if(NOT MSVC) + set(WARNING_LIST -Wall -W -Wstrict-prototypes -Wextra -Wcast-align -Wnested-externs -Wshadow) + include(CheckCCompilerFlag) + foreach(WARNING_FLAG ${WARNING_LIST}) + string(REPLACE - "" WARNING_VAR ${WARNING_FLAG}) + check_c_compiler_flag(${WARNING_FLAG} ${WARNING_VAR}_SUPPORTED) + if(${WARNING_VAR}_SUPPORTED) + add_compile_options(${WARNING_FLAG}) + endif() + endforeach() +endif() + +add_compile_options(-Wno-maybe-uninitialized -Wno-stringop-overread) diff --git a/lib/opus/cmake/OpusConfig.cmake.in b/lib/opus/cmake/OpusConfig.cmake.in new file mode 100644 index 00000000..0b21231d --- /dev/null +++ b/lib/opus/cmake/OpusConfig.cmake.in @@ -0,0 +1,20 @@ +set(OPUS_VERSION @PROJECT_VERSION@) +set(OPUS_VERSION_STRING @PROJECT_VERSION@) +set(OPUS_VERSION_MAJOR @PROJECT_VERSION_MAJOR@) +set(OPUS_VERSION_MINOR @PROJECT_VERSION_MINOR@) +set(OPUS_VERSION_PATCH @PROJECT_VERSION_PATCH@) + +@PACKAGE_INIT@ + +set_and_check(OPUS_INCLUDE_DIR "@PACKAGE_INCLUDE_INSTALL_DIR@") +set(OPUS_INCLUDE_DIR ${OPUS_INCLUDE_DIR};${OPUS_INCLUDE_DIR}/opus) +set(OPUS_INCLUDE_DIRS "@PACKAGE_INCLUDE_INSTALL_DIR@;@PACKAGE_INCLUDE_INSTALL_DIR@/opus") + +include(${CMAKE_CURRENT_LIST_DIR}/OpusTargets.cmake) + +set(OPUS_LIBRARY Opus::opus) +set(OPUS_LIBRARIES Opus::opus) + +check_required_components(Opus) + +set(OPUS_FOUND 1) diff --git a/lib/opus/cmake/OpusFunctions.cmake b/lib/opus/cmake/OpusFunctions.cmake new file mode 100644 index 00000000..3f22ad81 --- /dev/null +++ b/lib/opus/cmake/OpusFunctions.cmake @@ -0,0 +1,229 @@ +if(__opus_functions) + return() +endif() +set(__opus_functions INCLUDED) + +function(get_library_version OPUS_LIBRARY_VERSION OPUS_LIBRARY_VERSION_MAJOR) + file(STRINGS configure.ac opus_lt_current_string + LIMIT_COUNT 1 + REGEX "OPUS_LT_CURRENT=") + string(REGEX MATCH + "OPUS_LT_CURRENT=([0-9]*)" + _ + ${opus_lt_current_string}) + set(OPUS_LT_CURRENT ${CMAKE_MATCH_1}) + + file(STRINGS configure.ac opus_lt_revision_string + LIMIT_COUNT 1 + REGEX "OPUS_LT_REVISION=") + string(REGEX MATCH + "OPUS_LT_REVISION=([0-9]*)" + _ + ${opus_lt_revision_string}) + set(OPUS_LT_REVISION ${CMAKE_MATCH_1}) + + file(STRINGS configure.ac opus_lt_age_string + LIMIT_COUNT 1 + REGEX "OPUS_LT_AGE=") + string(REGEX MATCH + "OPUS_LT_AGE=([0-9]*)" + _ + ${opus_lt_age_string}) + set(OPUS_LT_AGE ${CMAKE_MATCH_1}) + + math(EXPR OPUS_LIBRARY_VERSION_MAJOR "${OPUS_LT_CURRENT} - ${OPUS_LT_AGE}") + set(OPUS_LIBRARY_VERSION_MINOR ${OPUS_LT_AGE}) + set(OPUS_LIBRARY_VERSION_PATCH ${OPUS_LT_REVISION}) + set( + OPUS_LIBRARY_VERSION + "${OPUS_LIBRARY_VERSION_MAJOR}.${OPUS_LIBRARY_VERSION_MINOR}.${OPUS_LIBRARY_VERSION_PATCH}" + PARENT_SCOPE) + set(OPUS_LIBRARY_VERSION_MAJOR ${OPUS_LIBRARY_VERSION_MAJOR} PARENT_SCOPE) +endfunction() + +function(check_flag NAME FLAG) + include(CheckCCompilerFlag) + check_c_compiler_flag(${FLAG} ${NAME}_SUPPORTED) +endfunction() + +include(CheckIncludeFile) +# function to check if compiler supports SSE, SSE2, SSE4.1 and AVX if target +# systems may not have SSE support then use OPUS_MAY_HAVE_SSE option if target +# system is guaranteed to have SSE support then OPUS_PRESUME_SSE can be used to +# skip SSE runtime check +function(opus_detect_sse COMPILER_SUPPORT_SIMD) + message(STATUS "Check SIMD support by compiler") + check_include_file(xmmintrin.h HAVE_XMMINTRIN_H) # SSE1 + if(HAVE_XMMINTRIN_H) + if(MSVC) + # different arch options for 32 and 64 bit target for MSVC + if(CMAKE_SIZEOF_VOID_P EQUAL 4) + check_flag(SSE1 /arch:SSE) + else() + set(SSE1_SUPPORTED + 1 + PARENT_SCOPE) + endif() + else() + check_flag(SSE1 -msse) + endif() + else() + set(SSE1_SUPPORTED + 0 + PARENT_SCOPE) + endif() + + check_include_file(emmintrin.h HAVE_EMMINTRIN_H) # SSE2 + if(HAVE_EMMINTRIN_H) + if(MSVC) + if(CMAKE_SIZEOF_VOID_P EQUAL 4) + check_flag(SSE2 /arch:SSE2) + else() + set(SSE2_SUPPORTED + 1 + PARENT_SCOPE) + endif() + else() + check_flag(SSE2 -msse2) + endif() + else() + set(SSE2_SUPPORTED + 0 + PARENT_SCOPE) + endif() + + check_include_file(smmintrin.h HAVE_SMMINTRIN_H) # SSE4.1 + if(HAVE_SMMINTRIN_H) + if(MSVC) + if(CMAKE_SIZEOF_VOID_P EQUAL 4) + check_flag(SSE4_1 /arch:SSE2) # SSE2 and above + else() + set(SSE4_1_SUPPORTED + 1 + PARENT_SCOPE) + endif() + else() + check_flag(SSE4_1 -msse4.1) + endif() + else() + set(SSE4_1_SUPPORTED + 0 + PARENT_SCOPE) + endif() + + check_include_file(immintrin.h HAVE_IMMINTRIN_H) # AVX + if(HAVE_IMMINTRIN_H) + if(MSVC) + check_flag(AVX /arch:AVX) + else() + check_flag(AVX -mavx) + endif() + else() + set(AVX_SUPPORTED + 0 + PARENT_SCOPE) + endif() + + if(SSE1_SUPPORTED OR SSE2_SUPPORTED OR SSE4_1_SUPPORTED OR AVX_SUPPORTED) + set(COMPILER_SUPPORT_SIMD 1 PARENT_SCOPE) + else() + message(STATUS "No SIMD support in compiler") + endif() +endfunction() + +function(opus_detect_neon COMPILER_SUPPORT_NEON) + if(CMAKE_SYSTEM_PROCESSOR MATCHES "(arm|aarch64)") + message(STATUS "Check NEON support by compiler") + check_include_file(arm_neon.h HAVE_ARM_NEON_H) + if(HAVE_ARM_NEON_H) + set(COMPILER_SUPPORT_NEON ${HAVE_ARM_NEON_H} PARENT_SCOPE) + endif() + endif() +endfunction() + +function(opus_supports_cpu_detection RUNTIME_CPU_CAPABILITY_DETECTION) + set(RUNTIME_CPU_CAPABILITY_DETECTION 0 PARENT_SCOPE) + if(OPUS_CPU_X86 OR OPUS_CPU_X64) + if(MSVC) + check_include_file(intrin.h HAVE_INTRIN_H) + if(HAVE_INTRIN_H) + # if intrin.h is available we assume __cpuid is there + set(RUNTIME_CPU_CAPABILITY_DETECTION 1 PARENT_SCOPE) + endif() + else() + include(CFeatureCheck) + c_feature_check(CPU_INFO_BY_ASM) + set(CPU_INFO_BY_ASM_SUPPORTED ${CPU_INFO_BY_ASM_SUPPORTED} PARENT_SCOPE) + check_include_file(cpuid.h HAVE_CPUID_H) + if(HAVE_CPUID_H) + c_feature_check(CPU_INFO_BY_C) + set(CPU_INFO_BY_C_SUPPORTED ${CPU_INFO_BY_C_SUPPORTED} PARENT_SCOPE) + endif() + if(CPU_INFO_BY_ASM_SUPPORTED OR CPU_INFO_BY_C_SUPPORTED) + set(RUNTIME_CPU_CAPABILITY_DETECTION 1 PARENT_SCOPE) + endif() + endif() + elseif(OPUS_CPU_ARM) + # ARM cpu detection is implemented for Windows and anything + # using a Linux kernel (such as Android). + if (CMAKE_SYSTEM_NAME MATCHES "(Windows|Linux|Android)") + set(RUNTIME_CPU_CAPABILITY_DETECTION 1 PARENT_SCOPE) + endif () + else() + set(RUNTIME_CPU_CAPABILITY_DETECTION 0 PARENT_SCOPE) + endif() +endfunction() + +function(add_sources_group target group) + target_sources(${target} PRIVATE ${ARGN}) + source_group(${group} FILES ${ARGN}) +endfunction() + +function(get_opus_sources SOURCE_GROUP MAKE_FILE SOURCES) + # read file, each item in list is one group + file(STRINGS ${MAKE_FILE} opus_sources) + + # add wildcard for regex match + string(CONCAT SOURCE_GROUP ${SOURCE_GROUP} ".*$") + + # find group + foreach(val IN LISTS opus_sources) + if(val MATCHES ${SOURCE_GROUP}) + list(LENGTH val list_length) + if(${list_length} EQUAL 1) + # for tests split by '=' and clean up the rest into a list + string(FIND ${val} "=" index) + math(EXPR index "${index} + 1") + string(SUBSTRING ${val} + ${index} + -1 + sources) + string(REPLACE " " + ";" + sources + ${sources}) + else() + # discard the group + list(REMOVE_AT val 0) + set(sources ${val}) + endif() + break() + endif() + endforeach() + + list(LENGTH sources list_length) + if(${list_length} LESS 1) + message( + FATAL_ERROR + "No files parsed succesfully from ${SOURCE_GROUP} in ${MAKE_FILE}") + endif() + + # remove trailing whitespaces + set(list_var "") + foreach(source ${sources}) + string(STRIP "${source}" source) + list(APPEND list_var "${source}") + endforeach() + + set(${SOURCES} ${list_var} PARENT_SCOPE) +endfunction() diff --git a/lib/opus/cmake/OpusPackageVersion.cmake b/lib/opus/cmake/OpusPackageVersion.cmake new file mode 100644 index 00000000..447ce3b1 --- /dev/null +++ b/lib/opus/cmake/OpusPackageVersion.cmake @@ -0,0 +1,70 @@ +if(__opus_version) + return() +endif() +set(__opus_version INCLUDED) + +function(get_package_version PACKAGE_VERSION PROJECT_VERSION) + + find_package(Git) + if(GIT_FOUND AND EXISTS "${CMAKE_CURRENT_LIST_DIR}/.git") + execute_process(COMMAND ${GIT_EXECUTABLE} + --git-dir=${CMAKE_CURRENT_LIST_DIR}/.git describe + --tags --match "v*" OUTPUT_VARIABLE OPUS_PACKAGE_VERSION) + if(OPUS_PACKAGE_VERSION) + string(STRIP ${OPUS_PACKAGE_VERSION}, OPUS_PACKAGE_VERSION) + string(REPLACE \n + "" + OPUS_PACKAGE_VERSION + ${OPUS_PACKAGE_VERSION}) + string(REPLACE , + "" + OPUS_PACKAGE_VERSION + ${OPUS_PACKAGE_VERSION}) + + string(SUBSTRING ${OPUS_PACKAGE_VERSION} + 1 + -1 + OPUS_PACKAGE_VERSION) + message(STATUS "Opus package version from git repo: ${OPUS_PACKAGE_VERSION}") + endif() + + elseif(EXISTS "${CMAKE_CURRENT_LIST_DIR}/package_version" + AND NOT OPUS_PACKAGE_VERSION) + # Not a git repo, lets' try to parse it from package_version file if exists + file(STRINGS package_version OPUS_PACKAGE_VERSION + LIMIT_COUNT 1 + REGEX "PACKAGE_VERSION=") + string(REPLACE "PACKAGE_VERSION=" + "" + OPUS_PACKAGE_VERSION + ${OPUS_PACKAGE_VERSION}) + string(REPLACE "\"" + "" + OPUS_PACKAGE_VERSION + ${OPUS_PACKAGE_VERSION}) + # In case we have a unknown dist here we just replace it with 0 + string(REPLACE "unknown" + "0" + OPUS_PACKAGE_VERSION + ${OPUS_PACKAGE_VERSION}) + message(STATUS "Opus package version from package_version file: ${OPUS_PACKAGE_VERSION}") + endif() + + if(OPUS_PACKAGE_VERSION) + string(REGEX + REPLACE "^([0-9]+.[0-9]+\\.?([0-9]+)?).*" + "\\1" + OPUS_PROJECT_VERSION + ${OPUS_PACKAGE_VERSION}) + else() + # fail to parse version from git and package version + message(WARNING "Could not get package version.") + set(OPUS_PACKAGE_VERSION 0) + set(OPUS_PROJECT_VERSION 0) + endif() + + message(STATUS "Opus project version: ${OPUS_PROJECT_VERSION}") + + set(PACKAGE_VERSION ${OPUS_PACKAGE_VERSION} PARENT_SCOPE) + set(PROJECT_VERSION ${OPUS_PROJECT_VERSION} PARENT_SCOPE) +endfunction() diff --git a/lib/opus/cmake/OpusSources.cmake b/lib/opus/cmake/OpusSources.cmake new file mode 100644 index 00000000..b47f8c69 --- /dev/null +++ b/lib/opus/cmake/OpusSources.cmake @@ -0,0 +1,49 @@ +if(__opus_sources) + return() +endif() +set(__opus_sources INCLUDED) + +include(OpusFunctions) + +get_opus_sources(SILK_HEAD silk_headers.mk silk_headers) +get_opus_sources(SILK_SOURCES silk_sources.mk silk_sources) +get_opus_sources(SILK_SOURCES_FLOAT silk_sources.mk silk_sources_float) +get_opus_sources(SILK_SOURCES_FIXED silk_sources.mk silk_sources_fixed) +get_opus_sources(SILK_SOURCES_X86_RTCD silk_sources.mk silk_sources_x86_rtcd) +get_opus_sources(SILK_SOURCES_SSE4_1 silk_sources.mk silk_sources_sse4_1) +get_opus_sources(SILK_SOURCES_FIXED_SSE4_1 silk_sources.mk + silk_sources_fixed_sse4_1) +get_opus_sources(SILK_SOURCES_ARM_RTCD silk_sources.mk silk_sources_arm_rtcd) +get_opus_sources(SILK_SOURCES_ARM_NEON_INTR silk_sources.mk + silk_sources_arm_neon_intr) +get_opus_sources(SILK_SOURCES_FIXED_ARM_NEON_INTR silk_sources.mk + silk_sources_fixed_arm_neon_intr) + +get_opus_sources(OPUS_HEAD opus_headers.mk opus_headers) +get_opus_sources(OPUS_SOURCES opus_sources.mk opus_sources) +get_opus_sources(OPUS_SOURCES_FLOAT opus_sources.mk opus_sources_float) + +get_opus_sources(CELT_HEAD celt_headers.mk celt_headers) +get_opus_sources(CELT_SOURCES celt_sources.mk celt_sources) +get_opus_sources(CELT_SOURCES_X86_RTCD celt_sources.mk celt_sources_x86_rtcd) +get_opus_sources(CELT_SOURCES_SSE celt_sources.mk celt_sources_sse) +get_opus_sources(CELT_SOURCES_SSE2 celt_sources.mk celt_sources_sse2) +get_opus_sources(CELT_SOURCES_SSE4_1 celt_sources.mk celt_sources_sse4_1) +get_opus_sources(CELT_SOURCES_ARM_RTCD celt_sources.mk celt_sources_arm_rtcd) +get_opus_sources(CELT_SOURCES_ARM_ASM celt_sources.mk celt_sources_arm_asm) +get_opus_sources(CELT_AM_SOURCES_ARM_ASM celt_sources.mk + celt_am_sources_arm_asm) +get_opus_sources(CELT_SOURCES_ARM_NEON_INTR celt_sources.mk + celt_sources_arm_neon_intr) +get_opus_sources(CELT_SOURCES_ARM_NE10 celt_sources.mk celt_sources_arm_ne10) + +get_opus_sources(opus_demo_SOURCES Makefile.am opus_demo_sources) +get_opus_sources(opus_custom_demo_SOURCES Makefile.am opus_custom_demo_sources) +get_opus_sources(opus_compare_SOURCES Makefile.am opus_compare_sources) +get_opus_sources(tests_test_opus_api_SOURCES Makefile.am test_opus_api_sources) +get_opus_sources(tests_test_opus_encode_SOURCES Makefile.am + test_opus_encode_sources) +get_opus_sources(tests_test_opus_decode_SOURCES Makefile.am + test_opus_decode_sources) +get_opus_sources(tests_test_opus_padding_SOURCES Makefile.am + test_opus_padding_sources) diff --git a/lib/opus/cmake/README.md b/lib/opus/cmake/README.md new file mode 100644 index 00000000..d5f03375 --- /dev/null +++ b/lib/opus/cmake/README.md @@ -0,0 +1,132 @@ +# Using CMake for the Opus Project + +This guide provides instructions for using CMake to build the Opus project with various configuration options. CMake is a widely used build system generator that helps manage the build process across different platforms. + +Note: Please keep in mind that software documentation can sometimes go out of date as new versions are released. It is always recommended to refer to the official CMake documentation for the most up-to-date and accurate information. You can find the official CMake documentation at [cmake.org/documentation](https://cmake.org/documentation/). + +## Prerequisites + +Before proceeding, make sure you have the following prerequisites installed: + +- CMake +- Git (optional, but recommended for version control integration) +- Working C compiler + +## Build Instructions + +Follow the steps below to build the Opus project using CMake: + +1. Clone the Opus repository using Git: + + ```shell + git clone https://gitlab.xiph.org/xiph/opus + ``` + +2. Create a build directory within the Opus repository: + + ```shell + cd opus + mkdir build + cd build + ``` + +3. Configure the build with CMake. You can set the desired configuration options using CMake's `-D` flag. Here are some available options: + + - `OPUS_BUILD_SHARED_LIBRARY`: build shared library. + - `OPUS_BUILD_TESTING`: build tests. + - `OPUS_BUILD_PROGRAMS`: build programs. + - `OPUS_CUSTOM_MODES`, enable non-Opus modes, e.g. 44.1 kHz & 2^n frames. + + For example, to enable the custom modes and build programs, use the following command: + + ```shell + cmake .. -DOPUS_BUILD_PROGRAMS=ON -DOPUS_BUILD_TESTING=ON + ``` + +4. Build the Opus project: + + ```shell + cmake --build . + ``` + +5. After a successful build, you can find the compiled Opus library and associated files in the build directory. + +## Testing with CTest + +Opus provides a comprehensive test suite to ensure the functionality and correctness of the project. You can execute the tests using CTest, a part of the CMake build system. CTest allows for automated testing and provides useful features for managing and evaluating the test results. + +To run the Opus tests using CTest, follow these steps: + +1. Navigate to the build directory after configuring and building the project with CMake: + + ```shell + cd build + ``` + +2. Execute the tests using CTest: + + ```shell + ctest + ``` + +Note: For Windows you need to specify which configuration to test + +```shell +ctest -C Debug +``` + +## Platform Support and Bug Reporting + +CMake aims to provide broad platform support, allowing the Opus project to be built and used on major operating systems and platforms. The supported platforms include: + +- Windows +- macOS +- Linux +- Android +- iOS + +CMake achieves platform support by generating platform-specific build files (e.g., Makefiles, Visual Studio projects) based on the target platform. This allows developers to build and configure the Opus project consistently across different operating systems and environments. + +While CMake strives to ensure compatibility and stability across platforms, bugs or issues may still arise in specific configurations. If you encounter any problems during the configuration process or while building the Opus project, we encourage you to file an issue in the [project's issue tracker](https://gitlab.xiph.org/xiph/opus/-/issues). + +When reporting an issue, please provide the following information to help us understand and reproduce the configuration problem effectively: + +1. Detailed description of the issue, including any error messages or unexpected behavior observed. +2. Steps to reproduce the problem, including the CMake command and any specific configuration options used. +3. Operating system and version (e.g., Windows 10, macOS Big Sur, Ubuntu 20.04). +4. CMake version (e.g., CMake 3.21.1). +5. Any relevant information about the platform, toolchain, or dependencies used. +6. Additional context or details that might assist in troubleshooting the issue. + +By providing thorough information when reporting configuration issues, you contribute to improving the Opus project's compatibility and reliability across different platforms. + +We appreciate your help in identifying and addressing any configuration-related problems, ensuring a better experience for all users of the Opus project. + +## Platform Specific Examples + +Note: Examples can go out of date. Always refer to documentation for latest reference. + +### Cross compiling for Android + +```shell +cmake .. -DCMAKE_TOOLCHAIN_FILE=${ANDROID_HOME}/ndk/25.2.9519653/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a +``` + +For more information about cross compiling for android, you can refer to the [Cross compiling for Android documentation](https://cmake.org/cmake/help/latest/manual/cmake-toolchains.7.html#cross-compiling-for-android). + +### Cross compiling for iOS + +```shell +cmake .. -G "Unix Makefiles" -DCMAKE_SYSTEM_NAME=iOS -DCMAKE_OSX_ARCHITECTURES=arm64 +``` + +For more information about cross compilation for iOS, you can refer to the [Cross compiling for iOS documentation](https://cmake.org/cmake/help/latest/manual/cmake-toolchains.7.html#cross-compiling-for-ios-tvos-or-watchos). + + +### Windows Visual Studio + +```shell +cmake .. -G "Visual Studio 17 2022" -A x64 +``` + +For more information about the Visual Studio generator options and additional customization, you can refer to the [Visual Studio Generator documentation](https://cmake.org/cmake/help/latest/generator/Visual%20Studio%2017%202022.html). diff --git a/lib/opus/cmake/RunTest.cmake b/lib/opus/cmake/RunTest.cmake new file mode 100644 index 00000000..f6f8b4a2 --- /dev/null +++ b/lib/opus/cmake/RunTest.cmake @@ -0,0 +1,61 @@ +if(NOT EXISTS ${TEST_EXECUTABLE}) + message(FATAL_ERROR "Error could not find ${TEST_EXECUTABLE}, ensure that you built the test binary") +endif() + +if(CMAKE_SYSTEM_NAME STREQUAL "Android") + + # support to run plain old binary on android devices + # requires android debug bridge to be installed + + find_program(adb_executable adb) + if(NOT adb_executable) + message(FATAL_ERROR "Error could not find adb") + endif() + + # check if any device emulator is attached + execute_process(COMMAND ${adb_executable} shell echo RESULT_VARIABLE CMD_RESULT) + if(CMD_RESULT) + message(FATAL_ERROR "Error adb: no devices/emulators found") + endif() + + # push binary + set(android_path /data/local/tmp) + execute_process(COMMAND ${adb_executable} push ${TEST_EXECUTABLE} ${android_path} RESULT_VARIABLE CMD_RESULT) + if(CMD_RESULT) + message(FATAL_ERROR "Error running ${adb_executable} push ${TEST_EXECUTABLE} ${android_path} failed with result ${CMD_RESULT}") + endif() + + # set permissions + get_filename_component(test_executable ${TEST_EXECUTABLE} NAME) + set(test_executable_on_android /data/local/tmp/${test_executable}) + execute_process(COMMAND ${adb_executable} shell chmod 555 ${test_executable_on_android} RESULT_VARIABLE CMD_RESULT) + if(CMD_RESULT) + message(FATAL_ERROR "Error running ${adb_executable} shell chmod 555 ${test_executable_on_android} failed with result ${CMD_RESULT}") + endif() + + # run executable + execute_process(COMMAND ${adb_executable} shell ${test_executable_on_android} RESULT_VARIABLE CMD_RESULT) + if(CMD_RESULT) + message(FATAL_ERROR "Error running ${adb_executable} shell ${test_executable_on_android} failed with result ${CMD_RESULT}") + endif() + + # clean up binary + execute_process(COMMAND ${adb_executable} shell rm ${test_executable_on_android} RESULT_VARIABLE CMD_RESULT) + if(CMD_RESULT) + message(FATAL_ERROR "Error running ${adb_executable} shell rm ${test_executable_on_android} failed with result ${CMD_RESULT}") + endif() + +elseif(CMAKE_SYSTEM_NAME STREQUAL "iOS") + # CTest doesn't support iOS + + message(FATAL_ERROR "Error CTest is not supported on iOS") + +else() + # for other platforms just execute test binary on host + + execute_process(COMMAND ${TEST_EXECUTABLE} RESULT_VARIABLE CMD_RESULT) + if(CMD_RESULT) + message(FATAL_ERROR "Error running ${TEST_EXECUTABLE} failed with result ${CMD_RESULT}") + endif() + +endif() \ No newline at end of file diff --git a/lib/opus/cmake/config.h.cmake.in b/lib/opus/cmake/config.h.cmake.in new file mode 100644 index 00000000..5550842c --- /dev/null +++ b/lib/opus/cmake/config.h.cmake.in @@ -0,0 +1 @@ +#cmakedefine PACKAGE_VERSION "@PACKAGE_VERSION@" \ No newline at end of file diff --git a/lib/opus/cmake/cpu_info_by_asm.c b/lib/opus/cmake/cpu_info_by_asm.c new file mode 100644 index 00000000..1a70a815 --- /dev/null +++ b/lib/opus/cmake/cpu_info_by_asm.c @@ -0,0 +1,31 @@ +#include +int main() { + unsigned int CPUInfo0; + unsigned int CPUInfo1; + unsigned int CPUInfo2; + unsigned int CPUInfo3; + unsigned int InfoType; +#if defined(__i386__) && defined(__PIC__) +/* %ebx is PIC register in 32-bit, so mustn't clobber it. */ + __asm__ __volatile__ ( + "xchg %%ebx, %1\n" + "cpuid\n" + "xchg %%ebx, %1\n": + "=a" (CPUInfo0), + "=r" (CPUInfo1), + "=c" (CPUInfo2), + "=d" (CPUInfo3) : + "0" (InfoType), "2" (0) + ); +#else + __asm__ __volatile__ ( + "cpuid": + "=a" (CPUInfo0), + "=b" (CPUInfo1), + "=c" (CPUInfo2), + "=d" (CPUInfo3) : + "0" (InfoType), "2" (0) + ); +#endif + return 0; +} diff --git a/lib/opus/cmake/cpu_info_by_c.c b/lib/opus/cmake/cpu_info_by_c.c new file mode 100644 index 00000000..117084eb --- /dev/null +++ b/lib/opus/cmake/cpu_info_by_c.c @@ -0,0 +1,9 @@ +#include +int main() { + unsigned int CPUInfo0; + unsigned int CPUInfo1; + unsigned int CPUInfo2; + unsigned int CPUInfo3; + unsigned int InfoType; + return __get_cpuid_count(InfoType, 0, &CPUInfo0, &CPUInfo1, &CPUInfo2, &CPUInfo3); +} diff --git a/lib/opus/cmake/vla.c b/lib/opus/cmake/vla.c new file mode 100644 index 00000000..05b21197 --- /dev/null +++ b/lib/opus/cmake/vla.c @@ -0,0 +1,7 @@ +int main() { + static int x; + char a[++x]; + a[sizeof a - 1] = 0; + int N; + return a[0]; +} \ No newline at end of file diff --git a/lib/opus/configure.ac b/lib/opus/configure.ac new file mode 100644 index 00000000..da2a260f --- /dev/null +++ b/lib/opus/configure.ac @@ -0,0 +1,951 @@ +dnl Process this file with autoconf to produce a configure script. -*-m4-*- + +dnl The package_version file will be automatically synced to the git revision +dnl by the update_version script when configured in the repository, but will +dnl remain constant in tarball releases unless it is manually edited. +m4_define([CURRENT_VERSION], + m4_esyscmd([ ./update_version 2>/dev/null || true + if test -e package_version; then + . ./package_version + printf "$PACKAGE_VERSION" + else + printf "unknown" + fi ])) + +AC_INIT([opus],[CURRENT_VERSION],[opus@xiph.org]) + +AC_CONFIG_SRCDIR(src/opus_encoder.c) +AC_CONFIG_MACRO_DIR([m4]) + +dnl enable silent rules on automake 1.11 and later +m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])]) + +# For libtool. +dnl Please update these for releases. +OPUS_LT_CURRENT=9 +OPUS_LT_REVISION=0 +OPUS_LT_AGE=9 + +AC_SUBST(OPUS_LT_CURRENT) +AC_SUBST(OPUS_LT_REVISION) +AC_SUBST(OPUS_LT_AGE) + +AM_INIT_AUTOMAKE([no-define]) +AM_MAINTAINER_MODE([enable]) + +AC_CANONICAL_HOST +AC_MINGW32 +AM_PROG_LIBTOOL +AM_PROG_CC_C_O + +AC_PROG_CC_C99 +AC_C_CONST +AC_C_INLINE + +AM_PROG_AS + +AC_DEFINE([OPUS_BUILD], [], [This is a build of OPUS]) + +#Use a hacked up version of autoconf's AC_C_RESTRICT because it's not +#strong enough a test to detect old buggy versions of GCC (e.g. 2.95.3) +#Note: Both this and the test for variable-size arrays below are also +# done by AC_PROG_CC_C99, but not thoroughly enough apparently. +AC_CACHE_CHECK([for C/C++ restrict keyword], ac_cv_c_restrict, + [ac_cv_c_restrict=no + # The order here caters to the fact that C++ does not require restrict. + for ac_kw in __restrict __restrict__ _Restrict restrict; do + AC_COMPILE_IFELSE([AC_LANG_PROGRAM( + [[typedef int * int_ptr; + int foo (int_ptr $ac_kw ip, int * $ac_kw baz[]) { + return ip[0]; + }]], + [[int s[1]; + int * $ac_kw t = s; + t[0] = 0; + return foo(t, (void *)0)]])], + [ac_cv_c_restrict=$ac_kw]) + test "$ac_cv_c_restrict" != no && break + done + ]) + +AH_VERBATIM([restrict], +[/* Define to the equivalent of the C99 'restrict' keyword, or to + nothing if this is not supported. Do not define if restrict is + supported directly. */ +#undef restrict +/* Work around a bug in Sun C++: it does not support _Restrict or + __restrict__, even though the corresponding Sun C compiler ends up with + "#define restrict _Restrict" or "#define restrict __restrict__" in the + previous line. Perhaps some future version of Sun C++ will work with + restrict; if so, hopefully it defines __RESTRICT like Sun C does. */ +#if defined __SUNPRO_CC && !defined __RESTRICT +# define _Restrict +# define __restrict__ +#endif]) + +case $ac_cv_c_restrict in + restrict) ;; + no) AC_DEFINE([restrict], []) ;; + *) AC_DEFINE_UNQUOTED([restrict], [$ac_cv_c_restrict]) ;; +esac + +AC_MSG_CHECKING(for C99 variable-size arrays) +AC_COMPILE_IFELSE([AC_LANG_PROGRAM([], + [[static int x; char a[++x]; a[sizeof a - 1] = 0; int N; return a[0];]])], + [ has_var_arrays=yes + use_alloca="no (using var arrays)" + AC_DEFINE([VAR_ARRAYS], [1], [Use C99 variable-size arrays]) + ],[ + has_var_arrays=no + ]) +AC_MSG_RESULT([$has_var_arrays]) + +AS_IF([test "$has_var_arrays" = "no"], + [ + AC_CHECK_HEADERS([alloca.h]) + AC_MSG_CHECKING(for alloca) + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include ]], + [[int foo=10; int *array = alloca(foo);]])], + [ use_alloca=yes; + AC_DEFINE([USE_ALLOCA], [], [Make use of alloca]) + ],[ + use_alloca=no + ]) + AC_MSG_RESULT([$use_alloca]) + ]) + +LT_LIB_M + +AC_ARG_ENABLE([fixed-point], + [AS_HELP_STRING([--enable-fixed-point], + [compile without floating point (for machines without a fast enough FPU)])],, + [enable_fixed_point=no]) + +AS_IF([test "$enable_fixed_point" = "yes"],[ + enable_float="no" + AC_DEFINE([FIXED_POINT], [1], [Compile as fixed-point (for machines without a fast enough FPU)]) + PC_BUILD="fixed-point" +],[ + enable_float="yes"; + PC_BUILD="floating-point" +]) + +AM_CONDITIONAL([FIXED_POINT], [test "$enable_fixed_point" = "yes"]) + +AC_ARG_ENABLE([fixed-point-debug], + [AS_HELP_STRING([--enable-fixed-point-debug], [debug fixed-point implementation])],, + [enable_fixed_point_debug=no]) + +AS_IF([test "$enable_fixed_point_debug" = "yes"],[ + AC_DEFINE([FIXED_DEBUG], [1], [Debug fixed-point implementation]) +]) + +AC_ARG_ENABLE([float_api], + [AS_HELP_STRING([--disable-float-api], + [compile without the floating point API (for machines with no float library)])],, + [enable_float_api=yes]) + +AM_CONDITIONAL([DISABLE_FLOAT_API], [test "$enable_float_api" = "no"]) + +AS_IF([test "$enable_float_api" = "no"],[ + AC_DEFINE([DISABLE_FLOAT_API], [1], [Do not build the float API]) +]) + +AC_ARG_ENABLE([custom-modes], + [AS_HELP_STRING([--enable-custom-modes], [enable non-Opus modes, e.g. 44.1 kHz & 2^n frames])],, + [enable_custom_modes=no]) + +AS_IF([test "$enable_custom_modes" = "yes"],[ + AC_DEFINE([CUSTOM_MODES], [1], [Custom modes]) + PC_BUILD="$PC_BUILD, custom modes" +]) + +AM_CONDITIONAL([CUSTOM_MODES], [test "$enable_custom_modes" = "yes"]) + +has_float_approx=no +#case "$host_cpu" in +#i[[3456]]86 | x86_64 | powerpc64 | powerpc32 | ia64) +# has_float_approx=yes +# ;; +#esac + +AC_ARG_ENABLE([float-approx], + [AS_HELP_STRING([--enable-float-approx], [enable fast approximations for floating point])], + [if test "$enable_float_approx" = "yes"; then + AC_WARN([Floating point approximations are not supported on all platforms.]) + fi + ], + [enable_float_approx=$has_float_approx]) + +AS_IF([test "$enable_float_approx" = "yes"],[ + AC_DEFINE([FLOAT_APPROX], [1], [Float approximations]) +]) + +AC_ARG_ENABLE([asm], + [AS_HELP_STRING([--disable-asm], [Disable assembly optimizations])],, + [enable_asm=yes]) + +AC_ARG_ENABLE([rtcd], + [AS_HELP_STRING([--disable-rtcd], [Disable run-time CPU capabilities detection])],, + [enable_rtcd=yes]) + +AC_ARG_ENABLE([intrinsics], + [AS_HELP_STRING([--disable-intrinsics], [Disable intrinsics optimizations])],, + [enable_intrinsics=yes]) + +rtcd_support=no +cpu_arm=no +cpu_x86=no + +AS_IF([test x"${enable_asm}" = x"yes"],[ + inline_optimization="No inline ASM for your platform, please send patches" + case $host_cpu in + arm*) + dnl Currently we only have asm for fixed-point + AS_IF([test "$enable_float" != "yes"],[ + cpu_arm=yes + AC_DEFINE([OPUS_ARM_ASM], [], [Make use of ARM asm optimization]) + AS_GCC_INLINE_ASSEMBLY( + [inline_optimization="ARM"], + [inline_optimization="disabled"] + ) + AS_ASM_ARM_EDSP([OPUS_ARM_INLINE_EDSP=1],[OPUS_ARM_INLINE_EDSP=0]) + AS_ASM_ARM_MEDIA([OPUS_ARM_INLINE_MEDIA=1], + [OPUS_ARM_INLINE_MEDIA=0]) + AS_ASM_ARM_NEON([OPUS_ARM_INLINE_NEON=1],[OPUS_ARM_INLINE_NEON=0]) + AS_IF([test x"$inline_optimization" = x"ARM"],[ + AM_CONDITIONAL([OPUS_ARM_INLINE_ASM],[true]) + AC_DEFINE([OPUS_ARM_INLINE_ASM], 1, + [Use generic ARMv4 inline asm optimizations]) + AS_IF([test x"$OPUS_ARM_INLINE_EDSP" = x"1"],[ + AC_DEFINE([OPUS_ARM_INLINE_EDSP], [1], + [Use ARMv5E inline asm optimizations]) + inline_optimization="$inline_optimization (EDSP)" + ]) + AS_IF([test x"$OPUS_ARM_INLINE_MEDIA" = x"1"],[ + AC_DEFINE([OPUS_ARM_INLINE_MEDIA], [1], + [Use ARMv6 inline asm optimizations]) + inline_optimization="$inline_optimization (Media)" + ]) + AS_IF([test x"$OPUS_ARM_INLINE_NEON" = x"1"],[ + AC_DEFINE([OPUS_ARM_INLINE_NEON], 1, + [Use ARM NEON inline asm optimizations]) + inline_optimization="$inline_optimization (NEON)" + ]) + ]) + dnl We need Perl to translate RVCT-syntax asm to gas syntax. + AC_CHECK_PROG([HAVE_PERL], perl, yes, no) + AS_IF([test x"$HAVE_PERL" = x"yes"],[ + AM_CONDITIONAL([OPUS_ARM_EXTERNAL_ASM],[true]) + asm_optimization="ARM" + AS_IF([test x"$OPUS_ARM_INLINE_EDSP" = x"1"], [ + OPUS_ARM_PRESUME_EDSP=1 + OPUS_ARM_MAY_HAVE_EDSP=1 + ], + [ + OPUS_ARM_PRESUME_EDSP=0 + OPUS_ARM_MAY_HAVE_EDSP=0 + ]) + AS_IF([test x"$OPUS_ARM_INLINE_MEDIA" = x"1"], [ + OPUS_ARM_PRESUME_MEDIA=1 + OPUS_ARM_MAY_HAVE_MEDIA=1 + ], + [ + OPUS_ARM_PRESUME_MEDIA=0 + OPUS_ARM_MAY_HAVE_MEDIA=0 + ]) + AS_IF([test x"$OPUS_ARM_INLINE_NEON" = x"1"], [ + OPUS_ARM_PRESUME_NEON=1 + OPUS_ARM_MAY_HAVE_NEON=1 + ], + [ + OPUS_ARM_PRESUME_NEON=0 + OPUS_ARM_MAY_HAVE_NEON=0 + ]) + AS_IF([test x"$enable_rtcd" = x"yes"],[ + AS_IF([test x"$OPUS_ARM_MAY_HAVE_EDSP" != x"1"],[ + AC_MSG_NOTICE( + [Trying to force-enable armv5e EDSP instructions...]) + AS_ASM_ARM_EDSP_FORCE([OPUS_ARM_MAY_HAVE_EDSP=1]) + ]) + AS_IF([test x"$OPUS_ARM_MAY_HAVE_MEDIA" != x"1"],[ + AC_MSG_NOTICE( + [Trying to force-enable ARMv6 media instructions...]) + AS_ASM_ARM_MEDIA_FORCE([OPUS_ARM_MAY_HAVE_MEDIA=1]) + ]) + AS_IF([test x"$OPUS_ARM_MAY_HAVE_NEON" != x"1"],[ + AC_MSG_NOTICE( + [Trying to force-enable NEON instructions...]) + AS_ASM_ARM_NEON_FORCE([OPUS_ARM_MAY_HAVE_NEON=1]) + ]) + ]) + rtcd_support= + AS_IF([test x"$OPUS_ARM_MAY_HAVE_EDSP" = x"1"],[ + AC_DEFINE(OPUS_ARM_MAY_HAVE_EDSP, 1, + [Define if assembler supports EDSP instructions]) + AS_IF([test x"$OPUS_ARM_PRESUME_EDSP" = x"1"],[ + AC_DEFINE(OPUS_ARM_PRESUME_EDSP, 1, + [Define if binary requires EDSP instruction support]) + asm_optimization="$asm_optimization (EDSP)" + ], + [rtcd_support="$rtcd_support (EDSP)"] + ) + ]) + AC_SUBST(OPUS_ARM_MAY_HAVE_EDSP) + AS_IF([test x"$OPUS_ARM_MAY_HAVE_MEDIA" = x"1"],[ + AC_DEFINE(OPUS_ARM_MAY_HAVE_MEDIA, 1, + [Define if assembler supports ARMv6 media instructions]) + AS_IF([test x"$OPUS_ARM_PRESUME_MEDIA" = x"1"],[ + AC_DEFINE(OPUS_ARM_PRESUME_MEDIA, 1, + [Define if binary requires ARMv6 media instruction support]) + asm_optimization="$asm_optimization (Media)" + ], + [rtcd_support="$rtcd_support (Media)"] + ) + ]) + AC_SUBST(OPUS_ARM_MAY_HAVE_MEDIA) + AS_IF([test x"$OPUS_ARM_MAY_HAVE_NEON" = x"1"],[ + AC_DEFINE(OPUS_ARM_MAY_HAVE_NEON, 1, + [Define if compiler supports NEON instructions]) + AS_IF([test x"$OPUS_ARM_PRESUME_NEON" = x"1"], [ + AC_DEFINE(OPUS_ARM_PRESUME_NEON, 1, + [Define if binary requires NEON instruction support]) + asm_optimization="$asm_optimization (NEON)" + ], + [rtcd_support="$rtcd_support (NEON)"] + ) + ]) + AC_SUBST(OPUS_ARM_MAY_HAVE_NEON) + dnl Make sure turning on RTCD gets us at least one + dnl instruction set. + AS_IF([test x"$rtcd_support" != x""], + [rtcd_support=ARM"$rtcd_support"], + [rtcd_support="no"] + ) + AC_MSG_CHECKING([for apple style tools]) + AC_PREPROC_IFELSE([AC_LANG_PROGRAM([ +#ifndef __APPLE__ +#error 1 +#endif],[])], + [AC_MSG_RESULT([yes]); ARM2GNU_PARAMS="--apple"], + [AC_MSG_RESULT([no]); ARM2GNU_PARAMS=""]) + AC_SUBST(ARM2GNU_PARAMS) + ], + [ + AC_MSG_WARN( + [*** ARM assembly requires perl -- disabling optimizations]) + asm_optimization="(missing perl dependency for ARM)" + ]) + ]) + ;; + esac +],[ + inline_optimization="disabled" + asm_optimization="disabled" +]) + +AM_CONDITIONAL([OPUS_ARM_INLINE_ASM], + [test x"${inline_optimization%% *}" = x"ARM"]) +AM_CONDITIONAL([OPUS_ARM_EXTERNAL_ASM], + [test x"${asm_optimization%% *}" = x"ARM"]) + +AM_CONDITIONAL([HAVE_SSE], [false]) +AM_CONDITIONAL([HAVE_SSE2], [false]) +AM_CONDITIONAL([HAVE_SSE4_1], [false]) +AM_CONDITIONAL([HAVE_AVX], [false]) + +m4_define([DEFAULT_X86_SSE_CFLAGS], [-msse]) +m4_define([DEFAULT_X86_SSE2_CFLAGS], [-msse2]) +m4_define([DEFAULT_X86_SSE4_1_CFLAGS], [-msse4.1]) +m4_define([DEFAULT_X86_AVX_CFLAGS], [-mavx]) +m4_define([DEFAULT_ARM_NEON_INTR_CFLAGS], [-mfpu=neon]) +# With GCC on ARM32 softfp architectures (e.g. Android, or older Ubuntu) you need to specify +# -mfloat-abi=softfp for -mfpu=neon to work. However, on ARM32 hardfp architectures (e.g. newer Ubuntu), +# this option will break things. + +# As a heuristic, if host matches arm*eabi* but not arm*hf*, it's probably soft-float. +m4_define([DEFAULT_ARM_NEON_SOFTFP_INTR_CFLAGS], [-mfpu=neon -mfloat-abi=softfp]) + +AS_CASE([$host], + [arm*hf*], [AS_VAR_SET([RESOLVED_DEFAULT_ARM_NEON_INTR_CFLAGS], "DEFAULT_ARM_NEON_INTR_CFLAGS")], + [arm*eabi*], [AS_VAR_SET([RESOLVED_DEFAULT_ARM_NEON_INTR_CFLAGS], "DEFAULT_ARM_NEON_SOFTFP_INTR_CFLAGS")], + [AS_VAR_SET([RESOLVED_DEFAULT_ARM_NEON_INTR_CFLAGS], "DEFAULT_ARM_NEON_INTR_CFLAGS")]) + +AC_ARG_VAR([X86_SSE_CFLAGS], [C compiler flags to compile SSE intrinsics @<:@default=]DEFAULT_X86_SSE_CFLAGS[@:>@]) +AC_ARG_VAR([X86_SSE2_CFLAGS], [C compiler flags to compile SSE2 intrinsics @<:@default=]DEFAULT_X86_SSE2_CFLAGS[@:>@]) +AC_ARG_VAR([X86_SSE4_1_CFLAGS], [C compiler flags to compile SSE4.1 intrinsics @<:@default=]DEFAULT_X86_SSE4_1_CFLAGS[@:>@]) +AC_ARG_VAR([X86_AVX_CFLAGS], [C compiler flags to compile AVX intrinsics @<:@default=]DEFAULT_X86_AVX_CFLAGS[@:>@]) +AC_ARG_VAR([ARM_NEON_INTR_CFLAGS], [C compiler flags to compile ARM NEON intrinsics @<:@default=]DEFAULT_ARM_NEON_INTR_CFLAGS / DEFAULT_ARM_NEON_SOFTFP_INTR_CFLAGS[@:>@]) + +AS_VAR_SET_IF([X86_SSE_CFLAGS], [], [AS_VAR_SET([X86_SSE_CFLAGS], "DEFAULT_X86_SSE_CFLAGS")]) +AS_VAR_SET_IF([X86_SSE2_CFLAGS], [], [AS_VAR_SET([X86_SSE2_CFLAGS], "DEFAULT_X86_SSE2_CFLAGS")]) +AS_VAR_SET_IF([X86_SSE4_1_CFLAGS], [], [AS_VAR_SET([X86_SSE4_1_CFLAGS], "DEFAULT_X86_SSE4_1_CFLAGS")]) +AS_VAR_SET_IF([X86_AVX_CFLAGS], [], [AS_VAR_SET([X86_AVX_CFLAGS], "DEFAULT_X86_AVX_CFLAGS")]) +AS_VAR_SET_IF([ARM_NEON_INTR_CFLAGS], [], [AS_VAR_SET([ARM_NEON_INTR_CFLAGS], ["$RESOLVED_DEFAULT_ARM_NEON_INTR_CFLAGS"])]) + +AC_DEFUN([OPUS_PATH_NE10], + [ + AC_ARG_WITH(NE10, + AC_HELP_STRING([--with-NE10=PFX],[Prefix where libNE10 is installed (optional)]), + NE10_prefix="$withval", NE10_prefix="") + AC_ARG_WITH(NE10-libraries, + AC_HELP_STRING([--with-NE10-libraries=DIR], + [Directory where libNE10 library is installed (optional)]), + NE10_libraries="$withval", NE10_libraries="") + AC_ARG_WITH(NE10-includes, + AC_HELP_STRING([--with-NE10-includes=DIR], + [Directory where libNE10 header files are installed (optional)]), + NE10_includes="$withval", NE10_includes="") + + if test "x$NE10_libraries" != "x" ; then + NE10_LIBS="-L$NE10_libraries" + elif test "x$NE10_prefix" = "xno" || test "x$NE10_prefix" = "xyes" ; then + NE10_LIBS="" + elif test "x$NE10_prefix" != "x" ; then + NE10_LIBS="-L$NE10_prefix/lib" + elif test "x$prefix" != "xNONE" ; then + NE10_LIBS="-L$prefix/lib" + fi + + if test "x$NE10_prefix" != "xno" ; then + NE10_LIBS="$NE10_LIBS -lNE10" + fi + + if test "x$NE10_includes" != "x" ; then + NE10_CFLAGS="-I$NE10_includes" + elif test "x$NE10_prefix" = "xno" || test "x$NE10_prefix" = "xyes" ; then + NE10_CFLAGS="" + elif test "x$NE10_prefix" != "x" ; then + NE10_CFLAGS="-I$NE10_prefix/include" + elif test "x$prefix" != "xNONE"; then + NE10_CFLAGS="-I$prefix/include" + fi + + AC_MSG_CHECKING(for NE10) + save_CFLAGS="$CFLAGS"; CFLAGS="$CFLAGS $NE10_CFLAGS" + save_LIBS="$LIBS"; LIBS="$LIBS $NE10_LIBS $LIBM" + AC_LINK_IFELSE( + [ + AC_LANG_PROGRAM( + [[#include + ]], + [[ + ne10_fft_cfg_float32_t cfg; + cfg = ne10_fft_alloc_c2c_float32_neon(480); + ]] + ) + ],[ + HAVE_ARM_NE10=1 + AC_MSG_RESULT([yes]) + ],[ + HAVE_ARM_NE10=0 + AC_MSG_RESULT([no]) + NE10_CFLAGS="" + NE10_LIBS="" + ] + ) + CFLAGS="$save_CFLAGS"; LIBS="$save_LIBS" + #Now we know if libNE10 is installed or not + AS_IF([test x"$HAVE_ARM_NE10" = x"1"], + [ + AC_DEFINE([HAVE_ARM_NE10], 1, [NE10 library is installed on host. Make sure it is on target!]) + AC_SUBST(HAVE_ARM_NE10) + AC_SUBST(NE10_CFLAGS) + AC_SUBST(NE10_LIBS) + ] + ) + ] +) + +AS_IF([test x"$enable_intrinsics" = x"yes"],[ + intrinsics_support="" + AS_CASE([$host_cpu], + [arm*|aarch64*], + [ + cpu_arm=yes + OPUS_CHECK_INTRINSICS( + [ARM Neon], + [$ARM_NEON_INTR_CFLAGS], + [OPUS_ARM_MAY_HAVE_NEON_INTR], + [OPUS_ARM_PRESUME_NEON_INTR], + [[#include + ]], + [[ + static float32x4_t A0, A1, SUMM; + SUMM = vmlaq_f32(SUMM, A0, A1); + return (int)vgetq_lane_f32(SUMM, 0); + ]] + ) + AS_IF([test x"$OPUS_ARM_MAY_HAVE_NEON_INTR" = x"1" && test x"$OPUS_ARM_PRESUME_NEON_INTR" != x"1"], + [ + OPUS_ARM_NEON_INTR_CFLAGS="$ARM_NEON_INTR_CFLAGS" + AC_SUBST([OPUS_ARM_NEON_INTR_CFLAGS]) + ] + ) + + AS_IF([test x"$OPUS_ARM_MAY_HAVE_NEON_INTR" = x"1"], + [ + AC_DEFINE([OPUS_ARM_MAY_HAVE_NEON_INTR], 1, [Compiler supports ARMv7/Aarch64 Neon Intrinsics]) + intrinsics_support="$intrinsics_support (NEON)" + + AS_IF([test x"$enable_rtcd" != x"no" && test x"$OPUS_ARM_PRESUME_NEON_INTR" != x"1"], + [AS_IF([test x"$rtcd_support" = x"no"], + [rtcd_support="ARM (NEON Intrinsics)"], + [rtcd_support="$rtcd_support (NEON Intrinsics)"])]) + + AS_IF([test x"$OPUS_ARM_PRESUME_NEON_INTR" = x"1"], + [AC_DEFINE([OPUS_ARM_PRESUME_NEON_INTR], 1, [Define if binary requires NEON intrinsics support])]) + + OPUS_PATH_NE10() + AS_IF([test x"$NE10_LIBS" != x""], + [ + intrinsics_support="$intrinsics_support (NE10)" + AS_IF([test x"enable_rtcd" != x"" \ + && test x"$OPUS_ARM_PRESUME_NEON_INTR" != x"1"], + [rtcd_support="$rtcd_support (NE10)"]) + ]) + + OPUS_CHECK_INTRINSICS( + [Aarch64 Neon], + [$ARM_NEON_INTR_CFLAGS], + [OPUS_ARM_MAY_HAVE_AARCH64_NEON_INTR], + [OPUS_ARM_PRESUME_AARCH64_NEON_INTR], + [[#include + ]], + [[ + static int32_t IN; + static int16_t OUT; + OUT = vqmovns_s32(IN); + ]] + ) + + AS_IF([test x"$OPUS_ARM_PRESUME_AARCH64_NEON_INTR" = x"1"], + [ + AC_DEFINE([OPUS_ARM_PRESUME_AARCH64_NEON_INTR], 1, [Define if binary requires Aarch64 Neon Intrinsics]) + intrinsics_support="$intrinsics_support (NEON [Aarch64])" + ]) + + AS_IF([test x"$intrinsics_support" = x""], + [intrinsics_support=no], + [intrinsics_support="ARM$intrinsics_support"]) + ], + [ + AC_MSG_WARN([Compiler does not support ARM intrinsics]) + intrinsics_support=no + ]) + ], + [i?86|x86_64], + [ + cpu_x86=yes + OPUS_CHECK_INTRINSICS( + [SSE], + [$X86_SSE_CFLAGS], + [OPUS_X86_MAY_HAVE_SSE], + [OPUS_X86_PRESUME_SSE], + [[#include + #include + ]], + [[ + __m128 mtest; + mtest = _mm_set1_ps((float)time(NULL)); + mtest = _mm_mul_ps(mtest, mtest); + return _mm_cvtss_si32(mtest); + ]] + ) + AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE" = x"1" && test x"$OPUS_X86_PRESUME_SSE" != x"1"], + [ + OPUS_X86_SSE_CFLAGS="$X86_SSE_CFLAGS" + AC_SUBST([OPUS_X86_SSE_CFLAGS]) + ] + ) + OPUS_CHECK_INTRINSICS( + [SSE2], + [$X86_SSE2_CFLAGS], + [OPUS_X86_MAY_HAVE_SSE2], + [OPUS_X86_PRESUME_SSE2], + [[#include + #include + ]], + [[ + __m128i mtest; + mtest = _mm_set1_epi32((int)time(NULL)); + mtest = _mm_mul_epu32(mtest, mtest); + return _mm_cvtsi128_si32(mtest); + ]] + ) + AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE2" = x"1" && test x"$OPUS_X86_PRESUME_SSE2" != x"1"], + [ + OPUS_X86_SSE2_CFLAGS="$X86_SSE2_CFLAGS" + AC_SUBST([OPUS_X86_SSE2_CFLAGS]) + ] + ) + OPUS_CHECK_INTRINSICS( + [SSE4.1], + [$X86_SSE4_1_CFLAGS], + [OPUS_X86_MAY_HAVE_SSE4_1], + [OPUS_X86_PRESUME_SSE4_1], + [[#include + #include + ]], + [[ + __m128i mtest; + mtest = _mm_set1_epi32((int)time(NULL)); + mtest = _mm_mul_epi32(mtest, mtest); + return _mm_cvtsi128_si32(mtest); + ]] + ) + AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE4_1" = x"1" && test x"$OPUS_X86_PRESUME_SSE4_1" != x"1"], + [ + OPUS_X86_SSE4_1_CFLAGS="$X86_SSE4_1_CFLAGS" + AC_SUBST([OPUS_X86_SSE4_1_CFLAGS]) + ] + ) + OPUS_CHECK_INTRINSICS( + [AVX], + [$X86_AVX_CFLAGS], + [OPUS_X86_MAY_HAVE_AVX], + [OPUS_X86_PRESUME_AVX], + [[#include + #include + ]], + [[ + __m256 mtest; + mtest = _mm256_set1_ps((float)time(NULL)); + mtest = _mm256_addsub_ps(mtest, mtest); + return _mm_cvtss_si32(_mm256_extractf128_ps(mtest, 0)); + ]] + ) + AS_IF([test x"$OPUS_X86_MAY_HAVE_AVX" = x"1" && test x"$OPUS_X86_PRESUME_AVX" != x"1"], + [ + OPUS_X86_AVX_CFLAGS="$X86_AVX_CFLAGS" + AC_SUBST([OPUS_X86_AVX_CFLAGS]) + ] + ) + AS_IF([test x"$rtcd_support" = x"no"], [rtcd_support=""]) + AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE" = x"1"], + [ + AC_DEFINE([OPUS_X86_MAY_HAVE_SSE], 1, [Compiler supports X86 SSE Intrinsics]) + intrinsics_support="$intrinsics_support SSE" + + AS_IF([test x"$OPUS_X86_PRESUME_SSE" = x"1"], + [AC_DEFINE([OPUS_X86_PRESUME_SSE], 1, [Define if binary requires SSE intrinsics support])], + [rtcd_support="$rtcd_support SSE"]) + ], + [ + AC_MSG_WARN([Compiler does not support SSE intrinsics]) + ]) + + AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE2" = x"1"], + [ + AC_DEFINE([OPUS_X86_MAY_HAVE_SSE2], 1, [Compiler supports X86 SSE2 Intrinsics]) + intrinsics_support="$intrinsics_support SSE2" + + AS_IF([test x"$OPUS_X86_PRESUME_SSE2" = x"1"], + [AC_DEFINE([OPUS_X86_PRESUME_SSE2], 1, [Define if binary requires SSE2 intrinsics support])], + [rtcd_support="$rtcd_support SSE2"]) + ], + [ + AC_MSG_WARN([Compiler does not support SSE2 intrinsics]) + ]) + + AS_IF([test x"$OPUS_X86_MAY_HAVE_SSE4_1" = x"1"], + [ + AC_DEFINE([OPUS_X86_MAY_HAVE_SSE4_1], 1, [Compiler supports X86 SSE4.1 Intrinsics]) + intrinsics_support="$intrinsics_support SSE4.1" + + AS_IF([test x"$OPUS_X86_PRESUME_SSE4_1" = x"1"], + [AC_DEFINE([OPUS_X86_PRESUME_SSE4_1], 1, [Define if binary requires SSE4.1 intrinsics support])], + [rtcd_support="$rtcd_support SSE4.1"]) + ], + [ + AC_MSG_WARN([Compiler does not support SSE4.1 intrinsics]) + ]) + AS_IF([test x"$OPUS_X86_MAY_HAVE_AVX" = x"1"], + [ + AC_DEFINE([OPUS_X86_MAY_HAVE_AVX], 1, [Compiler supports X86 AVX Intrinsics]) + intrinsics_support="$intrinsics_support AVX" + + AS_IF([test x"$OPUS_X86_PRESUME_AVX" = x"1"], + [AC_DEFINE([OPUS_X86_PRESUME_AVX], 1, [Define if binary requires AVX intrinsics support])], + [rtcd_support="$rtcd_support AVX"]) + ], + [ + AC_MSG_WARN([Compiler does not support AVX intrinsics]) + ]) + + AS_IF([test x"$intrinsics_support" = x""], + [intrinsics_support=no], + [intrinsics_support="x86$intrinsics_support"] + ) + AS_IF([test x"$rtcd_support" = x""], + [rtcd_support=no], + [rtcd_support="x86$rtcd_support"], + ) + + AS_IF([test x"$enable_rtcd" = x"yes" && test x"$rtcd_support" != x""],[ + get_cpuid_by_asm="no" + AC_MSG_CHECKING([How to get X86 CPU Info]) + AC_LINK_IFELSE([AC_LANG_PROGRAM([[ + #include + ]],[[ + unsigned int CPUInfo0; + unsigned int CPUInfo1; + unsigned int CPUInfo2; + unsigned int CPUInfo3; + unsigned int InfoType; + #if defined(__i386__) && defined(__PIC__) + __asm__ __volatile__ ( + "xchg %%ebx, %1\n" + "cpuid\n" + "xchg %%ebx, %1\n": + "=a" (CPUInfo0), + "=r" (CPUInfo1), + "=c" (CPUInfo2), + "=d" (CPUInfo3) : + "a" (InfoType), "c" (0) + ); + #else + __asm__ __volatile__ ( + "cpuid": + "=a" (CPUInfo0), + "=b" (CPUInfo1), + "=c" (CPUInfo2), + "=d" (CPUInfo3) : + "a" (InfoType), "c" (0) + ); + #endif + ]])], + [get_cpuid_by_asm="yes" + AC_MSG_RESULT([Inline Assembly]) + AC_DEFINE([CPU_INFO_BY_ASM], [1], [Get CPU Info by asm method])], + [AC_LINK_IFELSE([AC_LANG_PROGRAM([[ + #include + ]],[[ + unsigned int CPUInfo0; + unsigned int CPUInfo1; + unsigned int CPUInfo2; + unsigned int CPUInfo3; + unsigned int InfoType; + __get_cpuid_count(InfoType, 0, &CPUInfo0, &CPUInfo1, &CPUInfo2, &CPUInfo3); + ]])], + [AC_MSG_RESULT([C method]) + AC_DEFINE([CPU_INFO_BY_C], [1], [Get CPU Info by c method])], + [AC_MSG_ERROR([no supported Get CPU Info method, please disable run-time CPU capabilities detection or intrinsics])])])]) + ], + [ + AC_MSG_WARN([No intrinsics support for your architecture]) + intrinsics_support="no" + ]) +], +[ + intrinsics_support="no" +]) + +AM_CONDITIONAL([CPU_ARM], [test "$cpu_arm" = "yes"]) +AM_CONDITIONAL([HAVE_ARM_NEON_INTR], + [test x"$OPUS_ARM_MAY_HAVE_NEON_INTR" = x"1"]) +AM_CONDITIONAL([HAVE_ARM_NE10], + [test x"$HAVE_ARM_NE10" = x"1"]) +AM_CONDITIONAL([CPU_X86], [test "$cpu_x86" = "yes"]) +AM_CONDITIONAL([HAVE_SSE], + [test x"$OPUS_X86_MAY_HAVE_SSE" = x"1"]) +AM_CONDITIONAL([HAVE_SSE2], + [test x"$OPUS_X86_MAY_HAVE_SSE2" = x"1"]) +AM_CONDITIONAL([HAVE_SSE4_1], + [test x"$OPUS_X86_MAY_HAVE_SSE4_1" = x"1"]) +AM_CONDITIONAL([HAVE_AVX], + [test x"$OPUS_X86_MAY_HAVE_AVX" = x"1"]) + +AM_CONDITIONAL([HAVE_RTCD], + [test x"$enable_rtcd" = x"yes" -a x"$rtcd_support" != x"no"]) +AS_IF([test x"$enable_rtcd" = x"yes"],[ + AS_IF([test x"$rtcd_support" != x"no"],[ + AC_DEFINE([OPUS_HAVE_RTCD], [1], + [Use run-time CPU capabilities detection]) + OPUS_HAVE_RTCD=1 + AC_SUBST(OPUS_HAVE_RTCD) + ]) +],[ + rtcd_support="disabled" +]) + +AC_ARG_ENABLE([assertions], + [AS_HELP_STRING([--enable-assertions],[enable additional software error checking])],, + [enable_assertions=no]) + +AS_IF([test "$enable_assertions" = "yes"], [ + AC_DEFINE([ENABLE_ASSERTIONS], [1], [Assertions]) +]) + +AC_ARG_ENABLE([hardening], + [AS_HELP_STRING([--disable-hardening],[disable run-time checks that are cheap and safe for use in production])],, + [enable_hardening=yes]) + +AS_IF([test "$enable_hardening" = "yes"], [ + AC_DEFINE([ENABLE_HARDENING], [1], [Hardening]) +]) + +AC_ARG_ENABLE([fuzzing], + [AS_HELP_STRING([--enable-fuzzing],[causes the encoder to make random decisions (do not use in production)])],, + [enable_fuzzing=no]) + +AS_IF([test "$enable_fuzzing" = "yes"], [ + AC_DEFINE([FUZZING], [1], [Fuzzing]) +]) + +AC_ARG_ENABLE([check-asm], + [AS_HELP_STRING([--enable-check-asm], + [enable bit-exactness checks between optimized and c implementations])],, + [enable_check_asm=no]) + +AS_IF([test "$enable_check_asm" = "yes"], [ + AC_DEFINE([OPUS_CHECK_ASM], [1], [Run bit-exactness checks between optimized and c implementations]) +]) + +AC_ARG_ENABLE([doc], + [AS_HELP_STRING([--disable-doc], [Do not build API documentation])],, + [enable_doc=yes]) + +AS_IF([test "$enable_doc" = "yes"], [ + AC_CHECK_PROG(HAVE_DOXYGEN, [doxygen], [yes], [no]) + AC_CHECK_PROG(HAVE_DOT, [dot], [yes], [no]) +],[ + HAVE_DOXYGEN=no +]) + +AM_CONDITIONAL([HAVE_DOXYGEN], [test "$HAVE_DOXYGEN" = "yes"]) + +AC_ARG_ENABLE([extra-programs], + [AS_HELP_STRING([--disable-extra-programs], [Do not build extra programs (demo and tests)])],, + [enable_extra_programs=yes]) + +AM_CONDITIONAL([EXTRA_PROGRAMS], [test "$enable_extra_programs" = "yes"]) + + +AC_ARG_ENABLE([rfc8251], + AS_HELP_STRING([--disable-rfc8251], [Disable bitstream fixes from RFC 8251]),, + [enable_rfc8251=yes]) + +AS_IF([test "$enable_rfc8251" = "no"], [ + AC_DEFINE([DISABLE_UPDATE_DRAFT], [1], [Disable bitstream fixes from RFC 8251]) +]) + + +saved_CFLAGS="$CFLAGS" +CFLAGS="$CFLAGS -fvisibility=hidden" +AC_MSG_CHECKING([if ${CC} supports -fvisibility=hidden]) +AC_COMPILE_IFELSE([AC_LANG_SOURCE([[char foo;]])], + [ AC_MSG_RESULT([yes]) ], + [ AC_MSG_RESULT([no]) + CFLAGS="$saved_CFLAGS" + ]) + +on_x86=no +case "$host_cpu" in +i[[3456]]86 | x86_64) + on_x86=yes + ;; +esac + +on_windows=no +case $host in +*cygwin*|*mingw*) + on_windows=yes + ;; +esac + +dnl Enable stack-protector-all only on x86 where it's well supported. +dnl on some platforms it causes crashes. Hopefully the OS's default's +dnl include this on platforms that work but have been missed here. +AC_ARG_ENABLE([stack-protector], + [AS_HELP_STRING([--disable-stack-protector],[Disable compiler stack hardening])],, + [ + AS_IF([test "$ac_cv_c_compiler_gnu" = "yes" && test "$on_x86" = "yes" && test "$on_windows" = "no"], + [enable_stack_protector=yes],[enable_stack_protector=no]) + ]) + +AS_IF([test "$enable_stack_protector" = "yes"], + [ + saved_CFLAGS="$CFLAGS" + CFLAGS="$CFLAGS -fstack-protector-strong" + AC_MSG_CHECKING([if ${CC} supports -fstack-protector-strong]) + AC_LINK_IFELSE([AC_LANG_PROGRAM([],[[char foo;]])], + [ AC_MSG_RESULT([yes]) ], + [ + AC_MSG_RESULT([no]) + enable_stack_protector=no + CFLAGS="$saved_CFLAGS" + ]) + ]) + +AS_IF([test x$ac_cv_c_compiler_gnu = xyes], + [AX_ADD_FORTIFY_SOURCE] +) + +CFLAGS="$CFLAGS -W -Os" + +warn_CFLAGS="-Wall -Wextra -Wcast-align -Wnested-externs -Wshadow -Wstrict-prototypes" +saved_CFLAGS="$CFLAGS" +CFLAGS="$CFLAGS $warn_CFLAGS" +AC_MSG_CHECKING([if ${CC} supports ${warn_CFLAGS}]) +AC_COMPILE_IFELSE([AC_LANG_SOURCE([[char foo;]])], + [ AC_MSG_RESULT([yes]) ], + [ AC_MSG_RESULT([no]) + CFLAGS="$saved_CFLAGS" + ]) + +saved_LIBS="$LIBS" +LIBS="$LIBS $LIBM" +AC_CHECK_FUNCS([lrintf]) +AC_CHECK_FUNCS([lrint]) +LIBS="$saved_LIBS" + +AC_CHECK_FUNCS([__malloc_hook]) + +AC_SUBST([PC_BUILD]) + +AC_CONFIG_FILES([ + Makefile + opus.pc + opus-uninstalled.pc + celt/arm/armopts.s + doc/Makefile + doc/Doxyfile +]) +AC_CONFIG_HEADERS([config.h]) + +AC_OUTPUT + +AC_MSG_NOTICE([ +------------------------------------------------------------------------ + $PACKAGE_NAME $PACKAGE_VERSION: Automatic configuration OK. + + Compiler support: + + C99 var arrays: ................ ${has_var_arrays} + C99 lrintf: .................... ${ac_cv_func_lrintf} + Use alloca: .................... ${use_alloca} + + General configuration: + + Floating point support: ........ ${enable_float} + Fast float approximations: ..... ${enable_float_approx} + Fixed point debugging: ......... ${enable_fixed_point_debug} + Inline Assembly Optimizations: . ${inline_optimization} + External Assembly Optimizations: ${asm_optimization} + Intrinsics Optimizations: ...... ${intrinsics_support} + Run-time CPU detection: ........ ${rtcd_support} + Custom modes: .................. ${enable_custom_modes} + Assertion checking: ............ ${enable_assertions} + Hardening: ..................... ${enable_hardening} + Fuzzing: ....................... ${enable_fuzzing} + Check ASM: ..................... ${enable_check_asm} + + API documentation: ............. ${enable_doc} + Extra programs: ................ ${enable_extra_programs} +------------------------------------------------------------------------ + + Type "make; make install" to compile and install + Type "make check" to run the test suite +]) + diff --git a/lib/opus/doc/Doxyfile.in b/lib/opus/doc/Doxyfile.in new file mode 100644 index 00000000..6eef650b --- /dev/null +++ b/lib/opus/doc/Doxyfile.in @@ -0,0 +1,340 @@ +# Doxyfile 1.8.18 + +# This file describes the settings to be used by the documentation system +# doxygen (www.doxygen.org) for a project. +# +# All text after a double hash (##) is considered a comment and is placed in +# front of the TAG it is preceding. +# +# All text after a single hash (#) is considered a comment and will be ignored. +# The format is: +# TAG = value [value, ...] +# For lists, items can also be appended using: +# TAG += value [value, ...] +# Values that contain spaces should be placed between quotes (\" \"). + +# Only non-default options are included below to improve portability +# between doxygen versions. +# +#--------------------------------------------------------------------------- +# Project related configuration options +#--------------------------------------------------------------------------- + +# The PROJECT_NAME tag is a single word (or a sequence of words surrounded by +# double-quotes, unless you are using Doxywizard) that should identify the +# project for which the documentation is generated. This name is used in the +# title of most generated pages and in a few other places. +# The default value is: My Project. + +PROJECT_NAME = Opus + +# The PROJECT_NUMBER tag can be used to enter a project or revision number. This +# could be handy for archiving the generated documentation or if some version +# control system is used. + +PROJECT_NUMBER = @VERSION@ + +# Using the PROJECT_BRIEF tag one can provide an optional one line description +# for a project that appears at the top of each page and should give viewer a +# quick idea about the purpose of the project. Keep the description short. + +PROJECT_BRIEF = "Opus audio codec (RFC 6716): API and operations manual" + +# With the PROJECT_LOGO tag one can specify a logo or an icon that is included +# in the documentation. The maximum height of the logo should not exceed 55 +# pixels and the maximum width should not exceed 200 pixels. Doxygen will copy +# the logo to the output directory. + +PROJECT_LOGO = + +# If the FULL_PATH_NAMES tag is set to YES, doxygen will prepend the full path +# before files name in the file list and in the header files. If set to NO the +# shortest path that makes the file name unique will be used +# The default value is: YES. + +FULL_PATH_NAMES = NO + +# If the JAVADOC_AUTOBRIEF tag is set to YES then doxygen will interpret the +# first line (until the first dot) of a Javadoc-style comment as the brief +# description. If set to NO, the Javadoc-style will behave just like regular Qt- +# style comments (thus requiring an explicit @brief command for a brief +# description.) +# The default value is: NO. + +JAVADOC_AUTOBRIEF = YES + +# The TAB_SIZE tag can be used to set the number of spaces in a tab. Doxygen +# uses this value to replace tabs by spaces in code fragments. +# Minimum value: 1, maximum value: 16, default value: 4. + +TAB_SIZE = 8 + +# Set the OPTIMIZE_OUTPUT_FOR_C tag to YES if your project consists of C sources +# only. Doxygen will then generate output that is more tailored for C. For +# instance, some of the names that are used will be different. The list of all +# members will be omitted, etc. +# The default value is: NO. + +OPTIMIZE_OUTPUT_FOR_C = YES + +#--------------------------------------------------------------------------- +# Build related configuration options +#--------------------------------------------------------------------------- + +# If the EXTRACT_ALL tag is set to YES, doxygen will assume all entities in +# documentation are documented, even if no documentation was available. Private +# class members and static file members will be hidden unless the +# EXTRACT_PRIVATE respectively EXTRACT_STATIC tags are set to YES. +# Note: This will also disable the warnings about undocumented members that are +# normally produced when WARNINGS is set to YES. +# The default value is: NO. + +EXTRACT_ALL = YES + +# If the EXTRACT_PRIVATE tag is set to YES, all private members of a class will +# be included in the documentation. +# The default value is: NO. + +EXTRACT_PRIVATE = NO + +# If the CASE_SENSE_NAMES tag is set to NO then doxygen will only generate file +# names in lower-case letters. If set to YES, upper-case letters are also +# allowed. This is useful if you have classes or files whose names only differ +# in case and if your file system supports case sensitive file names. Windows +# (including Cygwin) ands Mac users are advised to set this option to NO. +# The default value is: system dependent. + +CASE_SENSE_NAMES = YES + +# The ENABLED_SECTIONS tag can be used to enable conditional documentation +# sections, marked by \if ... \endif and \cond +# ... \endcond blocks. + +ENABLED_SECTIONS = + +#--------------------------------------------------------------------------- +# Configuration options related to warning and progress messages +#--------------------------------------------------------------------------- + +# The QUIET tag can be used to turn on/off the messages that are generated to +# standard output by doxygen. If QUIET is set to YES this implies that the +# messages are off. +# The default value is: NO. + +QUIET = YES + +# The WARNINGS tag can be used to turn on/off the warning messages that are +# generated to standard error (stderr) by doxygen. If WARNINGS is set to YES +# this implies that the warnings are on. +# +# Tip: Turn warnings on while writing the documentation. +# The default value is: YES. + +WARNINGS = YES + +#--------------------------------------------------------------------------- +# Configuration options related to the input files +#--------------------------------------------------------------------------- + +# The INPUT tag is used to specify the files and/or directories that contain +# documented source files. You may enter file names like myfile.cpp or +# directories like /usr/src/myproject. Separate the files or directories with +# spaces. See also FILE_PATTERNS and EXTENSION_MAPPING +# Note: If this tag is empty the current directory is searched. + +INPUT = @top_srcdir@/include/opus.h \ + @top_srcdir@/include/opus_types.h \ + @top_srcdir@/include/opus_defines.h \ + @top_srcdir@/include/opus_multistream.h \ + @top_srcdir@/include/opus_custom.h + +# The EXCLUDE tag can be used to specify files and/or directories that should be +# excluded from the INPUT source files. This way you can easily exclude a +# subdirectory from a directory tree whose root is specified with the INPUT tag. +# +# Note that relative paths are relative to the directory from which doxygen is +# run. + +EXCLUDE = + +#--------------------------------------------------------------------------- +# Configuration options related to the alphabetical class index +#--------------------------------------------------------------------------- + +# If the ALPHABETICAL_INDEX tag is set to YES, an alphabetical index of all +# compounds will be generated. Enable this if the project contains a lot of +# classes, structs, unions or interfaces. +# The default value is: YES. + +ALPHABETICAL_INDEX = NO + +#--------------------------------------------------------------------------- +# Configuration options related to the HTML output +#--------------------------------------------------------------------------- + +# The HTML_HEADER tag can be used to specify a user-defined HTML header file for +# each generated HTML page. If the tag is left blank doxygen will generate a +# standard header. +# +# To get valid HTML the header file that includes any scripts and style sheets +# that doxygen needs, which is dependent on the configuration options used (e.g. +# the setting GENERATE_TREEVIEW). It is highly recommended to start with a +# default header using +# doxygen -w html new_header.html new_footer.html new_stylesheet.css +# YourConfigFile +# and then modify the file new_header.html. See also section "Doxygen usage" +# for information on how to generate the default header that doxygen normally +# uses. +# Note: The header is subject to change so you typically have to regenerate the +# default header when upgrading to a newer version of doxygen. For a description +# of the possible markers and block names see the documentation. +# This tag requires that the tag GENERATE_HTML is set to YES. + +HTML_HEADER = @top_srcdir@/doc/header.html + +# The HTML_FOOTER tag can be used to specify a user-defined HTML footer for each +# generated HTML page. If the tag is left blank doxygen will generate a standard +# footer. See HTML_HEADER for more information on how to generate a default +# footer and what special commands can be used inside the footer. See also +# section "Doxygen usage" for information on how to generate the default footer +# that doxygen normally uses. +# This tag requires that the tag GENERATE_HTML is set to YES. + +HTML_FOOTER = @top_srcdir@/doc/footer.html + +# The HTML_STYLESHEET tag can be used to specify a user-defined cascading style +# sheet that is used by each HTML page. It can be used to fine-tune the look of +# the HTML output. If left blank doxygen will generate a default style sheet. +# See also section "Doxygen usage" for information on how to generate the style +# sheet that doxygen normally uses. +# Note: It is recommended to use HTML_EXTRA_STYLESHEET instead of this tag, as +# it is more robust and this tag (HTML_STYLESHEET) will in the future become +# obsolete. +# This tag requires that the tag GENERATE_HTML is set to YES. + +HTML_STYLESHEET = @top_srcdir@/doc/customdoxygen.css + +# The HTML_EXTRA_FILES tag can be used to specify one or more extra images or +# other source files which should be copied to the HTML output directory. Note +# that these files will be copied to the base HTML output directory. Use the +# $relpath^ marker in the HTML_HEADER and/or HTML_FOOTER files to load these +# files. In the HTML_STYLESHEET file, use the file name only. Also note that the +# files will be copied as-is; there are no commands or markers available. +# This tag requires that the tag GENERATE_HTML is set to YES. + +HTML_EXTRA_FILES = @top_srcdir@/doc/opus_logo.svg + +# The HTML_COLORSTYLE_SAT tag controls the purity (or saturation) of the colors +# in the HTML output. For a value of 0 the output will use grayscales only. A +# value of 255 will produce the most vivid colors. +# Minimum value: 0, maximum value: 255, default value: 100. +# This tag requires that the tag GENERATE_HTML is set to YES. + +HTML_COLORSTYLE_SAT = 0 + +# If the HTML_TIMESTAMP tag is set to YES then the footer of each generated HTML +# page will contain the date and time when the page was generated. Setting this +# to YES can help to show when doxygen was last run and thus if the +# documentation is up to date. +# The default value is: NO. +# This tag requires that the tag GENERATE_HTML is set to YES. + +HTML_TIMESTAMP = YES + +# When MathJax is enabled you need to specify the location relative to the HTML +# output directory using the MATHJAX_RELPATH option. The destination directory +# should contain the MathJax.js script. For instance, if the mathjax directory +# is located at the same level as the HTML output directory, then +# MATHJAX_RELPATH should be ../mathjax. The default value points to the MathJax +# Content Delivery Network so you can quickly see the result without installing +# MathJax. However, it is strongly recommended to install a local copy of +# MathJax from https://www.mathjax.org before deployment. +# The default value is: https://cdn.jsdelivr.net/npm/mathjax@2. +# This tag requires that the tag USE_MATHJAX is set to YES. + +MATHJAX_RELPATH = https://www.mathjax.org/mathjax + +#--------------------------------------------------------------------------- +# Configuration options related to the LaTeX output +#--------------------------------------------------------------------------- + +# The PAPER_TYPE tag can be used to set the paper type that is used by the +# printer. +# Possible values are: a4 (210 x 297 mm), letter (8.5 x 11 inches), legal (8.5 x +# 14 inches) and executive (7.25 x 10.5 inches). +# The default value is: a4. +# This tag requires that the tag GENERATE_LATEX is set to YES. + +PAPER_TYPE = letter + +#--------------------------------------------------------------------------- +# Configuration options related to the man page output +#--------------------------------------------------------------------------- + +# If the GENERATE_MAN tag is set to YES, doxygen will generate man pages for +# classes and files. +# The default value is: NO. + +GENERATE_MAN = YES + +#--------------------------------------------------------------------------- +# Configuration options related to the preprocessor +#--------------------------------------------------------------------------- + +# If the MACRO_EXPANSION tag is set to YES, doxygen will expand all macro names +# in the source code. If set to NO, only conditional compilation will be +# performed. Macro expansion can be done in a controlled way by setting +# EXPAND_ONLY_PREDEF to YES. +# The default value is: NO. +# This tag requires that the tag ENABLE_PREPROCESSING is set to YES. + +MACRO_EXPANSION = YES + +# If the EXPAND_ONLY_PREDEF and MACRO_EXPANSION tags are both set to YES then +# the macro expansion is limited to the macros specified with the PREDEFINED and +# EXPAND_AS_DEFINED tags. +# The default value is: NO. +# This tag requires that the tag ENABLE_PREPROCESSING is set to YES. + +EXPAND_ONLY_PREDEF = YES + +# The INCLUDE_PATH tag can be used to specify one or more directories that +# contain include files that are not input files but should be processed by the +# preprocessor. +# This tag requires that the tag SEARCH_INCLUDES is set to YES. + +INCLUDE_PATH = + +# The PREDEFINED tag can be used to specify one or more macro names that are +# defined before the preprocessor is started (similar to the -D option of e.g. +# gcc). The argument of the tag is a list of macros of the form: name or +# name=definition (no spaces). If the definition and the "=" are omitted, "=1" +# is assumed. To prevent a macro definition from being undefined via #undef or +# recursively expanded use the := operator instead of the = operator. +# This tag requires that the tag ENABLE_PREPROCESSING is set to YES. + +PREDEFINED = OPUS_EXPORT= \ + OPUS_CUSTOM_EXPORT= \ + OPUS_CUSTOM_EXPORT_STATIC= \ + OPUS_WARN_UNUSED_RESULT= \ + OPUS_ARG_NONNULL(_x)= + +#--------------------------------------------------------------------------- +# Configuration options related to the dot tool +#--------------------------------------------------------------------------- + +# If you set the HAVE_DOT tag to YES then doxygen will assume the dot tool is +# available from the path. This tool is part of Graphviz (see: +# http://www.graphviz.org/), a graph visualization toolkit from AT&T and Lucent +# Bell Labs. The other options in this section have no effect if this option is +# set to NO +# The default value is: NO. + +# Debian defaults to YES here, while Fedora and Homebrew default to NO. +# So we set this based on whether the graphviz package is available at +# configure time. +# +HAVE_DOT = @HAVE_DOT@ + +# move docs to the correct place +OUTPUT_DIRECTORY = @top_builddir@/doc diff --git a/lib/opus/doc/Makefile.am b/lib/opus/doc/Makefile.am new file mode 100644 index 00000000..31fddab4 --- /dev/null +++ b/lib/opus/doc/Makefile.am @@ -0,0 +1,45 @@ +## Process this file with automake to produce Makefile.in + +DOCINPUTS = $(top_srcdir)/include/opus.h \ + $(top_srcdir)/include/opus_multistream.h \ + $(top_srcdir)/include/opus_defines.h \ + $(top_srcdir)/include/opus_types.h \ + $(top_srcdir)/include/opus_custom.h \ + $(top_srcdir)/doc/header.html \ + $(top_srcdir)/doc/footer.html \ + $(top_srcdir)/doc/customdoxygen.css + +EXTRA_DIST = customdoxygen.css Doxyfile.in footer.html header.html \ + opus_logo.svg trivial_example.c + + +if HAVE_DOXYGEN + +all-local: doxygen-build.stamp + +doxygen-build.stamp: Doxyfile $(DOCINPUTS) + doxygen + touch $@ + +install-data-local: + $(INSTALL) -d $(DESTDIR)$(docdir)/html/search + for f in `find html -type f \! -name "installdox"`; do \ + $(INSTALL_DATA) $$f $(DESTDIR)$(docdir)/$$f; \ + done + + $(INSTALL) -d $(DESTDIR)$(mandir)/man3 + cd man && find man3 -type f -name opus_*.3 \ + -exec $(INSTALL_DATA) \{} $(DESTDIR)$(mandir)/man3 \; + +clean-local: + $(RM) -r html + $(RM) -r latex + $(RM) -r man + $(RM) doxygen-build.stamp + $(RM) doxygen_sqlite3.db + +uninstall-local: + $(RM) -r $(DESTDIR)$(docdir)/html + $(RM) $(DESTDIR)$(mandir)/man3/opus_*.3 $(DESTDIR)$(mandir)/man3/opus.h.3 + +endif diff --git a/lib/opus/doc/build_draft.sh b/lib/opus/doc/build_draft.sh new file mode 100755 index 00000000..14ae51dd --- /dev/null +++ b/lib/opus/doc/build_draft.sh @@ -0,0 +1,113 @@ +#!/bin/sh + +# Copyright (c) 2011-2012 Xiph.Org Foundation and Mozilla Corporation +# +# This file is extracted from RFC6716. Please see that RFC for additional +# information. +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions +# are met: +# +# - Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# - Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in the +# documentation and/or other materials provided with the distribution. +# +# - Neither the name of Internet Society, IETF or IETF Trust, nor the +# names of specific contributors, may be used to endorse or promote +# products derived from this software without specific prior written +# permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +# ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +# OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#Stop on errors +set -e +#Set the CWD to the location of this script +[ -n "${0%/*}" ] && cd "${0%/*}" + +toplevel=".." +destdir="opus_source" + +echo packaging source code +rm -rf "${destdir}" +mkdir "${destdir}" +mkdir "${destdir}/src" +mkdir "${destdir}/silk" +mkdir "${destdir}/silk/float" +mkdir "${destdir}/silk/fixed" +mkdir "${destdir}/silk/fixed/x86" +mkdir "${destdir}/silk/fixed/arm" +mkdir "${destdir}/silk/fixed/mips" +mkdir "${destdir}/silk/x86" +mkdir "${destdir}/silk/arm" +mkdir "${destdir}/silk/mips" +mkdir "${destdir}/celt" +mkdir "${destdir}/celt/x86" +mkdir "${destdir}/celt/arm" +mkdir "${destdir}/celt/mips" +mkdir "${destdir}/include" +for f in `cat "${toplevel}"/opus_sources.mk "${toplevel}"/celt_sources.mk \ + "${toplevel}"/silk_sources.mk "${toplevel}"/opus_headers.mk \ + "${toplevel}"/celt_headers.mk "${toplevel}"/silk_headers.mk \ + | grep '\.[ch]' | sed -e 's/^.*=//' -e 's/\\\\//'` ; do + cp -a "${toplevel}/${f}" "${destdir}/${f}" +done +cp -a "${toplevel}"/src/opus_demo.c "${destdir}"/src/ +cp -a "${toplevel}"/src/opus_compare.c "${destdir}"/src/ +cp -a "${toplevel}"/celt/opus_custom_demo.c "${destdir}"/celt/ +cp -a "${toplevel}"/Makefile.unix "${destdir}"/Makefile +cp -a "${toplevel}"/opus_sources.mk "${destdir}"/ +cp -a "${toplevel}"/celt_sources.mk "${destdir}"/ +cp -a "${toplevel}"/silk_sources.mk "${destdir}"/ +cp -a "${toplevel}"/README.draft "${destdir}"/README +cp -a "${toplevel}"/COPYING "${destdir}"/COPYING +cp -a "${toplevel}"/tests/run_vectors.sh "${destdir}"/ + +GZIP=-9 tar --owner=root --group=root --format=v7 -czf opus_source.tar.gz "${destdir}" +echo building base64 version +cat opus_source.tar.gz| base64 | tr -d '\n' | fold -w 64 | \ + sed -e 's/^/\###/' -e 's/$/\<\/spanx\>\/' > \ + opus_source.base64 + + +#echo '
' > opus_compare_escaped.c +#echo '' >> opus_compare_escaped.c +#echo '> opus_compare_escaped.c +#cat opus_compare.c >> opus_compare_escaped.c +#echo ']]>' >> opus_compare_escaped.c +#echo '' >> opus_compare_escaped.c +#echo '
' >> opus_compare_escaped.c + +if [ ! -d ../opus_testvectors ] ; then + echo "Downloading test vectors..." + wget 'http://opus-codec.org/testvectors/opus_testvectors.tar.gz' + tar -C .. -xvzf opus_testvectors.tar.gz +fi +echo '
' > testvectors_sha1 +echo '' >> testvectors_sha1 +echo '> testvectors_sha1 +(cd ../opus_testvectors; sha1sum *.bit *.dec) >> testvectors_sha1 +#cd opus_testvectors +#sha1sum *.bit *.dec >> ../testvectors_sha1 +#cd .. +echo ']]>' >> testvectors_sha1 +echo '' >> testvectors_sha1 +echo '
' >> testvectors_sha1 + +echo running xml2rfc +xml2rfc draft-ietf-codec-opus.xml draft-ietf-codec-opus.html & +xml2rfc draft-ietf-codec-opus.xml +wait diff --git a/lib/opus/doc/build_isobmff.sh b/lib/opus/doc/build_isobmff.sh new file mode 100755 index 00000000..676ca4e0 --- /dev/null +++ b/lib/opus/doc/build_isobmff.sh @@ -0,0 +1,50 @@ +#!/bin/sh + +# Copyright (c) 2014 Xiph.Org Foundation and Mozilla Foundation +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions +# are met: +# +# - Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# - Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in the +# documentation and/or other materials provided with the distribution. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +# ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +# OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#Stop on errors +set -e +#Set the CWD to the location of this script +[ -n "${0%/*}" ] && cd "${0%/*}" + +HTML=opus_in_isobmff.html + +echo downloading updates... +CSS=${HTML%%.html}.css +wget -q http://vfrmaniac.fushizen.eu/contents/${HTML} -O ${HTML} +wget -q http://vfrmaniac.fushizen.eu/style.css -O ${CSS} + +echo updating links... +cat ${HTML} | sed -e "s/\\.\\.\\/style.css/${CSS}/" > ${HTML}+ && mv ${HTML}+ ${HTML} + +echo stripping... +cat ${HTML} | sed -e 's///g' > ${HTML}+ && mv ${HTML}+ ${HTML} +cat ${HTML} | sed -e 's/ *$//g' > ${HTML}+ && mv ${HTML}+ ${HTML} +cat ${CSS} | sed -e 's/ *$//g' > ${CSS}+ && mv ${CSS}+ ${CSS} + + +VERSION=$(grep -F Version ${HTML} | sed 's/.*Version \([0-9]\.[0-9]\.[0-9]\).*/\1/') +echo Now at version ${VERSION} diff --git a/lib/opus/doc/build_oggdraft.sh b/lib/opus/doc/build_oggdraft.sh new file mode 100755 index 00000000..30ee534b --- /dev/null +++ b/lib/opus/doc/build_oggdraft.sh @@ -0,0 +1,52 @@ +#!/bin/sh + +# Copyright (c) 2012 Xiph.Org Foundation and Mozilla Corporation +# +# This file is extracted from RFC6716. Please see that RFC for additional +# information. +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions +# are met: +# +# - Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# - Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in the +# documentation and/or other materials provided with the distribution. +# +# - Neither the name of Internet Society, IETF or IETF Trust, nor the +# names of specific contributors, may be used to endorse or promote +# products derived from this software without specific prior written +# permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +# ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +# OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +#Stop on errors +set -e +#Set the CWD to the location of this script +[ -n "${0%/*}" ] && cd "${0%/*}" + +if test -z `which xml2rfc 2> /dev/null`; then + echo "Error: couldn't find xml2rfc." + echo + echo "Please install xml2rfc version 2 or later." + echo "E.g. 'pip install xml2rfc' or follow the instructions" + echo "on http://pypi.python.org/pypi/xml2rfc/ or tools.ietf.org." + exit 1 +fi + +echo running xml2rfc +# version 2 syntax +xml2rfc draft-ietf-codec-oggopus.xml --text --html diff --git a/lib/opus/doc/customdoxygen.css b/lib/opus/doc/customdoxygen.css new file mode 100644 index 00000000..70047787 --- /dev/null +++ b/lib/opus/doc/customdoxygen.css @@ -0,0 +1,1011 @@ +/* The standard CSS for doxygen */ + +body, table, div, p, dl { + font-family: Lucida Grande, Verdana, Geneva, Arial, sans-serif; + font-size: 13px; + line-height: 1.3; +} + +/* @group Heading Levels */ + +h1 { + font-size: 150%; +} + +.title { + font-size: 150%; + font-weight: bold; + margin: 10px 2px; +} + +h2 { + font-size: 120%; +} + +h3 { + font-size: 100%; +} + +dt { + font-weight: bold; +} + +div.multicol { + -moz-column-gap: 1em; + -webkit-column-gap: 1em; + -moz-column-count: 3; + -webkit-column-count: 3; +} + +p.startli, p.startdd, p.starttd { + margin-top: 2px; +} + +p.endli { + margin-bottom: 0px; +} + +p.enddd { + margin-bottom: 4px; +} + +p.endtd { + margin-bottom: 2px; +} + +/* @end */ + +caption { + font-weight: bold; +} + +span.legend { + font-size: 70%; + text-align: center; +} + +h3.version { + font-size: 90%; + text-align: center; +} + +div.qindex, div.navtab{ + background-color: #F1F1F1; + border: 1px solid #BDBDBD; + text-align: center; +} + +div.qindex, div.navpath { + width: 100%; + line-height: 140%; +} + +div.navtab { + margin-right: 15px; +} + +/* @group Link Styling */ + +a { + color: #646464; + font-weight: normal; + text-decoration: none; +} + +.contents a:visited { + color: #747474; +} + +a:hover { + text-decoration: underline; +} + +a.qindex { + font-weight: bold; +} + +a.qindexHL { + font-weight: bold; + background-color: #B8B8B8; + color: #ffffff; + border: 1px double #A8A8A8; +} + +.contents a.qindexHL:visited { + color: #ffffff; +} + +a.el { + font-weight: bold; +} + +a.elRef { +} + +a.code, a.code:visited { + color: #4665A2; +} + +a.codeRef, a.codeRef:visited { + color: #4665A2; +} + +/* @end */ + +dl.el { + margin-left: -1cm; +} + +.fragment { + font-family: monospace, fixed; + font-size: 105%; +} + +pre.fragment { + border: 1px solid #D5D5D5; + background-color: #FCFCFC; + padding: 4px 6px; + margin: 4px 8px 4px 2px; + overflow: auto; + word-wrap: break-word; + font-size: 9pt; + line-height: 125%; +} + +div.ah { + background-color: black; + font-weight: bold; + color: #ffffff; + margin-bottom: 3px; + margin-top: 3px; + padding: 0.2em; + border: solid thin #333; + border-radius: 0.5em; + -webkit-border-radius: .5em; + -moz-border-radius: .5em; + box-shadow: 2px 2px 3px #999; + -webkit-box-shadow: 2px 2px 3px #999; + -moz-box-shadow: rgba(0, 0, 0, 0.15) 2px 2px 2px; + background-image: -webkit-gradient(linear, left top, left bottom, from(#eee), to(#000),color-stop(0.3, #444)); + background-image: -moz-linear-gradient(center top, #eee 0%, #444 40%, #000); +} + +div.groupHeader { + margin-left: 16px; + margin-top: 12px; + font-weight: bold; +} + +div.groupText { + margin-left: 16px; + font-style: italic; +} + +body { + background-color: white; + color: black; + margin: 0; +} + +div.contents { + margin-top: 10px; + margin-left: 8px; + margin-right: 8px; +} + +td.indexkey { + background-color: #F1F1F1; + font-weight: bold; + border: 1px solid #D5D5D5; + margin: 2px 0px 2px 0; + padding: 2px 10px; + white-space: nowrap; + vertical-align: top; +} + +td.indexvalue { + background-color: #F1F1F1; + border: 1px solid #D5D5D5; + padding: 2px 10px; + margin: 2px 0px; +} + +tr.memlist { + background-color: #F2F2F2; +} + +p.formulaDsp { + text-align: center; +} + +img.formulaDsp { + +} + +img.formulaInl { + vertical-align: middle; +} + +div.center { + text-align: center; + margin-top: 0px; + margin-bottom: 0px; + padding: 0px; +} + +div.center img { + border: 0px; +} + +address.footer { + text-align: right; + padding-right: 12px; +} + +img.footer { + border: 0px; + vertical-align: middle; +} + +/* @group Code Colorization */ + +span.keyword { + color: #008000 +} + +span.keywordtype { + color: #604020 +} + +span.keywordflow { + color: #e08000 +} + +span.comment { + color: #800000 +} + +span.preprocessor { + color: #806020 +} + +span.stringliteral { + color: #002080 +} + +span.charliteral { + color: #008080 +} + +span.vhdldigit { + color: #ff00ff +} + +span.vhdlchar { + color: #000000 +} + +span.vhdlkeyword { + color: #700070 +} + +span.vhdllogic { + color: #ff0000 +} + +blockquote { + background-color: #F9F9F9; + border-left: 2px solid #B8B8B8; + margin: 0 24px 0 4px; + padding: 0 12px 0 16px; +} + +/* @end */ + +/* +.search { + color: #003399; + font-weight: bold; +} + +form.search { + margin-bottom: 0px; + margin-top: 0px; +} + +input.search { + font-size: 75%; + color: #000080; + font-weight: normal; + background-color: #e8eef2; +} +*/ + +td.tiny { + font-size: 75%; +} + +.dirtab { + padding: 4px; + border-collapse: collapse; + border: 1px solid #BDBDBD; +} + +th.dirtab { + background: #F1F1F1; + font-weight: bold; +} + +hr { + height: 0px; + border: none; + border-top: 1px solid #7A7A7A; +} + +hr.footer { + height: 1px; +} + +/* @group Member Descriptions */ + +table.memberdecls { + border-spacing: 0px; + padding: 0px; +} + +.mdescLeft, .mdescRight, +.memItemLeft, .memItemRight, +.memTemplItemLeft, .memTemplItemRight, .memTemplParams { + background-color: #FAFAFA; + border: none; + margin: 4px; + padding: 1px 0 0 8px; +} + +.mdescLeft, .mdescRight { + padding: 0px 8px 4px 8px; + color: #555; +} + +.memItemLeft, .memItemRight, .memTemplParams { + border-top: 1px solid #D5D5D5; +} + +.memItemLeft, .memTemplItemLeft { + white-space: nowrap; +} + +.memItemRight { + width: 100%; +} + +.memTemplParams { + color: #747474; + white-space: nowrap; +} + +/* @end */ + +/* @group Member Details */ + +/* Styles for detailed member documentation */ + +.memtemplate { + font-size: 80%; + color: #747474; + font-weight: normal; + margin-left: 9px; +} + +.memnav { + background-color: #F1F1F1; + border: 1px solid #BDBDBD; + text-align: center; + margin: 2px; + margin-right: 15px; + padding: 2px; +} + +.mempage { + width: 100%; +} + +.memitem { + padding: 0; + margin-bottom: 10px; + margin-right: 5px; +} + +.memname { + white-space: nowrap; + font-weight: bold; + margin-left: 6px; +} + +.memproto, dl.reflist dt { + border-top: 1px solid #C0C0C0; + border-left: 1px solid #C0C0C0; + border-right: 1px solid #C0C0C0; + padding: 6px 0px 6px 0px; + color: #3D3D3D; + font-weight: bold; + text-shadow: 0px 1px 1px rgba(255, 255, 255, 0.9); + /* opera specific markup */ + box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15); + border-top-right-radius: 8px; + border-top-left-radius: 8px; + /* firefox specific markup */ + -moz-box-shadow: rgba(0, 0, 0, 0.15) 5px 5px 5px; + -moz-border-radius-topright: 8px; + -moz-border-radius-topleft: 8px; + /* webkit specific markup */ + -webkit-box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15); + -webkit-border-top-right-radius: 8px; + -webkit-border-top-left-radius: 8px; + background-image:url('nav_f.png'); + background-repeat:repeat-x; + background-color: #EAEAEA; + +} + +.memdoc, dl.reflist dd { + border-bottom: 1px solid #C0C0C0; + border-left: 1px solid #C0C0C0; + border-right: 1px solid #C0C0C0; + padding: 2px 5px; + background-color: #FCFCFC; + border-top-width: 0; + /* opera specific markup */ + border-bottom-left-radius: 8px; + border-bottom-right-radius: 8px; + box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15); + /* firefox specific markup */ + -moz-border-radius-bottomleft: 8px; + -moz-border-radius-bottomright: 8px; + -moz-box-shadow: rgba(0, 0, 0, 0.15) 5px 5px 5px; + background-image: -moz-linear-gradient(center top, #FFFFFF 0%, #FFFFFF 60%, #F9F9F9 95%, #F2F2F2); + /* webkit specific markup */ + -webkit-border-bottom-left-radius: 8px; + -webkit-border-bottom-right-radius: 8px; + -webkit-box-shadow: 5px 5px 5px rgba(0, 0, 0, 0.15); + background-image: -webkit-gradient(linear,center top,center bottom,from(#FFFFFF), color-stop(0.6,#FFFFFF), color-stop(0.60,#FFFFFF), color-stop(0.95,#F9F9F9), to(#F2F2F2)); +} + +dl.reflist dt { + padding: 5px; +} + +dl.reflist dd { + margin: 0px 0px 10px 0px; + padding: 5px; +} + +.paramkey { + text-align: right; +} + +.paramtype { + white-space: nowrap; +} + +.paramname { + color: #602020; + white-space: nowrap; +} +.paramname em { + font-style: normal; +} + +.params, .retval, .exception, .tparams { + border-spacing: 6px 2px; +} + +.params .paramname, .retval .paramname { + font-weight: bold; + vertical-align: top; +} + +.params .paramtype { + font-style: italic; + vertical-align: top; +} + +.params .paramdir { + font-family: "courier new",courier,monospace; + vertical-align: top; +} + + + + +/* @end */ + +/* @group Directory (tree) */ + +/* for the tree view */ + +.ftvtree { + font-family: sans-serif; + margin: 0px; +} + +/* these are for tree view when used as main index */ + +.directory { + font-size: 9pt; + font-weight: bold; + margin: 5px; +} + +.directory h3 { + margin: 0px; + margin-top: 1em; + font-size: 11pt; +} + +/* +The following two styles can be used to replace the root node title +with an image of your choice. Simply uncomment the next two styles, +specify the name of your image and be sure to set 'height' to the +proper pixel height of your image. +*/ + +/* +.directory h3.swap { + height: 61px; + background-repeat: no-repeat; + background-image: url("yourimage.gif"); +} +.directory h3.swap span { + display: none; +} +*/ + +.directory > h3 { + margin-top: 0; +} + +.directory p { + margin: 0px; + white-space: nowrap; +} + +.directory div { + display: none; + margin: 0px; +} + +.directory img { + vertical-align: -30%; +} + +/* these are for tree view when not used as main index */ + +.directory-alt { + font-size: 100%; + font-weight: bold; +} + +.directory-alt h3 { + margin: 0px; + margin-top: 1em; + font-size: 11pt; +} + +.directory-alt > h3 { + margin-top: 0; +} + +.directory-alt p { + margin: 0px; + white-space: nowrap; +} + +.directory-alt div { + display: none; + margin: 0px; +} + +.directory-alt img { + vertical-align: -30%; +} + +/* @end */ + +div.dynheader { + margin-top: 8px; +} + +address { + font-style: normal; + color: #464646; +} + +table.doxtable { + border-collapse:collapse; + margin-top: 4px; + margin-bottom: 4px; +} + +table.doxtable td, table.doxtable th { + border: 1px solid #4A4A4A; + padding: 3px 7px 2px; +} + +table.doxtable th { + background-color: #5B5B5B; + color: #FFFFFF; + font-size: 110%; + padding-bottom: 4px; + padding-top: 5px; +} + +table.fieldtable { + width: 100%; + margin-bottom: 10px; + border: 1px solid #C0C0C0; + border-spacing: 0px; + -moz-border-radius: 4px; + -webkit-border-radius: 4px; + border-radius: 4px; + -moz-box-shadow: rgba(0, 0, 0, 0.15) 2px 2px 2px; + -webkit-box-shadow: 2px 2px 2px rgba(0, 0, 0, 0.15); + box-shadow: 2px 2px 2px rgba(0, 0, 0, 0.15); +} + +.fieldtable td, .fieldtable th { + padding: 3px 7px 2px; +} + +.fieldtable td.fieldtype, .fieldtable td.fieldname { + white-space: nowrap; + border-right: 1px solid #C0C0C0; + border-bottom: 1px solid #C0C0C0; + vertical-align: top; +} + +.fieldtable td.fielddoc { + border-bottom: 1px solid #C0C0C0; + width: 100%; +} + +.fieldtable tr:last-child td { + border-bottom: none; +} + +.fieldtable th { + background-image:url('nav_f.png'); + background-repeat:repeat-x; + background-color: #EAEAEA; + font-size: 90%; + color: #3D3D3D; + padding-bottom: 4px; + padding-top: 5px; + text-align:left; + -moz-border-radius-topleft: 4px; + -moz-border-radius-topright: 4px; + -webkit-border-top-left-radius: 4px; + -webkit-border-top-right-radius: 4px; + border-top-left-radius: 4px; + border-top-right-radius: 4px; + border-bottom: 1px solid #C0C0C0; +} + + +.tabsearch { + top: 0px; + left: 10px; + height: 36px; + background-image: url('tab_b.png'); + z-index: 101; + overflow: hidden; + font-size: 13px; +} + +.navpath ul +{ + font-size: 11px; + background-image:url('tab_b.png'); + background-repeat:repeat-x; + height:30px; + line-height:30px; + color:#ABABAB; + border:solid 1px #D3D3D3; + overflow:hidden; + margin:0px; + padding:0px; +} + +.navpath li +{ + list-style-type:none; + float:left; + padding-left:10px; + padding-right:15px; + background-image:url('bc_s.png'); + background-repeat:no-repeat; + background-position:right; + color:#595959; +} + +.navpath li.navelem a +{ + height:32px; + display:block; + text-decoration: none; + outline: none; +} + +.navpath li.navelem a:hover +{ + color:#929292; +} + +.navpath li.footer +{ + list-style-type:none; + float:right; + padding-left:10px; + padding-right:15px; + background-image:none; + background-repeat:no-repeat; + background-position:right; + color:#595959; + font-size: 8pt; +} + + +div.summary +{ + float: right; + font-size: 8pt; + padding-right: 5px; + width: 50%; + text-align: right; +} + +div.summary a +{ + white-space: nowrap; +} + +div.ingroups +{ + margin-left: 5px; + font-size: 8pt; + padding-left: 5px; + width: 50%; + text-align: left; +} + +div.ingroups a +{ + white-space: nowrap; +} + +div.header +{ + background-image:url('nav_h.png'); + background-repeat:repeat-x; + background-color: #FAFAFA; + margin: 0px; + border-bottom: 1px solid #D5D5D5; +} + +div.headertitle +{ + padding: 5px 5px 5px 7px; +} + +dl +{ + padding: 0 0 0 10px; +} + +/* dl.note, dl.warning, dl.attention, dl.pre, dl.post, dl.invariant, dl.deprecated, dl.todo, dl.test, dl.bug */ +dl.section +{ + border-left:4px solid; + padding: 0 0 0 6px; +} + +dl.note +{ + border-color: #D0C000; +} + +dl.warning, dl.attention +{ + border-color: #FF0000; +} + +dl.pre, dl.post, dl.invariant +{ + border-color: #00D000; +} + +dl.deprecated +{ + border-color: #505050; +} + +dl.todo +{ + border-color: #00C0E0; +} + +dl.test +{ + border-color: #3030E0; +} + +dl.bug +{ + border-color: #C08050; +} + +dl.section dd { + margin-bottom: 6px; +} + + +#projectlogo +{ + text-align: center; + vertical-align: bottom; + border-collapse: separate; +} + +#projectlogo img +{ + border: 0px none; +} + +#projectname +{ + font: 300% Tahoma, Arial,sans-serif; + margin: 0px; + padding: 2px 0px; +} + +#projectbrief +{ + font: 120% Tahoma, Arial,sans-serif; + margin: 0px; + padding: 0px; +} + +#projectnumber +{ + font: 100% Tahoma, Arial,sans-serif; + margin: 0px; + padding: 0px; +} + +#titlearea +{ + padding: 0px; + margin: 0px; + width: 100%; + border-bottom: 1px solid #848484; +} + +.image +{ + text-align: center; +} + +.dotgraph +{ + text-align: center; +} + +.mscgraph +{ + text-align: center; +} + +.caption +{ + font-weight: bold; +} + +div.zoom +{ + border: 1px solid #AFAFAF; +} + +dl.citelist { + margin-bottom:50px; +} + +dl.citelist dt { + color:#545454; + float:left; + font-weight:bold; + margin-right:10px; + padding:5px; +} + +dl.citelist dd { + margin:2px 0; + padding:5px 0; +} + +div.toc { + padding: 14px 25px; + background-color: #F7F7F7; + border: 1px solid #E3E3E3; + border-radius: 7px 7px 7px 7px; + float: right; + height: auto; + margin: 0 20px 10px 10px; + width: 200px; +} + +div.toc li { + background: url("bdwn.png") no-repeat scroll 0 5px transparent; + font: 10px/1.2 Verdana,DejaVu Sans,Geneva,sans-serif; + margin-top: 5px; + padding-left: 10px; + padding-top: 2px; +} + +div.toc h3 { + font: bold 12px/1.2 Arial,FreeSans,sans-serif; + color: #747474; + border-bottom: 0 none; + margin: 0; +} + +div.toc ul { + list-style: none outside none; + border: medium none; + padding: 0px; +} + +div.toc li.level1 { + margin-left: 0px; +} + +div.toc li.level2 { + margin-left: 15px; +} + +div.toc li.level3 { + margin-left: 30px; +} + +div.toc li.level4 { + margin-left: 45px; +} + + +@media print +{ + #top { display: none; } + #side-nav { display: none; } + #nav-path { display: none; } + body { overflow:visible; } + h1, h2, h3, h4, h5, h6 { page-break-after: avoid; } + .summary { display: none; } + .memitem { page-break-inside: avoid; } + #doc-content + { + margin-left:0 !important; + height:auto !important; + width:auto !important; + overflow:inherit; + display:inline; + } + pre.fragment + { + overflow: visible; + text-wrap: unrestricted; + white-space: -moz-pre-wrap; /* Moz */ + white-space: -pre-wrap; /* Opera 4-6 */ + white-space: -o-pre-wrap; /* Opera 7 */ + white-space: pre-wrap; /* CSS3 */ + word-wrap: break-word; /* IE 5.5+ */ + } +} diff --git a/lib/opus/doc/draft-ietf-codec-oggopus.xml b/lib/opus/doc/draft-ietf-codec-oggopus.xml new file mode 100644 index 00000000..128816e9 --- /dev/null +++ b/lib/opus/doc/draft-ietf-codec-oggopus.xml @@ -0,0 +1,1873 @@ + + + + + + + + + + + + +]> + + + + + +Ogg Encapsulation for the Opus Audio Codec + +Mozilla Corporation +
+ +650 Castro Street +Mountain View +CA +94041 +USA + ++1 650 903-0800 +tterribe@xiph.org +
+ + + +Voicetronix +
+ +246 Pulteney Street, Level 1 +Adelaide +SA +5000 +Australia + ++61 8 8232 9112 +ron@debian.org +
+ + + +Mozilla Corporation +
+ +163 West Hastings Street +Vancouver +BC +V6B 1H5 +Canada + ++1 778 785 1540 +giles@xiph.org +
+ + + +RAI +codec + + + +This document defines the Ogg encapsulation for the Opus interactive speech and + audio codec. +This allows data encoded in the Opus format to be stored in an Ogg logical + bitstream. + + + + + +
+ +The IETF Opus codec is a low-latency audio codec optimized for both voice and + general-purpose audio. +See for technical details. +This document defines the encapsulation of Opus in a continuous, logical Ogg + bitstream . +Ogg encapsulation provides Opus with a long-term storage format supporting + all of the essential features, including metadata, fast and accurate seeking, + corruption detection, recapture after errors, low overhead, and the ability to + multiplex Opus with other codecs (including video) with minimal buffering. +It also provides a live streamable format, capable of delivery over a reliable + stream-oriented transport, without requiring all the data, or even the total + length of the data, up-front, in a form that is identical to the on-disk + storage format. + + +Ogg bitstreams are made up of a series of 'pages', each of which contains data + from one or more 'packets'. +Pages are the fundamental unit of multiplexing in an Ogg stream. +Each page is associated with a particular logical stream and contains a capture + pattern and checksum, flags to mark the beginning and end of the logical + stream, and a 'granule position' that represents an absolute position in the + stream, to aid seeking. +A single page can contain up to 65,025 octets of packet data from up to 255 + different packets. +Packets can be split arbitrarily across pages, and continued from one page to + the next (allowing packets much larger than would fit on a single page). +Each page contains 'lacing values' that indicate how the data is partitioned + into packets, allowing a demultiplexer (demuxer) to recover the packet + boundaries without examining the encoded data. +A packet is said to 'complete' on a page when the page contains the final + lacing value corresponding to that packet. + + +This encapsulation defines the contents of the packet data, including + the necessary headers, the organization of those packets into a logical + stream, and the interpretation of the codec-specific granule position field. +It does not attempt to describe or specify the existing Ogg container format. +Readers unfamiliar with the basic concepts mentioned above are encouraged to + review the details in . + + +
+ +
+ +The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", + "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this + document are to be interpreted as described in . + + +
+ +
+ +An Ogg Opus stream is organized as follows (see + for an example). + + +
+ +
+ + +There are two mandatory header packets. +The first packet in the logical Ogg bitstream MUST contain the identification + (ID) header, which uniquely identifies a stream as Opus audio. +The format of this header is defined in . +It is placed alone (without any other packet data) on the first page of + the logical Ogg bitstream, and completes on that page. +This page has its 'beginning of stream' flag set. + + +The second packet in the logical Ogg bitstream MUST contain the comment header, + which contains user-supplied metadata. +The format of this header is defined in . +It MAY span multiple pages, beginning on the second page of the logical + stream. +However many pages it spans, the comment header packet MUST finish the page on + which it completes. + + +All subsequent pages are audio data pages, and the Ogg packets they contain are + audio data packets. +Each audio data packet contains one Opus packet for each of N different + streams, where N is typically one for mono or stereo, but MAY be greater than + one for multichannel audio. +The value N is specified in the ID header (see + ), and is fixed over the entire length of the + logical Ogg bitstream. + + +The first (N - 1) Opus packets, if any, are packed one after another + into the Ogg packet, using the self-delimiting framing from Appendix B of + . +The remaining Opus packet is packed at the end of the Ogg packet using the + regular, undelimited framing from Section 3 of . +All of the Opus packets in a single Ogg packet MUST be constrained to have the + same duration. +An implementation of this specification SHOULD treat any Opus packet whose + duration is different from that of the first Opus packet in an Ogg packet as + if it were a malformed Opus packet with an invalid Table Of Contents (TOC) + sequence. + + +The TOC sequence at the beginning of each Opus packet indicates the coding + mode, audio bandwidth, channel count, duration (frame size), and number of + frames per packet, as described in Section 3.1 + of . +The coding mode is one of SILK, Hybrid, or Constrained Energy Lapped Transform + (CELT). +The combination of coding mode, audio bandwidth, and frame size is referred to + as the configuration of an Opus packet. + + +Packets are placed into Ogg pages in order until the end of stream. +Audio data packets might span page boundaries. +The first audio data page could have the 'continued packet' flag set + (indicating the first audio data packet is continued from a previous page) if, + for example, it was a live stream joined mid-broadcast, with the headers + pasted on the front. +If a page has the 'continued packet' flag set and one of the following + conditions is also true: + +the previous page with packet data does not end in a continued packet (does + not end with a lacing value of 255) OR +the page sequence numbers are not consecutive, + + then a demuxer MUST NOT attempt to decode the data for the first packet on the + page unless the demuxer has some special knowledge that would allow it to + interpret this data despite the missing pieces. +An implementation MUST treat a zero-octet audio data packet as if it were a + malformed Opus packet as described in + Section 3.4 of . + + +A logical stream ends with a page with the 'end of stream' flag set, but + implementations need to be prepared to deal with truncated streams that do not + have a page marked 'end of stream'. +There is no reason for the final packet on the last page to be a continued + packet, i.e., for the final lacing value to be 255. +However, demuxers might encounter such streams, possibly as the result of a + transfer that did not complete or of corruption. +If a packet continues onto a subsequent page (i.e., when the page ends with a + lacing value of 255) and one of the following conditions is also true: + +the next page with packet data does not have the 'continued packet' flag + set OR +there is no next page with packet data OR +the page sequence numbers are not consecutive, + + then a demuxer MUST NOT attempt to decode the data from that packet unless the + demuxer has some special knowledge that would allow it to interpret this data + despite the missing pieces. +There MUST NOT be any more pages in an Opus logical bitstream after a page + marked 'end of stream'. + +
+ +
+ +The granule position MUST be zero for the ID header page and the + page where the comment header completes. +That is, the first page in the logical stream, and the last header + page before the first audio data page both have a granule position of zero. + + +The granule position of an audio data page encodes the total number of PCM + samples in the stream up to and including the last fully-decodable sample from + the last packet completed on that page. +The granule position of the first audio data page will usually be larger than + zero, as described in . + + + +A page that is entirely spanned by a single packet (that completes on a + subsequent page) has no granule position, and the granule position field is + set to the special value '-1' in two's complement. + + + +The granule position of an audio data page is in units of PCM audio samples at + a fixed rate of 48 kHz (per channel; a stereo stream's granule position + does not increment at twice the speed of a mono stream). +It is possible to run an Opus decoder at other sampling rates, + but all Opus packets encode samples at a sampling rate that evenly divides + 48 kHz. +Therefore, the value in the granule position field always counts samples + assuming a 48 kHz decoding rate, and the rest of this specification makes + the same assumption. + + + +The duration of an Opus packet as defined in can be + any multiple of 2.5 ms, up to a maximum of 120 ms. +This duration is encoded in the TOC sequence at the beginning of each packet. +The number of samples returned by a decoder corresponds to this duration + exactly, even for the first few packets. +For example, a 20 ms packet fed to a decoder running at 48 kHz will + always return 960 samples. +A demuxer can parse the TOC sequence at the beginning of each Ogg packet to + work backwards or forwards from a packet with a known granule position (i.e., + the last packet completed on some page) in order to assign granule positions + to every packet, or even every individual sample. +The one exception is the last page in the stream, as described below. + + + +All other pages with completed packets after the first MUST have a granule + position equal to the number of samples contained in packets that complete on + that page plus the granule position of the most recent page with completed + packets. +This guarantees that a demuxer can assign individual packets the same granule + position when working forwards as when working backwards. +For this to work, there cannot be any gaps. + + +
+ +In order to support capturing a real-time stream that has lost or not + transmitted packets, a multiplexer (muxer) SHOULD emit packets that explicitly + request the use of Packet Loss Concealment (PLC) in place of the missing + packets. +Implementations that fail to do so still MUST NOT increment the granule + position for a page by anything other than the number of samples contained in + packets that actually complete on that page. + + +Only gaps that are a multiple of 2.5 ms are repairable, as these are the + only durations that can be created by packet loss or discontinuous + transmission. +Muxers need not handle other gap sizes. +Creating the necessary packets involves synthesizing a TOC byte (defined in +Section 3.1 of )—and whatever + additional internal framing is needed—to indicate the packet duration + for each stream. +The actual length of each missing Opus frame inside the packet is zero bytes, + as defined in Section 3.2.1 of . + + + +Zero-byte frames MAY be packed into packets using any of codes 0, 1, + 2, or 3. +When successive frames have the same configuration, the higher code packings + reduce overhead. +Likewise, if the TOC configuration matches, the muxer MAY further combine the + empty frames with previous or subsequent non-zero-length frames (using + code 2 or VBR code 3). + + + + does not impose any requirements on the PLC, but this + section outlines choices that are expected to have a positive influence on + most PLC implementations, including the reference implementation. +Synthesized TOC sequences SHOULD maintain the same mode, audio bandwidth, + channel count, and frame size as the previous packet (if any). +This is the simplest and usually the most well-tested case for the PLC to + handle and it covers all losses that do not include a configuration switch, + as defined in Section 4.5 of . + + + +When a previous packet is available, keeping the audio bandwidth and channel + count the same allows the PLC to provide maximum continuity in the concealment + data it generates. +However, if the size of the gap is not a multiple of the most recent frame + size, then the frame size will have to change for at least some frames. +Such changes SHOULD be delayed as long as possible to simplify + things for PLC implementations. + + + +As an example, a 95 ms gap could be encoded as nineteen 5 ms frames + in two bytes with a single CBR code 3 packet. +If the previous frame size was 20 ms, using four 20 ms frames + followed by three 5 ms frames requires 4 bytes (plus an extra byte + of Ogg lacing overhead), but allows the PLC to use its well-tested steady + state behavior for as long as possible. +The total bitrate of the latter approach, including Ogg overhead, is about + 0.4 kbps, so the impact on file size is minimal. + + + +Changing modes is discouraged, since this causes some decoder implementations + to reset their PLC state. +However, SILK and Hybrid mode frames cannot fill gaps that are not a multiple + of 10 ms. +If switching to CELT mode is needed to match the gap size, a muxer SHOULD do + so at the end of the gap to allow the PLC to function for as long as possible. + + + +In the example above, if the previous frame was a 20 ms SILK mode frame, + the better solution is to synthesize a packet describing four 20 ms SILK + frames, followed by a packet with a single 10 ms SILK + frame, and finally a packet with a 5 ms CELT frame, to fill the 95 ms + gap. +This also requires four bytes to describe the synthesized packet data (two + bytes for a CBR code 3 and one byte each for two code 0 packets) but three + bytes of Ogg lacing overhead are needed to mark the packet boundaries. +At 0.6 kbps, this is still a minimal bitrate impact over a naive, low quality + solution. + + + +Since medium-band audio is an option only in the SILK mode, wideband frames + SHOULD be generated if switching from that configuration to CELT mode, to + ensure that any PLC implementation which does try to migrate state between + the modes will be able to preserve all of the available audio bandwidth. + + +
+ +
+ +There is some amount of latency introduced during the decoding process, to + allow for overlap in the CELT mode, stereo mixing in the SILK mode, and + resampling. +The encoder might have introduced additional latency through its own resampling + and analysis (though the exact amount is not specified). +Therefore, the first few samples produced by the decoder do not correspond to + real input audio, but are instead composed of padding inserted by the encoder + to compensate for this latency. +These samples need to be stored and decoded, as Opus is an asymptotically + convergent predictive codec, meaning the decoded contents of each frame depend + on the recent history of decoder inputs. +However, a player will want to skip these samples after decoding them. + + + +A 'pre-skip' field in the ID header (see ) signals + the number of samples that SHOULD be skipped (decoded but discarded) at the + beginning of the stream, though some specific applications might have a reason + for looking at that data. +This amount need not be a multiple of 2.5 ms, MAY be smaller than a single + packet, or MAY span the contents of several packets. +These samples are not valid audio. + + + +For example, if the first Opus frame uses the CELT mode, it will always + produce 120 samples of windowed overlap-add data. +However, the overlap data is initially all zeros (since there is no prior + frame), meaning this cannot, in general, accurately represent the original + audio. +The SILK mode requires additional delay to account for its analysis and + resampling latency. +The encoder delays the original audio to avoid this problem. + + + +The pre-skip field MAY also be used to perform sample-accurate cropping of + already encoded streams. +In this case, a value of at least 3840 samples (80 ms) provides + sufficient history to the decoder that it will have converged + before the stream's output begins. + + +
+ +
+ +The PCM sample position is determined from the granule position using the + formula + +
+ +
+ + +For example, if the granule position of the first audio data page is 59,971, + and the pre-skip is 11,971, then the PCM sample position of the last decoded + sample from that page is 48,000. + + +This can be converted into a playback time using the formula + +
+ +
+ + +The initial PCM sample position before any samples are played is normally '0'. +In this case, the PCM sample position of the first audio sample to be played + starts at '1', because it marks the time on the clock + after that sample has been played, and a stream + that is exactly one second long has a final PCM sample position of '48000', + as in the example here. + + + +Vorbis streams use a granule position smaller than the number of audio samples + contained in the first audio data page to indicate that some of those samples + are trimmed from the output (see ). +However, to do so, Vorbis requires that the first audio data page contains + exactly two packets, in order to allow the decoder to perform PCM position + adjustments before needing to return any PCM data. +Opus uses the pre-skip mechanism for this purpose instead, since the encoder + might introduce more than a single packet's worth of latency, and since very + large packets in streams with a very large number of channels might not fit + on a single page. + +
+ +
+ +The page with the 'end of stream' flag set MAY have a granule position that + indicates the page contains less audio data than would normally be returned by + decoding up through the final packet. +This is used to end the stream somewhere other than an even frame boundary. +The granule position of the most recent audio data page with completed packets + is used to make this determination, or '0' is used if there were no previous + audio data pages with a completed packet. +The difference between these granule positions indicates how many samples to + keep after decoding the packets that completed on the final page. +The remaining samples are discarded. +The number of discarded samples SHOULD be no larger than the number decoded + from the last packet. + +
+ +
+ +The granule position of the first audio data page with a completed packet MAY + be larger than the number of samples contained in packets that complete on + that page, however it MUST NOT be smaller, unless that page has the 'end of + stream' flag set. +Allowing a granule position larger than the number of samples allows the + beginning of a stream to be cropped or a live stream to be joined without + rewriting the granule position of all the remaining pages. +This means that the PCM sample position just before the first sample to be + played MAY be larger than '0'. +Synchronization when multiplexing with other logical streams still uses the PCM + sample position relative to '0' to compute sample times. +This does not affect the behavior of pre-skip: exactly 'pre-skip' samples + SHOULD be skipped from the beginning of the decoded output, even if the + initial PCM sample position is greater than zero. + + + +On the other hand, a granule position that is smaller than the number of + decoded samples prevents a demuxer from working backwards to assign each + packet or each individual sample a valid granule position, since granule + positions are non-negative. +An implementation MUST treat any stream as invalid if the granule position + is smaller than the number of samples contained in packets that complete on + the first audio data page with a completed packet, unless that page has the + 'end of stream' flag set. +It MAY defer this action until it decodes the last packet completed on that + page. + + + +If that page has the 'end of stream' flag set, a demuxer MUST treat any stream + as invalid if its granule position is smaller than the 'pre-skip' amount. +This would indicate that there are more samples to be skipped from the initial + decoded output than exist in the stream. +If the granule position is smaller than the number of decoded samples produced + by the packets that complete on that page, then a demuxer MUST use an initial + granule position of '0', and can work forwards from '0' to timestamp + individual packets. +If the granule position is larger than the number of decoded samples available, + then the demuxer MUST still work backwards as described above, even if the + 'end of stream' flag is set, to determine the initial granule position, and + thus the initial PCM sample position. +Both of these will be greater than '0' in this case. + +
+ +
+ +Seeking in Ogg files is best performed using a bisection search for a page + whose granule position corresponds to a PCM position at or before the seek + target. +With appropriately weighted bisection, accurate seeking can be performed in + just one or two bisections on average, even in multi-gigabyte files. +See for an example of general implementation guidance. + + + +When seeking within an Ogg Opus stream, an implementation SHOULD start decoding + (and discarding the output) at least 3840 samples (80 ms) prior to + the seek target in order to ensure that the output audio is correct by the + time it reaches the seek target. +This 'pre-roll' is separate from, and unrelated to, the 'pre-skip' used at the + beginning of the stream. +If the point 80 ms prior to the seek target comes before the initial PCM + sample position, an implementation SHOULD start decoding from the beginning of + the stream, applying pre-skip as normal, regardless of whether the pre-skip is + larger or smaller than 80 ms, and then continue to discard samples + to reach the seek target (if any). + +
+ +
+ +
+ +An Ogg Opus logical stream contains exactly two mandatory header packets: + an identification header and a comment header. + + +
+ +
+ +
+ + +The fields in the identification (ID) header have the following meaning: + +Magic Signature: + +This is an 8-octet (64-bit) field that allows codec identification and is + human-readable. +It contains, in order, the magic numbers: + +0x4F 'O' +0x70 'p' +0x75 'u' +0x73 's' +0x48 'H' +0x65 'e' +0x61 'a' +0x64 'd' + +Starting with "Op" helps distinguish it from audio data packets, as this is an + invalid TOC sequence. + + +Version (8 bits, unsigned): + +The version number MUST always be '1' for this version of the encapsulation + specification. +Implementations SHOULD treat streams where the upper four bits of the version + number match that of a recognized specification as backwards-compatible with + that specification. +That is, the version number can be split into "major" and "minor" version + sub-fields, with changes to the "minor" sub-field (in the lower four bits) + signaling compatible changes. +For example, an implementation of this specification SHOULD accept any stream + with a version number of '15' or less, and SHOULD assume any stream with a + version number '16' or greater is incompatible. +The initial version '1' was chosen to keep implementations from relying on this + octet as a null terminator for the "OpusHead" string. + + +Output Channel Count 'C' (8 bits, unsigned): + +This is the number of output channels. +This might be different than the number of encoded channels, which can change + on a packet-by-packet basis. +This value MUST NOT be zero. +The maximum allowable value depends on the channel mapping family, and might be + as large as 255. +See for details. + + +Pre-skip (16 bits, unsigned, little + endian): + +This is the number of samples (at 48 kHz) to discard from the decoder + output when starting playback, and also the number to subtract from a page's + granule position to calculate its PCM sample position. +When cropping the beginning of existing Ogg Opus streams, a pre-skip of at + least 3,840 samples (80 ms) is RECOMMENDED to ensure complete + convergence in the decoder. + + +Input Sample Rate (32 bits, unsigned, little + endian): + +This is the sample rate of the original input (before encoding), in Hz. +This field is not the sample rate to use for + playback of the encoded data. + +Opus can switch between internal audio bandwidths of 4, 6, 8, 12, and + 20 kHz. +Each packet in the stream can have a different audio bandwidth. +Regardless of the audio bandwidth, the reference decoder supports decoding any + stream at a sample rate of 8, 12, 16, 24, or 48 kHz. +The original sample rate of the audio passed to the encoder is not preserved + by the lossy compression. + +An Ogg Opus player SHOULD select the playback sample rate according to the + following procedure: + +If the hardware supports 48 kHz playback, decode at 48 kHz. +Otherwise, if the hardware's highest available sample rate is a supported + rate, decode at this sample rate. +Otherwise, if the hardware's highest available sample rate is less than + 48 kHz, decode at the next higher Opus supported rate above the highest + available hardware rate and resample. +Otherwise, decode at 48 kHz and resample. + +However, the 'Input Sample Rate' field allows the muxer to pass the sample + rate of the original input stream as metadata. +This is useful when the user requires the output sample rate to match the + input sample rate. +For example, when not playing the output, an implementation writing PCM format + samples to disk might choose to resample the audio back to the original input + sample rate to reduce surprise to the user, who might reasonably expect to get + back a file with the same sample rate. + +A value of zero indicates 'unspecified'. +Muxers SHOULD write the actual input sample rate or zero, but implementations + which do something with this field SHOULD take care to behave sanely if given + crazy values (e.g., do not actually upsample the output to 10 MHz if + requested). +Implementations SHOULD support input sample rates between 8 kHz and + 192 kHz (inclusive). +Rates outside this range MAY be ignored by falling back to the default rate of + 48 kHz instead. + + +Output Gain (16 bits, signed, little endian): + +This is a gain to be applied when decoding. +It is 20*log10 of the factor by which to scale the decoder output to achieve + the desired playback volume, stored in a 16-bit, signed, two's complement + fixed-point value with 8 fractional bits (i.e., + Q7.8 ). + +To apply the gain, an implementation could use +
+ +
+ where output_gain is the raw 16-bit value from the header. + +Players and media frameworks SHOULD apply it by default. +If a player chooses to apply any volume adjustment or gain modification, such + as the R128_TRACK_GAIN (see ), the adjustment + MUST be applied in addition to this output gain in order to achieve playback + at the normalized volume. + +A muxer SHOULD set this field to zero, and instead apply any gain prior to + encoding, when this is possible and does not conflict with the user's wishes. +A nonzero output gain indicates the gain was adjusted after encoding, or that + a user wished to adjust the gain for playback while preserving the ability + to recover the original signal amplitude. + +Although the output gain has enormous range (+/- 128 dB, enough to amplify + inaudible sounds to the threshold of physical pain), most applications can + only reasonably use a small portion of this range around zero. +The large range serves in part to ensure that gain can always be losslessly + transferred between OpusHead and R128 gain tags (see below) without + saturating. + + +Channel Mapping Family (8 bits, unsigned): + +This octet indicates the order and semantic meaning of the output channels. + +Each currently specified value of this octet indicates a mapping family, which + defines a set of allowed channel counts, and the ordered set of channel names + for each allowed channel count. +The details are described in . + +Channel Mapping Table: +This table defines the mapping from encoded streams to output channels. +Its contents are specified in . + + + + + +All fields in the ID headers are REQUIRED, except for the channel mapping + table, which MUST be omitted when the channel mapping family is 0, but + is REQUIRED otherwise. +Implementations SHOULD treat a stream as invalid if it contains an ID header + that does not have enough data for these fields, even if it contain a valid + Magic Signature. +Future versions of this specification, even backwards-compatible versions, + might include additional fields in the ID header. +If an ID header has a compatible major version, but a larger minor version, + an implementation MUST NOT treat it as invalid for containing additional data + not specified here, provided it still completes on the first page. + + +
+ +An Ogg Opus stream allows mapping one number of Opus streams (N) to a possibly + larger number of decoded channels (M + N) to yet another number of + output channels (C), which might be larger or smaller than the number of + decoded channels. +The order and meaning of these channels are defined by a channel mapping, + which consists of the 'channel mapping family' octet and, for channel mapping + families other than family 0, a channel mapping table, as illustrated in + . + + +
+ +
+ + +The fields in the channel mapping table have the following meaning: + +Stream Count 'N' (8 bits, unsigned): + +This is the total number of streams encoded in each Ogg packet. +This value is necessary to correctly parse the packed Opus packets inside an + Ogg packet, as described in . +This value MUST NOT be zero, as without at least one Opus packet with a valid + TOC sequence, a demuxer cannot recover the duration of an Ogg packet. + +For channel mapping family 0, this value defaults to 1, and is not coded. + + +Coupled Stream Count 'M' (8 bits, unsigned): +This is the number of streams whose decoders are to be configured to produce + two channels (stereo). +This MUST be no larger than the total number of streams, N. + +Each packet in an Opus stream has an internal channel count of 1 or 2, which + can change from packet to packet. +This is selected by the encoder depending on the bitrate and the audio being + encoded. +The original channel count of the audio passed to the encoder is not + necessarily preserved by the lossy compression. + +Regardless of the internal channel count, any Opus stream can be decoded as + mono (a single channel) or stereo (two channels) by appropriate initialization + of the decoder. +The 'coupled stream count' field indicates that the decoders for the first M + Opus streams are to be initialized for stereo (two-channel) output, and the + remaining (N - M) decoders are to be initialized for mono (a single + channel) only. +The total number of decoded channels, (M + N), MUST be no larger than + 255, as there is no way to index more channels than that in the channel + mapping. + +For channel mapping family 0, this value defaults to (C - 1) + (i.e., 0 for mono and 1 for stereo), and is not coded. + + +Channel Mapping (8*C bits): +This contains one octet per output channel, indicating which decoded channel + is to be used for each one. +Let 'index' be the value of this octet for a particular output channel. +This value MUST either be smaller than (M + N), or be the special + value 255. +If 'index' is less than 2*M, the output MUST be taken from decoding stream + ('index'/2) as stereo and selecting the left channel if 'index' is even, and + the right channel if 'index' is odd. +If 'index' is 2*M or larger, but less than 255, the output MUST be taken from + decoding stream ('index' - M) as mono. +If 'index' is 255, the corresponding output channel MUST contain pure silence. + +The number of output channels, C, is not constrained to match the number of + decoded channels (M + N). +A single index value MAY appear multiple times, i.e., the same decoded channel + might be mapped to multiple output channels. +Some decoded channels might not be assigned to any output channel, as well. + +For channel mapping family 0, the first index defaults to 0, and if + C == 2, the second index defaults to 1. +Neither index is coded. + + + + + +After producing the output channels, the channel mapping family determines the + semantic meaning of each one. +There are three defined mapping families in this specification. + + +
+ +Allowed numbers of channels: 1 or 2. +RTP mapping. +This is the same channel interpretation as . + + + +1 channel: monophonic (mono). +2 channels: stereo (left, right). + +Special mapping: This channel mapping value also + indicates that the contents consists of a single Opus stream that is stereo if + and only if C == 2, with stream index 0 mapped to output + channel 0 (mono, or left channel) and stream index 1 mapped to + output channel 1 (right channel) if stereo. +When the 'channel mapping family' octet has this value, the channel mapping + table MUST be omitted from the ID header packet. + +
+ +
+ +Allowed numbers of channels: 1...8. +Vorbis channel order (see below). + + +Each channel is assigned to a speaker location in a conventional surround + arrangement. +Specific locations depend on the number of channels, and are given below + in order of the corresponding channel indices. + + 1 channel: monophonic (mono). + 2 channels: stereo (left, right). + 3 channels: linear surround (left, center, right) + 4 channels: quadraphonic (front left, front right, rear left, rear right). + 5 channels: 5.0 surround (front left, front center, front right, rear left, rear right). + 6 channels: 5.1 surround (front left, front center, front right, rear left, rear right, LFE). + 7 channels: 6.1 surround (front left, front center, front right, side left, side right, rear center, LFE). + 8 channels: 7.1 surround (front left, front center, front right, side left, side right, rear left, rear right, LFE) + + + +This set of surround options and speaker location orderings is the same + as those used by the Vorbis codec . +The ordering is different from the one used by the + WAVE and + Free Lossless Audio Codec (FLAC) formats, + so correct ordering requires permutation of the output channels when decoding + to or encoding from those formats. +'LFE' here refers to a Low Frequency Effects channel, often mapped to a + subwoofer with no particular spatial position. +Implementations SHOULD identify 'side' or 'rear' speaker locations with + 'surround' and 'back' as appropriate when interfacing with audio formats + or systems which prefer that terminology. + +
+ +
+ +Allowed numbers of channels: 1...255. +No defined channel meaning. + + +Channels are unidentified. +General-purpose players SHOULD NOT attempt to play these streams. +Offline implementations MAY deinterleave the output into separate PCM files, + one per channel. +Implementations SHOULD NOT produce output for channels mapped to stream index + 255 (pure silence) unless they have no other way to indicate the index of + non-silent channels. + +
+ +
+ +The remaining channel mapping families (2...254) are reserved. +A demuxer implementation encountering a reserved channel mapping family value + SHOULD act as though the value is 255. + +
+ +
+ +An Ogg Opus player MUST support any valid channel mapping with a channel + mapping family of 0 or 1, even if the number of channels does not match the + physically connected audio hardware. +Players SHOULD perform channel mixing to increase or reduce the number of + channels as needed. + + + +Implementations MAY use the matrices in + Figures  + through  to implement + downmixing from multichannel files using + Channel Mapping Family 1, which are + known to give acceptable results for stereo. +Matrices for 3 and 4 channels are normalized so each coefficient row sums + to 1 to avoid clipping. +For 5 or more channels they are normalized to 2 as a compromise between + clipping and dynamic range reduction. + + +In these matrices the front left and front right channels are generally +passed through directly. +When a surround channel is split between both the left and right stereo + channels, coefficients are chosen so their squares sum to 1, which + helps preserve the perceived intensity. +Rear channels are mixed more diffusely or attenuated to maintain focus + on the front channels. + + +
+ + +Exact coefficient values are 1 and 1/sqrt(2), multiplied by + 1/(1 + 1/sqrt(2)) for normalization. + +
+ +
+ + +Exact coefficient values are 1, sqrt(3)/2 and 1/2, multiplied by + 1/(1 + sqrt(3)/2 + 1/2) for normalization. + +
+ +
+ + +Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2 and 1/2, multiplied by + 2/(1 + 1/sqrt(2) + sqrt(3)/2 + 1/2) + for normalization. + +
+ +
+ + +Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2 and 1/2, multiplied by +2/(1 + 1/sqrt(2) + sqrt(3)/2 + 1/2 + 1/sqrt(2)) + for normalization. + +
+ +
+ + +Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2, 1/2 and + sqrt(3)/2/sqrt(2), multiplied by + 2/(1 + 1/sqrt(2) + sqrt(3)/2 + 1/2 + + sqrt(3)/2/sqrt(2) + 1/sqrt(2)) for normalization. +The coefficients are in the same order as in , + and the matrices above. + +
+ +
+ + +Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2 and 1/2, multiplied by + 2/(2 + 2/sqrt(2) + sqrt(3)) for normalization. +The coefficients are in the same order as in , + and the matrices above. + +
+ +
+ +
+ +
+ +
+ +
+ +
+ + +The comment header consists of a 64-bit magic signature, followed by data in + the same format as the header used in Ogg + Vorbis, except (like Ogg Theora and Speex) the final "framing bit" specified + in the Vorbis spec is not present. + +Magic Signature: + +This is an 8-octet (64-bit) field that allows codec identification and is + human-readable. +It contains, in order, the magic numbers: + +0x4F 'O' +0x70 'p' +0x75 'u' +0x73 's' +0x54 'T' +0x61 'a' +0x67 'g' +0x73 's' + +Starting with "Op" helps distinguish it from audio data packets, as this is an + invalid TOC sequence. + + +Vendor String Length (32 bits, unsigned, little endian): + +This field gives the length of the following vendor string, in octets. +It MUST NOT indicate that the vendor string is longer than the rest of the + packet. + + +Vendor String (variable length, UTF-8 vector): + +This is a simple human-readable tag for vendor information, encoded as a UTF-8 + string . +No terminating null octet is necessary. + +This tag is intended to identify the codec encoder and encapsulation + implementations, for tracing differences in technical behavior. +User-facing applications can use the 'ENCODER' user comment tag to identify + themselves. + + +User Comment List Length (32 bits, unsigned, little endian): + +This field indicates the number of user-supplied comments. +It MAY indicate there are zero user-supplied comments, in which case there are + no additional fields in the packet. +It MUST NOT indicate that there are so many comments that the comment string + lengths would require more data than is available in the rest of the packet. + + +User Comment #i String Length (32 bits, unsigned, little endian): + +This field gives the length of the following user comment string, in octets. +There is one for each user comment indicated by the 'user comment list length' + field. +It MUST NOT indicate that the string is longer than the rest of the packet. + + +User Comment #i String (variable length, UTF-8 vector): + +This field contains a single user comment encoded as a UTF-8 + string . +There is one for each user comment indicated by the 'user comment list length' + field. + + + + + +The vendor string length and user comment list length are REQUIRED, and + implementations SHOULD treat a stream as invalid if it contains a comment + header that does not have enough data for these fields, or that does not + contain enough data for the corresponding vendor string or user comments they + describe. +Making this check before allocating the associated memory to contain the data + helps prevent a possible Denial-of-Service (DoS) attack from small comment + headers that claim to contain strings longer than the entire packet or more + user comments than than could possibly fit in the packet. + + + +Immediately following the user comment list, the comment header MAY + contain zero-padding or other binary data which is not specified here. +If the least-significant bit of the first byte of this data is 1, then editors + SHOULD preserve the contents of this data when updating the tags, but if this + bit is 0, all such data MAY be treated as padding, and truncated or discarded + as desired. +This allows informal experimentation with the format of this binary data until + it can be specified later. + + + +The comment header can be arbitrarily large and might be spread over a large + number of Ogg pages. +Implementations MUST avoid attempting to allocate excessive amounts of memory + when presented with a very large comment header. +To accomplish this, implementations MAY treat a stream as invalid if it has a + comment header larger than 125,829,120 octets (120 MB), and MAY + ignore individual comments that are not fully contained within the first + 61,440 octets of the comment header. + + +
+ +The user comment strings follow the NAME=value format described by + with the same recommended tag names: + ARTIST, TITLE, DATE, ALBUM, and so on. + + +Two new comment tags are introduced here: + + +First, an optional gain for track normalization: +
+ +
+ + representing the volume shift needed to normalize the track's volume + during isolated playback, in random shuffle, and so on. +The gain is a Q7.8 fixed point number in dB, as in the ID header's 'output + gain' field. +This tag is similar to the REPLAYGAIN_TRACK_GAIN tag in + Vorbis , except that the normal volume + reference is the standard. + +Second, an optional gain for album normalization: +
+ +
+ + representing the volume shift needed to normalize the overall volume when + played as part of a particular collection of tracks. +The gain is also a Q7.8 fixed point number in dB, as in the ID header's + 'output gain' field. +The values '-573' and '111' given here are just examples. + + +An Ogg Opus stream MUST NOT have more than one of each of these tags, and if + present their values MUST be an integer from -32768 to 32767, inclusive, + represented in ASCII as a base 10 number with no whitespace. +A leading '+' or '-' character is valid. +Leading zeros are also permitted, but the value MUST be represented by + no more than 6 characters. +Other non-digit characters MUST NOT be present. + + +If present, R128_TRACK_GAIN and R128_ALBUM_GAIN MUST correctly represent + the R128 normalization gain relative to the 'output gain' field specified + in the ID header. +If a player chooses to make use of the R128_TRACK_GAIN tag or the + R128_ALBUM_GAIN tag, it MUST apply those gains + in addition to the 'output gain' value. +If a tool modifies the ID header's 'output gain' field, it MUST also update or + remove the R128_TRACK_GAIN and R128_ALBUM_GAIN comment tags if present. +A muxer SHOULD place the gain it wants other tools to use by default into the + 'output gain' field, and not the comment tag. + + +To avoid confusion with multiple normalization schemes, an Opus comment header + SHOULD NOT contain any of the REPLAYGAIN_TRACK_GAIN, REPLAYGAIN_TRACK_PEAK, + REPLAYGAIN_ALBUM_GAIN, or REPLAYGAIN_ALBUM_PEAK tags, unless they are only + to be used in some context where there is guaranteed to be no such confusion. + normalization is preferred to the earlier + REPLAYGAIN schemes because of its clear definition and adoption by industry. +Peak normalizations are difficult to calculate reliably for lossy codecs + because of variation in excursion heights due to decoder differences. +In the authors' investigations they were not applied consistently or broadly + enough to merit inclusion here. + +
+
+ +
+ +
+ +Technically, valid Opus packets can be arbitrarily large due to the padding + format, although the amount of non-padding data they can contain is bounded. +These packets might be spread over a similarly enormous number of Ogg pages. +When encoding, implementations SHOULD limit the use of padding in audio data + packets to no more than is necessary to make a variable bitrate (VBR) stream + constant bitrate (CBR), unless they have no reasonable way to determine what + is necessary. +Demuxers SHOULD treat audio data packets as invalid (treat them as if they were + malformed Opus packets with an invalid TOC sequence) if they are larger than + 61,440 octets per Opus stream, unless they have a specific reason for + allowing extra padding. +Such packets necessarily contain more padding than needed to make a stream CBR. +Demuxers MUST avoid attempting to allocate excessive amounts of memory when + presented with a very large packet. +Demuxers MAY treat audio data packets as invalid or partially process them if + they are larger than 61,440 octets in an Ogg Opus stream with channel + mapping families 0 or 1. +Demuxers MAY treat audio data packets as invalid or partially process them in + any Ogg Opus stream if the packet is larger than 61,440 octets and also + larger than 7,680 octets per Opus stream. +The presence of an extremely large packet in the stream could indicate a + memory exhaustion attack or stream corruption. + + +In an Ogg Opus stream, the largest possible valid packet that does not use + padding has a size of (61,298*N - 2) octets. +With 255 streams, this is 15,630,988 octets and can + span up to 61,298 Ogg pages, all but one of which will have a granule + position of -1. +This is of course a very extreme packet, consisting of 255 streams, each + containing 120 ms of audio encoded as 2.5 ms frames, each frame + using the maximum possible number of octets (1275) and stored in the least + efficient manner allowed (a VBR code 3 Opus packet). +Even in such a packet, most of the data will be zeros as 2.5 ms frames + cannot actually use all 1275 octets. + + +The largest packet consisting of entirely useful data is + (15,326*N - 2) octets. +This corresponds to 120 ms of audio encoded as 10 ms frames in either + SILK or Hybrid mode, but at a data rate of over 1 Mbps, which makes little + sense for the quality achieved. + + +A more reasonable limit is (7,664*N - 2) octets. +This corresponds to 120 ms of audio encoded as 20 ms stereo CELT mode + frames, with a total bitrate just under 511 kbps (not counting the Ogg + encapsulation overhead). +For channel mapping family 1, N=8 provides a reasonable upper bound, as it + allows for each of the 8 possible output channels to be decoded from a + separate stereo Opus stream. +This gives a size of 61,310 octets, which is rounded up to a multiple of + 1,024 octets to yield the audio data packet size of 61,440 octets + that any implementation is expected to be able to process successfully. + +
+ +
+ +When encoding Opus streams, Ogg muxers SHOULD take into account the + algorithmic delay of the Opus encoder. + + +In encoders derived from the reference + implementation , the number of samples can be + queried with: + +
+ +
+ +To achieve good quality in the very first samples of a stream, implementations + MAY use linear predictive coding (LPC) extrapolation to generate at least 120 + extra samples at the beginning to avoid the Opus encoder having to encode a + discontinuous signal. +For more information on linear prediction, see + . +For an input file containing 'length' samples, the implementation SHOULD set + the pre-skip header value to (delay_samples + extra_samples), encode + at least (length + delay_samples + extra_samples) + samples, and set the granule position of the last page to + (length + delay_samples + extra_samples). +This ensures that the encoded file has the same duration as the original, with + no time offset. The best way to pad the end of the stream is to also use LPC + extrapolation, but zero-padding is also acceptable. + + +
+ +The first step in LPC extrapolation is to compute linear prediction + coefficients. +When extending the end of the signal, order-N (typically with N ranging from 8 + to 40) LPC analysis is performed on a window near the end of the signal. +The last N samples are used as memory to an infinite impulse response (IIR) + filter. + + +The filter is then applied on a zero input to extrapolate the end of the signal. +Let a(k) be the kth LPC coefficient and x(n) be the nth sample of the signal, + each new sample past the end of the signal is computed as: + +
+ +
+ +The process is repeated independently for each channel. +It is possible to extend the beginning of the signal by applying the same + process backward in time. +When extending the beginning of the signal, it is best to apply a "fade in" to + the extrapolated signal, e.g. by multiplying it by a half-Hanning window + . + + +
+ +
+ +In some applications, such as Internet radio, it is desirable to cut a long + stream into smaller chains, e.g. so the comment header can be updated. +This can be done simply by separating the input streams into segments and + encoding each segment independently. +The drawback of this approach is that it creates a small discontinuity + at the boundary due to the lossy nature of Opus. +A muxer MAY avoid this discontinuity by using the following procedure: + +Encode the last frame of the first segment as an independent frame by + turning off all forms of inter-frame prediction. +De-emphasis is allowed. +Set the granule position of the last page to a point near the end of the + last frame. +Begin the second segment with a copy of the last frame of the first + segment. +Set the pre-skip value of the second stream in such a way as to properly + join the two streams. +Continue the encoding process normally from there, without any reset to + the encoder. + + + +In encoders derived from the reference implementation, inter-frame prediction + can be turned off by calling: + +
+ +
+ +For best results, this implementation requires that prediction be explicitly + enabled again before resuming normal encoding, even after a reset. + + +
+ +
+ +
+ +A brief summary of major implementations of this draft is available + at , + along with their status. + + +[Note to RFC Editor: please remove this entire section before + final publication per , along with + its references.] + +
+ +
+ +Implementations of the Opus codec need to take appropriate security + considerations into account, as outlined in . +This is just as much a problem for the container as it is for the codec itself. +Malicious payloads and/or input streams can be used to attack codec + implementations. +Implementations MUST NOT overrun their allocated memory nor consume excessive + resources when decoding payloads or processing input streams. +Although problems in encoding applications are typically rarer, this still + applies to a muxer, as vulnerabilities would allow an attacker to attack + transcoding gateways. + + + +Header parsing code contains the most likely area for potential overruns. +It is important for implementations to ensure their buffers contain enough + data for all of the required fields before attempting to read it (for example, + for all of the channel map data in the ID header). +Implementations would do well to validate the indices of the channel map, also, + to ensure they meet all of the restrictions outlined in + , in order to avoid attempting to read data + from channels that do not exist. + + + +To avoid excessive resource usage, we advise implementations to be especially + wary of streams that might cause them to process far more data than was + actually transmitted. +For example, a relatively small comment header may contain values for the + string lengths or user comment list length that imply that it is many + gigabytes in size. +Even computing the size of the required buffer could overflow a 32-bit integer, + and actually attempting to allocate such a buffer before verifying it would be + a reasonable size is a bad idea. +After reading the user comment list length, implementations might wish to + verify that the header contains at least the minimum amount of data for that + many comments (4 additional octets per comment, to indicate each has a + length of zero) before proceeding any further, again taking care to avoid + overflow in these calculations. +If allocating an array of pointers to point at these strings, the size of the + pointers may be larger than 4 octets, potentially requiring a separate + overflow check. + + + +Another bug in this class we have observed more than once involves the handling + of invalid data at the end of a stream. +Often, implementations will seek to the end of a stream to locate the last + timestamp in order to compute its total duration. +If they do not find a valid capture pattern and Ogg page from the desired + logical stream, they will back up and try again. +If care is not taken to avoid re-scanning data that was already scanned, this + search can quickly devolve into something with a complexity that is quadratic + in the amount of invalid data. + + + +In general when seeking, implementations will wish to be cautious about the + effects of invalid granule position values, and ensure all algorithms will + continue to make progress and eventually terminate, even if these are missing + or out-of-order. + + + +Like most other container formats, Ogg Opus streams SHOULD NOT be used with + insecure ciphers or cipher modes that are vulnerable to known-plaintext + attacks. +Elements such as the Ogg page capture pattern and the magic signatures in the + ID header and the comment header all have easily predictable values, in + addition to various elements of the codec data itself. + +
+ +
+ +An "Ogg Opus file" consists of one or more sequentially multiplexed segments, + each containing exactly one Ogg Opus stream. +The RECOMMENDED mime-type for Ogg Opus files is "audio/ogg". + + + +If more specificity is desired, one MAY indicate the presence of Opus streams + using the codecs parameter defined in and + , e.g., + +
+ +
+ + for an Ogg Opus file. + + + +The RECOMMENDED filename extension for Ogg Opus files is '.opus'. + + + +When Opus is concurrently multiplexed with other streams in an Ogg container, + one SHOULD use one of the "audio/ogg", "video/ogg", or "application/ogg" + mime-types, as defined in . +Such streams are not strictly "Ogg Opus files" as described above, + since they contain more than a single Opus stream per sequentially + multiplexed segment, e.g. video or multiple audio tracks. +In such cases the the '.opus' filename extension is NOT RECOMMENDED. + + + +In either case, this document updates + to add 'opus' as a codecs parameter value with char[8]: 'OpusHead' + as Codec Identifier. + +
+ +
+ +This document updates the IANA Media Types registry to add .opus + as a file extension for "audio/ogg", and to add itself as a reference + alongside for "audio/ogg", "video/ogg", and + "application/ogg" Media Types. + + +This document defines a new registry "Opus Channel Mapping Families" to + indicate how the semantic meanings of the channels in a multi-channel Opus + stream are described. +IANA is requested to create a new name space of "Opus Channel Mapping + Families". +This will be a new registry on the IANA Matrix, and not a subregistry of an + existing registry. +Modifications to this registry follow the "Specification Required" registration + policy as defined in . +Each registry entry consists of a Channel Mapping Family Number, which is + specified in decimal in the range 0 to 255, inclusive, and a Reference (or + list of references) +Each Reference must point to sufficient documentation to describe what + information is coded in the Opus identification header for this channel + mapping family, how a demuxer determines the Stream Count ('N') and Coupled + Stream Count ('M') from this information, and how it determines the proper + interpretation of each of the decoded channels. + + +This document defines three initial assignments for this registry. + + +ValueReference +0[RFCXXXX] +1[RFCXXXX] +255[RFCXXXX] + + +The designated expert will determine if the Reference points to a specification + that meets the requirements for permanence and ready availability laid out + in  and that it specifies the information + described above with sufficient clarity to allow interoperable + implementations. + +
+ +
+ +Thanks to Ben Campbell, Joel M. Halpern, Mark Harris, Greg Maxwell, + Christopher "Monty" Montgomery, Jean-Marc Valin, Stephan Wenger, and Mo Zanaty + for their valuable contributions to this document. +Additional thanks to Andrew D'Addesio, Greg Maxwell, and Vincent Penquerc'h for + their feedback based on early implementations. + +
+ +
+ +In , "RFCXXXX" is to be replaced with the RFC number + assigned to this draft. + +
+ + + + + &rfc2119; + &rfc3533; + &rfc3629; + &rfc5226; + &rfc5334; + &rfc6381; + &rfc6716; + + + + Loudness Recommendation EBU R128 + + EBU Technical Committee + + + + + + + +Ogg Vorbis I Format Specification: Comment Field and Header + Specification + + + + + + + + + + + &rfc4732; + &rfc6982; + &rfc7587; + + + + FLAC - Free Lossless Audio Codec Format Description + + + + + + + + Hann window + + Wikipedia + + + + + + + + Linear Predictive Coding + + Wikipedia + + + + + + + + Autocorrelation LPC coeff generation algorithm + (Vorbis source code) + + + + + + + + +Q (number format) +Wikipedia + + + + + + +VorbisComment: Replay Gain + + + + + + + + +Granulepos Encoding and How Seeking Really Works + + + + + + + + + +The Vorbis I Specification, Section 4.3.9 Output Channel Order + + + + + + + + The Vorbis I Specification, Appendix A: Embedding Vorbis + into an Ogg stream + + + + + + + + Multiple Channel Audio Data and WAVE Files + + Microsoft Corporation + + + + + + + + + diff --git a/lib/opus/doc/draft-ietf-codec-opus-update.xml b/lib/opus/doc/draft-ietf-codec-opus-update.xml new file mode 100644 index 00000000..3124e22c --- /dev/null +++ b/lib/opus/doc/draft-ietf-codec-opus-update.xml @@ -0,0 +1,513 @@ + + + + + + + + + + + + + + + Updates to the Opus Audio Codec + + +Mozilla Corporation +
+ +331 E. Evelyn Avenue +Mountain View +CA +94041 +USA + ++1 650 903-0800 +jmvalin@jmvalin.ca +
+ + + +vocTone +
+ + + + + + + + +koenvos74@gmail.com +
+ + + + + + + + This document addresses minor issues that were found in the specification + of the Opus audio codec in RFC 6716. It updates the normative decoder implementation + included in the appendix of RFC 6716. The changes fixes real and potential security-related + issues, as well minor quality-related issues. + + + + +
+ This document addresses minor issues that were discovered in the reference + implementation of the Opus codec. Unlike most IETF specifications, Opus is defined + in RFC 6716 in terms of a normative reference + decoder implementation rather than from the associated text description. + That RFC includes the reference decoder implementation as Appendix A. + That's why only issues affecting the decoder are + listed here. An up-to-date implementation of the Opus encoder can be found at + . + + Some of the changes in this document update normative behaviour in a way that requires + new test vectors. The English text of the specification is unaffected, only + the C implementation is. The updated specification remains fully compatible with + the original specification. + + + + Note: due to RFC formatting conventions, lines exceeding the column width + in the patch are split using a backslash character. The backslashes + at the end of a line and the white space at the beginning + of the following line are not part of the patch. A properly formatted patch + including all changes is available at + + and has a SHA-1 hash of 029e3aa88fc342c91e67a21e7bfbc9458661cd5f. + + +
+ +
+ The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", + "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this + document are to be interpreted as described in RFC 2119. +
+ +
+ The reference implementation does not reinitialize the stereo state + during a mode switch. The old stereo memory can produce a brief impulse + (i.e. single sample) in the decoded audio. This can be fixed by changing + silk/dec_API.c at line 72: + +
+ + for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { + ret = silk_init_decoder( &channel_state[ n ] ); + } ++ silk_memset(&((silk_decoder *)decState)->sStereo, 0, ++ sizeof(((silk_decoder *)decState)->sStereo)); ++ /* Not strictly needed, but it's cleaner that way */ ++ ((silk_decoder *)decState)->prev_decode_only_middle = 0; + + return ret; + } + +]]> +
+ + This change affects the normative output of the decoder, but the + amount of change is within the tolerance and too small to make the testvector check fail. + +
+ +
+ It was discovered that some invalid packets of very large size could trigger + an out-of-bounds read in the Opus packet parsing code responsible for padding. + This is due to an integer overflow if the signaled padding exceeds 2^31-1 bytes + (the actual packet may be smaller). The code can be fixed by decrementing the + (signed) len value, instead of incrementing a separate padding counter. + This is done by applying the following changes at line 596 of src/opus_decoder.c: + +
+ + /* Padding flag is bit 6 */ + if (ch&0x40) + { +- int padding=0; + int p; + do { + if (len<=0) + return OPUS_INVALID_PACKET; + p = *data++; + len--; +- padding += p==255 ? 254: p; ++ len -= p==255 ? 254: p; + } while (p==255); +- len -= padding; + } + +]]> +
+ This packet parsing issue is limited to reading memory up + to about 60 kB beyond the compressed buffer. This can only be triggered + by a compressed packet more than about 16 MB long, so it's not a problem + for RTP. In theory, it could crash a file + decoder (e.g. Opus in Ogg) if the memory just after the incoming packet + is out-of-range, but our attempts to trigger such a crash in a production + application built using an affected version of the Opus decoder failed. +
+ +
+ The SILK resampler had the following issues: + + The calls to memcpy() were using sizeof(opus_int32), but the type of the + local buffer was opus_int16. + Because the size was wrong, this potentially allowed the source + and destination regions of the memcpy() to overlap on the copy from "buf" to "buf". + We believe that nSamplesIn (number of input samples) is at least fs_in_khZ (sampling rate in kHz), + which is at least 8. + Since RESAMPLER_ORDER_FIR_12 is only 8, that should not be a problem once + the type size is fixed. + The size of the buffer used RESAMPLER_MAX_BATCH_SIZE_IN, but the + data stored in it was actually twice the input batch size + (nSamplesIn<<1). + + The code can be fixed by applying the following changes to line 78 of silk/resampler_private_IIR_FIR.c: + +
+ + ) + { + silk_resampler_state_struct *S = \ +(silk_resampler_state_struct *)SS; + opus_int32 nSamplesIn; + opus_int32 max_index_Q16, index_increment_Q16; +- opus_int16 buf[ RESAMPLER_MAX_BATCH_SIZE_IN + \ +RESAMPLER_ORDER_FIR_12 ]; ++ opus_int16 buf[ 2*RESAMPLER_MAX_BATCH_SIZE_IN + \ +RESAMPLER_ORDER_FIR_12 ]; + + /* Copy buffered samples to start of buffer */ +- silk_memcpy( buf, S->sFIR, RESAMPLER_ORDER_FIR_12 \ +* sizeof( opus_int32 ) ); ++ silk_memcpy( buf, S->sFIR, RESAMPLER_ORDER_FIR_12 \ +* sizeof( opus_int16 ) ); + + /* Iterate over blocks of frameSizeIn input samples */ + index_increment_Q16 = S->invRatio_Q16; + while( 1 ) { + nSamplesIn = silk_min( inLen, S->batchSize ); + + /* Upsample 2x */ + silk_resampler_private_up2_HQ( S->sIIR, &buf[ \ +RESAMPLER_ORDER_FIR_12 ], in, nSamplesIn ); + + max_index_Q16 = silk_LSHIFT32( nSamplesIn, 16 + 1 \ +); /* + 1 because 2x upsampling */ + out = silk_resampler_private_IIR_FIR_INTERPOL( out, \ +buf, max_index_Q16, index_increment_Q16 ); + in += nSamplesIn; + inLen -= nSamplesIn; + + if( inLen > 0 ) { + /* More iterations to do; copy last part of \ +filtered signal to beginning of buffer */ +- silk_memcpy( buf, &buf[ nSamplesIn << 1 ], \ +RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); ++ silk_memmove( buf, &buf[ nSamplesIn << 1 ], \ +RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + } else { + break; + } + } + + /* Copy last part of filtered signal to the state for \ +the next call */ +- silk_memcpy( S->sFIR, &buf[ nSamplesIn << 1 ], \ +RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); ++ silk_memcpy( S->sFIR, &buf[ nSamplesIn << 1 ], \ +RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + } + +]]> +
+
+ +
+ + It was discovered through decoder fuzzing that some bitstreams could produce + integer values exceeding 32-bits in LPC_inverse_pred_gain_QA(), causing + a wrap-around. The C standard considers + this behavior as undefined. The following patch to line 87 of silk/LPC_inv_pred_gain.c + detects values that do not fit in a 32-bit integer and considers the corresponding filters unstable: + +
+ + /* Update AR coefficient */ + for( n = 0; n < k; n++ ) { +- tmp_QA = Aold_QA[ n ] - MUL32_FRAC_Q( \ +Aold_QA[ k - n - 1 ], rc_Q31, 31 ); +- Anew_QA[ n ] = MUL32_FRAC_Q( tmp_QA, rc_mult2 , mult2Q ); ++ opus_int64 tmp64; ++ tmp_QA = silk_SUB_SAT32( Aold_QA[ n ], MUL32_FRAC_Q( \ +Aold_QA[ k - n - 1 ], rc_Q31, 31 ) ); ++ tmp64 = silk_RSHIFT_ROUND64( silk_SMULL( tmp_QA, \ +rc_mult2 ), mult2Q); ++ if( tmp64 > silk_int32_MAX || tmp64 < silk_int32_MIN ) { ++ return 0; ++ } ++ Anew_QA[ n ] = ( opus_int32 )tmp64; + } + +]]> +
+
+ +
+ + It was discovered -- also from decoder fuzzing -- that an integer wrap-around could + occur when decoding bitstreams with extremely large values for the high LSF parameters. + The end result of the wrap-around is an illegal read access on the stack, which + the authors do not believe is exploitable but should nonetheless be fixed. The following + patch to line 137 of silk/NLSF_stabilize.c prevents the problem: + +
+ + /* Keep delta_min distance between the NLSFs */ + for( i = 1; i < L; i++ ) +- NLSF_Q15[i] = silk_max_int( NLSF_Q15[i], \ +NLSF_Q15[i-1] + NDeltaMin_Q15[i] ); ++ NLSF_Q15[i] = silk_max_int( NLSF_Q15[i], \ +silk_ADD_SAT16( NLSF_Q15[i-1], NDeltaMin_Q15[i] ) ); + + /* Last NLSF should be no higher than 1 - NDeltaMin[L] */ + +]]> +
+ +
+ +
+ On extreme bit-streams, it is possible for log-domain band energy levels + to exceed the maximum single-precision floating point value once converted + to a linear scale. This would later cause the decoded values to be NaN (not a number), + possibly causing problems in the software using the PCM values. This can be + avoided with the following patch to line 552 of celt/quant_bands.c: + +
+ + { + opus_val16 lg = ADD16(oldEBands[i+c*m->nbEBands], + SHL16((opus_val16)eMeans[i],6)); ++ lg = MIN32(QCONST32(32.f, 16), lg); + eBands[i+c*m->nbEBands] = PSHR32(celt_exp2(lg),4); + } + for (;inbEBands;i++) + +]]> +
+
+ +
+ When encoding in hybrid mode at low bitrate, we sometimes only have + enough bits to code a single CELT band (8 - 9.6 kHz). When that happens, + the second band (CELT band 18, from 9.6 to 12 kHz) cannot use folding + because it is wider than the amount already coded, and falls back to + white noise. Because it can also happen on transients (e.g. stops), it + can cause audible pre-echo. + + + To address the issue, we change the folding behavior so that it is + never forced to fall back to LCG due to the first band not containing + enough coefficients to fold onto the second band. This + is achieved by simply repeating part of the first band in the folding + of the second band. This changes the code in celt/bands.c around line 1237: + +
+ + b = 0; + } + +- if (resynth && M*eBands[i]-N >= M*eBands[start] && \ +(update_lowband || lowband_offset==0)) ++ if (resynth && (M*eBands[i]-N >= M*eBands[start] || \ +i==start+1) && (update_lowband || lowband_offset==0)) + lowband_offset = i; + ++ if (i == start+1) ++ { ++ int n1, n2; ++ int offset; ++ n1 = M*(eBands[start+1]-eBands[start]); ++ n2 = M*(eBands[start+2]-eBands[start+1]); ++ offset = M*eBands[start]; ++ /* Duplicate enough of the first band folding data to \ +be able to fold the second band. ++ Copies no data for CELT-only mode. */ ++ OPUS_COPY(&norm[offset+n1], &norm[offset+2*n1 - n2], n2-n1); ++ if (C==2) ++ OPUS_COPY(&norm2[offset+n1], &norm2[offset+2*n1 - n2], \ +n2-n1); ++ } ++ + tf_change = tf_res[i]; + if (i>=m->effEBands) + { + +]]> +
+ + + as well as line 1260: + + +
+ + fold_start = lowband_offset; + while(M*eBands[--fold_start] > effective_lowband); + fold_end = lowband_offset-1; +- while(M*eBands[++fold_end] < effective_lowband+N); ++ while(++fold_end < i && M*eBands[fold_end] < \ +effective_lowband+N); + x_cm = y_cm = 0; + fold_i = fold_start; do { + x_cm |= collapse_masks[fold_i*C+0]; + + +]]> +
+ + The fix does not impact compatibility, because the improvement does + not depend on the encoder doing anything special. There is also no + reasonable way for an encoder to use the original behavior to + improve quality over the proposed change. + +
+ +
+ The last issue is not strictly a bug, but it is an issue that has been reported + when downmixing an Opus decoded stream to mono, whether this is done inside the decoder + or as a post-processing step on the stereo decoder output. Opus intensity stereo allows + optionally coding the two channels 180-degrees out of phase on a per-band basis. + This provides better stereo quality than forcing the two channels to be in phase, + but when the output is downmixed to mono, the energy in the affected bands is cancelled + sometimes resulting in audible artifacts. + + As a work-around for this issue, the decoder MAY choose not to apply the 180-degree + phase shift. This can be useful when downmixing to mono inside or + outside of the decoder (e.g. user-controllable). + +
+ + +
+ Changes in and have + sufficient impact on the testvectors to make them fail. For this reason, + this document also updates the Opus test vectors. The new test vectors now + include two decoded outputs for the same bitstream. The outputs with + suffix 'm' do not apply the CELT 180-degree phase shift as allowed in + , while the outputs without the suffix do. An + implementation is compliant as long as it passes either set of vectors. + + + Any Opus implementation + that passes either the original test vectors from RFC 6716 + or one of the new sets of test vectors is compliant with the Opus specification. However, newer implementations + SHOULD be based on the new test vectors rather than the old ones. + + The new test vectors are located at + . + The SHA-1 hashes of the test vectors are: +
+ + + +
+ Note that the decoder input bitstream files (.bit) are unchanged. + +
+ +
+ This document fixes two security issues reported on Opus and that affect the + reference implementation in RFC 6716: CVE-2013-0899 + + and CVE-2017-0381 . + CVE- 2013-0899 theoretically could have caused an information leak. The leaked + information would have gone through the decoder process before being accessible + to the attacker. It is fixed by . + CVE-2017-0381 could have resulted in a 16-bit out-of-bounds read from a fixed + location. It is fixed in . + Beyond the two fixed CVEs, this document adds no new security considerations on top of + RFC 6716. + +
+ +
+ This document makes no request of IANA. + + Note to RFC Editor: this section may be removed on publication as an + RFC. +
+ +
+ We would like to thank Juri Aedla for reporting the issue with the parsing of + the Opus padding. Thanks to Felicia Lim for reporting the LSF integer overflow issue. + Also, thanks to Tina le Grand, Jonathan Lennox, and Mark Harris for their + feedback on this document. +
+ + + + + + + + + + + diff --git a/lib/opus/doc/draft-ietf-codec-opus.xml b/lib/opus/doc/draft-ietf-codec-opus.xml new file mode 100644 index 00000000..334cad97 --- /dev/null +++ b/lib/opus/doc/draft-ietf-codec-opus.xml @@ -0,0 +1,8276 @@ + + + + + + + +Definition of the Opus Audio Codec + + + +Mozilla Corporation +
+ +650 Castro Street +Mountain View +CA +94041 +USA + ++1 650 903-0800 +jmvalin@jmvalin.ca +
+ + + +Skype Technologies S.A. +
+ +Soder Malarstrand 43 +Stockholm + +11825 +SE + ++46 73 085 7619 +koen.vos@skype.net +
+ + + +Mozilla Corporation +
+ +650 Castro Street +Mountain View +CA +94041 +USA + ++1 650 903-0800 +tterriberry@mozilla.com +
+ + + + +General + + + + + +This document defines the Opus interactive speech and audio codec. +Opus is designed to handle a wide range of interactive audio applications, + including Voice over IP, videoconferencing, in-game chat, and even live, + distributed music performances. +It scales from low bitrate narrowband speech at 6 kb/s to very high quality + stereo music at 510 kb/s. +Opus uses both linear prediction (LP) and the Modified Discrete Cosine + Transform (MDCT) to achieve good compression of both speech and music. + + + + + + +
+ +The Opus codec is a real-time interactive audio codec designed to meet the requirements +described in . +It is composed of a linear + prediction (LP)-based layer and a Modified Discrete Cosine Transform + (MDCT)-based layer. +The main idea behind using two layers is that in speech, linear prediction + techniques (such as Code-Excited Linear Prediction, or CELP) code low frequencies more efficiently than transform + (e.g., MDCT) domain techniques, while the situation is reversed for music and + higher speech frequencies. +Thus a codec with both layers available can operate over a wider range than + either one alone and, by combining them, achieve better quality than either + one individually. + + + +The primary normative part of this specification is provided by the source code + in . +Only the decoder portion of this software is normative, though a + significant amount of code is shared by both the encoder and decoder. + provides a decoder conformance test. +The decoder contains a great deal of integer and fixed-point arithmetic which + needs to be performed exactly, including all rounding considerations, so any + useful specification requires domain-specific symbolic language to adequately + define these operations. +Additionally, any +conflict between the symbolic representation and the included reference +implementation must be resolved. For the practical reasons of compatibility and +testability it would be advantageous to give the reference implementation +priority in any disagreement. The C language is also one of the most +widely understood human-readable symbolic representations for machine +behavior. +For these reasons this RFC uses the reference implementation as the sole + symbolic representation of the codec. + + +While the symbolic representation is unambiguous and complete it is not +always the easiest way to understand the codec's operation. For this reason +this document also describes significant parts of the codec in English and +takes the opportunity to explain the rationale behind many of the more +surprising elements of the design. These descriptions are intended to be +accurate and informative, but the limitations of common English sometimes +result in ambiguity, so it is expected that the reader will always read +them alongside the symbolic representation. Numerous references to the +implementation are provided for this purpose. The descriptions sometimes +differ from the reference in ordering or through mathematical simplification +wherever such deviation makes an explanation easier to understand. +For example, the right shift and left shift operations in the reference +implementation are often described using division and multiplication in the text. +In general, the text is focused on the "what" and "why" while the symbolic +representation most clearly provides the "how". + + +
+ +The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", + "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be + interpreted as described in RFC 2119 . + + +Various operations in the codec require bit-exact fixed-point behavior, even + when writing a floating point implementation. +The notation "Q<n>", where n is an integer, denotes the number of binary + digits to the right of the decimal point in a fixed-point number. +For example, a signed Q14 value in a 16-bit word can represent values from + -2.0 to 1.99993896484375, inclusive. +This notation is for informational purposes only. +Arithmetic, when described, always operates on the underlying integer. +E.g., the text will explicitly indicate any shifts required after a + multiplication. + + +Expressions, where included in the text, follow C operator rules and + precedence, with the exception that the syntax "x**y" indicates x raised to + the power y. +The text also makes use of the following functions: + + +
+ +The smallest of two values x and y. + +
+ +
+ +The largest of two values x and y. + +
+ +
+
+ +
+ +With this definition, if lo > hi, the lower bound is the one that + is enforced. + +
+ +
+ +The sign of x, i.e., +
+ 0 . +]]> +
+ +
+ +
+ +The absolute value of x, i.e., +
+ +
+ +
+ +
+ +The largest integer z such that z <= f. + +
+ +
+ +The smallest integer z such that z >= f. + +
+ +
+ +The integer z nearest to f, with ties rounded towards negative infinity, + i.e., +
+ +
+ +
+ +
+ +The base-two logarithm of f. + +
+ +
+ +The minimum number of bits required to store a positive integer n in two's + complement notation, or 0 for a non-positive integer n. +
+ 0 +]]> +
+Examples: + +ilog(-1) = 0 +ilog(0) = 0 +ilog(1) = 1 +ilog(2) = 2 +ilog(3) = 2 +ilog(4) = 3 +ilog(7) = 3 + + +
+ +
+ +
+ +
+ + +The Opus codec scales from 6 kb/s narrowband mono speech to 510 kb/s + fullband stereo music, with algorithmic delays ranging from 5 ms to + 65.2 ms. +At any given time, either the LP layer, the MDCT layer, or both, may be active. +It can seamlessly switch between all of its various operating modes, giving it + a great deal of flexibility to adapt to varying content and network + conditions without renegotiating the current session. +The codec allows input and output of various audio bandwidths, defined as + follows: + + +Abbreviation +Audio Bandwidth +Sample Rate (Effective) +NB (narrowband) 4 kHz 8 kHz +MB (medium-band) 6 kHz 12 kHz +WB (wideband) 8 kHz 16 kHz +SWB (super-wideband) 12 kHz 24 kHz +FB (fullband) 20 kHz (*) 48 kHz + + +(*) Although the sampling theorem allows a bandwidth as large as half the + sampling rate, Opus never codes audio above 20 kHz, as that is the + generally accepted upper limit of human hearing. + + + +Opus defines super-wideband (SWB) with an effective sample rate of 24 kHz, + unlike some other audio coding standards that use 32 kHz. +This was chosen for a number of reasons. +The band layout in the MDCT layer naturally allows skipping coefficients for + frequencies over 12 kHz, but does not allow cleanly dropping just those + frequencies over 16 kHz. +A sample rate of 24 kHz also makes resampling in the MDCT layer easier, + as 24 evenly divides 48, and when 24 kHz is sufficient, it can save + computation in other processing, such as Acoustic Echo Cancellation (AEC). +Experimental changes to the band layout to allow a 16 kHz cutoff + (32 kHz effective sample rate) showed potential quality degradations at + other sample rates, and at typical bitrates the number of bits saved by using + such a cutoff instead of coding in fullband (FB) mode is very small. +Therefore, if an application wishes to process a signal sampled at 32 kHz, + it should just use FB. + + + +The LP layer is based on the SILK codec + . +It supports NB, MB, or WB audio and frame sizes from 10 ms to 60 ms, + and requires an additional 5 ms look-ahead for noise shaping estimation. +A small additional delay (up to 1.5 ms) may be required for sampling rate + conversion. +Like Vorbis and many other modern codecs, SILK is inherently designed for + variable-bitrate (VBR) coding, though the encoder can also produce + constant-bitrate (CBR) streams. +The version of SILK used in Opus is substantially modified from, and not + compatible with, the stand-alone SILK codec previously deployed by Skype. +This document does not serve to define that format, but those interested in the + original SILK codec should see instead. + + + +The MDCT layer is based on the CELT codec . +It supports NB, WB, SWB, or FB audio and frame sizes from 2.5 ms to + 20 ms, and requires an additional 2.5 ms look-ahead due to the + overlapping MDCT windows. +The CELT codec is inherently designed for CBR coding, but unlike many CBR + codecs it is not limited to a set of predetermined rates. +It internally allocates bits to exactly fill any given target budget, and an + encoder can produce a VBR stream by varying the target on a per-frame basis. +The MDCT layer is not used for speech when the audio bandwidth is WB or less, + as it is not useful there. +On the other hand, non-speech signals are not always adequately coded using + linear prediction, so for music only the MDCT layer should be used. + + + +A "Hybrid" mode allows the use of both layers simultaneously with a frame size + of 10 or 20 ms and a SWB or FB audio bandwidth. +The LP layer codes the low frequencies by resampling the signal down to WB. +The MDCT layer follows, coding the high frequency portion of the signal. +The cutoff between the two lies at 8 kHz, the maximum WB audio bandwidth. +In the MDCT layer, all bands below 8 kHz are discarded, so there is no + coding redundancy between the two layers. + + + +The sample rate (in contrast to the actual audio bandwidth) can be chosen + independently on the encoder and decoder side, e.g., a fullband signal can be + decoded as wideband, or vice versa. +This approach ensures a sender and receiver can always interoperate, regardless + of the capabilities of their actual audio hardware. +Internally, the LP layer always operates at a sample rate of twice the audio + bandwidth, up to a maximum of 16 kHz, which it continues to use for SWB + and FB. +The decoder simply resamples its output to support different sample rates. +The MDCT layer always operates internally at a sample rate of 48 kHz. +Since all the supported sample rates evenly divide this rate, and since the + the decoder may easily zero out the high frequency portion of the spectrum in + the frequency domain, it can simply decimate the MDCT layer output to achieve + the other supported sample rates very cheaply. + + + +After conversion to the common, desired output sample rate, the decoder simply + adds the output from the two layers together. +To compensate for the different look-ahead required by each layer, the CELT + encoder input is delayed by an additional 2.7 ms. +This ensures that low frequencies and high frequencies arrive at the same time. +This extra delay may be reduced by an encoder by using less look-ahead for noise + shaping or using a simpler resampler in the LP layer, but this will reduce + quality. +However, the base 2.5 ms look-ahead in the CELT layer cannot be reduced in + the encoder because it is needed for the MDCT overlap, whose size is fixed by + the decoder. + + + +Both layers use the same entropy coder, avoiding any waste from "padding bits" + between them. +The hybrid approach makes it easy to support both CBR and VBR coding. +Although the LP layer is VBR, the bit allocation of the MDCT layer can produce + a final stream that is CBR by using all the bits left unused by the LP layer. + + +
+ +The Opus codec includes a number of control parameters which can be changed dynamically during +regular operation of the codec, without interrupting the audio stream from the encoder to the decoder. +These parameters only affect the encoder since any impact they have on the bit-stream is signaled +in-band such that a decoder can decode any Opus stream without any out-of-band signaling. Any Opus +implementation can add or modify these control parameters without affecting interoperability. The most +important encoder control parameters in the reference encoder are listed below. + + +
+ +Opus supports all bitrates from 6 kb/s to 510 kb/s. All other parameters being +equal, higher bitrate results in higher quality. For a frame size of 20 ms, these +are the bitrate "sweet spots" for Opus in various configurations: + +8-12 kb/s for NB speech, +16-20 kb/s for WB speech, +28-40 kb/s for FB speech, +48-64 kb/s for FB mono music, and +64-128 kb/s for FB stereo music. + + +
+ +
+ +Opus can transmit either mono or stereo frames within a single stream. +When decoding a mono frame in a stereo decoder, the left and right channels are + identical, and when decoding a stereo frame in a mono decoder, the mono output + is the average of the left and right channels. +In some cases, it is desirable to encode a stereo input stream in mono (e.g., + because the bitrate is too low to encode stereo with sufficient quality). +The number of channels encoded can be selected in real-time, but by default the + reference encoder attempts to make the best decision possible given the + current bitrate. + +
+ +
+ +The audio bandwidths supported by Opus are listed in + . +Just like for the number of channels, any decoder can decode audio encoded at + any bandwidth. +For example, any Opus decoder operating at 8 kHz can decode a FB Opus + frame, and any Opus decoder operating at 48 kHz can decode a NB frame. +Similarly, the reference encoder can take a 48 kHz input signal and + encode it as NB. +The higher the audio bandwidth, the higher the required bitrate to achieve + acceptable quality. +The audio bandwidth can be explicitly specified in real-time, but by default + the reference encoder attempts to make the best bandwidth decision possible + given the current bitrate. + +
+ + +
+ +Opus can encode frames of 2.5, 5, 10, 20, 40 or 60 ms. +It can also combine multiple frames into packets of up to 120 ms. +For real-time applications, sending fewer packets per second reduces the + bitrate, since it reduces the overhead from IP, UDP, and RTP headers. +However, it increases latency and sensitivity to packet losses, as losing one + packet constitutes a loss of a bigger chunk of audio. +Increasing the frame duration also slightly improves coding efficiency, but the + gain becomes small for frame sizes above 20 ms. +For this reason, 20 ms frames are a good choice for most applications. + +
+ +
+ +There are various aspects of the Opus encoding process where trade-offs +can be made between CPU complexity and quality/bitrate. In the reference +encoder, the complexity is selected using an integer from 0 to 10, where +0 is the lowest complexity and 10 is the highest. Examples of +computations for which such trade-offs may occur are: + +The order of the pitch analysis whitening filter , +The order of the short-term noise shaping filter, +The number of states in delayed decision quantization of the +residual signal, and +The use of certain bit-stream features such as variable time-frequency +resolution and the pitch post-filter. + + +
+ +
+ +Audio codecs often exploit inter-frame correlations to reduce the +bitrate at a cost in error propagation: after losing one packet +several packets need to be received before the decoder is able to +accurately reconstruct the speech signal. The extent to which Opus +exploits inter-frame dependencies can be adjusted on the fly to +choose a trade-off between bitrate and amount of error propagation. + +
+ +
+ + Another mechanism providing robustness against packet loss is the in-band + Forward Error Correction (FEC). Packets that are determined to + contain perceptually important speech information, such as onsets or + transients, are encoded again at a lower bitrate and this re-encoded + information is added to a subsequent packet. + +
+ +
+ +Opus is more efficient when operating with variable bitrate (VBR), which is +the default. However, in some (rare) applications, constant bitrate (CBR) +is required. There are two main reasons to operate in CBR mode: + +When the transport only supports a fixed size for each compressed frame +When encryption is used for an audio stream that is either highly constrained + (e.g. yes/no, recorded prompts) or highly sensitive + + +When low-latency transmission is required over a relatively slow connection, then +constrained VBR can also be used. This uses VBR in a way that simulates a +"bit reservoir" and is equivalent to what MP3 (MPEG 1, Layer 3) and +AAC (Advanced Audio Coding) call CBR (i.e., not true +CBR due to the bit reservoir). + +
+ +
+ + Discontinuous Transmission (DTX) reduces the bitrate during silence + or background noise. When DTX is enabled, only one frame is encoded + every 400 milliseconds. + +
+ +
+ +
+ +
+ + +The Opus encoder produces "packets", which are each a contiguous set of bytes + meant to be transmitted as a single unit. +The packets described here do not include such things as IP, UDP, or RTP + headers which are normally found in a transport-layer packet. +A single packet may contain multiple audio frames, so long as they share a + common set of parameters, including the operating mode, audio bandwidth, frame + size, and channel count (mono vs. stereo). +This section describes the possible combinations of these parameters and the + internal framing used to pack multiple frames into a single packet. +This framing is not self-delimiting. +Instead, it assumes that a higher layer (such as UDP or RTP +or Ogg or Matroska ) + will communicate the length, in bytes, of the packet, and it uses this + information to reduce the framing overhead in the packet itself. +A decoder implementation MUST support the framing described in this section. +An alternative, self-delimiting variant of the framing is described in + . +Support for that variant is OPTIONAL. + + + +All bit diagrams in this document number the bits so that bit 0 is the most + significant bit of the first byte, and bit 7 is the least significant. +Bit 8 is thus the most significant bit of the second byte, etc. +Well-formed Opus packets obey certain requirements, marked [R1] through [R7] + below. +These are summarized in along with + appropriate means of handling malformed packets. + + +
+ +A well-formed Opus packet MUST contain at least one byte [R1]. +This byte forms a table-of-contents (TOC) header that signals which of the + various modes and configurations a given packet uses. +It is composed of a configuration number, "config", a stereo flag, "s", and a + frame count code, "c", arranged as illustrated in + . +A description of each of these fields follows. + + +
+ +
+ + +The top five bits of the TOC byte, labeled "config", encode one of 32 possible + configurations of operating mode, audio bandwidth, and frame size. +As described, the LP (SILK) layer and MDCT (CELT) layer can be combined in three possible + operating modes: + +A SILK-only mode for use in low bitrate connections with an audio bandwidth + of WB or less, +A Hybrid (SILK+CELT) mode for SWB or FB speech at medium bitrates, and +A CELT-only mode for very low delay speech transmission as well as music + transmission (NB to FB). + +The 32 possible configurations each identify which one of these operating modes + the packet uses, as well as the audio bandwidth and the frame size. + lists the parameters for each configuration. + + +Configuration Number(s) +Mode +Bandwidth +Frame Sizes +0...3 SILK-only NB 10, 20, 40, 60 ms +4...7 SILK-only MB 10, 20, 40, 60 ms +8...11 SILK-only WB 10, 20, 40, 60 ms +12...13 Hybrid SWB 10, 20 ms +14...15 Hybrid FB 10, 20 ms +16...19 CELT-only NB 2.5, 5, 10, 20 ms +20...23 CELT-only WB 2.5, 5, 10, 20 ms +24...27 CELT-only SWB 2.5, 5, 10, 20 ms +28...31 CELT-only FB 2.5, 5, 10, 20 ms + + +The configuration numbers in each range (e.g., 0...3 for NB SILK-only) + correspond to the various choices of frame size, in the same order. +For example, configuration 0 has a 10 ms frame size and configuration 3 + has a 60 ms frame size. + + + +One additional bit, labeled "s", signals mono vs. stereo, with 0 indicating + mono and 1 indicating stereo. + + + +The remaining two bits of the TOC byte, labeled "c", code the number of frames + per packet (codes 0 to 3) as follows: + +0: 1 frame in the packet +1: 2 frames in the packet, each with equal compressed size +2: 2 frames in the packet, with different compressed sizes +3: an arbitrary number of frames in the packet + +This draft refers to a packet as a code 0 packet, code 1 packet, etc., based on + the value of "c". + + +
+ +
+ + +This section describes how frames are packed according to each possible value + of "c" in the TOC byte. + + +
+ +When a packet contains multiple VBR frames (i.e., code 2 or 3), the compressed + length of one or more of these frames is indicated with a one- or two-byte + sequence, with the meaning of the first byte as follows: + +0: No frame (discontinuous transmission (DTX) or lost packet) +1...251: Length of the frame in bytes +252...255: A second byte is needed. The total length is (second_byte*4)+first_byte + + + + +The special length 0 indicates that no frame is available, either because it + was dropped during transmission by some intermediary or because the encoder + chose not to transmit it. +Any Opus frame in any mode MAY have a length of 0. + + + +The maximum representable length is 255*4+255=1275 bytes. +For 20 ms frames, this represents a bitrate of 510 kb/s, which is + approximately the highest useful rate for lossily compressed fullband stereo + music. +Beyond this point, lossless codecs are more appropriate. +It is also roughly the maximum useful rate of the MDCT layer, as shortly + thereafter quality no longer improves with additional bits due to limitations + on the codebook sizes. + + + +No length is transmitted for the last frame in a VBR packet, or for any of the + frames in a CBR packet, as it can be inferred from the total size of the + packet and the size of all other data in the packet. +However, the length of any individual frame MUST NOT exceed + 1275 bytes [R2], to allow for repacketization by gateways, + conference bridges, or other software. + +
+ +
+ + +For code 0 packets, the TOC byte is immediately followed by N-1 bytes + of compressed data for a single frame (where N is the size of the packet), + as illustrated in . + +
+ +
+
+ +
+ +For code 1 packets, the TOC byte is immediately followed by the + (N-1)/2 bytes of compressed data for the first frame, followed by + (N-1)/2 bytes of compressed data for the second frame, as illustrated in + . +The number of payload bytes available for compressed data, N-1, MUST be even + for all code 1 packets [R3]. + +
+ +
+
+ +
+ +For code 2 packets, the TOC byte is followed by a one- or two-byte sequence + indicating the length of the first frame (marked N1 in ), + followed by N1 bytes of compressed data for the first frame. +The remaining N-N1-2 or N-N1-3 bytes are the compressed data for the + second frame. +This is illustrated in . +A code 2 packet MUST contain enough bytes to represent a valid length. +For example, a 1-byte code 2 packet is always invalid, and a 2-byte code 2 + packet whose second byte is in the range 252...255 is also invalid. +The length of the first frame, N1, MUST also be no larger than the size of the + payload remaining after decoding that length for all code 2 packets [R4]. +This makes, for example, a 2-byte code 2 packet with a second byte in the range + 1...251 invalid as well (the only valid 2-byte code 2 packet is one where the + length of both frames is zero). + +
+ +
+
+ +
+ +Code 3 packets signal the number of frames, as well as additional + padding, called "Opus padding" to indicate that this padding is added at the + Opus layer, rather than at the transport layer. +Code 3 packets MUST have at least 2 bytes [R6,R7]. +The TOC byte is followed by a byte encoding the number of frames in the packet + in bits 2 to 7 (marked "M" in ), with bit 1 indicating whether + or not Opus padding is inserted (marked "p" in ), and bit 0 + indicating VBR (marked "v" in ). +M MUST NOT be zero, and the audio duration contained within a packet MUST NOT + exceed 120 ms [R5]. +This limits the maximum frame count for any frame size to 48 (for 2.5 ms + frames), with lower limits for longer frame sizes. + illustrates the layout of the frame count + byte. + +
+ +
+ +When Opus padding is used, the number of bytes of padding is encoded in the + bytes following the frame count byte. +Values from 0...254 indicate that 0...254 bytes of padding are included, + in addition to the byte(s) used to indicate the size of the padding. +If the value is 255, then the size of the additional padding is 254 bytes, + plus the padding value encoded in the next byte. +There MUST be at least one more byte in the packet in this case [R6,R7]. +The additional padding bytes appear at the end of the packet, and MUST be set + to zero by the encoder to avoid creating a covert channel. +The decoder MUST accept any value for the padding bytes, however. + + +Although this encoding provides multiple ways to indicate a given number of + padding bytes, each uses a different number of bytes to indicate the padding + size, and thus will increase the total packet size by a different amount. +For example, to add 255 bytes to a packet, set the padding bit, p, to 1, insert + a single byte after the frame count byte with a value of 254, and append 254 + padding bytes with the value zero to the end of the packet. +To add 256 bytes to a packet, set the padding bit to 1, insert two bytes after + the frame count byte with the values 255 and 0, respectively, and append 254 + padding bytes with the value zero to the end of the packet. +By using the value 255 multiple times, it is possible to create a packet of any + specific, desired size. +Let P be the number of header bytes used to indicate the padding size plus the + number of padding bytes themselves (i.e., P is the total number of bytes added + to the packet). +Then P MUST be no more than N-2 [R6,R7]. + + +In the CBR case, let R=N-2-P be the number of bytes remaining in the packet + after subtracting the (optional) padding. +Then the compressed length of each frame in bytes is equal to R/M. +The value R MUST be a non-negative integer multiple of M [R6]. +The compressed data for all M frames follows, each of size + R/M bytes, as illustrated in . + + +
+ +
+ + +In the VBR case, the (optional) padding length is followed by M-1 frame + lengths (indicated by "N1" to "N[M-1]" in ), each encoded in a + one- or two-byte sequence as described above. +The packet MUST contain enough data for the M-1 lengths after removing the + (optional) padding, and the sum of these lengths MUST be no larger than the + number of bytes remaining in the packet after decoding them [R7]. +The compressed data for all M frames follows, each frame consisting of the + indicated number of bytes, with the final frame consuming any remaining bytes + before the final padding, as illustrated in . +The number of header bytes (TOC byte, frame count byte, padding length bytes, + and frame length bytes), plus the signaled length of the first M-1 frames themselves, + plus the signaled length of the padding MUST be no larger than N, the total size of the + packet. + + +
+ +
+
+
+ +
+ +Simplest case, one NB mono 20 ms SILK frame: + + +
+ +
+ + +Two FB mono 5 ms CELT frames of the same compressed size: + + +
+ +
+ + +Two FB mono 20 ms Hybrid frames of different compressed size: + + +
+ +
+ + +Four FB stereo 20 ms CELT frames of the same compressed size: + + +
+ +
+
+ +
+ +A receiver MUST NOT process packets which violate any of the rules above as + normal Opus packets. +They are reserved for future applications, such as in-band headers (containing + metadata, etc.). +Packets which violate these constraints may cause implementations of + this specification to treat them as malformed, and + discard them. + + +These constraints are summarized here for reference: + +Packets are at least one byte. +No implicit frame length is larger than 1275 bytes. +Code 1 packets have an odd total length, N, so that (N-1)/2 is an + integer. +Code 2 packets have enough bytes after the TOC for a valid frame + length, and that length is no larger than the number of bytes remaining in the + packet. +Code 3 packets contain at least one frame, but no more than 120 ms + of audio total. +The length of a CBR code 3 packet, N, is at least two bytes, the number of + bytes added to indicate the padding size plus the trailing padding bytes + themselves, P, is no more than N-2, and the frame count, M, satisfies + the constraint that (N-2-P) is a non-negative integer multiple of M. +VBR code 3 packets are large enough to contain all the header bytes (TOC + byte, frame count byte, any padding length bytes, and any frame length bytes), + plus the length of the first M-1 frames, plus any trailing padding bytes. + + +
+ +
+ +
+ +The Opus decoder consists of two main blocks: the SILK decoder and the CELT + decoder. +At any given time, one or both of the SILK and CELT decoders may be active. +The output of the Opus decode is the sum of the outputs from the SILK and CELT + decoders with proper sample rate conversion and delay compensation on the SILK + side, and optional decimation (when decoding to sample rates less than + 48 kHz) on the CELT side, as illustrated in the block diagram below. + +
+ +| Decoder |--->| Rate |----+ +Bit- +---------+ | | | | Conversion | v +stream | Range |---+ +---------+ +------------+ /---\ Audio +------->| Decoder | | + |------> + | |---+ +---------+ +------------+ \---/ + +---------+ | | CELT | | Decimation | ^ + +->| Decoder |--->| (Optional) |----+ + | | | | + +---------+ +------------+ +]]> + +
+ +
+ +Opus uses an entropy coder based on range coding +, +which is itself a rediscovery of the FIFO arithmetic code introduced by . +It is very similar to arithmetic encoding, except that encoding is done with +digits in any base instead of with bits, +so it is faster when using larger bases (i.e., a byte). All of the +calculations in the range coder must use bit-exact integer arithmetic. + + +Symbols may also be coded as "raw bits" packed directly into the bitstream, + bypassing the range coder. +These are packed backwards starting at the end of the frame, as illustrated in + . +This reduces complexity and makes the stream more resilient to bit errors, as + corruption in the raw bits will not desynchronize the decoding process, unlike + corruption in the input to the range decoder. +Raw bits are only used in the CELT layer. + + +
+ : ++ + +: : ++ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +: | <- Boundary occurs at an arbitrary bit position : ++-+-+-+ + +: <- Raw bits data (packed LSB to MSB) | ++-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +]]> +
+ + +Each symbol coded by the range coder is drawn from a finite alphabet and coded + in a separate "context", which describes the size of the alphabet and the + relative frequency of each symbol in that alphabet. + + +Suppose there is a context with n symbols, identified with an index that ranges + from 0 to n-1. +The parameters needed to encode or decode symbol k in this context are + represented by a three-tuple (fl[k], fh[k], ft), with + 0 <= fl[k] < fh[k] <= ft <= 65535. +The values of this tuple are derived from the probability model for the + symbol, represented by traditional "frequency counts". +Because Opus uses static contexts these are not updated as symbols are decoded. +Let f[i] be the frequency of symbol i. +Then the three-tuple corresponding to symbol k is given by + +
+ +
+ +The range decoder extracts the symbols and integers encoded using the range + encoder in . +The range decoder maintains an internal state vector composed of the two-tuple + (val, rng), representing the difference between the high end of the + current range and the actual coded value, minus one, and the size of the + current range, respectively. +Both val and rng are 32-bit unsigned integer values. + + +
+ +Let b0 be the first input byte (or zero if there are no bytes in this Opus + frame). +The decoder initializes rng to 128 and initializes val to + (127 - (b0>>1)), where (b0>>1) is the top 7 bits of the + first input byte. +It saves the remaining bit, (b0&1), for use in the renormalization + procedure described in , which the + decoder invokes immediately after initialization to read additional bits and + establish the invariant that rng > 2**23. + +
+ +
+ +Decoding a symbol is a two-step process. +The first step determines a 16-bit unsigned value fs, which lies within the + range of some symbol in the current context. +The second step updates the range decoder state with the three-tuple + (fl[k], fh[k], ft) corresponding to that symbol. + + +The first step is implemented by ec_decode() (entdec.c), which computes +
+ +
+The divisions here are integer division. + + +The decoder then identifies the symbol in the current context corresponding to + fs; i.e., the value of k whose three-tuple (fl[k], fh[k], ft) + satisfies fl[k] <= fs < fh[k]. +It uses this tuple to update val according to +
+ +
+If fl[k] is greater than zero, then the decoder updates rng using +
+ +
+Otherwise, it updates rng using +
+ +
+ + +Using a special case for the first symbol (rather than the last symbol, as is + commonly done in other arithmetic coders) ensures that all the truncation + error from the finite precision arithmetic accumulates in symbol 0. +This makes the cost of coding a 0 slightly smaller, on average, than its + estimated probability indicates and makes the cost of coding any other symbol + slightly larger. +When contexts are designed so that 0 is the most probable symbol, which is + often the case, this strategy minimizes the inefficiency introduced by the + finite precision. +It also makes some of the special-case decoding routines in + particularly simple. + + +After the updates, implemented by ec_dec_update() (entdec.c), the decoder + normalizes the range using the procedure in the next section, and returns the + index k. + + +
+ +To normalize the range, the decoder repeats the following process, implemented + by ec_dec_normalize() (entdec.c), until rng > 2**23. +If rng is already greater than 2**23, the entire process is skipped. +First, it sets rng to (rng<<8). +Then it reads the next byte of the Opus frame and forms an 8-bit value sym, + using the left-over bit buffered from the previous byte as the high bit + and the top 7 bits of the byte just read as the other 7 bits of sym. +The remaining bit in the byte just read is buffered for use in the next + iteration. +If no more input bytes remain, it uses zero bits instead. +See for the initialization used to process + the first byte. +Then, it sets +
+ +
+ + +It is normal and expected that the range decoder will read several bytes + into the raw bits data (if any) at the end of the packet by the time the frame + is completely decoded, as illustrated in . +This same data MUST also be returned as raw bits when requested. +The encoder is expected to terminate the stream in such a way that the decoder + will decode the intended values regardless of the data contained in the raw + bits. + describes a procedure for doing this. +If the range decoder consumes all of the bytes belonging to the current frame, + it MUST continue to use zero when any further input bytes are required, even + if there is additional data in the current packet from padding or other + frames. + + +
+ | : ++-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + ^ ^ + | End of data buffered by the range coder | +...-----------------------------------------------+ + | + | End of data consumed by raw bits + +-------------------------------------------------------... +]]> +
+
+
+ +
+ +The reference implementation uses three additional decoding methods that are + exactly equivalent to the above, but make assumptions and simplifications that + allow for a more efficient implementation. + +
+ +The first is ec_decode_bin() (entdec.c), defined using the parameter ftb + instead of ft. +It is mathematically equivalent to calling ec_decode() with + ft = (1<<ftb), but avoids one of the divisions. + +
+
+ +The next is ec_dec_bit_logp() (entdec.c), which decodes a single binary symbol, + replacing both the ec_decode() and ec_dec_update() steps. +The context is described by a single parameter, logp, which is the absolute + value of the base-2 logarithm of the probability of a "1". +It is mathematically equivalent to calling ec_decode() with + ft = (1<<logp), followed by ec_dec_update() with + the 3-tuple (fl[k] = 0, + fh[k] = (1<<logp) - 1, + ft = (1<<logp)) if the returned value + of fs is less than (1<<logp) - 1 (a "0" was decoded), and with + (fl[k] = (1<<logp) - 1, + fh[k] = ft = (1<<logp)) otherwise (a "1" was + decoded). +The implementation requires no multiplications or divisions. + +
+
+ +The last is ec_dec_icdf() (entdec.c), which decodes a single symbol with a + table-based context of up to 8 bits, also replacing both the ec_decode() and + ec_dec_update() steps, as well as the search for the decoded symbol in between. +The context is described by two parameters, an icdf + ("inverse" cumulative distribution function) table and ftb. +As with ec_decode_bin(), (1<<ftb) is equivalent to ft. +idcf[k], on the other hand, stores (1<<ftb)-fh[k], which is equal to + (1<<ftb) - fl[k+1]. +fl[0] is assumed to be 0, and the table is terminated by a value of 0 (where + fh[k] == ft). + + +The function is mathematically equivalent to calling ec_decode() with + ft = (1<<ftb), using the returned value fs to search the table + for the first entry where fs < (1<<ftb)-icdf[k], and + calling ec_dec_update() with + fl[k] = (1<<ftb) - icdf[k-1] (or 0 + if k == 0), fh[k] = (1<<ftb) - idcf[k], + and ft = (1<<ftb). +Combining the search with the update allows the division to be replaced by a + series of multiplications (which are usually much cheaper), and using an + inverse CDF allows the use of an ftb as large as 8 in an 8-bit table without + any special cases. +This is the primary interface with the range decoder in the SILK layer, though + it is used in a few places in the CELT layer as well. + + +Although icdf[k] is more convenient for the code, the frequency counts, f[k], + are a more natural representation of the probability distribution function + (PDF) for a given symbol. +Therefore this draft lists the latter, not the former, when describing the + context in which a symbol is coded as a list, e.g., {4, 4, 4, 4}/16 for a + uniform context with four possible values and ft = 16. +The value of ft after the slash is always the sum of the entries in the PDF, + but is included for convenience. +Contexts with identical probabilities, f[k]/ft, but different values of ft + (or equivalently, ftb) are not the same, and cannot, in general, be used in + place of one another. +An icdf table is also not capable of representing a PDF where the first symbol + has 0 probability. +In such contexts, ec_dec_icdf() can decode the symbol by using a table that + drops the entries for any initial zero-probability values and adding the + constant offset of the first value with a non-zero probability to its return + value. + +
+
+ +
+ +The raw bits used by the CELT layer are packed at the end of the packet, with + the least significant bit of the first value packed in the least significant + bit of the last byte, filling up to the most significant bit in the last byte, + continuing on to the least significant bit of the penultimate byte, and so on. +The reference implementation reads them using ec_dec_bits() (entdec.c). +Because the range decoder must read several bytes ahead in the stream, as + described in , the input consumed by the + raw bits may overlap with the input consumed by the range coder, and a decoder + MUST allow this. +The format should render it impossible to attempt to read more raw bits than + there are actual bits in the frame, though a decoder may wish to check for + this and report an error. + +
+ +
+ +The function ec_dec_uint() (entdec.c) decodes one of ft equiprobable values in + the range 0 to (ft - 1), inclusive, each with a frequency of 1, + where ft may be as large as (2**32 - 1). +Because ec_decode() is limited to a total frequency of (2**16 - 1), + it splits up the value into a range coded symbol representing up to 8 of the + high bits, and, if necessary, raw bits representing the remainder of the + value. +The limit of 8 bits in the range coded symbol is a trade-off between + implementation complexity, modeling error (since the symbols no longer truly + have equal coding cost), and rounding error introduced by the range coder + itself (which gets larger as more bits are included). +Using raw bits reduces the maximum number of divisions required in the worst + case, but means that it may be possible to decode a value outside the range + 0 to (ft - 1), inclusive. + + + +ec_dec_uint() takes a single, positive parameter, ft, which is not necessarily + a power of two, and returns an integer, t, whose value lies between 0 and + (ft - 1), inclusive. +Let ftb = ilog(ft - 1), i.e., the number of bits required + to store (ft - 1) in two's complement notation. +If ftb is 8 or less, then t is decoded with t = ec_decode(ft), and + the range coder state is updated using the three-tuple (t, t + 1, + ft). + + +If ftb is greater than 8, then the top 8 bits of t are decoded using +
+> (ftb - 8)) + 1) , +]]> +
+ the decoder state is updated using the three-tuple + (t, t + 1, + ((ft - 1) >> (ftb - 8)) + 1), + and the remaining bits are decoded as raw bits, setting +
+ +
+If, at this point, t >= ft, then the current frame is corrupt. +In that case, the decoder should assume there has been an error in the coding, + decoding, or transmission and SHOULD take measures to conceal the + error and/or report to the application that the error has occurred. + + +
+ +
+ +The bit allocation routines in the CELT decoder need a conservative upper bound + on the number of bits that have been used from the current frame thus far, + including both range coder bits and raw bits. +This drives allocation decisions that must match those made in the encoder. +The upper bound is computed in the reference implementation to whole-bit + precision by the function ec_tell() (entcode.h) and to fractional 1/8th bit + precision by the function ec_tell_frac() (entcode.c). +Like all operations in the range coder, it must be implemented in a bit-exact + manner, and must produce exactly the same value returned by the same functions + in the encoder after encoding the same symbols. + + +ec_tell() is guaranteed to return ceil(ec_tell_frac()/8.0). +In various places the codec will check to ensure there is enough room to + contain a symbol before attempting to decode it. +In practice, although the number of bits used so far is an upper bound, + decoding a symbol whose probability model suggests it has a worst-case cost of + p 1/8th bits may actually advance the return value of ec_tell_frac() by + p-1, p, or p+1 1/8th bits, due to approximation error in that upper bound, + truncation error in the range coder, and for large values of ft, modeling + error in ec_dec_uint(). + + +However, this error is bounded, and periodic calls to ec_tell() or + ec_tell_frac() at precisely defined points in the decoding process prevent it + from accumulating. +For a range coder symbol that requires a whole number of bits (i.e., + for which ft/(fh[k] - fl[k]) is a power of two), where there are at + least p 1/8th bits available, decoding the symbol will never cause ec_tell() or + ec_tell_frac() to exceed the size of the frame ("bust the budget"). +In this case the return value of ec_tell_frac() will only advance by more than + p 1/8th bits if there was an additional, fractional number of bits remaining, + and it will never advance beyond the next whole-bit boundary, which is safe, + since frames always contain a whole number of bits. +However, when p is not a whole number of bits, an extra 1/8th bit is required + to ensure that decoding the symbol will not bust the budget. + + +The reference implementation keeps track of the total number of whole bits that + have been processed by the decoder so far in the variable nbits_total, + including the (possibly fractional) number of bits that are currently + buffered, but not consumed, inside the range coder. +nbits_total is initialized to 9 just before the initial range renormalization + process completes (or equivalently, it can be initialized to 33 after the + first renormalization). +The extra two bits over the actual amount buffered by the range coder + guarantees that it is an upper bound and that there is enough room for the + encoder to terminate the stream. +Each iteration through the range coder's renormalization loop increases + nbits_total by 8. +Reading raw bits increases nbits_total by the number of raw bits read. + + +
+ +The whole number of bits buffered in rng may be estimated via lg = ilog(rng). +ec_tell() then becomes a simple matter of removing these bits from the total. +It returns (nbits_total - lg). + + +In a newly initialized decoder, before any symbols have been read, this reports + that 1 bit has been used. +This is the bit reserved for termination of the encoder. + +
+ +
+ +ec_tell_frac() estimates the number of bits buffered in rng to fractional + precision. +Since rng must be greater than 2**23 after renormalization, lg must be at least + 24. +Let +
+ +> (lg-16) , +]]> +
+ so that 32768 <= r_Q15 < 65536, an unsigned Q15 value representing the + fractional part of rng. +Then the following procedure can be used to add one bit of precision to lg. +First, update +
+ +> 15 . +]]> +
+Then add the 16th bit of r_Q15 to lg via +
+ +> 16) . +]]> +
+Finally, if this bit was a 1, reduce r_Q15 by a factor of two via +
+ +> 1 , +]]> +
+ so that it once again lies in the range 32768 <= r_Q15 < 65536. + + +This procedure is repeated three times to extend lg to 1/8th bit precision. +ec_tell_frac() then returns (nbits_total*8 - lg). + +
+ +
+ +
+ +
+ +The decoder's LP layer uses a modified version of the SILK codec (herein simply + called "SILK"), which runs a decoded excitation signal through adaptive + long-term and short-term prediction synthesis filters. +It runs at NB, MB, and WB sample rates internally. +When used in a SWB or FB Hybrid frame, the LP layer itself still only runs in + WB. + + +
+ +An overview of the decoder is given in . + +
+ +| Range |--->| Decode |---------------------------+ + 1 | Decoder | 2 | Parameters |----------+ 5 | + +---------+ +------------+ 4 | | + 3 | | | + \/ \/ \/ + +------------+ +------------+ +------------+ + | Generate |-->| LTP |-->| LPC | + | Excitation | | Synthesis | | Synthesis | + +------------+ +------------+ +------------+ + ^ | + | | + +-------------------+----------------+ + | 6 + | +------------+ +-------------+ + +-->| Stereo |-->| Sample Rate |--> + | Unmixing | 7 | Conversion | 8 + +------------+ +-------------+ + +1: Range encoded bitstream +2: Coded parameters +3: Pulses, LSBs, and signs +4: Pitch lags, Long-Term Prediction (LTP) coefficients +5: Linear Predictive Coding (LPC) coefficients and gains +6: Decoded signal (mono or mid-side stereo) +7: Unmixed signal (mono or left-right stereo) +8: Resampled signal +]]> + +
+ + +The decoder feeds the bitstream (1) to the range decoder from + , and then decodes the parameters in it (2) + using the procedures detailed in + Sections  + through . +These parameters (3, 4, 5) are used to generate an excitation signal (see + ), which is fed to an optional + long-term prediction (LTP) filter (voiced frames only, see + ) and then a short-term prediction filter + (see ), producing the decoded signal (6). +For stereo streams, the mid-side representation is converted to separate left + and right channels (7). +The result is finally resampled to the desired output sample rate (e.g., + 48 kHz) so that the resampled signal (8) can be mixed with the CELT + layer. + + +
+ +
+ + +Internally, the LP layer of a single Opus frame is composed of either a single + 10 ms regular SILK frame or between one and three 20 ms regular SILK + frames. +A stereo Opus frame may double the number of regular SILK frames (up to a total + of six), since it includes separate frames for a mid channel and, optionally, + a side channel. +Optional Low Bit-Rate Redundancy (LBRR) frames, which are reduced-bitrate + encodings of previous SILK frames, may be included to aid in recovery from + packet loss. +If present, these appear before the regular SILK frames. +They are in most respects identical to regular, active SILK frames, except that + they are usually encoded with a lower bitrate. +This draft uses "SILK frame" to refer to either one and "regular SILK frame" if + it needs to draw a distinction between the two. + + +Logically, each SILK frame is in turn composed of either two or four 5 ms + subframes. +Various parameters, such as the quantization gain of the excitation and the + pitch lag and filter coefficients can vary on a subframe-by-subframe basis. +Physically, the parameters for each subframe are interleaved in the bitstream, + as described in the relevant sections for each parameter. + + +All of these frames and subframes are decoded from the same range coder, with + no padding between them. +Thus packing multiple SILK frames in a single Opus frame saves, on average, + half a byte per SILK frame. +It also allows some parameters to be predicted from prior SILK frames in the + same Opus frame, since this does not degrade packet loss robustness (beyond + any penalty for merely using fewer, larger packets to store multiple frames). + + + +Stereo support in SILK uses a variant of mid-side coding, allowing a mono + decoder to simply decode the mid channel. +However, the data for the two channels is interleaved, so a mono decoder must + still unpack the data for the side channel. +It would be required to do so anyway for Hybrid Opus frames, or to support + decoding individual 20 ms frames. + + + + summarizes the overall grouping of the contents of + the LP layer. +Figures  + and  illustrate + the ordering of the various SILK frames for a 60 ms Opus frame, for both + mono and stereo, respectively. + + + +Symbol(s) +PDF(s) +Condition + +Voice Activity Detection (VAD) flags +{1, 1}/2 + + +LBRR flag +{1, 1}/2 + + +Per-frame LBRR flags + + + +LBRR Frame(s) + + + +Regular SILK Frame(s) + + + + + +
+ +
+ +
+ +
+ +
+ +
+ +The LP layer begins with two to eight header bits, decoded in silk_Decode() + (dec_API.c). +These consist of one Voice Activity Detection (VAD) bit per frame (up to 3), + followed by a single flag indicating the presence of LBRR frames. +For a stereo packet, these first flags correspond to the mid channel, and a + second set of flags is included for the side channel. + + +Because these are the first symbols decoded by the range coder and because they + are coded as binary values with uniform probability, they can be extracted + directly from the most significant bits of the first byte of compressed data. +Thus, a receiver can determine if an Opus frame contains any active SILK frames + without the overhead of using the range decoder. + +
+ +
+ +For Opus frames longer than 20 ms, a set of LBRR flags is + decoded for each channel that has its LBRR flag set. +Each set contains one flag per 20 ms SILK frame. +40 ms Opus frames use the 2-frame LBRR flag PDF from + , and 60 ms Opus frames use the + 3-frame LBRR flag PDF. +For each channel, the resulting 2- or 3-bit integer contains the corresponding + LBRR flag for each frame, packed in order from the LSB to the MSB. + + + +Frame Size +PDF +40 ms {0, 53, 53, 150}/256 +60 ms {0, 41, 20, 29, 41, 15, 28, 82}/256 + + + +A 10 or 20 ms Opus frame does not contain any per-frame LBRR flags, + as there may be at most one LBRR frame per channel. +The global LBRR flag in the header bits (see ) + is already sufficient to indicate the presence of that single LBRR frame. + + +
+ +
+ +The LBRR frames, if present, contain an encoded representation of the signal + immediately prior to the current Opus frame as if it were encoded with the + current mode, frame size, audio bandwidth, and channel count, even if those + differ from the prior Opus frame. +When one of these parameters changes from one Opus frame to the next, this + implies that the LBRR frames of the current Opus frame may not be simple + drop-in replacements for the contents of the previous Opus frame. + + + +For example, when switching from 20 ms to 60 ms, the 60 ms Opus + frame may contain LBRR frames covering up to three prior 20 ms Opus + frames, even if those frames already contained LBRR frames covering some of + the same time periods. +When switching from 20 ms to 10 ms, the 10 ms Opus frame can + contain an LBRR frame covering at most half the prior 20 ms Opus frame, + potentially leaving a hole that needs to be concealed from even a single + packet loss (see ). +When switching from mono to stereo, the LBRR frames in the first stereo Opus + frame MAY contain a non-trivial side channel. + + + +In order to properly produce LBRR frames under all conditions, an encoder might + need to buffer up to 60 ms of audio and re-encode it during these + transitions. +However, the reference implementation opts to disable LBRR frames at the + transition point for simplicity. +Since transitions are relatively infrequent in normal usage, this does not have + a significant impact on packet loss robustness. + + + +The LBRR frames immediately follow the LBRR flags, prior to any regular SILK + frames. + describes their exact contents. +LBRR frames do not include their own separate VAD flags. +LBRR frames are only meant to be transmitted for active speech, thus all LBRR + frames are treated as active. + + + +In a stereo Opus frame longer than 20 ms, although the per-frame LBRR + flags for the mid channel are coded as a unit before the per-frame LBRR flags + for the side channel, the LBRR frames themselves are interleaved. +The decoder parses an LBRR frame for the mid channel of a given 20 ms + interval (if present) and then immediately parses the corresponding LBRR + frame for the side channel (if present), before proceeding to the next + 20 ms interval. + +
+ +
+ +The regular SILK frame(s) follow the LBRR frames (if any). + describes their contents, as well. +Unlike the LBRR frames, a regular SILK frame is coded for each time interval in + an Opus frame, even if the corresponding VAD flags are unset. +For stereo Opus frames longer than 20 ms, the regular mid and side SILK + frames for each 20 ms interval are interleaved, just as with the LBRR + frames. +The side frame may be skipped by coding an appropriate flag, as detailed in + . + +
+ +
+ +Each SILK frame includes a set of side information that encodes + +The frame type and quantization type (), +Quantization gains (), +Short-term prediction filter coefficients (), +A Line Spectral Frequencies (LSF) interpolation weight (), + +Long-term prediction filter lags and gains (), + and + +A linear congruential generator (LCG) seed (). + +The quantized excitation signal (see ) follows + these at the end of the frame. + details the overall organization of a + SILK frame. + + + +Symbol(s) +PDF(s) +Condition + +Stereo Prediction Weights + + + +Mid-only Flag + + + +Frame Type + + + +Subframe Gains + + + +Normalized LSF Stage-1 Index + + + +Normalized LSF Stage-2 Residual + + + +Normalized LSF Interpolation Weight + +20 ms frame + +Primary Pitch Lag + +Voiced frame + +Subframe Pitch Contour + +Voiced frame + +Periodicity Index + +Voiced frame + +LTP Filter + +Voiced frame + +LTP Scaling + + + +LCG Seed + + + +Excitation Rate Level + + + +Excitation Pulse Counts + + + +Excitation Pulse Locations + +Non-zero pulse count + +Excitation LSBs + + + +Excitation Signs + + + + + +
+ +A SILK frame corresponding to the mid channel of a stereo Opus frame begins + with a pair of side channel prediction weights, designed such that zeros + indicate normal mid-side coupling. +Since these weights can change on every frame, the first portion of each frame + linearly interpolates between the previous weights and the current ones, using + zeros for the previous weights if none are available. +These prediction weights are never included in a mono Opus frame, and the + previous weights are reset to zeros on any transition from mono to stereo. +They are also not included in an LBRR frame for the side channel, even if the + LBRR flags indicate the corresponding mid channel was not coded. +In that case, the previous weights are used, again substituting in zeros if no + previous weights are available since the last decoder reset + (see ). + + + +To summarize, these weights are coded if and only if + +This is a stereo Opus frame (), and +The current SILK frame corresponds to the mid channel. + + + + +The prediction weights are coded in three separate pieces, which are decoded + by silk_stereo_decode_pred() (decode_stereo_pred.c). +The first piece jointly codes the high-order part of a table index for both + weights. +The second piece codes the low-order part of each table index. +The third piece codes an offset used to linearly interpolate between table + indices. +The details are as follows. + + + +Let n be an index decoded with the 25-element stage-1 PDF in + . +Then let i0 and i1 be indices decoded with the stage-2 and stage-3 PDFs in + , respectively, and let i2 and i3 + be two more indices decoded with the stage-2 and stage-3 PDFs, all in that + order. + + + +Stage +PDF +Stage 1 +{7, 2, 1, 1, 1, + 10, 24, 8, 1, 1, + 3, 23, 92, 23, 3, + 1, 1, 8, 24, 10, + 1, 1, 1, 2, 7}/256 + +Stage 2 +{85, 86, 85}/256 + +Stage 3 +{51, 51, 52, 51, 51}/256 + + + +Then use n, i0, and i2 to form two table indices, wi0 and wi1, according to +
+ +
+ where the division is integer division. +The range of these indices is 0 to 14, inclusive. +Let w[i] be the i'th weight from . +Then the two prediction weights, w0_Q13 and w1_Q13, are +
+> 16)*(2*i3 + 1) + +w0_Q13 = w_Q13[wi0] + + ((w_Q13[wi0+1] - w_Q13[wi0])*6554) >> 16)*(2*i1 + 1) + - w1_Q13 +]]> +
+N.b., w1_Q13 is computed first here, because w0_Q13 depends on it. +The constant 6554 is approximately 0.1 in Q16. +Although wi0 and wi1 only have 15 possible values, + contains 16 entries to allow + interpolation between entry wi0 and (wi0 + 1) (and likewise for wi1). + + + +Index +Weight (Q13) + 0 -13732 + 1 -10050 + 2 -8266 + 3 -7526 + 4 -6500 + 5 -5000 + 6 -2950 + 7 -820 + 8 820 + 9 2950 +10 5000 +11 6500 +12 7526 +13 8266 +14 10050 +15 13732 + + +
+ +
+ +A flag appears after the stereo prediction weights that indicates if only the + mid channel is coded for this time interval. +It appears only when + +This is a stereo Opus frame (see ), +The current SILK frame corresponds to the mid channel, and +Either + +This is a regular SILK frame where the VAD flags + (see ) indicate that the corresponding side + channel is not active. + +This is an LBRR frame where the LBRR flags + (see and ) + indicate that the corresponding side channel is not coded. + + + + +It is omitted when there are no stereo weights, for all of the same reasons. +It is also omitted for a regular SILK frame when the VAD flag of the + corresponding side channel frame is set (indicating it is active). +The side channel must be coded in this case, making the mid-only flag + redundant. +It is also omitted for an LBRR frame when the corresponding LBRR flags + indicate the side channel is coded. + + + +When the flag is present, the decoder reads a single value using the PDF in + , as implemented in + silk_stereo_decode_mid_only() (decode_stereo_pred.c). +If the flag is set, then there is no corresponding SILK frame for the side + channel, the entire decoding process for the side channel is skipped, and + zeros are fed to the stereo unmixing process (see + ) instead. +As stated above, LBRR frames still include this flag when the LBRR flag + indicates that the side channel is not coded. +In that case, if this flag is zero (indicating that there should be a side + channel), then Packet Loss Concealment (PLC, see + ) SHOULD be invoked to recover a + side channel signal. +Otherwise, the stereo image will collapse. + + + +PDF +{192, 64}/256 + + +
+ +
+ +Each SILK frame contains a single "frame type" symbol that jointly codes the + signal type and quantization offset type of the corresponding frame. +If the current frame is a regular SILK frame whose VAD bit was not set (an + "inactive" frame), then the frame type symbol takes on a value of either 0 or + 1 and is decoded using the first PDF in . +If the frame is an LBRR frame or a regular SILK frame whose VAD flag was set + (an "active" frame), then the value of the symbol may range from 2 to 5, + inclusive, and is decoded using the second PDF in + . + translates between the value of the + frame type symbol and the corresponding signal type and quantization offset + type. + + + +VAD Flag +PDF +Inactive {26, 230, 0, 0, 0, 0}/256 +Active {0, 0, 24, 74, 148, 10}/256 + + + +Frame Type +Signal Type +Quantization Offset Type +0 Inactive Low +1 Inactive High +2 Unvoiced Low +3 Unvoiced High +4 Voiced Low +5 Voiced High + + +
+ +
+ +A separate quantization gain is coded for each 5 ms subframe. +These gains control the step size between quantization levels of the excitation + signal and, therefore, the quality of the reconstruction. +They are independent of and unrelated to the pitch contours coded for voiced + frames. +The quantization gains are themselves uniformly quantized to 6 bits on a + log scale, giving them a resolution of approximately 1.369 dB and a range + of approximately 1.94 dB to 88.21 dB. + + +The subframe gains are either coded independently, or relative to the gain from + the most recent coded subframe in the same channel. +Independent coding is used if and only if + + +This is the first subframe in the current SILK frame, and + +Either + + +This is the first SILK frame of its type (LBRR or regular) for this channel in + the current Opus frame, or + + +The previous SILK frame of the same type (LBRR or regular) for this channel in + the same Opus frame was not coded. + + + + + + + +In an independently coded subframe gain, the 3 most significant bits of the + quantization gain are decoded using a PDF selected from + based on the decoded signal + type (see ). + + + +Signal Type +PDF +Inactive {32, 112, 68, 29, 12, 1, 1, 1}/256 +Unvoiced {2, 17, 45, 60, 62, 47, 19, 4}/256 +Voiced {1, 3, 26, 71, 94, 50, 9, 2}/256 + + + +The 3 least significant bits are decoded using a uniform PDF: + + +PDF +{32, 32, 32, 32, 32, 32, 32, 32}/256 + + + +These 6 bits are combined to form a value, gain_index, between 0 and 63. +When the gain for the previous subframe is available, then the current gain is + limited as follows: +
+ +
+This may help some implementations limit the change in precision of their + internal LTP history. +The indices which this clamp applies to cannot simply be removed from the + codebook, because previous_log_gain will not be available after packet loss. +The clamping is skipped after a decoder reset, and in the side channel if the + previous frame in the side channel was not coded, since there is no value for + previous_log_gain available. +It MAY also be skipped after packet loss. + + + +For subframes which do not have an independent gain (including the first + subframe of frames not listed as using independent coding above), the + quantization gain is coded relative to the gain from the previous subframe (in + the same channel). +The PDF in yields a delta_gain_index value + between 0 and 40, inclusive. + + +PDF +{6, 5, 11, 31, 132, 21, 8, 4, + 3, 2, 2, 2, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1}/256 + + +The following formula translates this index into a quantization gain for the + current subframe using the gain from the previous subframe: +
+ +
+ + +silk_gains_dequant() (gain_quant.c) dequantizes log_gain for the k'th subframe + and converts it into a linear Q16 scale factor via +
+>16) + 2090) +]]> +
+ + +The function silk_log2lin() (log2lin.c) computes an approximation of + 2**(inLog_Q7/128.0), where inLog_Q7 is its Q7 input. +Let i = inLog_Q7>>7 be the integer part of inLogQ7 and + f = inLog_Q7&127 be the fractional part. +Then +
+>16)+f)*((1<>7) +]]> +
+ yields the approximate exponential. +The final Q16 gain values lies between 81920 and 1686110208, inclusive + (representing scale factors of 1.25 to 25728, respectively). + +
+ +
+ +A set of normalized Line Spectral Frequency (LSF) coefficients follow the + quantization gains in the bitstream, and represent the Linear Predictive + Coding (LPC) coefficients for the current SILK frame. +Once decoded, the normalized LSFs form an increasing list of Q15 values between + 0 and 1. +These represent the interleaved zeros on the upper half of the unit circle + (between 0 and pi, hence "normalized") in the standard decomposition + of the LPC filter into a symmetric part + and an anti-symmetric part (P and Q in ). +Because of non-linear effects in the decoding process, an implementation SHOULD + match the fixed-point arithmetic described in this section exactly. +An encoder SHOULD also use the same process. + + +The normalized LSFs are coded using a two-stage vector quantizer (VQ) + ( and ). +NB and MB frames use an order-10 predictor, while WB frames use an order-16 + predictor, and thus have different sets of tables. +After reconstructing the normalized LSFs + (), the decoder runs them through a + stabilization process (), interpolates + them between frames (), converts them + back into LPC coefficients (), and then runs + them through further processes to limit the range of the coefficients + () and the gain of the filter + (). +All of this is necessary to ensure the reconstruction process is stable. + + +
+ +The first VQ stage uses a 32-element codebook, coded with one of the PDFs in + , depending on the audio bandwidth and + the signal type of the current SILK frame. +This yields a single index, I1, for the entire frame, which + +Indexes an element in a coarse codebook, +Selects the PDFs for the second stage of the VQ, and +Selects the prediction weights used to remove intra-frame redundancy from + the second stage. + +The actual codebook elements are listed in + and + , but they are not needed until the last + stages of reconstructing the LSF coefficients. + + + +Audio Bandwidth +Signal Type +PDF +NB or MB Inactive or unvoiced + +{44, 34, 30, 19, 21, 12, 11, 3, + 3, 2, 16, 2, 2, 1, 5, 2, + 1, 3, 3, 1, 1, 2, 2, 2, + 3, 1, 9, 9, 2, 7, 2, 1}/256 + +NB or MB Voiced + +{1, 10, 1, 8, 3, 8, 8, 14, +13, 14, 1, 14, 12, 13, 11, 11, +12, 11, 10, 10, 11, 8, 9, 8, + 7, 8, 1, 1, 6, 1, 6, 5}/256 + +WB Inactive or unvoiced + +{31, 21, 3, 17, 1, 8, 17, 4, + 1, 18, 16, 4, 2, 3, 1, 10, + 1, 3, 16, 11, 16, 2, 2, 3, + 2, 11, 1, 4, 9, 8, 7, 3}/256 + +WB Voiced + +{1, 4, 16, 5, 18, 11, 5, 14, +15, 1, 3, 12, 13, 14, 14, 6, +14, 12, 2, 6, 1, 12, 12, 11, +10, 3, 10, 5, 1, 1, 1, 3}/256 + + + +
+ +
+ +A total of 16 PDFs are available for the LSF residual in the second stage: the + 8 (a...h) for NB and MB frames given in + , and the 8 (i...p) for WB frames + given in . +Which PDF is used for which coefficient is driven by the index, I1, + decoded in the first stage. + lists the letter of the + corresponding PDF for each normalized LSF coefficient for NB and MB, and + lists the same information for WB. + + + +Codebook +PDF +a {1, 1, 1, 15, 224, 11, 1, 1, 1}/256 +b {1, 1, 2, 34, 183, 32, 1, 1, 1}/256 +c {1, 1, 4, 42, 149, 55, 2, 1, 1}/256 +d {1, 1, 8, 52, 123, 61, 8, 1, 1}/256 +e {1, 3, 16, 53, 101, 74, 6, 1, 1}/256 +f {1, 3, 17, 55, 90, 73, 15, 1, 1}/256 +g {1, 7, 24, 53, 74, 67, 26, 3, 1}/256 +h {1, 1, 18, 63, 78, 58, 30, 6, 1}/256 + + + +Codebook +PDF +i {1, 1, 1, 9, 232, 9, 1, 1, 1}/256 +j {1, 1, 2, 28, 186, 35, 1, 1, 1}/256 +k {1, 1, 3, 42, 152, 53, 2, 1, 1}/256 +l {1, 1, 10, 49, 126, 65, 2, 1, 1}/256 +m {1, 4, 19, 48, 100, 77, 5, 1, 1}/256 +n {1, 1, 14, 54, 100, 72, 12, 1, 1}/256 +o {1, 1, 15, 61, 87, 61, 25, 4, 1}/256 +p {1, 7, 21, 50, 77, 81, 17, 1, 1}/256 + + + +I1 +Coefficient + +0 1 2 3 4 5 6 7 8 9 + 0 +a a a a a a a a a a + 1 +b d b c c b c b b b + 2 +c b b b b b b b b b + 3 +b c c c c b c b b b + 4 +c d d d d c c c c c + 5 +a f d d c c c c b b + g +a c c c c c c c c b + 7 +c d g e e e f e f f + 8 +c e f f e f e g e e + 9 +c e e h e f e f f e +10 +e d d d c d c c c c +11 +b f f g e f e f f f +12 +c h e g f f f f f f +13 +c h f f f f f g f e +14 +d d f e e f e f e e +15 +c d d f f e e e e e +16 +c e e g e f e f f f +17 +c f e g f f f e f e +18 +c h e f e f e f f f +19 +c f e g h g f g f e +20 +d g h e g f f g e f +21 +c h g e e e f e f f +22 +e f f e g g f g f e +23 +c f f g f g e g e e +24 +e f f f d h e f f e +25 +c d e f f g e f f e +26 +c d c d d e c d d d +27 +b b c c c c c d c c +28 +e f f g g g f g e f +29 +d f f e e e e d d c +30 +c f d h f f e e f e +31 +e e f e f g f g f e + + + +I1 +Coefficient + +0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 + 0 +i  i  i  i  i  i  i  i  i  i  i  i  i  i  i  i + 1 +k  l  l  l  l  l  k  k  k  k  k  j  j  j  i  l + 2 +k  n  n  l  p  m  m  n  k  n  m  n  n  m  l  l + 3 +i  k  j  k  k  j  j  j  j  j  i  i  i  i  i  j + 4 +i  o  n  m  o  m  p  n  m  m  m  n  n  m  m  l + 5 +i  l  n  n  m  l  l  n  l  l  l  l  l  l  k  m + 6 +i  i  i  i  i  i  i  i  i  i  i  i  i  i  i  i + 7 +i  k  o  l  p  k  n  l  m  n  n  m  l  l  k  l + 8 +i  o  k  o  o  m  n  m  o  n  m  m  n  l  l  l + 9 +k  j  i  i  i  i  i  i  i  i  i  i  i  i  i  i +10 +i  j  i  i  i  i  i  i  i  i  i  i  i  i  i  j +11 +k  k  l  m  n  l  l  l  l  l  l  l  k  k  j  l +12 +k  k  l  l  m  l  l  l  l  l  l  l  l  k  j  l +13 +l  m  m  m  o  m  m  n  l  n  m  m  n  m  l  m +14 +i  o  m  n  m  p  n  k  o  n  p  m  m  l  n  l +15 +i  j  i  j  j  j  j  j  j  j  i  i  i  i  j  i +16 +j  o  n  p  n  m  n  l  m  n  m  m  m  l  l  m +17 +j  l  l  m  m  l  l  n  k  l  l  n  n  n  l  m +18 +k  l  l  k  k  k  l  k  j  k  j  k  j  j  j  m +19 +i  k  l  n  l  l  k  k  k  j  j  i  i  i  i  i +20 +l  m  l  n  l  l  k  k  j  j  j  j  j  k  k  m +21 +k  o  l  p  p  m  n  m  n  l  n  l  l  k  l  l +22 +k  l  n  o  o  l  n  l  m  m  l  l  l  l  k  m +23 +j  l  l  m  m  m  m  l  n  n  n  l  j  j  j  j +24 +k  n  l  o  o  m  p  m  m  n  l  m  m  l  l  l +25 +i  o  j  j  i  i  i  i  i  i  i  i  i  i  i  i +26 +i  o  o  l  n  k  n  n  l  m  m  p  p  m  m  m +27 +l  l  p  l  n  m  l  l  l  k  k  l  l  l  k  l +28 +i  i  j  i  i  i  k  j  k  j  j  k  k  k  j  j +29 +i  l  k  n  l  l  k  l  k  j  i  i  j  i  i  j +30 +l  n  n  m  p  n  l  l  k  l  k  k  j  i  j  i +31 +k  l  n  l  m  l  l  l  k  j  k  o  m  i  i  i + + + +Decoding the second stage residual proceeds as follows. +For each coefficient, the decoder reads a symbol using the PDF corresponding to + I1 from either or + , and subtracts 4 from the result + to give an index in the range -4 to 4, inclusive. +If the index is either -4 or 4, it reads a second symbol using the PDF in + , and adds the value of this second symbol + to the index, using the same sign. +This gives the index, I2[k], a total range of -10 to 10, inclusive. + + + +PDF +{156, 60, 24, 9, 4, 2, 1}/256 + + + +The decoded indices from both stages are translated back into normalized LSF + coefficients in silk_NLSF_decode() (NLSF_decode.c). +The stage-2 indices represent residuals after both the first stage of the VQ + and a separate backwards-prediction step. +The backwards prediction process in the encoder subtracts a prediction from + each residual formed by a multiple of the coefficient that follows it. +The decoder must undo this process. + contains lists of prediction weights + for each coefficient. +There are two lists for NB and MB, and another two lists for WB, giving two + possible prediction weights for each coefficient. + + + +Coefficient +A +B +C +D + 0 179 116 175 68 + 1 138 67 148 62 + 2 140 82 160 66 + 3 148 59 176 60 + 4 151 92 178 72 + 5 149 72 173 117 + 6 153 100 174 85 + 7 151 89 164 90 + 8 163 92 177 118 + 9 174 136 +10 196 151 +11 182 142 +12 198 160 +13 192 142 +14 182 155 + + + +The prediction is undone using the procedure implemented in + silk_NLSF_residual_dequant() (NLSF_decode.c), which is as follows. +Each coefficient selects its prediction weight from one of the two lists based + on the stage-1 index, I1. + gives the selections for each + coefficient for NB and MB, and gives + the selections for WB. +Let d_LPC be the order of the codebook, i.e., 10 for NB and MB, and 16 for WB, + and let pred_Q8[k] be the weight for the k'th coefficient selected by this + process for 0 <= k < d_LPC-1. +Then, the stage-2 residual for each coefficient is computed via +
+>8 : 0) + + ((((I2[k]<<10) - sign(I2[k])*102)*qstep)>>16) , +]]> +
+ where qstep is the Q16 quantization step size, which is 11796 for NB and MB + and 9830 for WB (representing step sizes of approximately 0.18 and 0.15, + respectively). + + + +I1 +Coefficient + +0 1 2 3 4 5 6 7 8 + 0 +A B A A A A A A A + 1 +B A A A A A A A A + 2 +A A A A A A A A A + 3 +B B B A A A A B A + 4 +A B A A A A A A A + 5 +A B A A A A A A A + 6 +B A B B A A A B A + 7 +A B B A A B B A A + 8 +A A B B A B A B B + 9 +A A B B A A B B B +10 +A A A A A A A A A +11 +A B A B B B B B A +12 +A B A B B B B B A +13 +A B B B B B B B A +14 +B A B B A B B B B +15 +A B B B B B A B A +16 +A A B B A B A B A +17 +A A B B B A B B B +18 +A B B A A B B B A +19 +A A A B B B A B A +20 +A B B A A B A B A +21 +A B B A A A B B A +22 +A A A A A B B B B +23 +A A B B A A A B B +24 +A A A B A B B B B +25 +A B B B B B B B A +26 +A A A A A A A A A +27 +A A A A A A A A A +28 +A A B A B B A B A +29 +B A A B A A A A A +30 +A A A B B A B A B +31 +B A B B A B B B B + + + +I1 +Coefficient + +0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 + 0 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  D + 1 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  C + 2 +C  C  D  C  C  D  D  D  C  D  D  D  D  C  C + 3 +C  C  C  C  C  C  C  C  C  C  C  C  D  C  C + 4 +C  D  D  C  D  C  D  D  C  D  D  D  D  D  C + 5 +C  C  D  C  C  C  C  C  C  C  C  C  C  C  C + 6 +D  C  C  C  C  C  C  C  C  C  C  D  C  D  C + 7 +C  D  D  C  C  C  D  C  D  D  D  C  D  C  D + 8 +C  D  C  D  D  C  D  C  D  C  D  D  D  D  D + 9 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  D +10 +C  D  C  C  C  C  C  C  C  C  C  C  C  C  C +11 +C  C  D  C  D  D  D  D  D  D  D  C  D  C  C +12 +C  C  D  C  C  D  C  D  C  D  C  C  D  C  C +13 +C  C  C  C  D  D  C  D  C  D  D  D  D  C  C +14 +C  D  C  C  C  D  D  C  D  D  D  C  D  D  D +15 +C  C  D  D  C  C  C  C  C  C  C  C  D  D  C +16 +C  D  D  C  D  C  D  D  D  D  D  C  D  C  C +17 +C  C  D  C  C  C  C  D  C  C  D  D  D  C  C +18 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  D +19 +C  C  C  C  C  C  C  C  C  C  C  C  D  C  C +20 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  C +21 +C  D  C  D  C  D  D  C  D  C  D  C  D  D  C +22 +C  C  D  D  D  D  C  D  D  C  C  D  D  C  C +23 +C  D  D  C  D  C  D  C  D  C  C  C  C  D  C +24 +C  C  C  D  D  C  D  C  D  D  D  D  D  D  D +25 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  D +26 +C  D  D  C  C  C  D  D  C  C  D  D  D  D  D +27 +C  C  C  C  C  D  C  D  D  D  D  C  D  D  D +28 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  D +29 +C  C  C  C  C  C  C  C  C  C  C  C  C  C  D +30 +D  C  C  C  C  C  C  C  C  C  C  D  C  C  C +31 +C  C  D  C  C  D  D  D  C  C  D  C  C  D  C + + +
+ +
+ +Once the stage-1 index I1 and the stage-2 residual res_Q10[] have been decoded, + the final normalized LSF coefficients can be reconstructed. + + +The spectral distortion introduced by the quantization of each LSF coefficient + varies, so the stage-2 residual is weighted accordingly, using the + low-complexity Inverse Harmonic Mean Weighting (IHMW) function proposed in + . +The weights are derived directly from the stage-1 codebook vector. +Let cb1_Q8[k] be the k'th entry of the stage-1 codebook vector from + or + . +Then for 0 <= k < d_LPC the following expression + computes the square of the weight as a Q18 value: +
+ + + +
+ where cb1_Q8[-1] = 0 and cb1_Q8[d_LPC] = 256, and the + division is integer division. +This is reduced to an unsquared, Q9 value using the following square-root + approximation: +
+>(i-8)) & 127 +y = ((i&1) ? 32768 : 46214) >> ((32-i)>>1) +w_Q9[k] = y + ((213*f*y)>>16) +]]> +
+The constant 46214 here is approximately the square root of 2 in Q15. +The cb1_Q8[] vector completely determines these weights, and they may be + tabulated and stored as 13-bit unsigned values (with a range of 1819 to 5227, + inclusive) to avoid computing them when decoding. +The reference implementation already requires code to compute these weights on + unquantized coefficients in the encoder, in silk_NLSF_VQ_weights_laroia() + (NLSF_VQ_weights_laroia.c) and its callers, so it reuses that code in the + decoder instead of using a pre-computed table to reduce the amount of ROM + required. + + + +I1 +Codebook (Q8) + + 0   1   2   3   4   5   6   7   8   9 +0 +12  35  60  83 108 132 157 180 206 228 +1 +15  32  55  77 101 125 151 175 201 225 +2 +19  42  66  89 114 137 162 184 209 230 +3 +12  25  50  72  97 120 147 172 200 223 +4 +26  44  69  90 114 135 159 180 205 225 +5 +13  22  53  80 106 130 156 180 205 228 +6 +15  25  44  64  90 115 142 168 196 222 +7 +19  24  62  82 100 120 145 168 190 214 +8 +22  31  50  79 103 120 151 170 203 227 +9 +21  29  45  65 106 124 150 171 196 224 +10 +30  49  75  97 121 142 165 186 209 229 +11 +19  25  52  70  93 116 143 166 192 219 +12 +26  34  62  75  97 118 145 167 194 217 +13 +25  33  56  70  91 113 143 165 196 223 +14 +21  34  51  72  97 117 145 171 196 222 +15 +20  29  50  67  90 117 144 168 197 221 +16 +22  31  48  66  95 117 146 168 196 222 +17 +24  33  51  77 116 134 158 180 200 224 +18 +21  28  70  87 106 124 149 170 194 217 +19 +26  33  53  64  83 117 152 173 204 225 +20 +27  34  65  95 108 129 155 174 210 225 +21 +20  26  72  99 113 131 154 176 200 219 +22 +34  43  61  78  93 114 155 177 205 229 +23 +23  29  54  97 124 138 163 179 209 229 +24 +30  38  56  89 118 129 158 178 200 231 +25 +21  29  49  63  85 111 142 163 193 222 +26 +27  48  77 103 133 158 179 196 215 232 +27 +29  47  74  99 124 151 176 198 220 237 +28 +33  42  61  76  93 121 155 174 207 225 +29 +29  53  87 112 136 154 170 188 208 227 +30 +24  30  52  84 131 150 166 186 203 229 +31 +37  48  64  84 104 118 156 177 201 230 + + + +I1 +Codebook (Q8) + + 0  1  2  3  4   5   6   7   8   9  10  11  12  13  14  15 +0 + 7 23 38 54 69  85 100 116 131 147 162 178 193 208 223 239 +1 +13 25 41 55 69  83  98 112 127 142 157 171 187 203 220 236 +2 +15 21 34 51 61  78  92 106 126 136 152 167 185 205 225 240 +3 +10 21 36 50 63  79  95 110 126 141 157 173 189 205 221 237 +4 +17 20 37 51 59  78  89 107 123 134 150 164 184 205 224 240 +5 +10 15 32 51 67  81  96 112 129 142 158 173 189 204 220 236 +6 + 8 21 37 51 65  79  98 113 126 138 155 168 179 192 209 218 +7 +12 15 34 55 63  78  87 108 118 131 148 167 185 203 219 236 +8 +16 19 32 36 56  79  91 108 118 136 154 171 186 204 220 237 +9 +11 28 43 58 74  89 105 120 135 150 165 180 196 211 226 241 +10 + 6 16 33 46 60  75  92 107 123 137 156 169 185 199 214 225 +11 +11 19 30 44 57  74  89 105 121 135 152 169 186 202 218 234 +12 +12 19 29 46 57  71  88 100 120 132 148 165 182 199 216 233 +13 +17 23 35 46 56  77  92 106 123 134 152 167 185 204 222 237 +14 +14 17 45 53 63  75  89 107 115 132 151 171 188 206 221 240 +15 + 9 16 29 40 56  71  88 103 119 137 154 171 189 205 222 237 +16 +16 19 36 48 57  76  87 105 118 132 150 167 185 202 218 236 +17 +12 17 29 54 71  81  94 104 126 136 149 164 182 201 221 237 +18 +15 28 47 62 79  97 115 129 142 155 168 180 194 208 223 238 +19 + 8 14 30 45 62  78  94 111 127 143 159 175 192 207 223 239 +20 +17 30 49 62 79  92 107 119 132 145 160 174 190 204 220 235 +21 +14 19 36 45 61  76  91 108 121 138 154 172 189 205 222 238 +22 +12 18 31 45 60  76  91 107 123 138 154 171 187 204 221 236 +23 +13 17 31 43 53  70  83 103 114 131 149 167 185 203 220 237 +24 +17 22 35 42 58  78  93 110 125 139 155 170 188 206 224 240 +25 + 8 15 34 50 67  83  99 115 131 146 162 178 193 209 224 239 +26 +13 16 41 66 73  86  95 111 128 137 150 163 183 206 225 241 +27 +17 25 37 52 63  75  92 102 119 132 144 160 175 191 212 231 +28 +19 31 49 65 83 100 117 133 147 161 174 187 200 213 227 242 +29 +18 31 52 68 88 103 117 126 138 149 163 177 192 207 223 239 +30 +16 29 47 61 76  90 106 119 133 147 161 176 193 209 224 240 +31 +15 21 35 50 61  73  86  97 110 119 129 141 175 198 218 237 + + + +Given the stage-1 codebook entry cb1_Q8[], the stage-2 residual res_Q10[], and + their corresponding weights, w_Q9[], the reconstructed normalized LSF + coefficients are +
+ +
+ where the division is integer division. +However, nothing in either the reconstruction process or the + quantization process in the encoder thus far guarantees that the coefficients + are monotonically increasing and separated well enough to ensure a stable + filter . +When using the reference encoder, roughly 2% of frames violate this constraint. +The next section describes a stabilization procedure used to make these + guarantees. + + +
+ +
+ +The normalized LSF stabilization procedure is implemented in + silk_NLSF_stabilize() (NLSF_stabilize.c). +This process ensures that consecutive values of the normalized LSF + coefficients, NLSF_Q15[], are spaced some minimum distance apart + (predetermined to be the 0.01 percentile of a large training set). + gives the minimum spacings for NB and MB + and those for WB, where row k is the minimum allowed value of + NLSF_Q[k]-NLSF_Q[k-1]. +For the purposes of computing this spacing for the first and last coefficient, + NLSF_Q15[-1] is taken to be 0, and NLSF_Q15[d_LPC] is taken to be 32768. + + + +Coefficient +NB and MB +WB + 0 250 100 + 1 3 3 + 2 6 40 + 3 3 3 + 4 3 3 + 5 3 3 + 6 4 5 + 7 3 14 + 8 3 14 + 9 3 10 +10 461 11 +11 3 +12 8 +13 9 +14 7 +15 3 +16 347 + + + +The procedure starts off by trying to make small adjustments which attempt to + minimize the amount of distortion introduced. +After 20 such adjustments, it falls back to a more direct method which + guarantees the constraints are enforced but may require large adjustments. + + +Let NDeltaMin_Q15[k] be the minimum required spacing for the current audio + bandwidth from . +First, the procedure finds the index i where + NLSF_Q15[i] - NLSF_Q15[i-1] - NDeltaMin_Q15[i] is the + smallest, breaking ties by using the lower value of i. +If this value is non-negative, then the stabilization stops; the coefficients + satisfy all the constraints. +Otherwise, if i == 0, it sets NLSF_Q15[0] to NDeltaMin_Q15[0], and if + i == d_LPC, it sets NLSF_Q15[d_LPC-1] to + (32768 - NDeltaMin_Q15[d_LPC]). +For all other values of i, both NLSF_Q15[i-1] and NLSF_Q15[i] are updated as + follows: +
+>1) + \ NDeltaMin_Q15[k] + /_ + k=0 + d_LPC + __ + max_center_Q15 = 32768 - (NDeltaMin_Q15[i]>>1) - \ NDeltaMin_Q15[k] + /_ + k=i+1 +center_freq_Q15 = clamp(min_center_Q15[i], + (NLSF_Q15[i-1] + NLSF_Q15[i] + 1)>>1, + max_center_Q15[i]) + + NLSF_Q15[i-1] = center_freq_Q15 - (NDeltaMin_Q15[i]>>1) + + NLSF_Q15[i] = NLSF_Q15[i-1] + NDeltaMin_Q15[i] . +]]> +
+Then the procedure repeats again, until it has either executed 20 times or + has stopped because the coefficients satisfy all the constraints. + + +After the 20th repetition of the above procedure, the following fallback + procedure executes once. +First, the values of NLSF_Q15[k] for 0 <= k < d_LPC + are sorted in ascending order. +Then for each value of k from 0 to d_LPC-1, NLSF_Q15[k] is set to +
+ +
+Next, for each value of k from d_LPC-1 down to 0, NLSF_Q15[k] is set to +
+ +
+ + +
+ +
+ +For 20 ms SILK frames, the first half of the frame (i.e., the first two + subframes) may use normalized LSF coefficients that are interpolated between + the decoded LSFs for the most recent coded frame (in the same channel) and the + current frame. +A Q2 interpolation factor follows the LSF coefficient indices in the bitstream, + which is decoded using the PDF in . +This happens in silk_decode_indices() (decode_indices.c). +After either + +An uncoded regular SILK frame in the side channel, or +A decoder reset (see ), + + the decoder still decodes this factor, but ignores its value and always uses + 4 instead. +For 10 ms SILK frames, this factor is not stored at all. + + + +PDF +{13, 22, 29, 11, 181}/256 + + + +Let n2_Q15[k] be the normalized LSF coefficients decoded by the procedure in + , n0_Q15[k] be the LSF coefficients + decoded for the prior frame, and w_Q2 be the interpolation factor. +Then the normalized LSF coefficients used for the first half of a 20 ms + frame, n1_Q15[k], are +
+> 2) . +]]> +
+This interpolation is performed in silk_decode_parameters() + (decode_parameters.c). + +
+ +
+ +Any LPC filter A(z) can be split into a symmetric part P(z) and an + anti-symmetric part Q(z) such that +
+ +
+with +
+ +
+The even normalized LSF coefficients correspond to a pair of conjugate roots of + P(z), while the odd coefficients correspond to a pair of conjugate roots of + Q(z), all of which lie on the unit circle. +In addition, P(z) has a root at pi and Q(z) has a root at 0. +Thus, they may be reconstructed mathematically from a set of normalized LSF + coefficients, n[k], as +
+ +
+ + +However, SILK performs this reconstruction using a fixed-point approximation so + that all decoders can reproduce it in a bit-exact manner to avoid prediction + drift. +The function silk_NLSF2A() (NLSF2A.c) implements this procedure. + + +To start, it approximates cos(pi*n[k]) using a table lookup with linear + interpolation. +The encoder SHOULD use the inverse of this piecewise linear approximation, + rather than the true inverse of the cosine function, when deriving the + normalized LSF coefficients. +These values are also re-ordered to improve numerical accuracy when + constructing the LPC polynomials. + + + +Coefficient +NB and MB +WB + 0 0 0 + 1 9 15 + 2 6 8 + 3 3 7 + 4 4 4 + 5 5 11 + 6 8 12 + 7 1 3 + 8 2 2 + 9 7 13 +10 10 +11 5 +12 6 +13 9 +14 14 +15 1 + + + +The top 7 bits of each normalized LSF coefficient index a value in the table, + and the next 8 bits interpolate between it and the next value. +Let i = (n[k] >> 8) be the integer index and + f = (n[k] & 255) be the fractional part of a given + coefficient. +Then the re-ordered, approximated cosine, c_Q17[ordering[k]], is +
+> 3 , +]]> +
+ where ordering[k] is the k'th entry of the column of + corresponding to the current audio + bandwidth and cos_Q12[i] is the i'th entry of . + + + +i ++0 ++1 ++2 ++3 +0 + 4096 4095 4091 4085 +4 + 4076 4065 4052 4036 +8 + 4017 3997 3973 3948 +12 + 3920 3889 3857 3822 +16 + 3784 3745 3703 3659 +20 + 3613 3564 3513 3461 +24 + 3406 3349 3290 3229 +28 + 3166 3102 3035 2967 +32 + 2896 2824 2751 2676 +36 + 2599 2520 2440 2359 +40 + 2276 2191 2106 2019 +44 + 1931 1842 1751 1660 +48 + 1568 1474 1380 1285 +52 + 1189 1093 995 897 +56 + 799 700 601 501 +60 + 401 301 201 101 +64 + 0 -101 -201 -301 +68 + -401 -501 -601 -700 +72 + -799 -897 -995 -1093 +76 +-1189-1285-1380-1474 +80 +-1568-1660-1751-1842 +84 +-1931-2019-2106-2191 +88 +-2276-2359-2440-2520 +92 +-2599-2676-2751-2824 +96 +-2896-2967-3035-3102 +100 +-3166-3229-3290-3349 +104 +-3406-3461-3513-3564 +108 +-3613-3659-3703-3745 +112 +-3784-3822-3857-3889 +116 +-3920-3948-3973-3997 +120 +-4017-4036-4052-4065 +124 +-4076-4085-4091-4095 +128 +-4096 + + + +Given the list of cosine values, silk_NLSF2A_find_poly() (NLSF2A.c) + computes the coefficients of P and Q, described here via a simple recurrence. +Let p_Q16[k][j] and q_Q16[k][j] be the coefficients of the products of the + first (k+1) root pairs for P and Q, with j indexing the coefficient number. +Only the first (k+2) coefficients are needed, as the products are symmetric. +Let p_Q16[0][0] = q_Q16[0][0] = 1<<16, + p_Q16[0][1] = -c_Q17[0], q_Q16[0][1] = -c_Q17[1], and + d2 = d_LPC/2. +As boundary conditions, assume + p_Q16[k][j] = q_Q16[k][j] = 0 for all + j < 0. +Also, assume p_Q16[k][k+2] = p_Q16[k][k] and + q_Q16[k][k+2] = q_Q16[k][k] (because of the symmetry). +Then, for 0 < k < d2 and 0 <= j <= k+1, +
+>16) , + +q_Q16[k][j] = q_Q16[k-1][j] + q_Q16[k-1][j-2] + - ((c_Q17[2*k+1]*q_Q16[k-1][j-1] + 32768)>>16) . +]]> +
+The use of Q17 values for the cosine terms in an otherwise Q16 expression + implicitly scales them by a factor of 2. +The multiplications in this recurrence may require up to 48 bits of precision + in the result to avoid overflow. +In practice, each row of the recurrence only depends on the previous row, so an + implementation does not need to store all of them. + + +silk_NLSF2A() uses the values from the last row of this recurrence to + reconstruct a 32-bit version of the LPC filter (without the leading 1.0 + coefficient), a32_Q17[k], 0 <= k < d2: +
+ +
+The sum and difference of two terms from each of the p_Q16 and q_Q16 + coefficient lists reflect the (1 + z**-1) and + (1 - z**-1) factors of P and Q, respectively. +The promotion of the expression from Q16 to Q17 implicitly scales the result + by 1/2. + +
+ +
+ +The a32_Q17[] coefficients are too large to fit in a 16-bit value, which + significantly increases the cost of applying this filter in fixed-point + decoders. +Reducing them to Q12 precision doesn't incur any significant quality loss, + but still does not guarantee they will fit. +silk_NLSF2A() applies up to 10 rounds of bandwidth expansion to limit + the dynamic range of these coefficients. +Even floating-point decoders SHOULD perform these steps, to avoid mismatch. + + +For each round, the process first finds the index k such that abs(a32_Q17[k]) + is largest, breaking ties by choosing the lowest value of k. +Then, it computes the corresponding Q12 precision value, maxabs_Q12, subject to + an upper bound to avoid overflow in subsequent computations: +
+> 5, 163838) . +]]> +
+If this is larger than 32767, the procedure derives the chirp factor, + sc_Q16[0], to use in the bandwidth expansion as +
+> 2 +]]> +
+ where the division here is integer division. +This is an approximation of the chirp factor needed to reduce the target + coefficient to 32767, though it is both less than 0.999 and, for + k > 0 when maxabs_Q12 is much greater than 32767, still slightly + too large. +The upper bound on maxabs_Q12, 163838, was chosen because it is equal to + ((2**31 - 1) >> 14) + 32767, i.e., the + largest value of maxabs_Q12 that would not overflow the numerator in the + equation above when stored in a signed 32-bit integer. + + +silk_bwexpander_32() (bwexpander_32.c) performs the bandwidth expansion (again, + only when maxabs_Q12 is greater than 32767) using the following recurrence: +
+> 16 + +sc_Q16[k+1] = (sc_Q16[0]*sc_Q16[k] + 32768) >> 16 +]]> +
+The first multiply may require up to 48 bits of precision in the result to + avoid overflow. +The second multiply must be unsigned to avoid overflow with only 32 bits of + precision. +The reference implementation uses a slightly more complex formulation that + avoids the 32-bit overflow using signed multiplication, but is otherwise + equivalent. + + +After 10 rounds of bandwidth expansion are performed, they are simply saturated + to 16 bits: +
+> 5, 32767) << 5 . +]]> +
+Because this performs the actual saturation in the Q12 domain, but converts the + coefficients back to the Q17 domain for the purposes of prediction gain + limiting, this step must be performed after the 10th round of bandwidth + expansion, regardless of whether or not the Q12 version of any coefficient + still overflows a 16-bit integer. +This saturation is not performed if maxabs_Q12 drops to 32767 or less prior to + the 10th round. + +
+ +
+ +The prediction gain of an LPC synthesis filter is the square-root of the output + energy when the filter is excited by a unit-energy impulse. +Even if the Q12 coefficients would fit, the resulting filter may still have a + significant gain (especially for voiced sounds), making the filter unstable. +silk_NLSF2A() applies up to 18 additional rounds of bandwidth expansion to + limit the prediction gain. +Instead of controlling the amount of bandwidth expansion using the prediction + gain itself (which may diverge to infinity for an unstable filter), + silk_NLSF2A() uses silk_LPC_inverse_pred_gain_QA() (LPC_inv_pred_gain.c) to + compute the reflection coefficients associated with the filter. +The filter is stable if and only if the magnitude of these coefficients is + sufficiently less than one. +The reflection coefficients, rc[k], can be computed using a simple Levinson + recurrence, initialized with the LPC coefficients + a[d_LPC-1][n] = a[n], and then updated via +
+ +
+ + +However, silk_LPC_inverse_pred_gain_QA() approximates this using fixed-point + arithmetic to guarantee reproducible results across platforms and + implementations. +Since small changes in the coefficients can make a stable filter unstable, it + takes the real Q12 coefficients that will be used during reconstruction as + input. +Thus, let +
+> 5 +]]> +
+ be the Q12 version of the LPC coefficients that will eventually be used. +As a simple initial check, the decoder computes the DC response as +
+ +
+ and if DC_resp > 4096, the filter is unstable. + + +Increasing the precision of these Q12 coefficients to Q24 for intermediate + computations allows more accurate computation of the reflection coefficients, + so the decoder initializes the recurrence via +
+ +
+Then for each k from d_LPC-1 down to 0, if + abs(a32_Q24[k][k]) > 16773022, the filter is unstable and the + recurrence stops. +The constant 16773022 here is approximately 0.99975 in Q24. +Otherwise, row k-1 of a32_Q24 is computed from row k as +
+> 32) , + + b1[k] = ilog(div_Q30[k]) , + + b2[k] = b1[k] - 16 , + + (1<<29) - 1 + inv_Qb2[k] = ----------------------- , + div_Q30[k] >> (b2[k]+1) + + err_Q29[k] = (1<<29) + - ((div_Q30[k]<<(15-b2[k]))*inv_Qb2[k] >> 16) , + + gain_Qb1[k] = ((inv_Qb2[k] << 16) + + (err_Q29[k]*inv_Qb2[k] >> 13)) , + +num_Q24[k-1][n] = a32_Q24[k][n] + - ((a32_Q24[k][k-n-1]*rc_Q31[k] + (1<<30)) >> 31) , + +a32_Q24[k-1][n] = (num_Q24[k-1][n]*gain_Qb1[k] + + (1<<(b1[k]-1))) >> b1[k] , +]]> +
+ where 0 <= n < k. +Here, rc_Q30[k] are the reflection coefficients. +div_Q30[k] is the denominator for each iteration, and gain_Qb1[k] is its + multiplicative inverse (with b1[k] fractional bits, where b1[k] ranges from + 20 to 31). +inv_Qb2[k], which ranges from 16384 to 32767, is a low-precision version of + that inverse (with b2[k] fractional bits). +err_Q29[k] is the residual error, ranging from -32763 to 32392, which is used + to improve the accuracy. +The values t_Q24[k-1][n] for each n are the numerators for the next row of + coefficients in the recursion, and a32_Q24[k-1][n] is the final version of + that row. +Every multiply in this procedure except the one used to compute gain_Qb1[k] + requires more than 32 bits of precision, but otherwise all intermediate + results fit in 32 bits or less. +In practice, because each row only depends on the next one, an implementation + does not need to store them all. + + +If abs(a32_Q24[k][k]) <= 16773022 for + 0 <= k < d_LPC, then the filter is considered stable. +However, the problem of determining stability is ill-conditioned when the + filter contains several reflection coefficients whose magnitude is very close + to one. +This fixed-point algorithm is not mathematically guaranteed to correctly + classify filters as stable or unstable in this case, though it does very well + in practice. + + +On round i, 1 <= i <= 18, if the filter passes these + stability checks, then this procedure stops, and the final LPC coefficients to + use for reconstruction in are +
+> 5 . +]]> +
+Otherwise, a round of bandwidth expansion is applied using the same procedure + as in , with +
+ +
+During the 15th round, sc_Q16[0] becomes 0 in the above equation, so a_Q12[k] + is set to 0 for all k, guaranteeing a stable filter. + +
+ +
+ +
+ +After the normalized LSF indices and, for 20 ms frames, the LSF + interpolation index, voiced frames (see ) + include additional LTP parameters. +There is one primary lag index for each SILK frame, but this is refined to + produce a separate lag index per subframe using a vector quantizer. +Each subframe also gets its own prediction gain coefficient. + + +
+ +The primary lag index is coded either relative to the primary lag of the prior + frame in the same channel, or as an absolute index. +Absolute coding is used if and only if + + +This is the first SILK frame of its type (LBRR or regular) for this channel in + the current Opus frame, + + +The previous SILK frame of the same type (LBRR or regular) for this channel in + the same Opus frame was not coded, or + + +That previous SILK frame was coded, but was not voiced (see + ). + + + + + +With absolute coding, the primary pitch lag may range from 2 ms + (inclusive) up to 18 ms (exclusive), corresponding to pitches from + 500 Hz down to 55.6 Hz, respectively. +It is comprised of a high part and a low part, where the decoder reads the high + part using the 32-entry codebook in + and the low part using the codebook corresponding to the current audio + bandwidth from . +The final primary pitch lag is then +
+ +
+ where lag_high is the high part, lag_low is the low part, and lag_scale + and lag_min are the values from the "Scale" and "Minimum Lag" columns of + , respectively. + + + +PDF +{3, 3, 6, 11, 21, 30, 32, 19, + 11, 10, 12, 13, 13, 12, 11, 9, + 8, 7, 6, 4, 2, 2, 2, 1, + 1, 1, 1, 1, 1, 1, 1, 1}/256 + + + +Audio Bandwidth +PDF +Scale +Minimum Lag +Maximum Lag +NB {64, 64, 64, 64}/256 4 16 144 +MB {43, 42, 43, 43, 42, 43}/256 6 24 216 +WB {32, 32, 32, 32, 32, 32, 32, 32}/256 8 32 288 + + + +All frames that do not use absolute coding for the primary lag index use + relative coding instead. +The decoder reads a single delta value using the 21-entry PDF in + . +If the resulting value is zero, it falls back to the absolute coding procedure + from the prior paragraph. +Otherwise, the final primary pitch lag is then +
+ +
+ where previous_lag is the primary pitch lag from the most recent frame in the + same channel and delta_lag_index is the value just decoded. +This allows a per-frame change in the pitch lag of -8 to +11 samples. +The decoder does no clamping at this point, so this value can fall outside the + range of 2 ms to 18 ms, and the decoder must use this unclamped + value when using relative coding in the next SILK frame (if any). +However, because an Opus frame can use relative coding for at most two + consecutive SILK frames, integer overflow should not be an issue. + + + +PDF +{46, 2, 2, 3, 4, 6, 10, 15, + 26, 38, 30, 22, 15, 10, 7, 6, + 4, 4, 2, 2, 2}/256 + + + +After the primary pitch lag, a "pitch contour", stored as a single entry from + one of four small VQ codebooks, gives lag offsets for each subframe in the + current SILK frame. +The codebook index is decoded using one of the PDFs in + depending on the current frame size + and audio bandwidth. +Tables  + through  + give the corresponding offsets to apply to the primary pitch lag for each + subframe given the decoded codebook index. + + + +Audio Bandwidth +SILK Frame Size +Codebook Size +PDF +NB 10 ms 3 +{143, 50, 63}/256 +NB 20 ms 11 +{68, 12, 21, 17, 19, 22, 30, 24, + 17, 16, 10}/256 +MB or WB 10 ms 12 +{91, 46, 39, 19, 14, 12, 8, 7, + 6, 5, 5, 4}/256 +MB or WB 20 ms 34 +{33, 22, 18, 16, 15, 14, 14, 13, + 13, 10, 9, 9, 8, 6, 6, 6, + 5, 4, 4, 4, 3, 3, 3, 2, + 2, 2, 2, 2, 2, 2, 1, 1, + 1, 1}/256 + + + +Index +Subframe Offsets +0  0  0 +1  1  0 +2  0  1 + + + +Index +Subframe Offsets + 0  0  0  0  0 + 1  2  1  0 -1 + 2 -1  0  1  2 + 3 -1  0  0  1 + 4 -1  0  0  0 + 5  0  0  0  1 + 6  0  0  1  1 + 7  1  1  0  0 + 8  1  0  0  0 + 9  0  0  0 -1 +10  1  0  0 -1 + + + +Index +Subframe Offsets + 0  0  0 + 1  0  1 + 2  1  0 + 3 -1  1 + 4  1 -1 + 5 -1  2 + 6  2 -1 + 7 -2  2 + 8  2 -2 + 9 -2  3 +10  3 -2 +11 -3  3 + + + +Index +Subframe Offsets + 0  0  0  0  0 + 1  0  0  1  1 + 2  1  1  0  0 + 3 -1  0  0  0 + 4  0  0  0  1 + 5  1  0  0  0 + 6 -1  0  0  1 + 7  0  0  0 -1 + 8 -1  0  1  2 + 9  1  0  0 -1 +10 -2 -1  1  2 +11  2  1  0 -1 +12 -2  0  0  2 +13 -2  0  1  3 +14  2  1 -1 -2 +15 -3 -1  1  3 +16  2  0  0 -2 +17  3  1  0 -2 +18 -3 -1  2  4 +19 -4 -1  1  4 +20  3  1 -1 -3 +21 -4 -1  2  5 +22  4  2 -1 -3 +23  4  1 -1 -4 +24 -5 -1  2  6 +25  5  2 -1 -4 +26 -6 -2  2  6 +27 -5 -2  2  5 +28  6  2 -1 -5 +29 -7 -2  3  8 +30  6  2 -2 -6 +31  5  2 -2 -5 +32  8  3 -2 -7 +33 -9 -3  3  9 + + + +The final pitch lag for each subframe is assembled in silk_decode_pitch() + (decode_pitch.c). +Let lag be the primary pitch lag for the current SILK frame, contour_index be + index of the VQ codebook, and lag_cb[contour_index][k] be the corresponding + entry of the codebook from the appropriate table given above for the k'th + subframe. +Then the final pitch lag for that subframe is +
+ +
+ where lag_min and lag_max are the values from the "Minimum Lag" and + "Maximum Lag" columns of , + respectively. + + +
+ +
+ +SILK uses a separate 5-tap pitch filter for each subframe, selected from one + of three codebooks. +The three codebooks each represent different rate-distortion trade-offs, with + average rates of 1.61 bits/subframe, 3.68 bits/subframe, and + 4.85 bits/subframe, respectively. + + + +The importance of the filter coefficients generally depends on two factors: the + periodicity of the signal and relative energy between the current subframe and + the signal from one period earlier. +Greater periodicity and decaying energy both lead to more important filter + coefficients, and thus should be coded with lower distortion and higher rate. +These properties are relatively stable over the duration of a single SILK + frame, hence all of the subframes in a SILK frame choose their filter from the + same codebook. +This is signaled with an explicitly-coded "periodicity index". +This immediately follows the subframe pitch lags, and is coded using the + 3-entry PDF from . + + + +PDF +{77, 80, 99}/256 + + + +The indices of the filters for each subframe follow. +They are all coded using the PDF from + corresponding to the periodicity index. +Tables  + through  + contain the corresponding filter taps as signed Q7 integers. + + + +Periodicity Index +Codebook Size +PDF +0 8 {185, 15, 13, 13, 9, 9, 6, 6}/256 +1 16 {57, 34, 21, 20, 15, 13, 12, 13, + 10, 10, 9, 10, 9, 8, 7, 8}/256 +2 32 {15, 16, 14, 12, 12, 12, 11, 11, + 11, 10, 9, 9, 9, 9, 8, 8, + 8, 8, 7, 7, 6, 6, 5, 4, + 5, 4, 4, 4, 3, 4, 3, 2}/256 + + + +Index +Filter Taps (Q7) + 0 +  4   6  24   7   5 + 1 +  0   0   2   0   0 + 2 + 12  28  41  13  -4 + 3 + -9  15  42  25  14 + 4 +  1  -2  62  41  -9 + 5 +-10  37  65  -4   3 + 6 + -6   4  66   7  -8 + 7 + 16  14  38  -3  33 + + + +Index +Filter Taps (Q7) + + 0 + 13  22  39  23  12 + 1 + -1  36  64  27  -6 + 2 + -7  10  55  43  17 + 3 +  1   1   8   1   1 + 4 +  6 -11  74  53  -9 + 5 +-12  55  76 -12   8 + 6 + -3   3  93  27  -4 + 7 + 26  39  59   3  -8 + 8 +  2   0  77  11   9 + 9 + -8  22  44  -6   7 +10 + 40   9  26   3   9 +11 + -7  20 101  -7   4 +12 +  3  -8  42  26   0 +13 +-15  33  68   2  23 +14 + -2  55  46  -2  15 +15 +  3  -1  21  16  41 + + + +Index +Filter Taps (Q7) + 0 + -6  27  61  39   5 + 1 +-11  42  88   4   1 + 2 + -2  60  65   6  -4 + 3 + -1  -5  73  56   1 + 4 + -9  19  94  29  -9 + 5 +  0  12  99   6   4 + 6 +  8 -19 102  46 -13 + 7 +  3   2  13   3   2 + 8 +  9 -21  84  72 -18 + 9 +-11  46 104 -22   8 +10 + 18  38  48  23   0 +11 +-16  70  83 -21  11 +12 +  5 -11 117  22  -8 +13 + -6  23 117 -12   3 +14 +  3  -8  95  28   4 +15 +-10  15  77  60 -15 +16 + -1   4 124   2  -4 +17 +  3  38  84  24 -25 +18 +  2  13  42  13  31 +19 + 21  -4  56  46  -1 +20 + -1  35  79 -13  19 +21 + -7  65  88  -9 -14 +22 + 20   4  81  49 -29 +23 + 20   0  75   3 -17 +24 +  5  -9  44  92  -8 +25 +  1  -3  22  69  31 +26 + -6  95  41 -12   5 +27 + 39  67  16  -4   1 +28 +  0  -6 120  55 -36 +29 +-13  44 122   4 -24 +30 + 81   5  11   3   7 +31 +  2   0   9  10  88 + + +
+ +
+ +An LTP scaling parameter appears after the LTP filter coefficients if and only + if + +This is a voiced frame (see ), and +Either + + +This SILK frame corresponds to the first time interval of the + current Opus frame for its type (LBRR or regular), or + + +This is an LBRR frame where the LBRR flags (see + ) indicate the previous LBRR frame in the same + channel is not coded. + + + + +This allows the encoder to trade off the prediction gain between + packets against the recovery time after packet loss. +Unlike absolute-coding for pitch lags, regular SILK frames that are not at the + start of an Opus frame (i.e., that do not correspond to the first 20 ms + time interval in Opus frames of 40 or 60 ms) do not include this + field, even if the prior frame was not voiced, or (in the case of the side + channel) not even coded. +After an uncoded frame in the side channel, the LTP buffer (see + ) is cleared to zero, and is thus in a + known state. +In contrast, LBRR frames do include this field when the prior frame was not + coded, since the LTP buffer contains the output of the PLC, which is + non-normative. + + +If present, the decoder reads a value using the 3-entry PDF in + . +The three possible values represent Q14 scale factors of 15565, 12288, and + 8192, respectively (corresponding to approximately 0.95, 0.75, and 0.5). +Frames that do not code the scaling parameter use the default factor of 15565 + (approximately 0.95). + + + +PDF +{128, 64, 64}/256 + + +
+ +
+ +
+ +As described in , SILK uses a + linear congruential generator (LCG) to inject pseudorandom noise into the + quantized excitation. +To ensure synchronization of this process between the encoder and decoder, each + SILK frame stores a 2-bit seed after the LTP parameters (if any). +The encoder may consider the choice of seed during quantization, and the + flexibility of this choice lets it reduce distortion, helping to pay for the + bit cost required to signal it. +The decoder reads the seed using the uniform 4-entry PDF in + , yielding a value between 0 and 3, inclusive. + + + +PDF +{64, 64, 64, 64}/256 + + +
+ +
+ +SILK codes the excitation using a modified version of the Pyramid Vector + Quantization (PVQ) codebook . +The PVQ codebook is designed for Laplace-distributed values and consists of all + sums of K signed, unit pulses in a vector of dimension N, where two pulses at + the same position are required to have the same sign. +Thus the codebook includes all integer codevectors y of dimension N that + satisfy +
+ +
+Unlike regular PVQ, SILK uses a variable-length, rather than fixed-length, + encoding. +This encoding is better suited to the more Gaussian-like distribution of the + coefficient magnitudes and the non-uniform distribution of their signs (caused + by the quantization offset described below). +SILK also handles large codebooks by coding the least significant bits (LSBs) + of each coefficient directly. +This adds a small coding efficiency loss, but greatly reduces the computation + time and ROM size required for decoding, as implemented in + silk_decode_pulses() (decode_pulses.c). + + + +SILK fixes the dimension of the codebook to N = 16. +The excitation is made up of a number of "shell blocks", each 16 samples in + size. + lists the number of shell blocks + required for a SILK frame for each possible audio bandwidth and frame size. +10 ms MB frames nominally contain 120 samples (10 ms at + 12 kHz), which is not a multiple of 16. +This is handled by coding 8 shell blocks (128 samples) and discarding the final + 8 samples of the last block. +The decoder contains no special case that prevents an encoder from placing + pulses in these samples, and they must be correctly parsed from the bitstream + if present, but they are otherwise ignored. + + + +Audio Bandwidth +Frame Size +Number of Shell Blocks +NB 10 ms 5 +MB 10 ms 8 +WB 10 ms 10 +NB 20 ms 10 +MB 20 ms 15 +WB 20 ms 20 + + +
+ +The first symbol in the excitation is a "rate level", which is an index from 0 + to 8, inclusive, coded using the PDF in + corresponding to the signal type of the current frame (from + ). +The rate level selects the PDF used to decode the number of pulses in + the individual shell blocks. +It does not directly convey any information about the bitrate or the number of + pulses itself, but merely changes the probability of the symbols in + . +Level 0 provides a more efficient encoding at low rates generally, and + level 8 provides a more efficient encoding at high rates generally, + though the most efficient level for a particular SILK frame may depend on the + exact distribution of the coded symbols. +An encoder should, but is not required to, use the most efficient rate level. + + + +Signal Type +PDF +Inactive or Unvoiced +{15, 51, 12, 46, 45, 13, 33, 27, 14}/256 +Voiced +{33, 30, 36, 17, 34, 49, 18, 21, 18}/256 + + +
+ +
+ +The total number of pulses in each of the shell blocks follows the rate level. +The pulse counts for all of the shell blocks are coded consecutively, before + the content of any of the blocks. +Each block may have anywhere from 0 to 16 pulses, inclusive, coded using the + 18-entry PDF in corresponding to the + rate level from . +The special value 17 indicates that this block has one or more additional + LSBs to decode for each coefficient. +If the decoder encounters this value, it decodes another value for the actual + pulse count of the block, but uses the PDF corresponding to the special rate + level 9 instead of the normal rate level. +This process repeats until the decoder reads a value less than 17, and it then + sets the number of extra LSBs used to the number of 17's decoded for that + block. +If it reads the value 17 ten times, then the next iteration uses the special + rate level 10 instead of 9. +The probability of decoding a 17 when using the PDF for rate level 10 is + zero, ensuring that the number of LSBs for a block will not exceed 10. +The cumulative distribution for rate level 10 is just a shifted version of + that for 9 and thus does not require any additional storage. + + + +Rate Level +PDF +0 +{131, 74, 25, 8, 3, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}/256 +1 +{58, 93, 60, 23, 7, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}/256 +2 +{43, 51, 46, 33, 24, 16, 11, 8, 6, 3, 3, 3, 2, 1, 1, 2, 1, 2}/256 +3 +{17, 52, 71, 57, 31, 12, 5, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}/256 +4 +{6, 21, 41, 53, 49, 35, 21, 11, 6, 3, 2, 2, 1, 1, 1, 1, 1, 1}/256 +5 +{7, 14, 22, 28, 29, 28, 25, 20, 17, 13, 11, 9, 7, 5, 4, 4, 3, 10}/256 +6 +{2, 5, 14, 29, 42, 46, 41, 31, 19, 11, 6, 3, 2, 1, 1, 1, 1, 1}/256 +7 +{1, 2, 4, 10, 19, 29, 35, 37, 34, 28, 20, 14, 8, 5, 4, 2, 2, 2}/256 +8 +{1, 2, 2, 5, 9, 14, 20, 24, 27, 28, 26, 23, 20, 15, 11, 8, 6, 15}/256 +9 +{1, 1, 1, 6, 27, 58, 56, 39, 25, 14, 10, 6, 3, 3, 2, 1, 1, 2}/256 +10 +{2, 1, 6, 27, 58, 56, 39, 25, 14, 10, 6, 3, 3, 2, 1, 1, 2, 0}/256 + + +
+ +
+ +The locations of the pulses in each shell block follow the pulse counts, + as decoded by silk_shell_decoder() (shell_coder.c). +As with the pulse counts, these locations are coded for all the shell blocks + before any of the remaining information for each block. +Unlike many other codecs, SILK places no restriction on the distribution of + pulses within a shell block. +All of the pulses may be placed in a single location, or each one in a unique + location, or anything in between. + + + +The location of pulses is coded by recursively partitioning each block into + halves, and coding how many pulses fall on the left side of the split. +All remaining pulses must fall on the right side of the split. +The process then recurses into the left half, and after that returns, the + right half (preorder traversal). +The PDF to use is chosen by the size of the current partition (16, 8, 4, or 2) + and the number of pulses in the partition (1 to 16, inclusive). +Tables  + through  list the + PDFs used for each partition size and pulse count. +This process skips partitions without any pulses, i.e., where the initial pulse + count from was zero, or where the split in + the prior level indicated that all of the pulses fell on the other side. +These partitions have nothing to code, so they require no PDF. + + + +Pulse Count +PDF + 1 {126, 130}/256 + 2 {56, 142, 58}/256 + 3 {25, 101, 104, 26}/256 + 4 {12, 60, 108, 64, 12}/256 + 5 {7, 35, 84, 87, 37, 6}/256 + 6 {4, 20, 59, 86, 63, 21, 3}/256 + 7 {3, 12, 38, 72, 75, 42, 12, 2}/256 + 8 {2, 8, 25, 54, 73, 59, 27, 7, 1}/256 + 9 {2, 5, 17, 39, 63, 65, 42, 18, 4, 1}/256 +10 {1, 4, 12, 28, 49, 63, 54, 30, 11, 3, 1}/256 +11 {1, 4, 8, 20, 37, 55, 57, 41, 22, 8, 2, 1}/256 +12 {1, 3, 7, 15, 28, 44, 53, 48, 33, 16, 6, 1, 1}/256 +13 {1, 2, 6, 12, 21, 35, 47, 48, 40, 25, 12, 5, 1, 1}/256 +14 {1, 1, 4, 10, 17, 27, 37, 47, 43, 33, 21, 9, 4, 1, 1}/256 +15 {1, 1, 1, 8, 14, 22, 33, 40, 43, 38, 28, 16, 8, 1, 1, 1}/256 +16 {1, 1, 1, 1, 13, 18, 27, 36, 41, 41, 34, 24, 14, 1, 1, 1, 1}/256 + + + +Pulse Count +PDF + 1 {127, 129}/256 + 2 {53, 149, 54}/256 + 3 {22, 105, 106, 23}/256 + 4 {11, 61, 111, 63, 10}/256 + 5 {6, 35, 86, 88, 36, 5}/256 + 6 {4, 20, 59, 87, 62, 21, 3}/256 + 7 {3, 13, 40, 71, 73, 41, 13, 2}/256 + 8 {3, 9, 27, 53, 70, 56, 28, 9, 1}/256 + 9 {3, 8, 19, 37, 57, 61, 44, 20, 6, 1}/256 +10 {3, 7, 15, 28, 44, 54, 49, 33, 17, 5, 1}/256 +11 {1, 7, 13, 22, 34, 46, 48, 38, 28, 14, 4, 1}/256 +12 {1, 1, 11, 22, 27, 35, 42, 47, 33, 25, 10, 1, 1}/256 +13 {1, 1, 6, 14, 26, 37, 43, 43, 37, 26, 14, 6, 1, 1}/256 +14 {1, 1, 4, 10, 20, 31, 40, 42, 40, 31, 20, 10, 4, 1, 1}/256 +15 {1, 1, 3, 8, 16, 26, 35, 38, 38, 35, 26, 16, 8, 3, 1, 1}/256 +16 {1, 1, 2, 6, 12, 21, 30, 36, 38, 36, 30, 21, 12, 6, 2, 1, 1}/256 + + + +Pulse Count +PDF + 1 {127, 129}/256 + 2 {49, 157, 50}/256 + 3 {20, 107, 109, 20}/256 + 4 {11, 60, 113, 62, 10}/256 + 5 {7, 36, 84, 87, 36, 6}/256 + 6 {6, 24, 57, 82, 60, 23, 4}/256 + 7 {5, 18, 39, 64, 68, 42, 16, 4}/256 + 8 {6, 14, 29, 47, 61, 52, 30, 14, 3}/256 + 9 {1, 15, 23, 35, 51, 50, 40, 30, 10, 1}/256 +10 {1, 1, 21, 32, 42, 52, 46, 41, 18, 1, 1}/256 +11 {1, 6, 16, 27, 36, 42, 42, 36, 27, 16, 6, 1}/256 +12 {1, 5, 12, 21, 31, 38, 40, 38, 31, 21, 12, 5, 1}/256 +13 {1, 3, 9, 17, 26, 34, 38, 38, 34, 26, 17, 9, 3, 1}/256 +14 {1, 3, 7, 14, 22, 29, 34, 36, 34, 29, 22, 14, 7, 3, 1}/256 +15 {1, 2, 5, 11, 18, 25, 31, 35, 35, 31, 25, 18, 11, 5, 2, 1}/256 +16 {1, 1, 4, 9, 15, 21, 28, 32, 34, 32, 28, 21, 15, 9, 4, 1, 1}/256 + + + +Pulse Count +PDF + 1 {128, 128}/256 + 2 {42, 172, 42}/256 + 3 {21, 107, 107, 21}/256 + 4 {12, 60, 112, 61, 11}/256 + 5 {8, 34, 86, 86, 35, 7}/256 + 6 {8, 23, 55, 90, 55, 20, 5}/256 + 7 {5, 15, 38, 72, 72, 36, 15, 3}/256 + 8 {6, 12, 27, 52, 77, 47, 20, 10, 5}/256 + 9 {6, 19, 28, 35, 40, 40, 35, 28, 19, 6}/256 +10 {4, 14, 22, 31, 37, 40, 37, 31, 22, 14, 4}/256 +11 {3, 10, 18, 26, 33, 38, 38, 33, 26, 18, 10, 3}/256 +12 {2, 8, 13, 21, 29, 36, 38, 36, 29, 21, 13, 8, 2}/256 +13 {1, 5, 10, 17, 25, 32, 38, 38, 32, 25, 17, 10, 5, 1}/256 +14 {1, 4, 7, 13, 21, 29, 35, 36, 35, 29, 21, 13, 7, 4, 1}/256 +15 {1, 2, 5, 10, 17, 25, 32, 36, 36, 32, 25, 17, 10, 5, 2, 1}/256 +16 {1, 2, 4, 7, 13, 21, 28, 34, 36, 34, 28, 21, 13, 7, 4, 2, 1}/256 + + +
+ +
+ +After the decoder reads the pulse locations for all blocks, it reads the LSBs + (if any) for each block in turn. +Inside each block, it reads all the LSBs for each coefficient in turn, even + those where no pulses were allocated, before proceeding to the next one. +For 10 ms MB frames, it reads LSBs even for the extra 8 samples in + the last block. +The LSBs are coded from most significant to least significant, and they all use + the PDF in . + + + +PDF +{136, 120}/256 + + + +The number of LSBs read for each coefficient in a block is determined in + . +The magnitude of the coefficient is initially equal to the number of pulses + placed at that location in . +As each LSB is decoded, the magnitude is doubled, and then the value of the LSB + added to it, to obtain an updated magnitude. + +
+ +
+ +After decoding the pulse locations and the LSBs, the decoder knows the + magnitude of each coefficient in the excitation. +It then decodes a sign for all coefficients with a non-zero magnitude, using + one of the PDFs from . +If the value decoded is 0, then the coefficient magnitude is negated. +Otherwise, it remains positive. + + + +The decoder chooses the PDF for the sign based on the signal type and + quantization offset type (from ) and the + number of pulses in the block (from ). +The number of pulses in the block does not take into account any LSBs. +Most PDFs are skewed towards negative signs because of the quantization offset, + but the PDFs for zero pulses are highly skewed towards positive signs. +If a block contains many positive coefficients, it is sometimes beneficial to + code it solely using LSBs (i.e., with zero pulses), since the encoder may be + able to save enough bits on the signs to justify the less efficient + coefficient magnitude encoding. + + + +Signal Type +Quantization Offset Type +Pulse Count +PDF +Inactive Low 0 {2, 254}/256 +Inactive Low 1 {207, 49}/256 +Inactive Low 2 {189, 67}/256 +Inactive Low 3 {179, 77}/256 +Inactive Low 4 {174, 82}/256 +Inactive Low 5 {163, 93}/256 +Inactive Low 6 or more {157, 99}/256 +Inactive High 0 {58, 198}/256 +Inactive High 1 {245, 11}/256 +Inactive High 2 {238, 18}/256 +Inactive High 3 {232, 24}/256 +Inactive High 4 {225, 31}/256 +Inactive High 5 {220, 36}/256 +Inactive High 6 or more {211, 45}/256 +Unvoiced Low 0 {1, 255}/256 +Unvoiced Low 1 {210, 46}/256 +Unvoiced Low 2 {190, 66}/256 +Unvoiced Low 3 {178, 78}/256 +Unvoiced Low 4 {169, 87}/256 +Unvoiced Low 5 {162, 94}/256 +Unvoiced Low 6 or more {152, 104}/256 +Unvoiced High 0 {48, 208}/256 +Unvoiced High 1 {242, 14}/256 +Unvoiced High 2 {235, 21}/256 +Unvoiced High 3 {224, 32}/256 +Unvoiced High 4 {214, 42}/256 +Unvoiced High 5 {205, 51}/256 +Unvoiced High 6 or more {190, 66}/256 +Voiced Low 0 {1, 255}/256 +Voiced Low 1 {162, 94}/256 +Voiced Low 2 {152, 104}/256 +Voiced Low 3 {147, 109}/256 +Voiced Low 4 {144, 112}/256 +Voiced Low 5 {141, 115}/256 +Voiced Low 6 or more {138, 118}/256 +Voiced High 0 {8, 248}/256 +Voiced High 1 {203, 53}/256 +Voiced High 2 {187, 69}/256 +Voiced High 3 {176, 80}/256 +Voiced High 4 {168, 88}/256 +Voiced High 5 {161, 95}/256 +Voiced High 6 or more {154, 102}/256 + + +
+ +
+ + +After the signs have been read, there is enough information to reconstruct the + complete excitation signal. +This requires adding a constant quantization offset to each non-zero sample, + and then pseudorandomly inverting and offsetting every sample. +The constant quantization offset varies depending on the signal type and + quantization offset type (see ). + + + +Signal Type +Quantization Offset Type +Quantization Offset (Q23) +Inactive Low 25 +Inactive High 60 +Unvoiced Low 25 +Unvoiced High 60 +Voiced Low 8 +Voiced High 25 + + + +Let e_raw[i] be the raw excitation value at position i, with a magnitude + composed of the pulses at that location (see + ) combined with any additional LSBs (see + ), and with the corresponding sign decoded in + . +Additionally, let seed be the current pseudorandom seed, which is initialized + to the value decoded from for the first sample in + the current SILK frame, and updated for each subsequent sample according to + the procedure below. +Finally, let offset_Q23 be the quantization offset from + . +Then the following procedure produces the final reconstructed excitation value, + e_Q23[i]: +
+ +
+When e_raw[i] is zero, sign() returns 0 by the definition in + , so the factor of 20 does not get added. +The final e_Q23[i] value may require more than 16 bits per sample, but will not + require more than 23, including the sign. + + +
+ +
+ +
+ + +The remainder of the reconstruction process for the frame does not need to be + bit-exact, as small errors should only introduce proportionally small + distortions. +Although the reference implementation only includes a fixed-point version of + the remaining steps, this section describes them in terms of a floating-point + version for simplicity. +This produces a signal with a nominal range of -1.0 to 1.0. + + + +silk_decode_core() (decode_core.c) contains the code for the main + reconstruction process. +It proceeds subframe-by-subframe, since quantization gains, LTP parameters, and + (in 20 ms SILK frames) LPC coefficients can vary from one to the + next. + + + +Let a_Q12[k] be the LPC coefficients for the current subframe. +If this is the first or second subframe of a 20 ms SILK frame and the LSF + interpolation factor, w_Q2 (see ), is + less than 4, then these correspond to the final LPC coefficients produced by + from the interpolated LSF coefficients, + n1_Q15[k] (computed in ). +Otherwise, they correspond to the final LPC coefficients produced from the + uninterpolated LSF coefficients for the current frame, n2_Q15[k]. + + + +Also, let n be the number of samples in a subframe (40 for NB, 60 for MB, and + 80 for WB), s be the index of the current subframe in this SILK frame (0 or 1 + for 10 ms frames, or 0 to 3 for 20 ms frames), and j be the index of + the first sample in the residual corresponding to the current subframe. + + +
+ +Voiced SILK frames (see ) pass the excitation + through an LTP filter using the parameters decoded in + to produce an LPC residual. +The LTP filter requires LPC residual values from before the current subframe as + input. +However, since the LPC coefficients may have changed, it obtains this residual + by "rewhitening" the corresponding output signal using the LPC coefficients + from the current subframe. +Let out[i] for + (j - pitch_lags[s] - d_LPC - 2) <= i < j + be the fully reconstructed output signal from the last + (pitch_lags[s] + d_LPC + 2) samples of previous subframes + (see ), where pitch_lags[s] is the pitch + lag for the current subframe from . +During reconstruction of the first subframe for this channel after either + +An uncoded regular SILK frame (if this is the side channel), or +A decoder reset (see ), + + out[] is rewhitened into an LPC residual, + res[i], via +
+ +
+This requires storage to buffer up to 306 values of out[i] from previous + subframes. +This corresponds to WB with a maximum pitch lag of + 18 ms * 16 kHz samples, plus 16 samples for d_LPC, plus 2 + samples for the width of the LTP filter. + + + +Let e_Q23[i] for j <= i < (j + n) be the + excitation for the current subframe, and b_Q7[k] for + 0 <= k < 5 be the coefficients of the LTP filter + taken from the codebook entry in one of + Tables  + through  + corresponding to the index decoded for the current subframe in + . +Then for i such that j <= i < (j + n), + the LPC residual is +
+ +
+ + + +For unvoiced frames, the LPC residual for + j <= i < (j + n) is simply a normalized + copy of the excitation signal, i.e., +
+ +
+ +
+ +
+ +LPC synthesis uses the short-term LPC filter to predict the next output + coefficient. +For i such that (j - d_LPC) <= i < j, let + lpc[i] be the result of LPC synthesis from the last d_LPC samples of the + previous subframe, or zeros in the first subframe for this channel after + either + +An uncoded regular SILK frame (if this is the side channel), or +A decoder reset (see ). + +Then for i such that j <= i < (j + n), the + result of LPC synthesis for the current subframe is +
+ +
+The decoder saves the final d_LPC values, i.e., lpc[i] such that + (j + n - d_LPC) <= i < (j + n), + to feed into the LPC synthesis of the next subframe. +This requires storage for up to 16 values of lpc[i] (for WB frames). + + + +Then, the signal is clamped into the final nominal range: +
+ +
+This clamping occurs entirely after the LPC synthesis filter has run. +The decoder saves the unclamped values, lpc[i], to feed into the LPC filter for + the next subframe, but saves the clamped values, out[i], for rewhitening in + voiced frames. + +
+ +
+ +
+ +
+ +For stereo streams, after decoding a frame from each channel, the decoder must + convert the mid-side (MS) representation into a left-right (LR) + representation. +The function silk_stereo_MS_to_LR (stereo_MS_to_LR.c) implements this process. +In it, the decoder predicts the side channel using a) a simple low-passed + version of the mid channel, and b) the unfiltered mid channel, using the + prediction weights decoded in . +This simple low-pass filter imposes a one-sample delay, and the unfiltered +mid channel is also delayed by one sample. +In order to allow seamless switching between stereo and mono, mono streams must + also impose the same one-sample delay. +The encoder requires an additional one-sample delay for both mono and stereo + streams, though an encoder may omit the delay for mono if it knows it will + never switch to stereo. + + + +The unmixing process operates in two phases. +The first phase lasts for 8 ms, during which it interpolates the + prediction weights from the previous frame, prev_w0_Q13 and prev_w1_Q13, to + the values for the current frame, w0_Q13 and w1_Q13. +The second phase simply uses these weights for the remainder of the frame. + + + +Let mid[i] and side[i] be the contents of out[i] (from + ) for the current mid and side channels, + respectively, and let left[i] and right[i] be the corresponding stereo output + channels. +If the side channel is not coded (see ), + then side[i] is set to zero. +Also let j be defined as in , n1 be + the number of samples in phase 1 (64 for NB, 96 for MB, and 128 for WB), + and n2 be the total number of samples in the frame. +Then for i such that j <= i < (j + n2), + the left and right channel output is +
+ +
+These formulas require two samples prior to index j, the start of the + frame, for the mid channel, and one prior sample for the side channel. +For the first frame after a decoder reset, zeros are used instead. + + +
+ +
+ +After stereo unmixing (if any), the decoder applies resampling to convert the + decoded SILK output to the sample rate desired by the application. +This is necessary when decoding a Hybrid frame at SWB or FB sample rates, or + whenever the decoder wants the output at a different sample rate than the + internal SILK sampling rate (e.g., to allow a constant sample rate when the + audio bandwidth changes, or to allow mixing with audio from other + applications). +The resampler itself is non-normative, and a decoder can use any method it + wants to perform the resampling. + + + +However, a minimum amount of delay is imposed to allow the resampler to + operate, and this delay is normative, so that the corresponding delay can be + applied to the MDCT layer in the encoder. +A decoder is always free to use a resampler which requires more delay than + allowed for here (e.g., to improve quality), but it must then delay the output + of the MDCT layer by this extra amount. +Keeping as much delay as possible on the encoder side allows an encoder which + knows it will never use any of the SILK or Hybrid modes to skip this delay. +By contrast, if it were all applied by the decoder, then a decoder which + processes audio in fixed-size blocks would be forced to delay the output of + CELT frames just in case of a later switch to a SILK or Hybrid mode. + + + + gives the maximum resampler delay + in samples at 48 kHz for each SILK audio bandwidth. +Because the actual output rate may not be 48 kHz, it may not be possible + to achieve exactly these delays while using a whole number of input or output + samples. +The reference implementation is able to resample to any of the supported + output sampling rates (8, 12, 16, 24, or 48 kHz) within or near this + delay constraint. +Some resampling filters (including those used by the reference implementation) + may add a delay that is not an exact integer, or is not linear-phase, and so + cannot be represented by a single delay at all frequencies. +However, such deviations are unlikely to be perceptible, and the comparison + tool described in is designed to be relatively + insensitive to them. +The delays listed here are the ones that should be targeted by the encoder. + + + +Audio Bandwidth +Delay in millisecond +NB 0.538 +MB 0.692 +WB 0.706 + + + +NB is given a smaller decoder delay allocation than MB and WB to allow a + higher-order filter when resampling to 8 kHz in both the encoder and + decoder. +This implies that the audio content of two SILK frames operating at different + bandwidths are not perfectly aligned in time. +This is not an issue for any transitions described in + , because they all involve a SILK decoder reset. +When the decoder is reset, any samples remaining in the resampling buffer + are discarded, and the resampler is re-initialized with silence. + + +
+ +
+ + +
+ + +The CELT layer of Opus is based on the Modified Discrete Cosine Transform + with partially overlapping windows of 5 to 22.5 ms. +The main principle behind CELT is that the MDCT spectrum is divided into +bands that (roughly) follow the Bark scale, i.e., the scale of the ear's +critical bands . The normal CELT layer uses 21 of those bands, though Opus + Custom (see ) may use a different number of bands. +In Hybrid mode, the first 17 bands (up to 8 kHz) are not coded. +A band can contain as little as one MDCT bin per channel, and as many as 176 +bins per channel, as detailed in . +In each band, the gain (energy) is coded separately from +the shape of the spectrum. Coding the gain explicitly makes it easy to +preserve the spectral envelope of the signal. The remaining unit-norm shape +vector is encoded using a Pyramid Vector Quantizer (PVQ) . + + + +Frame Size: +2.5 ms +5 ms +10 ms +20 ms +Start Frequency +Stop Frequency +Band Bins: + 0 1 2 4 8 0 Hz 200 Hz + 1 1 2 4 8 200 Hz 400 Hz + 2 1 2 4 8 400 Hz 600 Hz + 3 1 2 4 8 600 Hz 800 Hz + 4 1 2 4 8 800 Hz 1000 Hz + 5 1 2 4 8 1000 Hz 1200 Hz + 6 1 2 4 8 1200 Hz 1400 Hz + 7 1 2 4 8 1400 Hz 1600 Hz + 8 2 4 8 16 1600 Hz 2000 Hz + 9 2 4 8 16 2000 Hz 2400 Hz +10 2 4 8 16 2400 Hz 2800 Hz +11 2 4 8 16 2800 Hz 3200 Hz +12 4 8 16 32 3200 Hz 4000 Hz +13 4 8 16 32 4000 Hz 4800 Hz +14 4 8 16 32 4800 Hz 5600 Hz +15 6 12 24 48 5600 Hz 6800 Hz +16 6 12 24 48 6800 Hz 8000 Hz +17 8 16 32 64 8000 Hz 9600 Hz +18 12 24 48 96 9600 Hz 12000 Hz +19 18 36 72 144 12000 Hz 15600 Hz +20 22 44 88 176 15600 Hz 20000 Hz + + + +Transients are notoriously difficult for transform codecs to code. +CELT uses two different strategies for them: + +Using multiple smaller MDCTs instead of a single large MDCT, and +Dynamic time-frequency resolution changes (See ). + +To improve quality on highly tonal and periodic signals, CELT includes +a prefilter/postfilter combination. The prefilter on the encoder side +attenuates the signal's harmonics. The postfilter on the decoder side +restores the original gain of the harmonics, while shaping the coding noise +to roughly follow the harmonics. Such noise shaping reduces the perception +of the noise. + + + +When coding a stereo signal, three coding methods are available: + +mid-side stereo: encodes the mean and the difference of the left and right channels, +intensity stereo: only encodes the mean of the left and right channels (discards the difference), +dual stereo: encodes the left and right channels separately. + + + + +An overview of the decoder is given in . + + +
+| decoder |----+ + | +---------+ | + | | + | +---------+ v + | | Fine | +---+ + +->| decoder |->| + | + | +---------+ +---+ + | ^ | ++---------+ | | | +| Range | | +----------+ v +| Decoder |-+ | Bit | +------+ ++---------+ | |Allocation| | 2**x | + | +----------+ +------+ + | | | + | v v +--------+ + | +---------+ +---+ +-------+ | pitch | + +->| PVQ |->| * |->| IMDCT |->| post- |---> + | | decoder | +---+ +-------+ | filter | + | +---------+ +--------+ + | ^ + +--------------------------------------+ +]]> +
+ + +The decoder is based on the following symbols and sets of symbols: + + + +Symbol(s) +PDF +Condition +silence {32767, 1}/32768 +post-filter {1, 1}/2 +octave uniform (6)post-filter +period raw bits (4+octave)post-filter +gain raw bits (3)post-filter +tapset {2, 1, 1}/4post-filter +transient {7, 1}/8 +intra {7, 1}/8 +coarse energy +tf_change +tf_select {1, 1}/2 +spread {7, 2, 21, 2}/32 +dyn. alloc. +alloc. trim {2, 2, 5, 10, 22, 46, 22, 10, 5, 2, 2}/128 +skip {1, 1}/2 +intensity uniform +dual {1, 1}/2 +fine energy +residual +anti-collapse{1, 1}/2 +finalize + + + +The decoder extracts information from the range-coded bitstream in the order +described in . In some circumstances, it is +possible for a decoded value to be out of range due to a very small amount of redundancy +in the encoding of large integers by the range coder. +In that case, the decoder should assume there has been an error in the coding, +decoding, or transmission and SHOULD take measures to conceal the error and/or report +to the application that a problem has occurred. Such out of range errors cannot occur +in the SILK layer. + + +
+ +The "transient" flag indicates whether the frame uses a single long MDCT or several short MDCTs. +When it is set, then the MDCT coefficients represent multiple +short MDCTs in the frame. When not set, the coefficients represent a single +long MDCT for the frame. The flag is encoded in the bitstream with a probability of 1/8. +In addition to the global transient flag is a per-band +binary flag to change the time-frequency (tf) resolution independently in each band. The +change in tf resolution is defined in tf_select_table[][] in celt.c and depends +on the frame size, whether the transient flag is set, and the value of tf_select. +The tf_select flag uses a 1/2 probability, but is only decoded +if it can have an impact on the result knowing the value of all per-band +tf_change flags. + +
+ +
+ + +It is important to quantize the energy with sufficient resolution because +any energy quantization error cannot be compensated for at a later +stage. Regardless of the resolution used for encoding the spectral shape of a band, +it is perceptually important to preserve the energy in each band. CELT uses a +three-step coarse-fine-fine strategy for encoding the energy in the base-2 log +domain, as implemented in quant_bands.c + +
+ +Coarse quantization of the energy uses a fixed resolution of 6 dB +(integer part of base-2 log). To minimize the bitrate, prediction is applied +both in time (using the previous frame) and in frequency (using the previous +bands). The part of the prediction that is based on the +previous frame can be disabled, creating an "intra" frame where the energy +is coded without reference to prior frames. The decoder first reads the intra flag +to determine what prediction is used. +The 2-D z-transform of +the prediction filter is: +
+ +
+where b is the band index and l is the frame index. The prediction coefficients +applied depend on the frame size in use when not using intra energy and are alpha=0, beta=4915/32768 +when using intra energy. +The time-domain prediction is based on the final fine quantization of the previous +frame, while the frequency domain (within the current frame) prediction is based +on coarse quantization only (because the fine quantization has not been computed +yet). The prediction is clamped internally so that fixed point implementations with +limited dynamic range always remain in the same state as floating point implementations. +We approximate the ideal +probability distribution of the prediction error using a Laplace distribution +with separate parameters for each frame size in intra- and inter-frame modes. These +parameters are held in the e_prob_model table in quant_bands.c. +The +coarse energy quantization is performed by unquant_coarse_energy() and +unquant_coarse_energy_impl() (quant_bands.c). The encoding of the Laplace-distributed values is +implemented in ec_laplace_decode() (laplace.c). + + +
+ +
+ +The number of bits assigned to fine energy quantization in each band is determined +by the bit allocation computation described in . +Let B_i be the number of fine energy bits +for band i; the refinement is an integer f in the range [0,2**B_i-1]. The mapping between f +and the correction applied to the coarse energy is equal to (f+1/2)/2**B_i - 1/2. Fine +energy quantization is implemented in quant_fine_energy() (quant_bands.c). + + +When some bits are left "unused" after all other flags have been decoded, these bits +are assigned to a "final" step of fine allocation. In effect, these bits are used +to add one extra fine energy bit per band per channel. The allocation process +determines two "priorities" for the final fine bits. +Any remaining bits are first assigned only to bands of priority 0, starting +from band 0 and going up. If all bands of priority 0 have received one bit per +channel, then bands of priority 1 are assigned an extra bit per channel, +starting from band 0. If any bits are left after this, they are left unused. +This is implemented in unquant_energy_finalise() (quant_bands.c). + + +
+ +
+ +
+ +Because the bit allocation drives the decoding of the range-coder +stream, it MUST be recovered exactly so that identical coding decisions are +made in the encoder and decoder. Any deviation from the reference's resulting +bit allocation will result in corrupted output, though implementers are +free to implement the procedure in any way which produces identical results. + +The per-band gain-shape structure of the CELT layer ensures that using + the same number of bits for the spectral shape of a band in every frame will + result in a roughly constant signal-to-noise ratio in that band. +This results in coding noise that has the same spectral envelope as the signal. +The masking curve produced by a standard psychoacoustic model also closely + follows the spectral envelope of the signal. +This structure means that the ideal allocation is more consistent from frame to + frame than it is for other codecs without an equivalent structure, and that a + fixed allocation provides fairly consistent perceptual + performance . + +Many codecs transmit significant amounts of side information to control the + bit allocation within a frame. +Often this control is only indirect, and must be exercised carefully to + achieve the desired rate constraints. +The CELT layer, however, can adapt over a very wide range of rates, and thus + has a large number of codebook sizes to choose from for each band. +Explicitly signaling the size of each of these codebooks would impose + considerable overhead, even though the allocation is relatively static from + frame to frame. +This is because all of the information required to compute these codebook sizes + must be derived from a single frame by itself, in order to retain robustness + to packet loss, so the signaling cannot take advantage of knowledge of the + allocation in neighboring frames. +This problem is exacerbated in low-latency (small frame size) applications, + which would include this overhead in every frame. + +For this reason, in the MDCT mode Opus uses a primarily implicit bit +allocation. The available bitstream capacity is known in advance to both +the encoder and decoder without additional signaling, ultimately from the +packet sizes expressed by a higher-level protocol. Using this information, +the codec interpolates an allocation from a hard-coded table. + +While the band-energy structure effectively models intra-band masking, +it ignores the weaker inter-band masking, band-temporal masking, and +other less significant perceptual effects. While these effects can +often be ignored, they can become significant for particular samples. One +mechanism available to encoders would be to simply increase the overall +rate for these frames, but this is not possible in a constant rate mode +and can be fairly inefficient. As a result three explicitly signaled +mechanisms are provided to alter the implicit allocation: + + + +Band boost +Allocation trim +Band skipping + + + +The first of these mechanisms, band boost, allows an encoder to boost +the allocation in specific bands. The second, allocation trim, works by +biasing the overall allocation towards higher or lower frequency bands. The third, band +skipping, selects which low-precision high frequency bands +will be allocated no shape bits at all. + +In stereo mode there are two additional parameters +potentially coded as part of the allocation procedure: a parameter to allow the +selective elimination of allocation for the 'side' (i.e., intensity stereo) in jointly coded bands, +and a flag to deactivate joint coding (i.e., dual stereo). These values are not signaled if +they would be meaningless in the overall context of the allocation. + +Because every signaled adjustment increases overhead and implementation +complexity, none were included speculatively: the reference encoder makes use +of all of these mechanisms. While the decision logic in the reference was +found to be effective enough to justify the overhead and complexity, further +analysis techniques may be discovered which increase the effectiveness of these +parameters. As with other signaled parameters, an encoder is free to choose the +values in any manner, but unless a technique is known to deliver superior +perceptual results the methods used by the reference implementation should be +used. + +The allocation process consists of the following steps: determining the per-band +maximum allocation vector, decoding the boosts, decoding the tilt, determining +the remaining capacity of the frame, searching the mode table for the +entry nearest but not exceeding the available space (subject to the tilt, boosts, band +maximums, and band minimums), linear interpolation, reallocation of +unused bits with concurrent skip decoding, determination of the +fine-energy vs. shape split, and final reallocation. This process results +in a per-band shape allocation (in 1/8th bit units), a per-band fine-energy +allocation (in 1 bit per channel units), a set of band priorities for +controlling the use of remaining bits at the end of the frame, and a +remaining balance of unallocated space, which is usually zero except +at very high rates. + + +The "static" bit allocation (in 1/8 bits) for a quality q, excluding the minimums, maximums, +tilt and boosts, is equal to channels*N*alloc[band][q]<<LM>>2, where +alloc[][] is given in and LM=log2(frame_size/120). The allocation +is obtained by linearly interpolating between two values of q (in steps of 1/64) to find the +highest allocation that does not exceed the number of bits remaining. + + + + Rows indicate the MDCT bands, columns are the different quality (q) parameters. The units are 1/32 bit per MDCT bin. +0 +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +090110118126134144152162172200 +080100110119127137145155165200 +07590103112120130138148158200 +0698493104114124132142152200 +063788695103113123133143200 +05671808997107117127137200 +04965758391101111121131200 +0405870788595105115125200 +034516572788898108118198 +029455966728292102112193 +02039536066768696106188 +01832475460708090100183 +0102640475464748494178 +002031394757677787173 +001223324151617181168 +00015253545556575163 +0004172939495969158 +0000122333435363153 +000011626364656148 +000001015203045129 +00000111120104 + + +The maximum allocation vector is an approximation of the maximum space +that can be used by each band for a given mode. The value is +approximate because the shape encoding is variable rate (due +to entropy coding of splitting parameters). Setting the maximum too low reduces the +maximum achievable quality in a band while setting it too high +may result in waste: bitstream capacity available at the end +of the frame which can not be put to any use. The maximums +specified by the codec reflect the average maximum. In the reference +implementation, the maximums in bits/sample are precomputed in a static table +(see cache_caps50[] in static_modes_float.h) for each band, +for each value of LM, and for both mono and stereo. + +Implementations are expected +to simply use the same table data, but the procedure for generating +this table is included in rate.c as part of compute_pulse_cache(). + +To convert the values in cache.caps into the actual maximums: first +set nbBands to the maximum number of bands for this mode, and stereo to +zero if stereo is not in use and one otherwise. For each band set N +to the number of MDCT bins covered by the band (for one channel), set LM +to the shift value for the frame size, +then set i to nbBands*(2*LM+stereo). Then set the maximum for the band to +the i-th index of cache.caps + 64 and multiply by the number of channels +in the current frame (one or two) and by N, then divide the result by 4 +using integer division. The resulting vector will be called +cap[]. The elements fit in signed 16-bit integers but do not fit in 8 bits. +This procedure is implemented in the reference in the function init_caps() in celt.c. + + +The band boosts are represented by a series of binary symbols which +are entropy coded with very low probability. Each band can potentially be boosted +multiple times, subject to the frame actually having enough room to obey +the boost and having enough room to code the boost symbol. The default +coding cost for a boost starts out at six bits (probability p=1/64), but subsequent boosts +in a band cost only a single bit and every time a band is boosted the +initial cost is reduced (down to a minimum of two bits, or p=1/4). Since the initial +cost of coding a boost is 6 bits, the coding cost of the boost symbols when +completely unused is 0.48 bits/frame for a 21 band mode (21*-log2(1-1/2**6)). + +To decode the band boosts: First set 'dynalloc_logp' to 6, the initial +amount of storage required to signal a boost in bits, 'total_bits' to the +size of the frame in 8th bits, 'total_boost' to zero, and 'tell' to the total number +of 8th bits decoded +so far. For each band from the coding start (0 normally, but 17 in Hybrid mode) +to the coding end (which changes depending on the signaled bandwidth), the boost quanta +in units of 1/8 bit is calculated as quanta = min(8*N, max(48, N)). +This represents a boost step size of six bits, subject to a lower limit of +1/8th bit/sample and an upper limit of 1 bit/sample. +Set 'boost' to zero and 'dynalloc_loop_logp' +to dynalloc_logp. While dynalloc_loop_log (the current worst case symbol cost) in +8th bits plus tell is less than total_bits plus total_boost and boost is less than cap[] for this +band: Decode a bit from the bitstream with a with dynalloc_loop_logp as the cost +of a one, update tell to reflect the current used capacity, if the decoded value +is zero break the loop otherwise add quanta to boost and total_boost, subtract quanta from +total_bits, and set dynalloc_loop_log to 1. When the while loop finishes +boost contains the boost for this band. If boost is non-zero and dynalloc_logp +is greater than 2, decrease dynalloc_logp. Once this process has been +executed on all bands, the band boosts have been decoded. This procedure +is implemented around line 2474 of celt.c. + +At very low rates it is possible that there won't be enough available +space to execute the inner loop even once. In these cases band boost +is not possible but its overhead is completely eliminated. Because of the +high cost of band boost when activated, a reasonable encoder should not be +using it at very low rates. The reference implements its dynalloc decision +logic around line 1304 of celt.c. + +The allocation trim is a integer value from 0-10. The default value of +5 indicates no trim. The trim parameter is entropy coded in order to +lower the coding cost of less extreme adjustments. Values lower than +5 bias the allocation towards lower frequencies and values above 5 +bias it towards higher frequencies. Like other signaled parameters, signaling +of the trim is gated so that it is not included if there is insufficient space +available in the bitstream. To decode the trim, first set +the trim value to 5, then if and only if the count of decoded 8th bits so far (ec_tell_frac) +plus 48 (6 bits) is less than or equal to the total frame size in 8th +bits minus total_boost (a product of the above band boost procedure), +decode the trim value using the PDF in . + + +PDF +{1, 1, 2, 5, 10, 22, 46, 22, 10, 5, 2, 2}/128 + + +For 10 ms and 20 ms frames using short blocks and that have at least LM+2 bits left prior to +the allocation process, then one anti-collapse bit is reserved in the allocation process so it can +be decoded later. Following the the anti-collapse reservation, one bit is reserved for skip if available. + +For stereo frames, bits are reserved for intensity stereo and for dual stereo. Intensity stereo +requires ilog2(end-start) bits. Those bits are reserved if there is enough bits left. Following this, one +bit is reserved for dual stereo if available. + + +The allocation computation begins by setting up some initial conditions. +'total' is set to the remaining available 8th bits, computed by taking the +size of the coded frame times 8 and subtracting ec_tell_frac(). From this value, one (8th bit) +is subtracted to ensure that the resulting allocation will be conservative. 'anti_collapse_rsv' +is set to 8 (8th bits) if and only if the frame is a transient, LM is greater than 1, and total is +greater than or equal to (LM+2) * 8. Total is then decremented by anti_collapse_rsv and clamped +to be equal to or greater than zero. 'skip_rsv' is set to 8 (8th bits) if total is greater than +8, otherwise it is zero. Total is then decremented by skip_rsv. This reserves space for the +final skipping flag. + +If the current frame is stereo, intensity_rsv is set to the conservative log2 in 8th bits +of the number of coded bands for this frame (given by the table LOG2_FRAC_TABLE in rate.c). If +intensity_rsv is greater than total then intensity_rsv is set to zero. Otherwise total is +decremented by intensity_rsv, and if total is still greater than 8, dual_stereo_rsv is +set to 8 and total is decremented by dual_stereo_rsv. + +The allocation process then computes a vector representing the hard minimum amounts allocation +any band will receive for shape. This minimum is higher than the technical limit of the PVQ +process, but very low rate allocations produce an excessively sparse spectrum and these bands +are better served by having no allocation at all. For each coded band, set thresh[band] to +twenty-four times the number of MDCT bins in the band and divide by 16. If 8 times the number +of channels is greater, use that instead. This sets the minimum allocation to one bit per channel +or 48 128th bits per MDCT bin, whichever is greater. The band-size dependent part of this +value is not scaled by the channel count, because at the very low rates where this limit is +applicable there will usually be no bits allocated to the side. + +The previously decoded allocation trim is used to derive a vector of per-band adjustments, +'trim_offsets[]'. For each coded band take the alloc_trim and subtract 5 and LM. Then multiply +the result by the number of channels, the number of MDCT bins in the shortest frame size for this mode, +the number of remaining bands, 2**LM, and 8. Then divide this value by 64. Finally, if the +number of MDCT bins in the band per channel is only one, 8 times the number of channels is subtracted +in order to diminish the allocation by one bit, because width 1 bands receive greater benefit +from the coarse energy coding. + + +
+ +
+ +In each band, the normalized "shape" is encoded +using a vector quantization scheme called a "pyramid vector quantizer". + + +In +the simplest case, the number of bits allocated in + is converted to a number of pulses as described +by . Knowing the number of pulses and the +number of samples in the band, the decoder calculates the size of the codebook +as detailed in . The size is used to decode +an unsigned integer (uniform probability model), which is the codeword index. +This index is converted into the corresponding vector as explained in +. This vector is then scaled to unit norm. + + +
+ +Although the allocation is performed in 1/8th bit units, the quantization requires +an integer number of pulses K. To do this, the encoder searches for the value +of K that produces the number of bits nearest to the allocated value +(rounding down if exactly halfway between two values), not to exceed +the total number of bits available. For efficiency reasons, the search is performed against a +precomputed allocation table which only permits some K values for each N. The number of +codebook entries can be computed as explained in . The difference +between the number of bits allocated and the number of bits used is accumulated to a +"balance" (initialized to zero) that helps adjust the +allocation for the next bands. One third of the balance is applied to the +bit allocation of each band to help achieve the target allocation. The only +exceptions are the band before the last and the last band, for which half the balance +and the whole balance are applied, respectively. + +
+ +
+ + +Decoding of PVQ vectors is implemented in decode_pulses() (cwrs.c). +The unique codeword index is decoded as a uniformly-distributed integer value between 0 and +V(N,K)-1, where V(N,K) is the number of possible combinations of K pulses in +N samples. The index is then converted to a vector in the same way specified in +. The indexing is based on the calculation of V(N,K) +(denoted N(L,K) in ). + + + + The number of combinations can be computed recursively as +V(N,K) = V(N-1,K) + V(N,K-1) + V(N-1,K-1), with V(N,0) = 1 and V(0,K) = 0, K != 0. +There are many different ways to compute V(N,K), including precomputed tables and direct +use of the recursive formulation. The reference implementation applies the recursive +formulation one line (or column) at a time to save on memory use, +along with an alternate, +univariate recurrence to initialize an arbitrary line, and direct +polynomial solutions for small N. All of these methods are +equivalent, and have different trade-offs in speed, memory usage, and +code size. Implementations MAY use any methods they like, as long as +they are equivalent to the mathematical definition. + + + +The decoded vector X is recovered as follows. +Let i be the index decoded with the procedure in + with ft = V(N,K), so that 0 <= i < V(N,K). +Let k = K. +Then for j = 0 to (N - 1), inclusive, do: + +Let p = (V(N-j-1,k) + V(N-j,k))/2. + +If i < p, then let sgn = 1, else let sgn = -1 + and set i = i - p. + +Let k0 = k and set p = p - V(N-j-1,k). + +While p > i, set k = k - 1 and + p = p - V(N-j-1,k). + + +Set X[j] = sgn*(k0 - k) and i = i - p. + + + + + +The decoded vector X is then normalized such that its +L2-norm equals one. + +
+ +
+ +The normalized vector decoded in is then rotated +for the purpose of avoiding tonal artifacts. The rotation gain is equal to +
+ +
+ +where N is the number of dimensions, K is the number of pulses, and f_r depends on +the value of the "spread" parameter in the bit-stream. + + + +Spread value +f_r + 0 infinite (no rotation) + 1 15 + 2 10 + 3 5 + + + +The rotation angle is then calculated as +
+ +
+A 2-D rotation R(i,j) between points x_i and x_j is defined as: +
+ +
+ +An N-D rotation is then achieved by applying a series of 2-D rotations back and forth, in the +following order: R(x_1, x_2), R(x_2, x_3), ..., R(x_N-2, X_N-1), R(x_N-1, X_N), +R(x_N-2, X_N-1), ..., R(x_1, x_2). + + + +If the decoded vector represents more +than one time block, then this spreading process is applied separately on each time block. +Also, if each block represents 8 samples or more, then another N-D rotation, by +(pi/2-theta), is applied before the rotation described above. This +extra rotation is applied in an interleaved manner with a stride equal to round(sqrt(N/nb_blocks)), +i.e., it is applied independently for each set of sample S_k = {stride*n + k}, n=0..N/stride-1. + +
+ +
+ +To avoid the need for multi-precision calculations when decoding PVQ codevectors, +the maximum size allowed for codebooks is 32 bits. When larger codebooks are +needed, the vector is instead split in two sub-vectors of size N/2. +A quantized gain parameter with precision +derived from the current allocation is entropy coded to represent the relative +gains of each side of the split, and the entire decoding process is recursively +applied. Multiple levels of splitting may be applied up to a limit of LM+1 splits. +The same recursive mechanism is applied for the joint coding +of stereo audio. + + +
+ +
+ +The time-frequency (TF) parameters are used to control the time-frequency resolution tradeoff +in each coded band. For each band, there are two possible TF choices. For the first +band coded, the PDF is {3, 1}/4 for frames marked as transient and {15, 1}/16 for +the other frames. For subsequent bands, the TF choice is coded relative to the +previous TF choice with probability {15, 1}/15 for transient frames and {31, 1}/32 +otherwise. The mapping between the decoded TF choices and the adjustment in TF +resolution is shown in the tables below. + + + +Frame size (ms) +0 +1 +2.5 0 -1 +5 0 -1 +10 0 -2 +20 0 -2 + + + +Frame size (ms) +0 +1 +2.5 0 -1 +5 0 -2 +10 0 -3 +20 0 -3 + + + + +Frame size (ms) +0 +1 +2.5 0 -1 +5 1 0 +10 2 0 +20 3 0 + + + +Frame size (ms) +0 +1 +2.5 0 -1 +5 1 -1 +10 1 -1 +20 1 -1 + + + +A negative TF adjustment means that the temporal resolution is increased, +while a positive TF adjustment means that the frequency resolution is increased. +Changes in TF resolution are implemented using the Hadamard transform . To increase +the time resolution by N, N "levels" of the Hadamard transform are applied to the +decoded vector for each interleaved MDCT vector. To increase the frequency resolution +(assumes a transient frame), then N levels of the Hadamard transform are applied +across the interleaved MDCT vector. In the case of increased +time resolution the decoder uses the "sequency order" because the input vector +is sorted in time. + +
+ + +
+ +
+ +The anti-collapse feature is designed to avoid the situation where the use of multiple +short MDCTs causes the energy in one or more of the MDCTs to be zero for +some bands, causing unpleasant artifacts. +When the frame has the transient bit set, an anti-collapse bit is decoded. +When anti-collapse is set, the energy in each small MDCT is prevented +from collapsing to zero. For each band of each MDCT where a collapse is +detected, a pseudo-random signal is inserted with an energy corresponding +to the minimum energy over the two previous frames. A renormalization step is +then required to ensure that the anti-collapse step did not alter the +energy preservation property. + +
+ +
+ +Just as each band was normalized in the encoder, the last step of the decoder before +the inverse MDCT is to denormalize the bands. Each decoded normalized band is +multiplied by the square root of the decoded energy. This is done by denormalise_bands() +(bands.c). + +
+ +
+ + +The inverse MDCT implementation has no special characteristics. The +input is N frequency-domain samples and the output is 2*N time-domain +samples, while scaling by 1/2. A "low-overlap" window reduces the algorithmic delay. +It is derived from a basic (full overlap) 240-sample version of the window used by the Vorbis codec: +
+ +
+The low-overlap window is created by zero-padding the basic window and inserting ones in the +middle, such that the resulting window still satisfies power complementarity . +The IMDCT and +windowing are performed by mdct_backward (mdct.c). + + +
+ +The output of the inverse MDCT (after weighted overlap-add) is sent to the +post-filter. Although the post-filter is applied at the end, the post-filter +parameters are encoded at the beginning, just after the silence flag. +The post-filter can be switched on or off using one bit (logp=1). +If the post-filter is enabled, then the octave is decoded as an integer value +between 0 and 6 of uniform probability. Once the octave is known, the fine pitch +within the octave is decoded using 4+octave raw bits. The final pitch period +is equal to (16<<octave)+fine_pitch-1 so it is bounded between 15 and 1022, +inclusively. Next, the gain is decoded as three raw bits and is equal to +G=3*(int_gain+1)/32. The set of post-filter taps is decoded last, using +a pdf equal to {2, 1, 1}/4. Tapset zero corresponds to the filter coefficients +g0 = 0.3066406250, g1 = 0.2170410156, g2 = 0.1296386719. Tapset one +corresponds to the filter coefficients g0 = 0.4638671875, g1 = 0.2680664062, +g2 = 0, and tapset two uses filter coefficients g0 = 0.7998046875, +g1 = 0.1000976562, g2 = 0. + + + +The post-filter response is thus computed as: +
+ + + +
+ +During a transition between different gains, a smooth transition is calculated +using the square of the MDCT window. It is important that values of y(n) be +interpolated one at a time such that the past value of y(n) used is interpolated. + +
+ +
+ +After the post-filter, +the signal is de-emphasized using the inverse of the pre-emphasis filter +used in the encoder: +
+ +
+where alpha_p=0.8500061035. + +
+ +
+ +
+ +
+ +Packet loss concealment (PLC) is an optional decoder-side feature that +SHOULD be included when receiving from an unreliable channel. Because +PLC is not part of the bitstream, there are many acceptable ways to +implement PLC with different complexity/quality trade-offs. + + + +The PLC in +the reference implementation depends on the mode of last packet received. +In CELT mode, the PLC finds a periodicity in the decoded +signal and repeats the windowed waveform using the pitch offset. The windowed +waveform is overlapped in such a way as to preserve the time-domain aliasing +cancellation with the previous frame and the next frame. This is implemented +in celt_decode_lost() (mdct.c). In SILK mode, the PLC uses LPC extrapolation +from the previous frame, implemented in silk_PLC() (PLC.c). + + +
+ +Clock drift refers to the gradual desynchronization of two endpoints +whose sample clocks run at different frequencies while they are streaming +live audio. Differences in clock frequencies are generally attributable to +manufacturing variation in the endpoints' clock hardware. For long-lived +streams, the time difference between sender and receiver can grow without +bound. + + + +When the sender's clock runs slower than the receiver's, the effect is similar +to packet loss: too few packets are received. The receiver can distinguish +between drift and loss if the transport provides packet timestamps. A receiver +for live streams SHOULD conceal the effects of drift, and MAY do so by invoking +the PLC. + + + +When the sender's clock runs faster than the receiver's, too many packets will +be received. The receiver MAY respond by skipping any packet (i.e., not +submitting the packet for decoding). This is likely to produce a less severe +artifact than if the frame were dropped after decoding. + + + +A decoder MAY employ a more sophisticated drift compensation method. For +example, the +NetEQ component +of the +Google WebRTC codebase +compensates for drift by adding or removing +one period when the signal is highly periodic. The reference implementation of +Opus allows a caller to learn whether the current frame's signal is highly +periodic, and if so what the period is, using the OPUS_GET_PITCH() request. + +
+ +
+ +
+ + +Switching between the Opus coding modes, audio bandwidths, and channel counts + requires careful consideration to avoid audible glitches. +Switching between any two configurations of the CELT-only mode, any two + configurations of the Hybrid mode, or from WB SILK to Hybrid mode does not + require any special treatment in the decoder, as the MDCT overlap will smooth + the transition. +Switching from Hybrid mode to WB SILK requires adding in the final contents + of the CELT overlap buffer to the first SILK-only packet. +This can be done by decoding a 2.5 ms silence frame with the CELT decoder + using the channel count of the SILK-only packet (and any choice of audio + bandwidth), which will correctly handle the cases when the channel count + changes as well. + + + +When changing the channel count for SILK-only or Hybrid packets, the encoder + can avoid glitches by smoothly varying the stereo width of the input signal + before or after the transition, and SHOULD do so. +However, other transitions between SILK-only packets or between NB or MB SILK + and Hybrid packets may cause glitches, because neither the LSF coefficients + nor the LTP, LPC, stereo unmixing, and resampler buffers are available at the + new sample rate. +These switches SHOULD be delayed by the encoder until quiet periods or + transients, where the inevitable glitches will be less audible. Additionally, + the bit-stream MAY include redundant side information ("redundancy"), in the + form of additional CELT frames embedded in each of the Opus frames around the + transition. + + + +The other transitions that cannot be easily handled are those where the lower + frequencies switch between the SILK LP-based model and the CELT MDCT model. +However, an encoder may not have an opportunity to delay such a switch to a + convenient point. +For example, if the content switches from speech to music, and the encoder does + not have enough latency in its analysis to detect this in advance, there may + be no convenient silence period during which to make the transition for quite + some time. +To avoid or reduce glitches during these problematic mode transitions, and + also between audio bandwidth changes in the SILK-only modes, transitions MAY + include redundant side information ("redundancy"), in the form of an + additional CELT frame embedded in the Opus frame. + + + +A transition between coding the lower frequencies with the LP model and the + MDCT model or a transition that involves changing the SILK bandwidth + is only normatively specified when it includes redundancy. +For those without redundancy, it is RECOMMENDED that the decoder use a + concealment technique (e.g., make use of a PLC algorithm) to "fill in" the + gap or discontinuity caused by the mode transition. +Therefore, PLC MUST NOT be applied during any normative transition, i.e., when + +A packet includes redundancy for this transition (as described below), +The transition is between any WB SILK packet and any Hybrid packet, or vice + versa, +The transition is between any two Hybrid mode packets, or +The transition is between any two CELT mode packets, + + unless there is actual packet loss. + + +
+ +Transitions with side information include an extra 5 ms "redundant" CELT + frame within the Opus frame. +This frame is designed to fill in the gap or discontinuity in the different + layers without requiring the decoder to conceal it. +For transitions from CELT-only to SILK-only or Hybrid, the redundant frame is + inserted in the first Opus frame after the transition (i.e., the first + SILK-only or Hybrid frame). +For transitions from SILK-only or Hybrid to CELT-only, the redundant frame is + inserted in the last Opus frame before the transition (i.e., the last + SILK-only or Hybrid frame). + + +
+ +The presence of redundancy is signaled in all SILK-only and Hybrid frames, not + just those involved in a mode transition. +This allows the frames to be decoded correctly even if an adjacent frame is + lost. +For SILK-only frames, this signaling is implicit, based on the size of the + of the Opus frame and the number of bits consumed decoding the SILK portion of + it. +After decoding the SILK portion of the Opus frame, the decoder uses ec_tell() + (see ) to check if there are at least 17 bits + remaining. +If so, then the frame contains redundancy. + + + +For Hybrid frames, this signaling is explicit. +After decoding the SILK portion of the Opus frame, the decoder uses ec_tell() + (see ) to ensure there are at least 37 bits remaining. +If so, it reads a symbol with the PDF in + , and if the value is 1, then the + frame contains redundancy. +Otherwise (if there were fewer than 37 bits left or the value was 0), the frame + does not contain redundancy. + + + +PDF +{4095, 1}/4096 + +
+ +
+ +Since the current frame is a SILK-only or a Hybrid frame, it must be at least + 10 ms. +Therefore, it needs an additional flag to indicate whether the redundant + 5 ms CELT frame should be mixed into the beginning of the current frame, + or the end. +After determining that a frame contains redundancy, the decoder reads a + 1 bit symbol with a uniform PDF + (). + + + +PDF +{1, 1}/2 + + + +If the value is zero, this is the first frame in the transition, and the + redundancy belongs at the end. +If the value is one, this is the second frame in the transition, and the + redundancy belongs at the beginning. +There is no way to specify that an Opus frame contains separate redundant CELT + frames at both the beginning and the end. + +
+ +
+ +Unlike the CELT portion of a Hybrid frame, the redundant CELT frame does not + use the same entropy coder state as the rest of the Opus frame, because this + would break the CELT bit allocation mechanism in Hybrid frames. +Thus, a redundant CELT frame always starts and ends on a byte boundary, even in + SILK-only frames, where this is not strictly necessary. + + + +For SILK-only frames, the number of bytes in the redundant CELT frame is simply + the number of whole bytes remaining, which must be at least 2, due to the + space check in . +For Hybrid frames, the number of bytes is equal to 2, plus a decoded unsigned + integer less than 256 (see ). +This may be more than the number of whole bytes remaining in the Opus frame, + in which case the frame is invalid. +However, a decoder is not required to ignore the entire frame, as this may be + the result of a bit error that desynchronized the range coder. +There may still be useful data before the error, and a decoder MAY keep any + audio decoded so far instead of invoking the PLC, but it is RECOMMENDED that + the decoder stop decoding and discard the rest of the current Opus frame. + + + +It would have been possible to avoid these invalid states in the design of Opus + by limiting the range of the explicit length decoded from Hybrid frames by the + actual number of whole bytes remaining. +However, this would require an encoder to determine the rate allocation for the + MDCT layer up front, before it began encoding that layer. +By allowing some invalid sizes, the encoder is able to defer that decision + until much later. +When encoding Hybrid frames which do not include redundancy, the encoder must + still decide up-front if it wishes to use the minimum 37 bits required to + trigger encoding of the redundancy flag, but this is a much looser + restriction. + + + +After determining the size of the redundant CELT frame, the decoder reduces + the size of the buffer currently in use by the range coder by that amount. +The CELT layer read any raw bits from the end of this reduced buffer, and all + calculations of the number of bits remaining in the buffer must be done using + this new, reduced size, rather than the original size of the Opus frame. + +
+ +
+ +The redundant frame is decoded like any other CELT-only frame, with the + exception that it does not contain a TOC byte. +The frame size is fixed at 5 ms, the channel count is set to that of the + current frame, and the audio bandwidth is also set to that of the current + frame, with the exception that for MB SILK frames, it is set to WB. + + + +If the redundancy belongs at the beginning (in a CELT-only to SILK-only or + Hybrid transition), the final reconstructed output uses the first 2.5 ms + of audio output by the decoder for the redundant frame as-is, discarding + the corresponding output from the SILK-only or Hybrid portion of the frame. +The remaining 2.5 ms is cross-lapped with the decoded SILK/Hybrid signal + using the CELT's power-complementary MDCT window to ensure a smooth + transition. + + + +If the redundancy belongs at the end (in a SILK-only or Hybrid to CELT-only + transition), only the second half (2.5 ms) of the audio output by the + decoder for the redundant frame is used. +In that case, the second half of the redundant frame is cross-lapped with the + end of the SILK/Hybrid signal, again using CELT's power-complementary MDCT + window to ensure a smooth transition. + +
+ +
+ +
+ +When a transition occurs, the state of the SILK or the CELT decoder (or both) + may need to be reset before decoding a frame in the new mode. +This avoids reusing "out of date" memory, which may not have been updated in + some time or may not be in a well-defined state due to, e.g., PLC. +The SILK state is reset before every SILK-only or Hybrid frame where the + previous frame was CELT-only. +The CELT state is reset every time the operating mode changes and the new mode + is either Hybrid or CELT-only, except when the transition uses redundancy as + described above. +When switching from SILK-only or Hybrid to CELT-only with redundancy, the CELT + state is reset before decoding the redundant CELT frame embedded in the + SILK-only or Hybrid frame, but it is not reset before decoding the following + CELT-only frame. +When switching from CELT-only mode to SILK-only or Hybrid mode with redundancy, + the CELT decoder is not reset for decoding the redundant CELT frame. + +
+ +
+ + + illustrates all of the normative + transitions involving a mode change, an audio bandwidth change, or both. +Each one uses an S, H, or C to represent an Opus frame in the corresponding + mode. +In addition, an R indicates the presence of redundancy in the Opus frame it is + cross-lapped with. +Its location in the first or last 5 ms is assumed to correspond to whether + it is the frame before or after the transition. +Other uses of redundancy are non-normative. +Finally, a c indicates the contents of the CELT overlap buffer after the + previously decoded frame (i.e., as extracted by decoding a silence frame). +
+ S -> S + & + !R -> R + & + ;S -> S -> S + +NB or MB SILK to Hybrid with Redundancy: S -> S -> S + & + !R ->;H -> H -> H + +WB SILK to Hybrid: S -> S -> S ->!H -> H -> H + +SILK to CELT with Redundancy: S -> S -> S + & + !R -> C -> C -> C + +Hybrid to NB or MB SILK with Redundancy: H -> H -> H + & + !R -> R + & + ;S -> S -> S + +Hybrid to WB SILK: H -> H -> H -> c + \ + + > S -> S -> S + +Hybrid to CELT with Redundancy: H -> H -> H + & + !R -> C -> C -> C + +CELT to SILK with Redundancy: C -> C -> C -> R + & + ;S -> S -> S + +CELT to Hybrid with Redundancy: C -> C -> C -> R + & + |H -> H -> H + +Key: +S SILK-only frame ; SILK decoder reset +H Hybrid frame | CELT and SILK decoder resets +C CELT-only frame ! CELT decoder reset +c CELT overlap + Direct mixing +R Redundant CELT frame & Windowed cross-lap +]]> +
+The first two and the last two Opus frames in each example are illustrative, + i.e., there is no requirement that a stream remain in the same configuration + for three consecutive frames before or after a switch. + + + +The behavior of transitions without redundancy where PLC is allowed is non-normative. +An encoder might still wish to use these transitions if, for example, it + doesn't want to add the extra bitrate required for redundancy or if it makes + a decision to switch after it has already transmitted the frame that would + have had to contain the redundancy. + illustrates the recommended + cross-lapping and decoder resets for these transitions. +
+ S -> S ;S -> S -> S + +NB or MB SILK to Hybrid: S -> S -> S |H -> H -> H + +SILK to CELT without Redundancy: S -> S -> S -> P + & + !C -> C -> C + +Hybrid to NB or MB SILK: H -> H -> H -> c + + + ;S -> S -> S + +Hybrid to CELT without Redundancy: H -> H -> H -> P + & + !C -> C -> C + +CELT to SILK without Redundancy: C -> C -> C -> P + & + ;S -> S -> S + +CELT to Hybrid without Redundancy: C -> C -> C -> P + & + |H -> H -> H + +Key: +S SILK-only frame ; SILK decoder reset +H Hybrid frame | CELT and SILK decoder resets +C CELT-only frame ! CELT decoder reset +c CELT overlap + Direct mixing +P Packet Loss Concealment & Windowed cross-lap +]]> +
+Encoders SHOULD NOT use other transitions, e.g., those that involve redundancy + in ways not illustrated in . + + +
+ +
+ +
+ + + + + + +
+ +Just like the decoder, the Opus encoder also normally consists of two main blocks: the +SILK encoder and the CELT encoder. However, unlike the case of the decoder, a valid +(though potentially suboptimal) Opus encoder is not required to support all modes and +may thus only include a SILK encoder module or a CELT encoder module. +The output bit-stream of the Opus encoding contains bits from the SILK and CELT + encoders, though these are not separable due to the use of a range coder. +A block diagram of the encoder is illustrated below. + +
+ +| Rate |--->| Encoder | V + +-----------+ | | Conversion | | | +---------+ + | Optional | | +------------+ +---------+ | Range | +->| High-pass |--+ | Encoder |----> + | Filter | | +--------------+ +---------+ | | Bit- + +-----------+ | | Delay | | CELT | +---------+ stream + +->| Compensation |->| Encoder | ^ + | | | |------+ + +--------------+ +---------+ +]]> + +
+ + + +For a normal encoder where both the SILK and the CELT modules are included, an optimal +encoder should select which coding mode to use at run-time depending on the conditions. +In the reference implementation, the frame size is selected by the application, but the +other configuration parameters (number of channels, bandwidth, mode) are automatically +selected (unless explicitly overridden by the application) depend on the following: + +Requested bitrate +Input sampling rate +Type of signal (speech vs music) +Frame size in use + + +The type of signal currently needs to be provided by the application (though it can be +changed in real-time). An Opus encoder implementation could also do automatic detection, +but since Opus is an interactive codec, such an implementation would likely have to either +delay the signal (for non-interactive applications) or delay the mode switching decisions (for +interactive applications). + + + +When the encoder is configured for voice over IP applications, the input signal is +filtered by a high-pass filter to remove the lowest part of the spectrum +that contains little speech energy and may contain background noise. This is a second order +Auto Regressive Moving Average (i.e., with poles and zeros) filter with a cut-off frequency around 50 Hz. +In the future, a music detector may also be used to lower the cut-off frequency when the +input signal is detected to be music rather than speech. + + +
+ +The range coder acts as the bit-packer for Opus. +It is used in three different ways: to encode + + +Entropy-coded symbols with a fixed probability model using ec_encode() + (entenc.c), + + +Integers from 0 to (2**M - 1) using ec_enc_uint() or ec_enc_bits() + (entenc.c), + +Integers from 0 to (ft - 1) (where ft is not a power of two) using + ec_enc_uint() (entenc.c). + + + + + +The range encoder maintains an internal state vector composed of the four-tuple + (val, rng, rem, ext) representing the low end of the current + range, the size of the current range, a single buffered output byte, and a + count of additional carry-propagating output bytes. +Both val and rng are 32-bit unsigned integer values, rem is a byte value or + less than 255 or the special value -1, and ext is an unsigned integer with at + least 11 bits. +This state vector is initialized at the start of each each frame to the value + (0, 2**31, -1, 0). +After encoding a sequence of symbols, the value of rng in the encoder should + exactly match the value of rng in the decoder after decoding the same sequence + of symbols. +This is a powerful tool for detecting errors in either an encoder or decoder + implementation. +The value of val, on the other hand, represents different things in the encoder + and decoder, and is not expected to match. + + + +The decoder has no analog for rem and ext. +These are used to perform carry propagation in the renormalization loop below. +Each iteration of this loop produces 9 bits of output, consisting of 8 data + bits and a carry flag. +The encoder cannot determine the final value of the output bytes until it + propagates these carry flags. +Therefore the reference implementation buffers a single non-propagating output + byte (i.e., one less than 255) in rem and keeps a count of additional + propagating (i.e., 255) output bytes in ext. +An implementation may choose to use any mathematically equivalent scheme to + perform carry propagation. + + +
+ +The main encoding function is ec_encode() (entenc.c), which encodes symbol k in + the current context using the same three-tuple (fl[k], fh[k], ft) + as the decoder to describe the range of the symbol (see + ). + + +ec_encode() updates the state of the encoder as follows. +If fl[k] is greater than zero, then +
+ +
+Otherwise, val is unchanged and +
+ +
+The divisions here are integer division. + + +
+ +After this update, the range is normalized using a procedure very similar to + that of , implemented by + ec_enc_normalize() (entenc.c). +The following process is repeated until rng > 2**23. +First, the top 9 bits of val, (val>>23), are sent to the carry buffer, + described in . +Then, the encoder sets +
+ +
+ +
+ +
+ +The function ec_enc_carry_out() (entenc.c) implements carry propagation and + output buffering. +It takes as input a 9-bit value, c, consisting of 8 data bits and an additional + carry bit. +If c is equal to the value 255, then ext is simply incremented, and no other + state updates are performed. +Otherwise, let b = (c>>8) be the carry bit. +Then, + + +If the buffered byte rem contains a value other than -1, the encoder outputs + the byte (rem + b). +Otherwise, if rem is -1, no byte is output. + + +If ext is non-zero, then the encoder outputs ext bytes---all with a value of 0 + if b is set, or 255 if b is unset---and sets ext to 0. + + +rem is set to the 8 data bits: +
+ +
+ + + +
+ +
+ +
+ +The reference implementation uses three additional encoding methods that are + exactly equivalent to the above, but make assumptions and simplifications that + allow for a more efficient implementation. + + +
+ +The first is ec_encode_bin() (entenc.c), defined using the parameter ftb + instead of ft. +It is mathematically equivalent to calling ec_encode() with + ft = (1<<ftb), but avoids using division. + +
+ +
+ +The next is ec_enc_bit_logp() (entenc.c), which encodes a single binary symbol. +The context is described by a single parameter, logp, which is the absolute + value of the base-2 logarithm of the probability of a "1". +It is mathematically equivalent to calling ec_encode() with the 3-tuple + (fl[k] = 0, fh[k] = (1<<logp) - 1, + ft = (1<<logp)) if k is 0 and with + (fl[k] = (1<<logp) - 1, + fh[k] = ft = (1<<logp)) if k is 1. +The implementation requires no multiplications or divisions. + +
+ +
+ +The last is ec_enc_icdf() (entenc.c), which encodes a single binary symbol with + a table-based context of up to 8 bits. +This uses the same icdf table as ec_dec_icdf() from + . +The function is mathematically equivalent to calling ec_encode() with + fl[k] = (1<<ftb) - icdf[k-1] (or 0 if + k == 0), fh[k] = (1<<ftb) - icdf[k], and + ft = (1<<ftb). +This only saves a few arithmetic operations over ec_encode_bin(), but allows + the encoder to use the same icdf tables as the decoder. + +
+ +
+ +
+ +The raw bits used by the CELT layer are packed at the end of the buffer using + ec_enc_bits() (entenc.c). +Because the raw bits may continue into the last byte output by the range coder + if there is room in the low-order bits, the encoder must be prepared to merge + these values into a single byte. +The procedure in does this in a way that + ensures both the range coded data and the raw bits can be decoded + successfully. + +
+ +
+ +The function ec_enc_uint() (entenc.c) encodes one of ft equiprobable symbols in + the range 0 to (ft - 1), inclusive, each with a frequency of 1, + where ft may be as large as (2**32 - 1). +Like the decoder (see ), it splits up the + value into a range coded symbol representing up to 8 of the high bits, and, if + necessary, raw bits representing the remainder of the value. + + +ec_enc_uint() takes a two-tuple (t, ft), where t is the value to be + encoded, 0 <= t < ft, and ft is not necessarily a + power of two. +Let ftb = ilog(ft - 1), i.e., the number of bits required + to store (ft - 1) in two's complement notation. +If ftb is 8 or less, then t is encoded directly using ec_encode() with the + three-tuple (t, t + 1, ft). + + +If ftb is greater than 8, then the top 8 bits of t are encoded using the + three-tuple (t>>(ftb - 8), + (t>>(ftb - 8)) + 1, + ((ft - 1)>>(ftb - 8)) + 1), and the + remaining bits, + (t & ((1<<(ftb - 8)) - 1), + are encoded as raw bits with ec_enc_bits(). + +
+ +
+ +After all symbols are encoded, the stream must be finalized by outputting a + value inside the current range. +Let end be the integer in the interval [val, val + rng) with the + largest number of trailing zero bits, b, such that + (end + (1<<b) - 1) is also in the interval + [val, val + rng). +This choice of end allows the maximum number of trailing bits to be set to + arbitrary values while still ensuring the range coded part of the buffer can + be decoded correctly. +Then, while end is not zero, the top 9 bits of end, i.e., (end>>23), are + passed to the carry buffer in accordance with the procedure in + , and end is updated via +
+ +
+Finally, if the buffered output byte, rem, is neither zero nor the special + value -1, or the carry count, ext, is greater than zero, then 9 zero bits are + sent to the carry buffer to flush it to the output buffer. +When outputting the final byte from the range coder, if it would overlap any + raw bits already packed into the end of the output buffer, they should be ORed + into the same byte. +The bit allocation routines in the CELT layer should ensure that this can be + done without corrupting the range coder data so long as end is chosen as + described above. +If there is any space between the end of the range coder data and the end of + the raw bits, it is padded with zero bits. +This entire process is implemented by ec_enc_done() (entenc.c). + +
+ +
+ + The bit allocation routines in Opus need to be able to determine a + conservative upper bound on the number of bits that have been used + to encode the current frame thus far. This drives allocation + decisions and ensures that the range coder and raw bits will not + overflow the output buffer. This is computed in the + reference implementation to whole-bit precision by + the function ec_tell() (entcode.h) and to fractional 1/8th bit + precision by the function ec_tell_frac() (entcode.c). + Like all operations in the range coder, it must be implemented in a + bit-exact manner, and must produce exactly the same value returned by + the same functions in the decoder after decoding the same symbols. + +
+ +
+ +
+ + In many respects the SILK encoder mirrors the SILK decoder described + in . + Details such as the quantization and range coder tables can be found + there, while this section describes the high-level design choices that + were made. + The diagram below shows the basic modules of the SILK encoder. +
+ +| Rate |--->| Mixing |--->| Core |----------> +Input |Conversion| | | | Encoder | Bitstream + +----------+ +--------+ +---------+ +]]> + +
+ + +
+ +The input signal's sampling rate is adjusted by a sample rate conversion +module so that it matches the SILK internal sampling rate. +The input to the sample rate converter is delayed by a number of samples +depending on the sample rate ratio, such that the overall delay is constant +for all input and output sample rates. + +
+ +
+ +The stereo mixer is only used for stereo input signals. +It converts a stereo left/right signal into an adaptive +mid/side representation. +The first step is to compute non-adaptive mid/side signals +as half the sum and difference between left and right signals. +The side signal is then minimized in energy by subtracting a +prediction of it based on the mid signal. +This prediction works well when the left and right signals +exhibit linear dependency, for instance for an amplitude-panned +input signal. +Like in the decoder, the prediction coefficients are linearly +interpolated during the first 8 ms of the frame. + The mid signal is always encoded, whereas the residual + side signal is only encoded if it has sufficient + energy compared to the mid signal's energy. + If it has not, + the "mid_only_flag" is set without encoding the side signal. + + +The predictor coefficients are coded regardless of whether +the side signal is encoded. +For each frame, two predictor coefficients are computed, one +that predicts between low-passed mid and side channels, and +one that predicts between high-passed mid and side channels. +The low-pass filter is a simple three-tap filter +and creates a delay of one sample. +The high-pass filtered signal is the difference between +the mid signal delayed by one sample and the low-passed +signal. Instead of explicitly computing the high-passed +signal, it is computationally more efficient to transform +the prediction coefficients before applying them to the +filtered mid signal, as follows +
+ + + +
+where w0 and w1 are the low-pass and high-pass prediction +coefficients, mid(n-1) is the mid signal delayed by one sample, +LP(n) and HP(n) are the low-passed and high-passed +signals and pred(n) is the prediction signal that is subtracted +from the side signal. + +
+ +
+ +What follows is a description of the core encoder and its components. +For simplicity, the core encoder is referred to simply as the encoder in +the remainder of this section. An overview of the encoder is given in +. + +
+ +| | + +---------+ | +---------+ | | + |Voice | | |LTP |12 | | + +-->|Activity |--+ +----->|Scaling |-----------+---->| | + | |Detector |3 | | |Control |<--+ | | | + | +---------+ | | +---------+ | | | | + | | | +---------+ | | | | + | | | |Gains | | | | | + | | | +-->|Processor|---|---+---|---->| R | + | | | | | |11 | | | | a | + | \/ | | +---------+ | | | | n | + | +---------+ | | +---------+ | | | | g | + | |Pitch | | | |LSF | | | | | e | + | +->|Analysis |---+ | |Quantizer|---|---|---|---->| | + | | | |4 | | | |8 | | | | E |--> + | | +---------+ | | +---------+ | | | | n | 2 + | | | | 9/\ 10| | | | | c | + | | | | | \/ | | | | o | + | | +---------+ | | +----------+ | | | | d | + | | |Noise | +--|-->|Prediction|--+---|---|---->| e | + | +->|Shaping |---|--+ |Analysis |7 | | | | r | + | | |Analysis |5 | | | | | | | | | + | | +---------+ | | +----------+ | | | | | + | | | | /\ | | | | | + | | +----------|--|--------+ | | | | | + | | | \/ \/ \/ \/ \/ | | + | | | +---------+ +------------+ | | + | | | | | |Noise | | | +-+-------+-----+------>|Prefilter|--------->|Shaping |-->| | +1 | | 6 |Quantization|13 | | + +---------+ +------------+ +---+ + +1: Input speech signal +2: Range encoded bitstream +3: Voice activity estimate +4: Pitch lags (per 5 ms) and voicing decision (per 20 ms) +5: Noise shaping quantization coefficients + - Short term synthesis and analysis + noise shaping coefficients (per 5 ms) + - Long term synthesis and analysis noise + shaping coefficients (per 5 ms and for voiced speech only) + - Noise shaping tilt (per 5 ms) + - Quantizer gain/step size (per 5 ms) +6: Input signal filtered with analysis noise shaping filters +7: Short and long term prediction coefficients + LTP (per 5 ms) and LPC (per 20 ms) +8: LSF quantization indices +9: LSF coefficients +10: Quantized LSF coefficients +11: Processed gains, and synthesis noise shape coefficients +12: LTP state scaling coefficient. Controlling error propagation + / prediction gain trade-off +13: Quantized signal +]]> + +
+ +
+ +The input signal is processed by a Voice Activity Detector (VAD) to produce +a measure of voice activity, spectral tilt, and signal-to-noise estimates for +each frame. The VAD uses a sequence of half-band filterbanks to split the +signal into four subbands: 0...Fs/16, Fs/16...Fs/8, Fs/8...Fs/4, and +Fs/4...Fs/2, where Fs is the sampling frequency (8, 12, 16, or 24 kHz). +The lowest subband, from 0 - Fs/16, is high-pass filtered with a first-order +moving average (MA) filter (with transfer function H(z) = 1-z**(-1)) to +reduce the energy at the lowest frequencies. For each frame, the signal +energy per subband is computed. +In each subband, a noise level estimator tracks the background noise level +and a Signal-to-Noise Ratio (SNR) value is computed as the logarithm of the +ratio of energy to noise level. +Using these intermediate variables, the following parameters are calculated +for use in other SILK modules: + + +Average SNR. The average of the subband SNR values. + + + +Smoothed subband SNRs. Temporally smoothed subband SNR values. + + + +Speech activity level. Based on the average SNR and a weighted average of the +subband energies. + + + +Spectral tilt. A weighted average of the subband SNRs, with positive weights +for the low subbands and negative weights for the high subbands. + + + +
+ +
+ +The input signal is processed by the open loop pitch estimator shown in +. +
+ +|sampling|->|Correlator| | + | | | | | |4 + | +--------+ +----------+ \/ + | | 2 +-------+ + | | +-->|Speech |5 + +---------+ +--------+ | \/ | |Type |-> + |LPC | |Down | | +----------+ | | + +->|Analysis | +->|sample |-+------------->|Time- | +-------+ + | | | | |to 8 kHz| |Correlator|-----------> + | +---------+ | +--------+ |__________| 6 + | | | |3 + | \/ | \/ + | +---------+ | +----------+ + | |Whitening| | |Time- | +-+->|Filter |-+--------------------------->|Correlator|-----------> +1 | | | | 7 + +---------+ +----------+ + +1: Input signal +2: Lag candidates from stage 1 +3: Lag candidates from stage 2 +4: Correlation threshold +5: Voiced/unvoiced flag +6: Pitch correlation +7: Pitch lags +]]> + +
+The pitch analysis finds a binary voiced/unvoiced classification, and, for +frames classified as voiced, four pitch lags per frame - one for each +5 ms subframe - and a pitch correlation indicating the periodicity of +the signal. +The input is first whitened using a Linear Prediction (LP) whitening filter, +where the coefficients are computed through standard Linear Prediction Coding +(LPC) analysis. The order of the whitening filter is 16 for best results, but +is reduced to 12 for medium complexity and 8 for low complexity modes. +The whitened signal is analyzed to find pitch lags for which the time +correlation is high. +The analysis consists of three stages for reducing the complexity: + +In the first stage, the whitened signal is downsampled to 4 kHz +(from 8 kHz) and the current frame is correlated to a signal delayed +by a range of lags, starting from a shortest lag corresponding to +500 Hz, to a longest lag corresponding to 56 Hz. + + +The second stage operates on an 8 kHz signal (downsampled from 12, 16, +or 24 kHz) and measures time correlations only near the lags +corresponding to those that had sufficiently high correlations in the first +stage. The resulting correlations are adjusted for a small bias towards +short lags to avoid ending up with a multiple of the true pitch lag. +The highest adjusted correlation is compared to a threshold depending on: + + +Whether the previous frame was classified as voiced + + +The speech activity level + + +The spectral tilt. + + +If the threshold is exceeded, the current frame is classified as voiced and +the lag with the highest adjusted correlation is stored for a final pitch +analysis of the highest precision in the third stage. + + +The last stage operates directly on the whitened input signal to compute time +correlations for each of the four subframes independently in a narrow range +around the lag with highest correlation from the second stage. + + + +
+ +
+ +The noise shaping analysis finds gains and filter coefficients used in the +prefilter and noise shaping quantizer. These parameters are chosen such that +they will fulfill several requirements: + + +Balancing quantization noise and bitrate. +The quantization gains determine the step size between reconstruction levels +of the excitation signal. Therefore, increasing the quantization gain +amplifies quantization noise, but also reduces the bitrate by lowering +the entropy of the quantization indices. + + +Spectral shaping of the quantization noise; the noise shaping quantizer is +capable of reducing quantization noise in some parts of the spectrum at the +cost of increased noise in other parts without substantially changing the +bitrate. +By shaping the noise such that it follows the signal spectrum, it becomes +less audible. In practice, best results are obtained by making the shape +of the noise spectrum slightly flatter than the signal spectrum. + + +De-emphasizing spectral valleys; by using different coefficients in the +analysis and synthesis part of the prefilter and noise shaping quantizer, +the levels of the spectral valleys can be decreased relative to the levels +of the spectral peaks such as speech formants and harmonics. +This reduces the entropy of the signal, which is the difference between the +coded signal and the quantization noise, thus lowering the bitrate. + + +Matching the levels of the decoded speech formants to the levels of the +original speech formants; an adjustment gain and a first order tilt +coefficient are computed to compensate for the effect of the noise +shaping quantization on the level and spectral tilt. + + + + +
+ + + Frequency + +1: Input signal spectrum +2: De-emphasized and level matched spectrum +3: Quantization noise spectrum +]]> + +
+ shows an example of an +input signal spectrum (1). +After de-emphasis and level matching, the spectrum has deeper valleys (2). +The quantization noise spectrum (3) more or less follows the input signal +spectrum, while having slightly less pronounced peaks. +The entropy, which provides a lower bound on the bitrate for encoding the +excitation signal, is proportional to the area between the de-emphasized +spectrum (2) and the quantization noise spectrum (3). Without de-emphasis, +the entropy is proportional to the area between input spectrum (1) and +quantization noise (3) - clearly higher. + + + +The transformation from input signal to de-emphasized signal can be +described as a filtering operation with a filter +
+ + + +
+having an adjustment gain G, a first order tilt adjustment filter with +tilt coefficient c_tilt, and where +
+ + + +
+is the analysis part of the de-emphasis filter, consisting of the short-term +shaping filter with coefficients a_ana(k), and the long-term shaping filter +with coefficients b_ana(k) and pitch lag L. +The parameter d determines the number of long-term shaping filter taps. + + + +Similarly, but without the tilt adjustment, the synthesis part can be written as +
+ + + +
+ + +All noise shaping parameters are computed and applied per subframe of 5 ms. +First, an LPC analysis is performed on a windowed signal block of 15 ms. +The signal block has a look-ahead of 5 ms relative to the current subframe, +and the window is an asymmetric sine window. The LPC analysis is done with the +autocorrelation method, with an order of between 8, in lowest-complexity mode, +and 16, for best quality. + + +Optionally the LPC analysis and noise shaping filters are warped by replacing +the delay elements by first-order allpass filters. +This increases the frequency resolution at low frequencies and reduces it at +high ones, which better matches the human auditory system and improves +quality. +The warped analysis and filtering comes at a cost in complexity +and is therefore only done in higher complexity modes. + + +The quantization gain is found by taking the square root of the residual energy +from the LPC analysis and multiplying it by a value inversely proportional +to the coding quality control parameter and the pitch correlation. + + +Next the two sets of short-term noise shaping coefficients a_ana(k) and +a_syn(k) are obtained by applying different amounts of bandwidth expansion to the +coefficients found in the LPC analysis. +This bandwidth expansion moves the roots of the LPC polynomial towards the +origin, using the formulas +
+ + + +
+where a(k) is the k'th LPC coefficient, and the bandwidth expansion factors +g_ana and g_syn are calculated as +
+ + + +
+where C is the coding quality control parameter between 0 and 1. +Applying more bandwidth expansion to the analysis part than to the synthesis +part gives the desired de-emphasis of spectral valleys in between formants. + + + +The long-term shaping is applied only during voiced frames. +It uses three filter taps, described by +
+ + + +
+For unvoiced frames these coefficients are set to 0. The multiplication factors +F_ana and F_syn are chosen between 0 and 1, depending on the coding quality +control parameter, as well as the calculated pitch correlation and smoothed +subband SNR of the lowest subband. By having F_ana less than F_syn, +the pitch harmonics are emphasized relative to the valleys in between the +harmonics. + + + +The tilt coefficient c_tilt is for unvoiced frames chosen as +
+ + + +
+and as +
+ + + +
+for voiced frames, where V is the voice activity level between 0 and 1. + + +The adjustment gain G serves to correct any level mismatch between the original +and decoded signals that might arise from the noise shaping and de-emphasis. +This gain is computed as the ratio of the prediction gain of the short-term +analysis and synthesis filter coefficients. The prediction gain of an LPC +synthesis filter is the square root of the output energy when the filter is +excited by a unit-energy impulse on the input. +An efficient way to compute the prediction gain is by first computing the +reflection coefficients from the LPC coefficients through the step-down +algorithm, and extracting the prediction gain from the reflection coefficients +as +
+ + + +
+where r_k is the k'th reflection coefficient. + + + +Initial values for the quantization gains are computed as the square-root of +the residual energy of the LPC analysis, adjusted by the coding quality control +parameter. +These quantization gains are later adjusted based on the results of the +prediction analysis. + +
+ +
+ +The prediction analysis is performed in one of two ways depending on how +the pitch estimator classified the frame. +The processing for voiced and unvoiced speech is described in + and + , respectively. + Inputs to this function include the pre-whitened signal from the + pitch estimator (see ). + + +
+ + For a frame of voiced speech the pitch pulses will remain dominant in the + pre-whitened input signal. + Further whitening is desirable as it leads to higher quality at the same + available bitrate. + To achieve this, a Long-Term Prediction (LTP) analysis is carried out to + estimate the coefficients of a fifth-order LTP filter for each of four + subframes. + The LTP coefficients are quantized using the method described in + , and the quantized LTP + coefficients are used to compute the LTP residual signal. + This LTP residual signal is the input to an LPC analysis where the LPC coefficients are + estimated using Burg's method , such that the residual energy is minimized. + The estimated LPC coefficients are converted to a Line Spectral Frequency (LSF) vector + and quantized as described in . +After quantization, the quantized LSF vector is converted back to LPC +coefficients using the full procedure in . +By using quantized LTP coefficients and LPC coefficients derived from the +quantized LSF coefficients, the encoder remains fully synchronized with the +decoder. +The quantized LPC and LTP coefficients are also used to filter the input +signal and measure residual energy for each of the four subframes. + +
+
+ +For a speech signal that has been classified as unvoiced, there is no need +for LTP filtering, as it has already been determined that the pre-whitened +input signal is not periodic enough within the allowed pitch period range +for LTP analysis to be worth the cost in terms of complexity and bitrate. +The pre-whitened input signal is therefore discarded, and instead the input +signal is used for LPC analysis using Burg's method. +The resulting LPC coefficients are converted to an LSF vector and quantized +as described in the following section. +They are then transformed back to obtain quantized LPC coefficients, which +are then used to filter the input signal and measure residual energy for +each of the four subframes. + +
+ +The main purpose of linear prediction in SILK is to reduce the bitrate by +minimizing the residual energy. +At least at high bitrates, perceptual aspects are handled +independently by the noise shaping filter. +Burg's method is used because it provides higher prediction gain +than the autocorrelation method and, unlike the covariance method, +produces stable filters (assuming numerical errors don't spoil +that). SILK's implementation of Burg's method is also computationally +faster than the autocovariance method. +The implementation of Burg's method differs from traditional +implementations in two aspects. +The first difference is that it +operates on autocorrelations, similar to the Schur algorithm , but +with a simple update to the autocorrelations after finding each +reflection coefficient to make the result identical to Burg's method. +This brings down the complexity of Burg's method to near that of +the autocorrelation method. +The second difference is that the signal in each subframe is scaled +by the inverse of the residual quantization step size. Subframes with +a small quantization step size will on average spend more bits for a +given amount of residual energy than subframes with a large step size. +Without scaling, Burg's method minimizes the total residual energy in +all subframes, which doesn't necessarily minimize the total number of +bits needed for coding the quantized residual. The residual energy +of the scaled subframes is a better measure for that number of +bits. + +
+
+
+ +
+ +Unlike many other speech codecs, SILK uses variable bitrate coding +for the LSFs. +This improves the average rate-distortion (R-D) tradeoff and reduces outliers. +The variable bitrate coding minimizes a linear combination of the weighted +quantization errors and the bitrate. +The weights for the quantization errors are the Inverse +Harmonic Mean Weighting (IHMW) function proposed by Laroia et al. +(see ). +These weights are referred to here as Laroia weights. + + +The LSF quantizer consists of two stages. +The first stage is an (unweighted) vector quantizer (VQ), with a +codebook size of 32 vectors. +The quantization errors for the codebook vector are sorted, and +for the N best vectors a second stage quantizer is run. +By varying the number N a tradeoff is made between R-D performance +and computational efficiency. +For each of the N codebook vectors the Laroia weights corresponding +to that vector (and not to the input vector) are calculated. +Then the residual between the input LSF vector and the codebook +vector is scaled by the square roots of these Laroia weights. +This scaling partially normalizes error sensitivity for the +residual vector, so that a uniform quantizer with fixed +step sizes can be used in the second stage without too much +performance loss. +And by scaling with Laroia weights determined from the first-stage +codebook vector, the process can be reversed in the decoder. + + +The second stage uses predictive delayed decision scalar +quantization. +The quantization error is weighted by Laroia weights determined +from the LSF input vector. +The predictor multiplies the previous quantized residual value +by a prediction coefficient that depends on the vector index from the +first stage VQ and on the location in the LSF vector. +The prediction is subtracted from the LSF residual value before +quantizing the result, and added back afterwards. +This subtraction can be interpreted as shifting the quantization levels +of the scalar quantizer, and as a result the quantization error of +each value depends on the quantization decision of the previous value. +This dependency is exploited by the delayed decision mechanism to +search for a quantization sequency with best R-D performance +with a Viterbi-like algorithm . +The quantizer processes the residual LSF vector in reverse order +(i.e., it starts with the highest residual LSF value). +This is done because the prediction works slightly +better in the reverse direction. + + +The quantization index of the first stage is entropy coded. +The quantization sequence from the second stage is also entropy +coded, where for each element the probability table is chosen +depending on the vector index from the first stage and the location +of that element in the LSF vector. + + +
+ +If the input is stable, finding the best candidate usually results in a +quantized vector that is also stable. Because of the two-stage approach, +however, it is possible that the best quantization candidate is unstable. +The encoder applies the same stabilization procedure applied by the decoder + (see to ensure the LSF parameters + are within their valid range, increasingly sorted, and have minimum + distances between each other and the border values. + +
+
+ +
+ +For voiced frames, the prediction analysis described in + resulted in four sets +(one set per subframe) of five LTP coefficients, plus four weighting matrices. +The LTP coefficients for each subframe are quantized using entropy constrained +vector quantization. +A total of three vector codebooks are available for quantization, with +different rate-distortion trade-offs. The three codebooks have 10, 20, and +40 vectors and average rates of about 3, 4, and 5 bits per vector, respectively. +Consequently, the first codebook has larger average quantization distortion at +a lower rate, whereas the last codebook has smaller average quantization +distortion at a higher rate. +Given the weighting matrix W_ltp and LTP vector b, the weighted rate-distortion +measure for a codebook vector cb_i with rate r_i is give by +
+ + + +
+where u is a fixed, heuristically-determined parameter balancing the distortion +and rate. +Which codebook gives the best performance for a given LTP vector depends on the +weighting matrix for that LTP vector. +For example, for a low valued W_ltp, it is advantageous to use the codebook +with 10 vectors as it has a lower average rate. +For a large W_ltp, on the other hand, it is often better to use the codebook +with 40 vectors, as it is more likely to contain the best codebook vector. +The weighting matrix W_ltp depends mostly on two aspects of the input signal. +The first is the periodicity of the signal; the more periodic, the larger W_ltp. +The second is the change in signal energy in the current subframe, relative to +the signal one pitch lag earlier. +A decaying energy leads to a larger W_ltp than an increasing energy. +Both aspects fluctuate relatively slowly, which causes the W_ltp matrices for +different subframes of one frame often to be similar. +Because of this, one of the three codebooks typically gives good performance +for all subframes, and therefore the codebook search for the subframe LTP +vectors is constrained to only allow codebook vectors to be chosen from the +same codebook, resulting in a rate reduction. + + + +To find the best codebook, each of the three vector codebooks is +used to quantize all subframe LTP vectors and produce a combined +weighted rate-distortion measure for each vector codebook. +The vector codebook with the lowest combined rate-distortion +over all subframes is chosen. The quantized LTP vectors are used +in the noise shaping quantizer, and the index of the codebook +plus the four indices for the four subframe codebook vectors +are passed on to the range encoder. + +
+ +
+ +In the prefilter the input signal is filtered using the spectral valley +de-emphasis filter coefficients from the noise shaping analysis +(see ). +By applying only the noise shaping analysis filter to the input signal, +it provides the input to the noise shaping quantizer. + +
+ +
+ +The noise shaping quantizer independently shapes the signal and coding noise +spectra to obtain a perceptually higher quality at the same bitrate. + + +The prefilter output signal is multiplied with a compensation gain G computed +in the noise shaping analysis. Then the output of a synthesis shaping filter +is added, and the output of a prediction filter is subtracted to create a +residual signal. +The residual signal is multiplied by the inverse quantized quantization gain +from the noise shaping analysis, and input to a scalar quantizer. +The quantization indices of the scalar quantizer represent a signal of pulses +that is input to the pyramid range encoder. +The scalar quantizer also outputs a quantization signal, which is multiplied +by the quantized quantization gain from the noise shaping analysis to create +an excitation signal. +The output of the prediction filter is added to the excitation signal to form +the quantized output signal y(n). +The quantized output signal y(n) is input to the synthesis shaping and +prediction filters. + + +Optionally the noise shaping quantizer operates in a delayed decision +mode. +In this mode it uses a Viterbi algorithm to keep track of +multiple rounding choices in the quantizer and select the best +one after a delay of 32 samples. This improves the rate/distortion +performance of the quantizer. + +
+ +
+ + SILK was designed to run in Variable Bitrate (VBR) mode. However + the reference implementation also has a Constant Bitrate (CBR) mode + for SILK. In CBR mode SILK will attempt to encode each packet with + no more than the allowed number of bits. The Opus wrapper code + then pads the bitstream if any unused bits are left in SILK mode, or + encodes the high band with the remaining number of bits in Hybrid mode. + The number of payload bits is adjusted by changing + the quantization gains and the rate/distortion tradeoff in the noise + shaping quantizer, in an iterative loop + around the noise shaping quantizer and entropy coding. + Compared to the SILK VBR mode, the CBR mode has lower + audio quality at a given average bitrate, and also has higher + computational complexity. + +
+ +
+ +
+ + +
+ +Most of the aspects of the CELT encoder can be directly derived from the description +of the decoder. For example, the filters and rotations in the encoder are simply the +inverse of the operation performed by the decoder. Similarly, the quantizers generally +optimize for the mean square error (because noise shaping is part of the bit-stream itself), +so no special search is required. For this reason, only the less straightforward aspects of the +encoder are described here. + + +
+The pitch prefilter is applied after the pre-emphasis. It is applied +in such a way as to be the inverse of the decoder's post-filter. The main non-obvious aspect of the +prefilter is the selection of the pitch period. The pitch search should be optimized for the +following criteria: + +continuity: it is important that the pitch period +does not change abruptly between frames; and +avoidance of pitch multiples: when the period used is a multiple of the real period +(lower frequency fundamental), the post-filter loses most of its ability to reduce noise + + +
+ +
+ +The MDCT output is divided into bands that are designed to match the ear's critical +bands for the smallest (2.5 ms) frame size. The larger frame sizes use integer +multiples of the 2.5 ms layout. For each band, the encoder +computes the energy that will later be encoded. Each band is then normalized by the +square root of the unquantized energy, such that each band now forms a unit vector X. +The energy and the normalization are computed by compute_band_energies() +and normalise_bands() (bands.c), respectively. + +
+ +
+ + +Energy quantization (both coarse and fine) can be easily understood from the decoding process. +For all useful bitrates, the coarse quantizer always chooses the quantized log energy value that +minimizes the error for each band. Only at very low rate does the encoder allow larger errors to +minimize the rate and avoid using more bits than are available. When the +available CPU requirements allow it, it is best to try encoding the coarse energy both with and without +inter-frame prediction such that the best prediction mode can be selected. The optimal mode depends on +the coding rate, the available bitrate, and the current rate of packet loss. + + +The fine energy quantizer always chooses the quantized log energy value that +minimizes the error for each band because the rate of the fine quantization depends only +on the bit allocation and not on the values that are coded. + +
+ +
+The encoder must use exactly the same bit allocation process as used by the decoder +and described in . The three mechanisms that can be used by the +encoder to adjust the bitrate on a frame-by-frame basis are band boost, allocation trim, +and band skipping. + + +
+The reference encoder makes a decision to boost a band when the energy of that band is significantly +higher than that of the neighboring bands. Let E_j be the log-energy of band j, we define + +D_j = 2*E_j - E_j-1 - E_j+1 + + +The allocation of band j is boosted once if D_j > t1 and twice if D_j > t2. For LM>=1, t1=2 and t2=4, +while for LM<1, t1=3 and t2=5. + + +
+ +
+The allocation trim is a value between 0 and 10 (inclusively) that controls the allocation +balance between the low and high frequencies. The encoder starts with a safe "default" of 5 +and deviates from that default in two different ways. First the trim can deviate by +/- 2 +depending on the spectral tilt of the input signal. For signals with more low frequencies, the +trim is increased by up to 2, while for signals with more high frequencies, the trim is +decreased by up to 2. +For stereo inputs, the trim value can +be decreased by up to 4 when the inter-channel correlation at low frequency (first 8 bands) +is high. +
+ +
+The encoder uses band skipping to ensure that the shape of the bands is only coded +if there is at least 1/2 bit per sample available for the PVQ. If not, then no bit is allocated +and folding is used instead. To ensure continuity in the allocation, some amount of hysteresis is +added to the process, such that a band that received PVQ bits in the previous frame only needs 7/16 +bit/sample to be coded for the current frame, while a band that did not receive PVQ bits in the +previous frames needs at least 9/16 bit/sample to be coded. +
+ +
+ +
+Because CELT applies mid-side stereo coupling in the normalized domain, it does not suffer from +important stereo image problems even when the two channels are completely uncorrelated. For this reason +it is always safe to use stereo coupling on any audio frame. That being said, there are some frames +for which dual (independent) stereo is still more efficient. This decision is made by comparing the estimated +entropy with and without coupling over the first 13 bands, taking into account the fact that all bands with +more than two MDCT bins require one extra degree of freedom when coded in mid-side. Let L1_ms and L1_lr +be the L1-norm of the mid-side vector and the L1-norm of the left-right vector, respectively. The decision +to use mid-side is made if and only if +
+ +
+where bins is the number of MDCT bins in the first 13 bands and E is the number of extra degrees of +freedom for mid-side coding. For LM>1, E=13, otherwise E=5. + + +The reference encoder decides on the intensity stereo threshold based on the bitrate alone. After +taking into account the frame size by subtracting 80 bits per frame for coarse energy, the first +band using intensity coding is as follows: + + + +bitrate (kb/s) +start band +<35 8 +35-50 12 +50-68 16 +84-84 18 +84-102 19 +102-130 20 +>130 disabled + + + +
+ +
+ +The choice of time-frequency resolution used in is based on +R-D optimization. The distortion is the L1-norm (sum of absolute values) of each band +after each TF resolution under consideration. The L1 norm is used because it represents the entropy +for a Laplacian source. The number of bits required to code a change in TF resolution between +two bands is higher than the cost of having those two bands use the same resolution, which is +what requires the R-D optimization. The optimal decision is computed using the Viterbi algorithm. +See tf_analysis() in celt/celt.c. + +
+ +
+ +The choice of the spreading value in has an +impact on the nature of the coding noise introduced by CELT. The larger the f_r value, the +lower the impact of the rotation, and the more tonal the coding noise. The +more tonal the signal, the more tonal the noise should be, so the CELT encoder determines +the optimal value for f_r by estimating how tonal the signal is. The tonality estimate +is based on discrete pdf (4-bin histogram) of each band. Bands that have a large number of small +values are considered more tonal and a decision is made by combining all bands with more than +8 samples. See spreading_decision() in celt/bands.c. + +
+ +
+CELT uses a Pyramid Vector Quantization (PVQ) +codebook for quantizing the details of the spectrum in each band that have not +been predicted by the pitch predictor. The PVQ codebook consists of all sums +of K signed pulses in a vector of N samples, where two pulses at the same position +are required to have the same sign. Thus the codebook includes +all integer codevectors y of N dimensions that satisfy sum(abs(y(j))) = K. + + + +In bands where there are sufficient bits allocated PVQ is used to encode +the unit vector that results from the normalization in + directly. Given a PVQ codevector y, +the unit vector X is obtained as X = y/||y||, where ||.|| denotes the +L2 norm. + + + +
+ + +The search for the best codevector y is performed by alg_quant() +(vq.c). There are several possible approaches to the +search, with a trade-off between quality and complexity. The method used in the reference +implementation computes an initial codeword y1 by projecting the normalized spectrum +X onto the codebook pyramid of K-1 pulses: + + +y0 = truncate_towards_zero( (K-1) * X / sum(abs(X))) + + + +Depending on N, K and the input data, the initial codeword y0 may contain from +0 to K-1 non-zero values. All the remaining pulses, with the exception of the last one, +are found iteratively with a greedy search that minimizes the normalized correlation +between y and X: +
+ +
+ + + +The search described above is considered to be a good trade-off between quality +and computational cost. However, there are other possible ways to search the PVQ +codebook and the implementers MAY use any other search methods. See alg_quant() in celt/vq.c. + +
+ +
+ + +The vector to encode, X, is converted into an index i such that + 0 <= i < V(N,K) as follows. +Let i = 0 and k = 0. +Then for j = (N - 1) down to 0, inclusive, do: + + +If k > 0, set + i = i + (V(N-j-1,k-1) + V(N-j,k-1))/2. + +Set k = k + abs(X[j]). + +If X[j] < 0, set + i = i + (V(N-j-1,k) + V(N-j,k))/2. + + + + + +The index i is then encoded using the procedure in + with ft = V(N,K). + + +
+ +
+ + + + + +
+ +
+ + +
+ + +It is our intention to allow the greatest possible choice of freedom in +implementing the specification. For this reason, outside of the exceptions +noted in this section, conformance is defined through the reference +implementation of the decoder provided in . +Although this document includes an English description of the codec, should +the description contradict the source code of the reference implementation, +the latter shall take precedence. + + + +Compliance with this specification means that in addition to following the normative keywords in this document, + a decoder's output MUST also be + within the thresholds specified by the opus_compare.c tool (included + with the code) when compared to the reference implementation for each of the + test vectors provided (see ) and for each output + sampling rate and channel count supported. In addition, a compliant + decoder implementation MUST have the same final range decoder state as that of the + reference decoder. It is therefore RECOMMENDED that the + decoder implement the same functional behavior as the reference. + + A decoder implementation is not required to support all output sampling + rates or all output channel counts. + + +
+ +Using the reference code provided in , +a test vector can be decoded with + +opus_demo -d <rate> <channels> testvectorX.bit testX.out + +where <rate> is the sampling rate and can be 8000, 12000, 16000, 24000, or 48000, and +<channels> is 1 for mono or 2 for stereo. + + + +If the range decoder state is incorrect for one of the frames, the decoder will exit with +"Error: Range coder state mismatch between encoder and decoder". If the decoder succeeds, then +the output can be compared with the "reference" output with + +opus_compare -s -r <rate> testvectorX.dec testX.out + +for stereo or + +opus_compare -r <rate> testvectorX.dec testX.out + +for mono. + + +In addition to indicating whether the test vector comparison passes, the opus_compare tool +outputs an "Opus quality metric" that indicates how well the tested decoder matches the +reference implementation. A quality of 0 corresponds to the passing threshold, while +a quality of 100 is the highest possible value and means that the output of the tested decoder is identical to the reference +implementation. The passing threshold (quality 0) was calibrated in such a way that it corresponds to +additive white noise with a 48 dB SNR (similar to what can be obtained on a cassette deck). +It is still possible for an implementation to sound very good with such a low quality measure +(e.g. if the deviation is due to inaudible phase distortion), but unless this is verified by +listening tests, it is RECOMMENDED that implementations achieve a quality above 90 for 48 kHz +decoding. For other sampling rates, it is normal for the quality metric to be lower +(typically as low as 50 even for a good implementation) because of harmless mismatch with +the delay and phase of the internal sampling rate conversion. + + + +On POSIX environments, the run_vectors.sh script can be used to verify all test +vectors. This can be done with + +run_vectors.sh <exec path> <vector path> <rate> + +where <exec path> is the directory where the opus_demo and opus_compare executables +are built and <vector path> is the directory containing the test vectors. + +
+ +
+ +Opus Custom is an OPTIONAL part of the specification that is defined to +handle special sample rates and frame rates that are not supported by the +main Opus specification. Use of Opus Custom is discouraged for all but very +special applications for which a frame size different from 2.5, 5, 10, or 20 ms is +needed (for either complexity or latency reasons). Because Opus Custom is +optional, streams encoded using Opus Custom cannot be expected to be decodable by all Opus +implementations. Also, because no in-band mechanism exists for specifying the sampling +rate and frame size of Opus Custom streams, out-of-band signaling is required. +In Opus Custom operation, only the CELT layer is available, using the opus_custom_* function +calls in opus_custom.h. + +
+ +
+ +
+ + +Implementations of the Opus codec need to take appropriate security considerations +into account, as outlined in . +It is extremely important for the decoder to be robust against malicious +payloads. +Malicious payloads must not cause the decoder to overrun its allocated memory + or to take an excessive amount of resources to decode. +Although problems +in encoders are typically rarer, the same applies to the encoder. Malicious +audio streams must not cause the encoder to misbehave because this would +allow an attacker to attack transcoding gateways. + + +The reference implementation contains no known buffer overflow or cases where + a specially crafted packet or audio segment could cause a significant increase + in CPU load. +However, on certain CPU architectures where denormalized floating-point + operations are much slower than normal floating-point operations, it is + possible for some audio content (e.g., silence or near-silence) to cause an + increase in CPU load. +Denormals can be introduced by reordering operations in the compiler and depend + on the target architecture, so it is difficult to guarantee that an implementation + avoids them. +For architectures on which denormals are problematic, adding very small + floating-point offsets to the affected signals to prevent significant numbers + of denormalized operations is RECOMMENDED. +Alternatively, it is often possible to configure the hardware to treat + denormals as zero (DAZ). +No such issue exists for the fixed-point reference implementation. + +The reference implementation was validated in the following conditions: + + +Sending the decoder valid packets generated by the reference encoder and + verifying that the decoder's final range coder state matches that of the + encoder. + + +Sending the decoder packets generated by the reference encoder and then + subjected to random corruption. + +Sending the decoder random packets. + +Sending the decoder packets generated by a version of the reference encoder + modified to make random coding decisions (internal fuzzing), including mode + switching, and verifying that the range coder final states match. + + +In all of the conditions above, both the encoder and the decoder were run + inside the Valgrind memory + debugger, which tracks reads and writes to invalid memory regions as well as + the use of uninitialized memory. +There were no errors reported on any of the tested conditions. + +
+ + +
+ +This document has no actions for IANA. + +
+ +
+ +Thanks to all other developers, including Raymond Chen, Soeren Skak Jensen, Gregory Maxwell, +Christopher Montgomery, and Karsten Vandborg Soerensen. We would also +like to thank Igor Dyakonov, Jan Skoglund, and Christian Hoene for their help with subjective testing of the +Opus codec. Thanks to Ralph Giles, John Ridges, Ben Schwartz, Keith Yan, Christian Hoene, Kat Walsh, and many others on the Opus and CELT mailing lists +for their bug reports and feedback. + +
+ +
+The authors agree to grant third parties the irrevocable right to copy, use and distribute +the work (excluding Code Components available under the simplified BSD license), with or +without modification, in any medium, without royalty, provided that, unless separate +permission is granted, redistributed modified works do not contain misleading author, version, +name of work, or endorsement information. +
+ + + + + + + + + +Key words for use in RFCs to Indicate Requirement Levels + + + + + + + + + + + +Requirements for an Internet Audio Codec + + + + + +IETF + + +This document provides specific requirements for an Internet audio + codec. These requirements address quality, sample rate, bitrate, + and packet-loss robustness, as well as other desirable properties. + + + + + + + + + + +SILK Speech Codec + + + + + + + + + + + + + + + + + +Robust and Efficient Quantization of Speech LSP Parameters Using Structured Vector Quantization + + + + + + + + + + + + + + + + +Constrained-Energy Lapped Transform (CELT) Codec + + + + + + + + + + + + + + + + + + +Guidelines for the use of Variable Bit Rate Audio with Secure RTP + + + + + + + + + + + + + + +Internet Denial-of-Service Considerations + + + + + +IAB + + +This document provides an overview of possible avenues for denial-of-service (DoS) attack on Internet systems. The aim is to encourage protocol designers and network engineers towards designs that are more robust. We discuss partial solutions that reduce the effectiveness of attacks, and how some solutions might inadvertently open up alternative vulnerabilities. This memo provides information for the Internet community. + + + + + + +Range encoding: An algorithm for removing redundancy from a digitised message + + + + + + + + +Source coding algorithms for fast data compression + + + + + + + + +A Pyramid Vector Quantizer + + + + + + + + +The Computation of Line Spectral Frequencies Using Chebyshev Polynomials + + + + + + + + + + +Valgrind website + + + + + + +Google NetEQ code + + + + + + +Google WebRTC code + + + + + + + +Opus Git Repository + + + + + + +Opus website + + + + + + +Vorbis website + + + + + + +Matroska website + + + + + + +Opus Testvectors (webside) + + + + + + +Opus Testvectors (proceedings) + + + + + + +Line Spectral Pairs +Wikipedia + + + + + +Range Coding +Wikipedia + + + + + +Hadamard Transform +Wikipedia + + + + + +Viterbi Algorithm +Wikipedia + + + + + +White Noise +Wikipedia + + + + + +Linear Prediction +Wikipedia + + + + + +Modified Discrete Cosine Transform +Wikipedia + + + + + +Fast Fourier Transform +Wikipedia + + + + + +Z-transform +Wikipedia + + + + + + +Maximum Entropy Spectral Analysis + + + + + + +A fixed point computation of partial correlation coefficients + + + + + + + + +Analysis/synthesis filter bank design based on time domain aliasing cancellation + + + + + + + + +A High-Quality Speech and Audio Codec With Less Than 10 ms delay + + + + + + + + + + + + +Subdivision of the audible frequency range into critical bands + + + + + + + + + +
+ +This appendix contains the complete source code for the +reference implementation of the Opus codec written in C. By default, +this implementation relies on floating-point arithmetic, but it can be +compiled to use only fixed-point arithmetic by defining the FIXED_POINT +macro. Information on building and using the reference implementation is +available in the README file. + + +The implementation can be compiled with either a C89 or a C99 +compiler. It is reasonably optimized for most platforms such that +only architecture-specific optimizations are likely to be useful. +The FFT used is a slightly modified version of the KISS-FFT library, +but it is easy to substitute any other FFT library. + + + +While the reference implementation does not rely on any +undefined behavior as defined by C89 or C99, +it relies on common implementation-defined behavior +for two's complement architectures: + +Right shifts of negative values are consistent with two's complement arithmetic, so that a>>b is equivalent to floor(a/(2**b)), +For conversion to a signed integer of N bits, the value is reduced modulo 2**N to be within range of the type, +The result of integer division of a negative value is truncated towards zero, and +The compiler provides a 64-bit integer type (a C99 requirement which is supported by most C89 compilers). + + + + +In its current form, the reference implementation also requires the following +architectural characteristics to obtain acceptable performance: + +Two's complement arithmetic, +At least a 16 bit by 16 bit integer multiplier (32-bit result), and +At least a 32-bit adder/accumulator. + + + + +
+ +The complete source code can be extracted from this draft, by running the +following command line: + + + opus_source.tar.gz +]]> + +tar xzvf opus_source.tar.gz + +cd opus_source +make + +On systems where the provided Makefile does not work, the following command line may be used to compile +the source code: + + + + + +On systems where the base64 utility is not present, the following commands can be used instead: + + opus.b64 +]]> +openssl base64 -d -in opus.b64 > opus_source.tar.gz + + + +
+ +
+ +As of the time of publication of this memo, an up-to-date implementation conforming to +this standard is available in a + Git repository. +Releases and other resources are available at + . However, although that implementation is expected to + remain conformant with the standard, it is the code in this document that shall + remain normative. + +
+ +
+ + + +
+ +
+ +Because of size constraints, the Opus test vectors are not distributed in this +draft. They are available in the proceedings of the 83th IETF meeting (Paris) and from the Opus codec website at +. These test vectors were created specifically to exercise +all aspects of the decoder and therefore the audio quality of the decoded output is +significantly lower than what Opus can achieve in normal operation. + + + +The SHA1 hash of the files in the test vector package are + + + +
+ +
+ +
+ +To use the internal framing described in , the decoder + must know the total length of the Opus packet, in bytes. +This section describes a simple variation of that framing which can be used + when the total length of the packet is not known. +Nothing in the encoding of the packet itself allows a decoder to distinguish + between the regular, undelimited framing and the self-delimiting framing + described in this appendix. +Which one is used and where must be established by context at the transport + layer. +It is RECOMMENDED that a transport layer choose exactly one framing scheme, + rather than allowing an encoder to signal which one it wants to use. + + + +For example, although a regular Opus stream does not support more than two + channels, a multi-channel Opus stream may be formed from several one- and + two-channel streams. +To pack an Opus packet from each of these streams together in a single packet + at the transport layer, one could use the self-delimiting framing for all but + the last stream, and then the regular, undelimited framing for the last one. +Reverting to the undelimited framing for the last stream saves overhead + (because the total size of the transport-layer packet will still be known), + and ensures that a "multi-channel" stream which only has a single Opus stream + uses the same framing as a regular Opus stream does. +This avoids the need for signaling to distinguish these two cases. + + + +The self-delimiting framing is identical to the regular, undelimited framing + from , except that each Opus packet contains one extra + length field, encoded using the same one- or two-byte scheme from + . +This extra length immediately precedes the compressed data of the first Opus + frame in the packet, and is interpreted in the various modes as follows: + + +Code 0 packets: It is the length of the single Opus frame (see + ). + + +Code 1 packets: It is the length used for both of the Opus frames (see + ). + + +Code 2 packets: It is the length of the second Opus frame (see + ). + +CBR Code 3 packets: It is the length used for all of the Opus frames (see + ). + +VBR Code 3 packets: It is the length of the last Opus frame (see + ). + + + + +
+ +
+ +
+ +
+ +
+ +
+ +
+ +
+ +
+ +
+ +
+ + + + diff --git a/lib/opus/doc/draft-ietf-payload-rtp-opus.xml b/lib/opus/doc/draft-ietf-payload-rtp-opus.xml new file mode 100644 index 00000000..c4eb2107 --- /dev/null +++ b/lib/opus/doc/draft-ietf-payload-rtp-opus.xml @@ -0,0 +1,960 @@ + + + + + + + + + + + + + + + + + + + + + + ]> + + + + + + + + + + + + + + + + + + RTP Payload Format for the Opus Speech and Audio Codec + + + +
+ jspittka@gmail.com +
+ + + + vocTone +
+ + + + + + + + koenvos74@gmail.com +
+ + + + Mozilla +
+ + 331 E. Evelyn Avenue + Mountain View + CA + 94041 + USA + + jmvalin@jmvalin.ca +
+ + + + + + + This document defines the Real-time Transport Protocol (RTP) payload + format for packetization of Opus encoded + speech and audio data necessary to integrate the codec in the + most compatible way. It also provides an applicability statement + for the use of Opus over RTP. Further, it describes media type registrations + for the RTP payload format. + + + + + +
+ + Opus is a speech and audio codec developed within the + IETF Internet Wideband Audio Codec working group. The codec + has a very low algorithmic delay and it + is highly scalable in terms of audio bandwidth, bitrate, and + complexity. Further, it provides different modes to efficiently encode speech signals + as well as music signals, thus making it the codec of choice for + various applications using the Internet or similar networks. + + + This document defines the Real-time Transport Protocol (RTP) + payload format for packetization + of Opus encoded speech and audio data necessary to + integrate Opus in the + most compatible way. It also provides an applicability statement + for the use of Opus over RTP. + Further, it describes media type registrations for + the RTP payload format. + +
+ +
+ The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", + "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this + document are to be interpreted as described in . + + + The range of audio frequecies being coded + Constant bitrate + Central Processing Unit + Discontinuous transmission + Forward error correction + Internet Protocol + Speech or audio samples (per channel) + Session Description Protocol + Variable bitrate + + + + Throughout this document, we refer to the following definitions: + + + Abbreviation + Name + Audio Bandwidth (Hz) + Sampling Rate (Hz) + NB + Narrowband + 0 - 4000 + 8000 + + MB + Mediumband + 0 - 6000 + 12000 + + WB + Wideband + 0 - 8000 + 16000 + + SWB + Super-wideband + 0 - 12000 + 24000 + + FB + Fullband + 0 - 20000 + 48000 + + + Audio bandwidth naming + + +
+ +
+ + Opus encodes speech + signals as well as general audio signals. Two different modes can be + chosen, a voice mode or an audio mode, to allow the most efficient coding + depending on the type of the input signal, the sampling frequency of the + input signal, and the intended application. + + + + The voice mode allows efficient encoding of voice signals at lower bit + rates while the audio mode is optimized for general audio signals at medium and + higher bitrates. + + + + Opus is highly scalable in terms of audio + bandwidth, bitrate, and complexity. Further, Opus allows + transmitting stereo signals with in-band signaling in the bit-stream. + + +
+ + Opus supports bitrates from 6 kb/s to 510 kb/s. + The bitrate can be changed dynamically within that range. + All + other parameters being + equal, higher bitrates result in higher audio quality. + +
+ + For a frame size of + 20 ms, these + are the bitrate "sweet spots" for Opus in various configurations: + + + 8-12 kb/s for NB speech, + 16-20 kb/s for WB speech, + 28-40 kb/s for FB speech, + 48-64 kb/s for FB mono music, and + 64-128 kb/s for FB stereo music. + + +
+
+ + For the same average bitrate, variable bitrate (VBR) can achieve higher audio quality + than constant bitrate (CBR). For the majority of voice transmission applications, VBR + is the best choice. One reason for choosing CBR is the potential + information leak that might occur when encrypting the + compressed stream. See for guidelines on when VBR is + appropriate for encrypted audio communications. In the case where an existing + VBR stream needs to be converted to CBR for security reasons, then the Opus padding + mechanism described in is the RECOMMENDED way to achieve padding + because the RTP padding bit is unencrypted. + + + The bitrate can be adjusted at any point in time. To avoid congestion, + the average bitrate SHOULD NOT exceed the available + network bandwidth. If no target bitrate is specified, the bitrates specified in + are RECOMMENDED. + + +
+ +
+ + + Opus can, as described in , + be operated with a variable bitrate. In that case, the encoder will + automatically reduce the bitrate for certain input signals, like periods + of silence. When using continuous transmission, it will reduce the + bitrate when the characteristics of the input signal permit, but + will never interrupt the transmission to the receiver. Therefore, the + received signal will maintain the same high level of audio quality over the + full duration of a transmission while minimizing the average bit + rate over time. + + + + In cases where the bitrate of Opus needs to be reduced even + further or in cases where only constant bitrate is available, + the Opus encoder can use discontinuous + transmission (DTX), where parts of the encoded signal that + correspond to periods of silence in the input speech or audio signal + are not transmitted to the receiver. A receiver can distinguish + between DTX and packet loss by looking for gaps in the sequence + number, as described by Section 4.1 + of . + + + + On the receiving side, the non-transmitted parts will be handled by a + frame loss concealment unit in the Opus decoder which generates a + comfort noise signal to replace the non transmitted parts of the + speech or audio signal. Use of Comfort + Noise (CN) with Opus is discouraged. + The transmitter MUST drop whole frames only, + based on the size of the last transmitted frame, + to ensure successive RTP timestamps differ by a multiple of 120 and + to allow the receiver to use whole frames for concealment. + + + + DTX can be used with both variable and constant bitrate. + It will have a slightly lower speech or audio + quality than continuous transmission. Therefore, using continuous + transmission is RECOMMENDED unless constraints on available network bandwidth + are severe. + + +
+ +
+ +
+ + + Complexity of the encoder can be scaled to optimize for CPU resources in real-time, mostly as + a trade-off between audio quality and bitrate. Also, different modes of Opus have different complexity. + + +
+ +
+ + + The voice mode of Opus allows for embedding "in-band" forward error correction (FEC) + data into the Opus bit stream. This FEC scheme adds + redundant information about the previous packet (N-1) to the current + output packet N. For + each frame, the encoder decides whether to use FEC based on (1) an + externally-provided estimate of the channel's packet loss rate; (2) an + externally-provided estimate of the channel's capacity; (3) the + sensitivity of the audio or speech signal to packet loss; (4) whether + the receiving decoder has indicated it can take advantage of "in-band" + FEC information. The decision to send "in-band" FEC information is + entirely controlled by the encoder and therefore no special precautions + for the payload have to be taken. + + + + On the receiving side, the decoder can take advantage of this + additional information when it loses a packet and the next packet + is available. In order to use the FEC data, the jitter buffer needs + to provide access to payloads with the FEC data. + Instead of performing loss concealment for a missing packet, the + receiver can then configure its decoder to decode the FEC data from the next packet. + + + + Any compliant Opus decoder is capable of ignoring + FEC information when it is not needed, so encoding with FEC cannot cause + interoperability problems. + However, if FEC cannot be used on the receiving side, then FEC + SHOULD NOT be used, as it leads to an inefficient usage of network + resources. Decoder support for FEC SHOULD be indicated at the time a + session is set up. + + +
+ +
+ + + Opus allows for transmission of stereo audio signals. This operation + is signaled in-band in the Opus bit-stream and no special arrangement + is needed in the payload format. An + Opus decoder is capable of handling a stereo encoding, but an + application might only be capable of consuming a single audio + channel. + + + If a decoder cannot take advantage of the benefits of a stereo signal + this SHOULD be indicated at the time a session is set up. In that case + the sending side SHOULD NOT send stereo signals as it leads to an + inefficient usage of network resources. + + +
+ +
+ +
+ The payload format for Opus consists of the RTP header and Opus payload + data. +
+ The format of the RTP header is specified in . + The use of the fields of the RTP header by the Opus payload format is + consistent with that specification. + + The payload length of Opus is an integer number of octets and + therefore no padding is necessary. The payload MAY be padded by an + integer number of octets according to , + although the Opus internal padding is preferred. + + The timestamp, sequence number, and marker bit (M) of the RTP header + are used in accordance with Section 4.1 + of . + + The RTP payload type for Opus is to be assigned dynamically. + + The receiving side MUST be prepared to receive duplicate RTP + packets. The receiver MUST provide at most one of those payloads to the + Opus decoder for decoding, and MUST discard the others. + + Opus supports 5 different audio bandwidths, which can be adjusted during + a stream. + The RTP timestamp is incremented with a 48000 Hz clock rate + for all modes of Opus and all sampling rates. + The unit + for the timestamp is samples per single (mono) channel. The RTP timestamp corresponds to the + sample time of the first encoded sample in the encoded frame. + For data encoded with sampling rates other than 48000 Hz, + the sampling rate has to be adjusted to 48000 Hz. + +
+ +
+ + The Opus encoder can output encoded frames representing 2.5, 5, 10, 20, + 40, or 60 ms of speech or audio data. Further, an arbitrary number of frames can be + combined into a packet, up to a maximum packet duration representing + 120 ms of speech or audio data. The grouping of one or more Opus + frames into a single Opus packet is defined in Section 3 of + . An RTP payload MUST contain exactly one + Opus packet as defined by that document. + + + shows the structure combined with the RTP header. + +
+ + + +
+ + + shows supported frame sizes in + milliseconds of encoded speech or audio data for the speech and audio modes + (Mode) and sampling rates (fs) of Opus and shows how the timestamp is + incremented for packetization (ts incr). If the Opus encoder + outputs multiple encoded frames into a single packet, the timestamp + increment is the sum of the increments for the individual frames. + + + + Mode + fs + 2.5 + 5 + 10 + 20 + 40 + 60 + ts incr + all + 120 + 240 + 480 + 960 + 1920 + 2880 + voice + NB/MB/WB/SWB/FB + x + x + o + o + o + o + audio + NB/WB/SWB/FB + o + o + o + o + x + x + + +
+ +
+ +
+ + The target bitrate of Opus can be adjusted at any point in time, thus + allowing efficient congestion control. Furthermore, the amount + of encoded speech or audio data encoded in a + single packet can be used for congestion control, since the transmission + rate is inversely proportional to the packet duration. A lower packet + transmission rate reduces the amount of header overhead, but at the same + time increases latency and loss sensitivity, so it ought to be used with + care. + + Since UDP does not provide congestion control, applications that use + RTP over UDP SHOULD implement their own congestion control above the + UDP layer . Work in the rmcat working group + describes the + interactions and conceptual interfaces necessary between the application + components that relate to congestion control, including the RTP layer, + the higher-level media codec control layer, and the lower-level + transport interface, as well as components dedicated to congestion + control functions. +
+ +
+ One media subtype (audio/opus) has been defined and registered as + described in the following section. + +
+ Media type registration is done according to and . + + Type name: audio + Subtype name: opus + + Required parameters: + + the RTP timestamp is incremented with a + 48000 Hz clock rate for all modes of Opus and all sampling + rates. For data encoded with sampling rates other than 48000 Hz, + the sampling rate has to be adjusted to 48000 Hz. + + + + Optional parameters: + + + + a hint about the maximum output sampling rate that the receiver is + capable of rendering in Hz. + The decoder MUST be capable of decoding + any audio bandwidth but due to hardware limitations only signals + up to the specified sampling rate can be played back. Sending signals + with higher audio bandwidth results in higher than necessary network + usage and encoding complexity, so an encoder SHOULD NOT encode + frequencies above the audio bandwidth specified by maxplaybackrate. + This parameter can take any value between 8000 and 48000, although + commonly the value will match one of the Opus bandwidths + (). + By default, the receiver is assumed to have no limitations, i.e. 48000. + + + + + a hint about the maximum input sampling rate that the sender is likely to produce. + This is not a guarantee that the sender will never send any higher bandwidth + (e.g. it could send a pre-recorded prompt that uses a higher bandwidth), but it + indicates to the receiver that frequencies above this maximum can safely be discarded. + This parameter is useful to avoid wasting receiver resources by operating the audio + processing pipeline (e.g. echo cancellation) at a higher rate than necessary. + This parameter can take any value between 8000 and 48000, although + commonly the value will match one of the Opus bandwidths + (). + By default, the sender is assumed to have no limitations, i.e. 48000. + + + + the maximum duration of media represented + by a packet (according to Section 6 of + ) that a decoder wants to receive, in + milliseconds rounded up to the next full integer value. + Possible values are 3, 5, 10, 20, 40, 60, or an arbitrary + multiple of an Opus frame size rounded up to the next full integer + value, up to a maximum value of 120, as + defined in . If no value is + specified, the default is 120. + + + the preferred duration of media represented + by a packet (according to Section 6 of + ) that a decoder wants to receive, in + milliseconds rounded up to the next full integer value. + Possible values are 3, 5, 10, 20, 40, 60, or an arbitrary + multiple of an Opus frame size rounded up to the next full integer + value, up to a maximum value of 120, as defined in . If no value is + specified, the default is 20. + + + specifies the maximum average + receive bitrate of a session in bits per second (b/s). The actual + value of the bitrate can vary, as it is dependent on the + characteristics of the media in a packet. Note that the maximum + average bitrate MAY be modified dynamically during a session. Any + positive integer is allowed, but values outside the range + 6000 to 510000 SHOULD be ignored. If no value is specified, the + maximum value specified in + for the corresponding mode of Opus and corresponding maxplaybackrate + is the default. + + + specifies whether the decoder prefers receiving stereo or mono signals. + Possible values are 1 and 0 where 1 specifies that stereo signals are preferred, + and 0 specifies that only mono signals are preferred. + Independent of the stereo parameter every receiver MUST be able to receive and + decode stereo signals but sending stereo signals to a receiver that signaled a + preference for mono signals may result in higher than necessary network + utilization and encoding complexity. If no value is specified, + the default is 0 (mono). + + + + specifies whether the sender is likely to produce stereo audio. + Possible values are 1 and 0, where 1 specifies that stereo signals are likely to + be sent, and 0 specifies that the sender will likely only send mono. + This is not a guarantee that the sender will never send stereo audio + (e.g. it could send a pre-recorded prompt that uses stereo), but it + indicates to the receiver that the received signal can be safely downmixed to mono. + This parameter is useful to avoid wasting receiver resources by operating the audio + processing pipeline (e.g. echo cancellation) in stereo when not necessary. + If no value is specified, the default is 0 + (mono). + + + + specifies if the decoder prefers the use of a constant bitrate versus + variable bitrate. Possible values are 1 and 0, where 1 specifies constant + bitrate and 0 specifies variable bitrate. If no value is specified, + the default is 0 (vbr). When cbr is 1, the maximum average bitrate can still + change, e.g. to adapt to changing network conditions. + + + specifies that the decoder has the capability to + take advantage of the Opus in-band FEC. Possible values are 1 and 0. + Providing 0 when FEC cannot be used on the receiving side is + RECOMMENDED. If no + value is specified, useinbandfec is assumed to be 0. + This parameter is only a preference and the receiver MUST be able to process + packets that include FEC information, even if it means the FEC part is discarded. + + + specifies if the decoder prefers the use of + DTX. Possible values are 1 and 0. If no value is specified, the + default is 0. + + + Encoding considerations: + + The Opus media type is framed and consists of binary data according + to Section 4.8 in . + + + Security considerations: + + See of this document. + + + Interoperability considerations: none + Published specification: RFC [XXXX] + Note to the RFC Editor: Replace [XXXX] with the number of the published + RFC. + + Applications that use this media type: + + Any application that requires the transport of + speech or audio data can use this media type. Some examples are, + but not limited to, audio and video conferencing, Voice over IP, + media streaming. + + + Fragment identifier considerations: N/A + + Person & email address to contact for further information: + + SILK Support silksupport@skype.net + Jean-Marc Valin jmvalin@jmvalin.ca + + + Intended usage: COMMON + + Restrictions on usage: + + + For transfer over RTP, the RTP payload format ( of this document) SHALL be + used. + + + Author: + + Julian Spittka jspittka@gmail.com + Koen Vos koenvos74@gmail.com + Jean-Marc Valin jmvalin@jmvalin.ca + + + Change controller: IETF Payload Working Group delegated from the IESG +
+
+ +
+ The information described in the media type specification has a + specific mapping to fields in the Session Description Protocol (SDP) + , which is commonly used to describe RTP + sessions. When SDP is used to specify sessions employing Opus, + the mapping is as follows: + + + + The media type ("audio") goes in SDP "m=" as the media name. + + The media subtype ("opus") goes in SDP "a=rtpmap" as the encoding + name. The RTP clock rate in "a=rtpmap" MUST be 48000 and the number of + channels MUST be 2. + + The OPTIONAL media type parameters "ptime" and "maxptime" are + mapped to "a=ptime" and "a=maxptime" attributes, respectively, in the + SDP. + + The OPTIONAL media type parameters "maxaveragebitrate", + "maxplaybackrate", "stereo", "cbr", "useinbandfec", and + "usedtx", when present, MUST be included in the "a=fmtp" attribute + in the SDP, expressed as a media type string in the form of a + semicolon-separated list of parameter=value pairs (e.g., + maxplaybackrate=48000). They MUST NOT be specified in an + SSRC-specific "fmtp" source-level attribute (as defined in + Section 6.3 of ). + + The OPTIONAL media type parameters "sprop-maxcapturerate", + and "sprop-stereo" MAY be mapped to the "a=fmtp" SDP attribute by + copying them directly from the media type parameter string as part + of the semicolon-separated list of parameter=value pairs (e.g., + sprop-stereo=1). These same OPTIONAL media type parameters MAY also + be specified using an SSRC-specific "fmtp" source-level attribute + as described in Section 6.3 of . + They MAY be specified in both places, in which case the parameter + in the source-level attribute overrides the one found on the + "a=fmtp" line. The value of any parameter which is not specified in + a source-level source attribute MUST be taken from the "a=fmtp" + line, if it is present there. + + + + + Below are some examples of SDP session descriptions for Opus: + + Example 1: Standard mono session with 48000 Hz clock rate +
+ + + +
+ + + Example 2: 16000 Hz clock rate, maximum packet size of 40 ms, + recommended packet size of 40 ms, maximum average bitrate of 20000 bps, + prefers to receive stereo but only plans to send mono, FEC is desired, + DTX is not desired + +
+ + + +
+ + Example 3: Two-way full-band stereo preferred + +
+ + + +
+ + +
+ + When using the offer-answer procedure described in to negotiate the use of Opus, the following + considerations apply: + + + + Opus supports several clock rates. For signaling purposes only + the highest, i.e. 48000, is used. The actual clock rate of the + corresponding media is signaled inside the payload and is not + restricted by this payload format description. The decoder MUST be + capable of decoding every received clock rate. An example + is shown below: + +
+ + + +
+ + + The "ptime" and "maxptime" parameters are unidirectional + receive-only parameters and typically will not compromise + interoperability; however, some values might cause application + performance to suffer. defines the SDP offer-answer handling of the + "ptime" parameter. The "maxptime" parameter MUST be handled in the + same way. + + + The "maxplaybackrate" parameter is a unidirectional receive-only + parameter that reflects limitations of the local receiver. When + sending to a single destination, a sender MUST NOT use an audio + bandwidth higher than necessary to make full use of audio sampled at + a sampling rate of "maxplaybackrate". Gateways or senders that + are sending the same encoded audio to multiple destinations + SHOULD NOT use an audio bandwidth higher than necessary to + represent audio sampled at "maxplaybackrate", as this would lead + to inefficient use of network resources. + The "maxplaybackrate" parameter does not + affect interoperability. Also, this parameter SHOULD NOT be used + to adjust the audio bandwidth as a function of the bitrate, as this + is the responsibility of the Opus encoder implementation. + + + The "maxaveragebitrate" parameter is a unidirectional receive-only + parameter that reflects limitations of the local receiver. The sender + of the other side MUST NOT send with an average bitrate higher than + "maxaveragebitrate" as it might overload the network and/or + receiver. The "maxaveragebitrate" parameter typically will not + compromise interoperability; however, some values might cause + application performance to suffer, and ought to be set with + care. + + The "sprop-maxcapturerate" and "sprop-stereo" parameters are + unidirectional sender-only parameters that reflect limitations of + the sender side. + They allow the receiver to set up a reduced-complexity audio + processing pipeline if the sender is not planning to use the full + range of Opus's capabilities. + Neither "sprop-maxcapturerate" nor "sprop-stereo" affect + interoperability and the receiver MUST be capable of receiving any signal. + + + + The "stereo" parameter is a unidirectional receive-only + parameter. When sending to a single destination, a sender MUST + NOT use stereo when "stereo" is 0. Gateways or senders that are + sending the same encoded audio to multiple destinations SHOULD + NOT use stereo when "stereo" is 0, as this would lead to + inefficient use of network resources. The "stereo" parameter does + not affect interoperability. + + + + The "cbr" parameter is a unidirectional receive-only + parameter. + + + The "useinbandfec" parameter is a unidirectional receive-only + parameter. + + The "usedtx" parameter is a unidirectional receive-only + parameter. + + Any unknown parameter in an offer MUST be ignored by the receiver + and MUST be removed from the answer. + + + + + The Opus parameters in an SDP Offer/Answer exchange are completely + orthogonal, and there is no relationship between the SDP Offer and + the Answer. + +
+ +
+ + For declarative use of SDP such as in Session Announcement Protocol + (SAP), , and RTSP, , for + Opus, the following needs to be considered: + + + + The values for "maxptime", "ptime", "maxplaybackrate", and + "maxaveragebitrate" ought to be selected carefully to ensure that a + reasonable performance can be achieved for the participants of a session. + + + The values for "maxptime", "ptime", and of the payload + format configuration are recommendations by the decoding side to ensure + the best performance for the decoder. + + + All other parameters of the payload format configuration are declarative + and a participant MUST use the configurations that are provided for + the session. More than one configuration can be provided if necessary + by declaring multiple RTP payload types; however, the number of types + ought to be kept small. + +
+
+ +
+ + Use of variable bitrate (VBR) is subject to the security considerations in + . + + RTP packets using the payload format defined in this specification + are subject to the security considerations discussed in the RTP + specification , and in any applicable RTP profile such as + RTP/AVP , RTP/AVPF , + RTP/SAVP or RTP/SAVPF . + However, as "Securing the RTP Protocol Framework: + Why RTP Does Not Mandate a Single Media Security Solution" + discusses, it is not an RTP payload + format's responsibility to discuss or mandate what solutions are used + to meet the basic security goals like confidentiality, integrity and + source authenticity for RTP in general. This responsibility lays on + anyone using RTP in an application. They can find guidance on + available security mechanisms and important considerations in Options + for Securing RTP Sessions [I-D.ietf-avtcore-rtp-security-options]. + Applications SHOULD use one or more appropriate strong security + mechanisms. + + This payload format and the Opus encoding do not exhibit any + significant non-uniformity in the receiver-end computational load and thus + are unlikely to pose a denial-of-service threat due to the receipt of + pathological datagrams. +
+ +
+ Many people have made useful comments and suggestions contributing to this document. + In particular, we would like to thank + Tina le Grand, Cullen Jennings, Jonathan Lennox, Gregory Maxwell, Colin Perkins, Jan Skoglund, + Timothy B. Terriberry, Martin Thompson, Justin Uberti, Magnus Westerlund, and Mo Zanaty. +
+ + + + + &rfc2119; + &rfc3389; + &rfc3550; + &rfc3711; + &rfc3551; + &rfc6838; + &rfc4855; + &rfc4566; + &rfc3264; + &rfc2326; + &rfc5576; + &rfc6562; + &rfc6716; + + + + &rfc2974; + &rfc4585; + &rfc5124; + &rfc5405; + &rfc7202; + + + + rmcat documents + + + + + + + + + + + + diff --git a/lib/opus/doc/footer.html b/lib/opus/doc/footer.html new file mode 100644 index 00000000..346c40ab --- /dev/null +++ b/lib/opus/doc/footer.html @@ -0,0 +1,32 @@ + + + + + + +
+ + + + + + + +
+For more information visit the Opus Website. + +
+ Generated by + doxygen + $doxygenversion +
+
+ + + diff --git a/lib/opus/doc/header.html b/lib/opus/doc/header.html new file mode 100644 index 00000000..babdcf6d --- /dev/null +++ b/lib/opus/doc/header.html @@ -0,0 +1,61 @@ + + + + + + + + +$projectname: $title +$title + + + +$treeview +$search +$mathjax + +$extrastylesheet + + +
+ + +
+ + + + + + + + + + + + + + + + + + + + + + +
+
Opus
+ 		+ + +
+
$projectbrief
+
$projectnumber +
+ 		+
$projectbrief
+ 		$searchbox
+
+ + diff --git a/lib/opus/doc/meson.build b/lib/opus/doc/meson.build new file mode 100644 index 00000000..ee42a2b0 --- /dev/null +++ b/lib/opus/doc/meson.build @@ -0,0 +1,33 @@ +have_dot = find_program('dot', required: false).found() + +doxyfile_conf = configuration_data() +doxyfile_conf.set('VERSION', opus_version) +doxyfile_conf.set('HAVE_DOT', have_dot) +doxyfile_conf.set('top_srcdir', top_srcdir) +doxyfile_conf.set('top_builddir', top_builddir) + +doxyfile = configure_file(input: 'Doxyfile.in', + output: 'Doxyfile', + configuration: doxyfile_conf, + install: false) + +docdir = join_paths(get_option('datadir'), get_option('docdir')) + +doc_inputs = [ + 'customdoxygen.css', + 'footer.html', + 'header.html', + 'opus_logo.svg', + top_srcdir + '/include/opus.h', + top_srcdir + '/include/opus_multistream.h', + top_srcdir + '/include/opus_defines.h', + top_srcdir + '/include/opus_types.h', + top_srcdir + '/include/opus_custom.h', +] + +custom_target('doc', + input: [ doxyfile ] + doc_inputs, + output: [ 'html' ], + command: [ doxygen, doxyfile ], + install_dir: docdir, + install: true) diff --git a/lib/opus/doc/opus_in_isobmff.css b/lib/opus/doc/opus_in_isobmff.css new file mode 100644 index 00000000..bffe8f45 --- /dev/null +++ b/lib/opus/doc/opus_in_isobmff.css @@ -0,0 +1,60 @@ +/* Normal links */ +.normal_link a:link +{ + color : yellow; +} +.normal_link a:visited +{ + color : green; +} + +/* Boxes */ +.pre +{ + white-space: pre; /* CSS 2.0 */ + white-space: pre-wrap; /* CSS 2.1 */ + white-space: -pre-wrap; /* Opera 4-6 */ + white-space: -o-pre-wrap; /* Opera 7 */ + white-space: -moz-pre-wrap; /* Mozilla */ + white-space: -hp-pre-wrap; /* HP Printers */ + word-wrap : break-word; /* IE 5+ */ +} + +.title_box +{ + width : 470px; + height : 70px; + margin : 2px 50px 2px 2px; + padding : 10px; + border : 1px solid black; + background-color : #666666; + white-space : pre; + float : left; + text-align : center; + color : #C0C0C0; + font-size : 50pt; + font-style : italic; +} + +.subindex_box +{ + margin : 5px; + padding : 14px 22px; + border : 1px solid black; + background-color : #778877; + float : left; + text-align : center; + color : #115555; + font-size : 32pt; +} + +.frame_box +{ + margin : 10px; + padding : 10px; + border : 0px; + background-color : #084040; + text-align : left; + color : #C0C0C0; + font-family : monospace; +} diff --git a/lib/opus/doc/opus_in_isobmff.html b/lib/opus/doc/opus_in_isobmff.html new file mode 100644 index 00000000..38aefbf2 --- /dev/null +++ b/lib/opus/doc/opus_in_isobmff.html @@ -0,0 +1,692 @@ + + + + + + Encapsulation of Opus in ISO Base Media File Format + + + Encapsulation of Opus in ISO Base Media File Format
+ last updated: August 28, 2018
+
+
+ + Encapsulation of Opus in ISO Base Media File Format + Version 0.8.1 (incomplete) + + +Table of Contents +1 Scope +2 Normative References +3 Terms and Definitions +4 Design Rules of Encapsulation + 4.1 File Type Identification + 4.2 Overview of Track Structure + 4.3 Definitions of Opus sample + 4.3.1 Sample entry format + 4.3.2 Opus Specific Box + 4.3.3 Sample format + 4.3.4 Duration of Opus sample + 4.3.5 Sub-sample + 4.3.6 Random Access + 4.3.6.1 Random Access Point + 4.3.6.2 Pre-roll + 4.4 Trimming of Actual Duration + 4.5 Channel Mapping + 4.5.1 ISO Base Media native Channel Mapping + 4.5.2 Composition on all active tracks (informative) + 4.6 Basic Structure (informative) + 4.6.2 Initial Movie + 4.6.3 Movie Fragments + 4.7 Example of Encapsulation (informative) +5 Author's Address + + +1 Scope + This specification specifies the fundamental way of the encapsulation of Opus coded bitstreams in ISO Base Media + file format and its derivatives. The encapsulation of Opus coded bitstreams in QuickTime file format is outside + the scope of this specification. + + +2 Normative References + [1] ISO/IEC 14496-12:2015 Corrected version + Information technology — Coding of audio-visual objects — Part 12: ISO base media file format + + [2] RFC 6716 + Definition of the Opus Audio Codec + + [3] RFC 7845 + Ogg Encapsulation for the Opus Audio Codec + + +3 Terms and Definitions + 3.1 active track + enabled track from the non-alternate group or selected track from alternate group + + 3.2 actual duration + duration constructed from valid samples + + 3.3 edit + entry in the Edit List Box + + 3.4 padded samples + PCM samples after decoding Opus sample(s) which are not valid samples + An Opus bitstream always contains them partially at the beginning and may contain them in part at the end, as + long as not physically removed yet at the beginning and/or the end. + + 3.5 priming samples + padded samples at the beginning of the Opus bitstream + + 3.6 sample-accurate + for any PCM sample, a timestamp exactly matching its sampling timestamp is present in the media timeline. + + 3.7 valid samples + PCM samples after decoding Opus sample(s) corresponding to input PCM samples + + +4 Design Rules of Encapsulation + 4.1 File Type Identification + This specification defines the brand 'Opus' to declare files are conformant to this specification. Additionally, + files conformant to this specification shall contain at least one brand, which supports the requirements and the + requirements described in this clause without contradiction, in the compatible brands list of the File Type Box. + As an example, the minimal support of the encapsulation of Opus bitstreams in ISO Base Media file format requires + the 'iso2' brand in the compatible brands list since support of roll groups is required. + + 4.2 Overview of Track Structure + This clause summarizes requirements of the encapsulation of Opus coded bitstream as media data in audio tracks + in file formats compliant with the ISO Base Media File Format. The details are described in clauses after this + clause. + + The handler_type field in the Handler Reference Box shall be set to 'soun'. + + The Media Information Box shall contain the Sound Media Header Box. + + The codingname of the sample entry is 'Opus'. + This specification does not define any encapsulation using MP4AudioSampleEntry with objectTypeIndication + specified by the MPEG-4 Registration Authority (http://www.mp4ra.org/). + See 4.3.1 Sample entry format to get the details about the sample entry. + + The 'dOps' box is added to the sample entry to convey initializing information for the decoder. + See 4.3.2 Opus Specific Box to get the details. + + An Opus sample is exactly one Opus packet for each of different Opus bitstreams. + See 4.3.3 Sample format to get the details. + + Every Opus sample is a sync sample but requires pre-roll for every random access to get correct output. + See 4.3.6 Random Access to get the details. + + 4.3 Definitions of Opus sample + 4.3.1 Sample entry format + For any track containing Opus bitstreams, at least one sample entry describing corresponding Opus bitstream + shall be present inside the Sample Table Box. This version of the specification defines only one sample + entry format named OpusSampleEntry whose codingname is 'Opus'. This sample entry includes exactly one Opus + Specific Box defined in 4.3.2 as a mandatory box and indicates that Opus samples described by this sample + entry are stored by the sample format described in 4.3.3. + + The syntax and semantics of the OpusSampleEntry is shown as follows. + + class OpusSampleEntry() extends AudioSampleEntry ('Opus') { + OpusSpecificBox(); + } + + + channelcount: + The channelcount field indicates the number of output channels and shall be set to the same value of + the OutputChannelCount in the OpusDecoderConfigurationRecord. The value of this field may be used in + the ChannelLayout if any as described in 4.5.1. + + samplesize: + The samplesize field shall be set to 16. + + samplerate: + The samplerate field shall be set to 48000<<16. + + OpusSpecificBox + This box contains initializing information for the decoder as defined in 4.3.2. + + 4.3.2 Opus Specific Box + Exactly one Opus Specific Box shall be present in each OpusSampleEntry. + The Opus Specific Box contains an OpusDecoderConfigurationRecord which contains the Version field and + this specification defines version 0 of this record. If incompatible changes occured in the fields after + the Version field within the OpusDecoderConfigurationRecord in the future versions of this specification, + another version will be defined. + This box refers to Ogg Opus [3] at many parts but all the data are stored as big-endian format. + + The syntax and semantics of the Opus Specific Box is shown as follows. + + class ChannelMappingTable (unsigned int(8) OutputChannelCount) { + unsigned int(8) StreamCount; + unsigned int(8) CoupledCount; + unsigned int(8 * OutputChannelCount) ChannelMapping; + } + + aligned(8) class OpusDecoderConfigurationRecord { + unsigned int(8) Version; + unsigned int(8) OutputChannelCount; + unsigned int(16) PreSkip; + unsigned int(32) InputSampleRate; + signed int(16) OutputGain; + unsigned int(8) ChannelMappingFamily; + if (ChannelMappingFamily != 0) { + ChannelMappingTable(OutputChannelCount); + } + } + + class OpusSpecificBox extends Box('dOps') { + OpusDecoderConfigurationRecord() OpusConfig; + } + + + Version: + The Version field shall be set to 0. + In the future versions of this specification, this field may be set to other values. And without support + of those values, the reader shall not read the fields after this within the OpusSpecificBox. + + OutputChannelCount: + The OutputChannelCount field shall be set to the same value as the *Output Channel Count* field in the + identification header defined in Ogg Opus [3]. + + PreSkip: + The PreSkip field indicates the number of the priming samples, that is, the number of samples at 48000 Hz + to discard from the decoder output when starting playback. The value of the PreSkip field shall be at least + 80 milliseconds' worth of PCM samples even when removing any number of Opus samples which may or may not + contain the priming samples. The PreSkip field is not used for discarding the priming samples at the whole + playback at all since it is informative only, and that task falls on the Edit List Box. + + InputSampleRate: + The InputSampleRate field shall be set to the same value as the *Input Sample Rate* field in the + identification header defined in Ogg Opus [3]. + + OutputGain: + The OutputGain field shall be set to the same value as the *Output Gain* field in the identification + header define in Ogg Opus [3]. Note that the value is stored as 8.8 fixed-point. + + ChannelMappingFamily: + The ChannelMappingFamily field shall be set to the same value as the *Channel Mapping Family* field in + the identification header defined in Ogg Opus [3]. Note that the value 255 may be used for an alternative + to map channels by ISO Base Media native mapping. The details are described in 4.5.1. + + StreamCount: + The StreamCount field shall be set to the same value as the *Stream Count* field in the identification + header defined in Ogg Opus [3]. + + CoupledCount: + The CoupledCount field shall be set to the same value as the *Coupled Count* field in the identification + header defined in Ogg Opus [3]. + + ChannelMapping: + The ChannelMapping field shall be set to the same octet string as *Channel Mapping* field in the identi- + fication header defined in Ogg Opus [3]. + + 4.3.3 Sample format + An Opus sample is exactly one Opus packet for each of different Opus bitstreams. Due to support more than + two channels, an Opus sample can contain frames from multiple Opus bitstreams but all Opus packets shall + share with the total of frame sizes in a single Opus sample. The way of how to pack an Opus packet from + each of Opus bitstreams into a single Opus sample follows Appendix B. in RFC 6716 [2]. + The endianness has nothing to do with any Opus sample since every Opus packet is processed byte-by-byte. + In this specification, 'sample' means 'Opus sample' except for 'padded samples', 'priming samples', 'valid + sample' and 'sample-accurate', i.e. 'sample' is 'sample' in the term defined in ISO/IEC 14496-12 [1]. + + +-----------------------------------------+-------------------------------------+ + | Opus packet 0 (self-delimiting framing) | Opus packet 1 (undelimited framing) | + +-----------------------------------------+-------------------------------------+ + |<---------------------------- the size of Opus sample ------------------------>| + + Figure 1 - Example structure of an Opus sample containing two Opus bitstreams + + 4.3.4 Duration of Opus sample + The duration of Opus sample is given by multiplying the total of frame sizes for a single Opus bitstream + expressed in seconds by the value of the timescale field in the Media Header Box. + Let's say an Opus sample consists of two Opus bitstreams, where the frame size of one bitstream is 40 milli- + seconds and the frame size of another is 60 milliseconds, and the timescale field in the Media Header Box + is set to 48000, then the duration of that Opus sample shall be 120 milliseconds since three 40 millisecond + frame and two 60 millisecond frames shall be contained because of the maximum duration of Opus packet, 120 + milliseconds, and 5760 in the timescale indicated in the Media Header Box. + + To indicate the valid samples excluding the padded samples at the end of Opus bitstream, the duration of + the last Opus sample of an Opus bitstream is given by multiplying the number of the valid samples by the + value produced by dividing the value of the timescale field in the Media Header Box by 48000. + + 4.3.5 Sub-sample + The structure of the last Opus packet in an Opus sample is different from the others in the same Opus sample, + and the others are invalid Opus packets as an Opus sample because of self-delimiting framing. To avoid + complexities, sub-sample is not defined for Opus sample in this specification. + + 4.3.6 Random Access + This subclause describes the nature of the random access of Opus sample. + + 4.3.6.1 Random Access Point + All Opus samples can be independently decoded i.e. every Opus sample is a sync sample. Therefore, the + Sync Sample Box shall not be present as long as there are no samples other than Opus samples in the same + track. And the sample_is_non_sync_sample field for Opus samples shall be set to 0. + + 4.3.6.2 Pre-roll + Opus bitstream requires at least 80 millisecond pre-roll after each random access to get correct output. + Pre-roll is indicated by the roll_distance field in AudioRollRecoveryEntry. AudioPreRollEntry shall not + be used since every Opus sample is a sync sample in Opus bitstream. Note that roll_distance is expressed + in sample units in a term of ISO Base Media File Format, and always takes negative values. + + For any track containing Opus bitstreams, at least one Sample Group Description Box and at least one + Sample to Group Box within the Sample Table Box shall be present and these have the grouping_type field + set to 'roll'. If any Opus sample is contained in a track fragment, the Sample to Group Box with the + grouping_type field set to 'roll' shall be present for that track fragment. + + For the requirement of AudioRollRecoveryEntry, the compatible_brands field in the File Type Box shall + contain at least one brand which requires support for roll groups. + + 4.4 Trimming of Actual Duration + Due to the priming samples (or the padding at the beginning) derived from the pre-roll for the startup and the + padded samples at the end, we need trim from media to get the actual duration. An edit in the Edit List Box can + achieve this demand, and the Edit Box and the Edit List Box shall be present. + + For sample-accurate trimming, proper timescale should be set to the timescale field in the Movie Header Box + and the Media Header Box inside Track Box(es) for Opus bitstream. The timescale field in the Media Header Box is + typically set to 48000. It is recommended that the timescale field in the Movie Header Box be set to the same + value of the timescale field in the Media Header Box in order to avoid the rounding problem when specifying + duration of edit if the timescales in all of the Media Header Boxes are set to the same value. + + For example, to indicate the actual duration of an Opus bitstream in a track with the timescale fields of both + the Movie Header Box and the Media Header Box set to 48000, we would use the following edit: + segment_duration = the number of the valid samples + media_time = the number of the priming samples + media_rate = 1 << 16 + + The Edit List Box is applied to whole movie including all movie fragments. Therefore, it is impossible to tell + the actual duration in the case producing movie fragments on the fly such as live-streaming. In such cases, + the duration of the last Opus sample may be helpful by setting zero to the segment_duration field since the + value 0 represents implicit duration equal to the sum of the duration of all samples. + + 4.5 Channel Mapping + 4.5.1 ISO Base Media native Channel Mapping + ISO Base Media File Format, that is ISO/IEC 14496-12 [1], defines an extension ChannelLayout to the + AudioSampleEntry, which conveys information of mapping channels to loudspeaker positions. The ChannelLayout + enables to specify the channel layout more flexibly than the predefined layouts of the ChannelMappingFamily. + + To utilize the ChannelLayout for OpusSampleEntry, the ChannelMappingFamily field should be set to 255. + Even when the ChannelMappingFamily field is set to another value, the assignment of each output channel to + loudspeaker position specified by the ChannelMappingFamily would be changed as specified by the ChannelLayout. + The procedure of the assignment is the following. + + 1. Decoded channels are mapped to output channels according to the ChannelMappingTable. + 2. Output channels are mapped to loudspeaker positions according to the ChannelLayout. + + In this way, the parameters of the Opus Specific Box are processed before the ChannelLayout, and the + ChannelLayout shall follow the Opus Specific Box. + + 4.5.2 Composition on all active tracks (informative) + By the application of alternate_group in the Track Header Box, whole audio channels in all active tracks from + non-alternate group and/or different alternate group from each other are composited into the presentation. If + an Opus sample consists of multiple Opus bitstreams, it can be splitted into individual Opus bitstreams and + reconstructed into new Opus samples as long as every Opus bitstream has the same total duration in each Opus + sample. This nature can be utilized to encapsulate a single Opus bitstream in each track without breaking the + original channel layout. + + As an example, let's say there is a following track: + OutputChannelCount = 6; + StreamCount = 4; + CoupledCount = 2; + ChannelMapping = {0, 4, 1, 2, 3, 5}; // front left, front center, front right, + // rear left, rear right, LFE + Here, to couple front left to front right channels into the first stream, and couple rear left to rear right + channels into the second stream, reordering is needed since coupled streams must precede any non-coupled + stream. You extract the four Opus bitstreams from this track and you encapsulate two of the four into a track + and the others into another track. The former track is as follows. + OutputChannelCount = 6; + StreamCount = 2; + CoupledCount = 2; + ChannelMapping = {0, 255, 1, 2, 3, 255}; // front left, front center, front right, + // rear left, rear right, LFE + And the latter track is as follows. + OutputChannelCount = 6; + StreamCount = 2; + CoupledCount = 0; + ChannelMapping = {255, 0, 255, 255, 255, 1}; // front left, front center, front right, + // rear left, rear right, LFE + In addition, the value of the alternate_group field in the both tracks is set to 0. As the result, the player + may play as if channels with 255 are not present, and play the presentation constructed from the both tracks + in the same channel layout as the one of the original track. Keep in mind that the way of the composition, i.e. + the mixing for playback, is not defined here, and maybe different results could occur except for the channel + layout of the original, depending on an implementation or the definition of a derived file format. + + Note that some derived file formats may specify the restriction to ignore alternate grouping. In the context + of such file formats, this application is not available. This unavailability does not mean incompatibilities + among file formats unless the restriction to the value of the alternate_group field is specified and brings + about any conflict among their definitions. + + 4.6 Basic Structure (informative) + 4.6.1 Initial Movie + This subclause shows a basic structure of the Movie Box as follows: + + +----+----+----+----+----+----+----+----+------------------------------+ + |moov| | | | | | | | Movie Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | |mvhd| | | | | | | Movie Header Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | |trak| | | | | | | Track Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |tkhd| | | | | | Track Header Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |edts| | | | | | Edit Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | |elst| | | | | Edit List Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |mdia| | | | | | Media Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | |mdhd| | | | | Media Header Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | |hdlr| | | | | Handler Reference Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | |minf| | | | | Media Information Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | |smhd| | | | Sound Media Header Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | |dinf| | | | Data Information Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |dref| | | Data Reference Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | | |url | | DataEntryUrlBox | + +----+----+----+----+----+----+ or +----+------------------------------+ + | | | | | | |urn | | DataEntryUrnBox | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | |stbl| | | | Sample Table | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |stsd| | | Sample Description Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | | |Opus| | OpusSampleEntry | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | | | |dOps| Opus Specific Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |stts| | | Decoding Time to Sample Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |stsc| | | Sample To Chunk Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |stsz| | | Sample Size Box | + +----+----+----+----+----+ or +----+----+------------------------------+ + | | | | | |stz2| | | Compact Sample Size Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |stco| | | Chunk Offset Box | + +----+----+----+----+----+ or +----+----+------------------------------+ + | | | | | |co64| | | Chunk Large Offset Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |sgpd| | | Sample Group Description Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | | | | |sbgp| | | Sample to Group Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | |mvex|* | | | | | | Movie Extends Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |trex|* | | | | | Track Extends Box | + +----+----+----+----+----+----+----+----+------------------------------+ + + Figure 2 - Basic structure of Movie Box + + It is strongly recommended that the order of boxes should follow the above structure. + Boxes marked with an asterisk (*) may be present. + For most boxes listed above, the definition is as is defined in ISO/IEC 14496-12 [1]. The additional boxes + and the additional requirements, restrictions and recommendations to the other boxes are described in this + specification. + + 4.6.2 Movie Fragments + This subclause shows a basic structure of the Movie Fragment Box as follows: + + +----+----+----+----+----+----+----+----+------------------------------+ + |moof| | | | | | | | Movie Fragment Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | |mfhd| | | | | | | Movie Fragment Header Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | |traf| | | | | | | Track Fragment Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |tfhd| | | | | | Track Fragment Header Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |trun| | | | | | Track Fragment Run Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |sgpd|* | | | | | Sample Group Description Box | + +----+----+----+----+----+----+----+----+------------------------------+ + | | |sbgp| | | | | | Sample to Group Box | + +----+----+----+----+----+----+----+----+------------------------------+ + + Figure 3 - Basic structure of Movie Fragment Box + + It is strongly recommended that the Movie Fragment Header Box and the Track Fragment Header Box be + placed first in their container. + Boxes marked with an asterisk (*) may be present. + For the boxes listed above, the definition is as is defined in ISO/IEC 14496-12 [1]. + + 4.7 Example of Encapsulation (informative) + [File] + size = 17757 + [ftyp: File Type Box] + position = 0 + size = 24 + major_brand = Opus : Opus audio coding + minor_version = 0 + compatible_brands + brand[0] = Opus : Opus audio coding + brand[1] = iso2 : ISO Base Media file format version 2 + [moov: Movie Box] + position = 24 + size = 757 + [mvhd: Movie Header Box] + position = 32 + size = 108 + version = 0 + flags = 0x000000 + creation_time = UTC 2014/12/12, 18:41:19 + modification_time = UTC 2014/12/12, 18:41:19 + timescale = 48000 + duration = 33600 (00:00:00.700) + rate = 1.000000 + volume = 1.000000 + reserved = 0x0000 + reserved = 0x00000000 + reserved = 0x00000000 + transformation matrix + | a, b, u | | 1.000000, 0.000000, 0.000000 | + | c, d, v | = | 0.000000, 1.000000, 0.000000 | + | x, y, w | | 0.000000, 0.000000, 1.000000 | + pre_defined = 0x00000000 + pre_defined = 0x00000000 + pre_defined = 0x00000000 + pre_defined = 0x00000000 + pre_defined = 0x00000000 + pre_defined = 0x00000000 + next_track_ID = 2 + [trak: Track Box] + position = 140 + size = 608 + [tkhd: Track Header Box] + position = 148 + size = 92 + version = 0 + flags = 0x000007 + Track enabled + Track in movie + Track in preview + creation_time = UTC 2014/12/12, 18:41:19 + modification_time = UTC 2014/12/12, 18:41:19 + track_ID = 1 + reserved = 0x00000000 + duration = 33600 (00:00:00.700) + reserved = 0x00000000 + reserved = 0x00000000 + layer = 0 + alternate_group = 0 + volume = 1.000000 + reserved = 0x0000 + transformation matrix + | a, b, u | | 1.000000, 0.000000, 0.000000 | + | c, d, v | = | 0.000000, 1.000000, 0.000000 | + | x, y, w | | 0.000000, 0.000000, 1.000000 | + width = 0.000000 + height = 0.000000 + [edts: Edit Box] + position = 240 + size = 36 + [elst: Edit List Box] + position = 281 + size = 28 + version = 0 + flags = 0x000000 + entry_count = 1 + entry[0] + segment_duration = 33600 + media_time = 312 + media_rate = 1.000000 + [mdia: Media Box] + position = 276 + size = 472 + [mdhd: Media Header Box] + position = 284 + size = 32 + version = 0 + flags = 0x000000 + creation_time = UTC 2014/12/12, 18:41:19 + modification_time = UTC 2014/12/12, 18:41:19 + timescale = 48000 + duration = 34560 (00:00:00.720) + language = und + pre_defined = 0x0000 + [hdlr: Handler Reference Box] + position = 316 + size = 51 + version = 0 + flags = 0x000000 + pre_defined = 0x00000000 + handler_type = soun + reserved = 0x00000000 + reserved = 0x00000000 + reserved = 0x00000000 + name = Xiph Audio Handler + [minf: Media Information Box] + position = 367 + size = 381 + [smhd: Sound Media Header Box] + position = 375 + size = 16 + version = 0 + flags = 0x000000 + balance = 0.000000 + reserved = 0x0000 + [dinf: Data Information Box] + position = 391 + size = 36 + [dref: Data Reference Box] + position = 399 + size = 28 + version = 0 + flags = 0x000000 + entry_count = 1 + [url : Data Entry Url Box] + position = 415 + size = 12 + version = 0 + flags = 0x000001 + location = in the same file + [stbl: Sample Table Box] + position = 427 + size = 321 + [stsd: Sample Description Box] + position = 435 + size = 79 + version = 0 + flags = 0x000000 + entry_count = 1 + [Opus: Audio Description] + position = 451 + size = 63 + reserved = 0x000000000000 + data_reference_index = 1 + reserved = 0x0000 + reserved = 0x0000 + reserved = 0x00000000 + channelcount = 6 + samplesize = 16 + pre_defined = 0 + reserved = 0 + samplerate = 48000.000000 + [dOps: Opus Specific Box] + position = 487 + size = 27 + Version = 0 + OutputChannelCount = 6 + PreSkip = 312 + InputSampleRate = 48000 + OutputGain = 0 + ChannelMappingFamily = 1 + StreamCount = 4 + CoupledCount = 2 + ChannelMapping + 0 -> 0: front left + 1 -> 4: fron center + 2 -> 1: front right + 3 -> 2: side left + 4 -> 3: side right + 5 -> 5: rear center + [stts: Decoding Time to Sample Box] + position = 514 + size = 24 + version = 0 + flags = 0x000000 + entry_count = 1 + entry[0] + sample_count = 18 + sample_delta = 1920 + [stsc: Sample To Chunk Box] + position = 538 + size = 40 + version = 0 + flags = 0x000000 + entry_count = 2 + entry[0] + first_chunk = 1 + samples_per_chunk = 13 + sample_description_index = 1 + entry[1] + first_chunk = 2 + samples_per_chunk = 5 + sample_description_index = 1 + [stsz: Sample Size Box] + position = 578 + size = 92 + version = 0 + flags = 0x000000 + sample_size = 0 (variable) + sample_count = 18 + entry_size[0] = 977 + entry_size[1] = 938 + entry_size[2] = 939 + entry_size[3] = 938 + entry_size[4] = 934 + entry_size[5] = 945 + entry_size[6] = 948 + entry_size[7] = 956 + entry_size[8] = 955 + entry_size[9] = 930 + entry_size[10] = 933 + entry_size[11] = 934 + entry_size[12] = 972 + entry_size[13] = 977 + entry_size[14] = 958 + entry_size[15] = 949 + entry_size[16] = 962 + entry_size[17] = 848 + [stco: Chunk Offset Box] + position = 670 + size = 24 + version = 0 + flags = 0x000000 + entry_count = 2 + chunk_offset[0] = 797 + chunk_offset[1] = 13096 + [sgpd: Sample Group Description Box] + position = 694 + size = 26 + version = 1 + flags = 0x000000 + grouping_type = roll + default_length = 2 (constant) + entry_count = 1 + roll_distance[0] = -2 + [sbgp: Sample to Group Box] + position = 720 + size = 28 + version = 0 + flags = 0x000000 + grouping_type = roll + entry_count = 1 + entry[0] + sample_count = 18 + group_description_index = 1 + [free: Free Space Box] + position = 748 + size = 8 + [mdat: Media Data Box] + position = 756 + size = 17001 + +5 Authors' Address + Yusuke Nakamura + Email: muken.the.vfrmaniac |at| gmail.com +
+ + diff --git a/lib/opus/doc/opus_logo.svg b/lib/opus/doc/opus_logo.svg new file mode 100644 index 00000000..db2879ef --- /dev/null +++ b/lib/opus/doc/opus_logo.svg @@ -0,0 +1,157 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/lib/opus/doc/opus_update.patch b/lib/opus/doc/opus_update.patch new file mode 100644 index 00000000..11f066c7 --- /dev/null +++ b/lib/opus/doc/opus_update.patch @@ -0,0 +1,244 @@ +diff --git a/celt/bands.c b/celt/bands.c +index 6962587..32e1de6 100644 +--- a/celt/bands.c ++++ b/celt/bands.c +@@ -1234,9 +1234,23 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end, + b = 0; + } + +- if (resynth && M*eBands[i]-N >= M*eBands[start] && (update_lowband || lowband_offset==0)) ++ if (resynth && (M*eBands[i]-N >= M*eBands[start] || i==start+1) && (update_lowband || lowband_offset==0)) + lowband_offset = i; + ++ if (i == start+1) ++ { ++ int n1, n2; ++ int offset; ++ n1 = M*(eBands[start+1]-eBands[start]); ++ n2 = M*(eBands[start+2]-eBands[start+1]); ++ offset = M*eBands[start]; ++ /* Duplicate enough of the first band folding data to be able to fold the second band. ++ Copies no data for CELT-only mode. */ ++ OPUS_COPY(&norm[offset+n1], &norm[offset+2*n1 - n2], n2-n1); ++ if (C==2) ++ OPUS_COPY(&norm2[offset+n1], &norm2[offset+2*n1 - n2], n2-n1); ++ } ++ + tf_change = tf_res[i]; + if (i>=m->effEBands) + { +@@ -1257,7 +1271,7 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end, + fold_start = lowband_offset; + while(M*eBands[--fold_start] > effective_lowband); + fold_end = lowband_offset-1; +- while(M*eBands[++fold_end] < effective_lowband+N); ++ while(++fold_end < i && M*eBands[fold_end] < effective_lowband+N); + x_cm = y_cm = 0; + fold_i = fold_start; do { + x_cm |= collapse_masks[fold_i*C+0]; +diff --git a/celt/quant_bands.c b/celt/quant_bands.c +index e5ed9ef..82fb823 100644 +--- a/celt/quant_bands.c ++++ b/celt/quant_bands.c +@@ -552,6 +552,7 @@ void log2Amp(const CELTMode *m, int start, int end, + { + opus_val16 lg = ADD16(oldEBands[i+c*m->nbEBands], + SHL16((opus_val16)eMeans[i],6)); ++ lg = MIN32(QCONST32(32.f, 16), lg); + eBands[i+c*m->nbEBands] = PSHR32(celt_exp2(lg),4); + } + for (;inbEBands;i++) +diff --git a/silk/LPC_inv_pred_gain.c b/silk/LPC_inv_pred_gain.c +index 60c439b..6c301da 100644 +--- a/silk/LPC_inv_pred_gain.c ++++ b/silk/LPC_inv_pred_gain.c +@@ -84,8 +84,13 @@ static opus_int32 LPC_inverse_pred_gain_QA( /* O Returns inver + + /* Update AR coefficient */ + for( n = 0; n < k; n++ ) { +- tmp_QA = Aold_QA[ n ] - MUL32_FRAC_Q( Aold_QA[ k - n - 1 ], rc_Q31, 31 ); +- Anew_QA[ n ] = MUL32_FRAC_Q( tmp_QA, rc_mult2 , mult2Q ); ++ opus_int64 tmp64; ++ tmp_QA = silk_SUB_SAT32( Aold_QA[ n ], MUL32_FRAC_Q( Aold_QA[ k - n - 1 ], rc_Q31, 31 ) ); ++ tmp64 = silk_RSHIFT_ROUND64( silk_SMULL( tmp_QA, rc_mult2 ), mult2Q); ++ if( tmp64 > silk_int32_MAX || tmp64 < silk_int32_MIN ) { ++ return 0; ++ } ++ Anew_QA[ n ] = ( opus_int32 )tmp64; + } + } + +diff --git a/silk/NLSF_stabilize.c b/silk/NLSF_stabilize.c +index 979aaba..2ef2398 100644 +--- a/silk/NLSF_stabilize.c ++++ b/silk/NLSF_stabilize.c +@@ -134,7 +134,7 @@ void silk_NLSF_stabilize( + + /* Keep delta_min distance between the NLSFs */ + for( i = 1; i < L; i++ ) +- NLSF_Q15[i] = silk_max_int( NLSF_Q15[i], NLSF_Q15[i-1] + NDeltaMin_Q15[i] ); ++ NLSF_Q15[i] = silk_max_int( NLSF_Q15[i], silk_ADD_SAT16( NLSF_Q15[i-1], NDeltaMin_Q15[i] ) ); + + /* Last NLSF should be no higher than 1 - NDeltaMin[L] */ + NLSF_Q15[L-1] = silk_min_int( NLSF_Q15[L-1], (1<<15) - NDeltaMin_Q15[L] ); +diff --git a/silk/dec_API.c b/silk/dec_API.c +index efd7918..21bb7e0 100644 +--- a/silk/dec_API.c ++++ b/silk/dec_API.c +@@ -72,6 +72,9 @@ opus_int silk_InitDecoder( /* O Returns error co + for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { + ret = silk_init_decoder( &channel_state[ n ] ); + } ++ silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo)); ++ /* Not strictly needed, but it's cleaner that way */ ++ ((silk_decoder *)decState)->prev_decode_only_middle = 0; + + return ret; + } +diff --git a/silk/resampler_private_IIR_FIR.c b/silk/resampler_private_IIR_FIR.c +index dbd6d9a..91a43aa 100644 +--- a/silk/resampler_private_IIR_FIR.c ++++ b/silk/resampler_private_IIR_FIR.c +@@ -75,10 +75,10 @@ void silk_resampler_private_IIR_FIR( + silk_resampler_state_struct *S = (silk_resampler_state_struct *)SS; + opus_int32 nSamplesIn; + opus_int32 max_index_Q16, index_increment_Q16; +- opus_int16 buf[ RESAMPLER_MAX_BATCH_SIZE_IN + RESAMPLER_ORDER_FIR_12 ]; ++ opus_int16 buf[ 2*RESAMPLER_MAX_BATCH_SIZE_IN + RESAMPLER_ORDER_FIR_12 ]; + + /* Copy buffered samples to start of buffer */ +- silk_memcpy( buf, S->sFIR, RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); ++ silk_memcpy( buf, S->sFIR, RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + + /* Iterate over blocks of frameSizeIn input samples */ + index_increment_Q16 = S->invRatio_Q16; +@@ -95,13 +95,13 @@ void silk_resampler_private_IIR_FIR( + + if( inLen > 0 ) { + /* More iterations to do; copy last part of filtered signal to beginning of buffer */ +- silk_memcpy( buf, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); ++ silk_memmove( buf, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + } else { + break; + } + } + + /* Copy last part of filtered signal to the state for the next call */ +- silk_memcpy( S->sFIR, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int32 ) ); ++ silk_memcpy( S->sFIR, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + } + +diff --git a/src/opus_decoder.c b/src/opus_decoder.c +index 0cc56f8..8a30fbc 100644 +--- a/src/opus_decoder.c ++++ b/src/opus_decoder.c +@@ -595,16 +595,14 @@ static int opus_packet_parse_impl(const unsigned char *data, int len, + /* Padding flag is bit 6 */ + if (ch&0x40) + { +- int padding=0; + int p; + do { + if (len<=0) + return OPUS_INVALID_PACKET; + p = *data++; + len--; +- padding += p==255 ? 254: p; ++ len -= p==255 ? 254: p; + } while (p==255); +- len -= padding; + } + if (len<0) + return OPUS_INVALID_PACKET; +diff --git a/run_vectors.sh b/run_vectors.sh +index 7cd23ed..4841b0a 100755 +--- a/run_vectors.sh ++++ b/run_vectors.sh +@@ -1,3 +1,5 @@ ++#!/bin/sh ++# + # Copyright (c) 2011-2012 IETF Trust, Jean-Marc Valin. All rights reserved. + # + # This file is extracted from RFC6716. Please see that RFC for additional +@@ -31,10 +33,8 @@ + # NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + # SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +-#!/bin/sh +- +-rm logs_mono.txt +-rm logs_stereo.txt ++rm -f logs_mono.txt logs_mono2.txt ++rm -f logs_stereo.txt logs_stereo2.txt + + if [ "$#" -ne "3" ]; then + echo "usage: run_vectors.sh " +@@ -45,18 +45,23 @@ CMD_PATH=$1 + VECTOR_PATH=$2 + RATE=$3 + +-OPUS_DEMO=$CMD_PATH/opus_demo +-OPUS_COMPARE=$CMD_PATH/opus_compare ++: ${OPUS_DEMO:=$CMD_PATH/opus_demo} ++: ${OPUS_COMPARE:=$CMD_PATH/opus_compare} + + if [ -d $VECTOR_PATH ]; then + echo Test vectors found in $VECTOR_PATH + else + echo No test vectors found +- #Don't make the test fail here because the test vectors will be +- #distributed separately ++ #Don't make the test fail here because the test vectors ++ #will be distributed separately + exit 0 + fi + ++if [ ! -x $OPUS_COMPARE ]; then ++ echo ERROR: Compare program not found: $OPUS_COMPARE ++ exit 1 ++fi ++ + if [ -x $OPUS_DEMO ]; then + echo Decoding with $OPUS_DEMO + else +@@ -82,9 +87,11 @@ do + echo ERROR: decoding failed + exit 1 + fi +- $OPUS_COMPARE -r $RATE $VECTOR_PATH/testvector$file.dec tmp.out >> logs_mono.txt 2>&1 ++ $OPUS_COMPARE -r $RATE $VECTOR_PATH/testvector${file}.dec tmp.out >> logs_mono.txt 2>&1 + float_ret=$? +- if [ "$float_ret" -eq "0" ]; then ++ $OPUS_COMPARE -r $RATE $VECTOR_PATH/testvector${file}m.dec tmp.out >> logs_mono2.txt 2>&1 ++ float_ret2=$? ++ if [ "$float_ret" -eq "0" ] || [ "$float_ret2" -eq "0" ]; then + echo output matches reference + else + echo ERROR: output does not match reference +@@ -111,9 +118,11 @@ do + echo ERROR: decoding failed + exit 1 + fi +- $OPUS_COMPARE -s -r $RATE $VECTOR_PATH/testvector$file.dec tmp.out >> logs_stereo.txt 2>&1 ++ $OPUS_COMPARE -s -r $RATE $VECTOR_PATH/testvector${file}.dec tmp.out >> logs_stereo.txt 2>&1 + float_ret=$? +- if [ "$float_ret" -eq "0" ]; then ++ $OPUS_COMPARE -s -r $RATE $VECTOR_PATH/testvector${file}m.dec tmp.out >> logs_stereo2.txt 2>&1 ++ float_ret2=$? ++ if [ "$float_ret" -eq "0" ] || [ "$float_ret2" -eq "0" ]; then + echo output matches reference + else + echo ERROR: output does not match reference +@@ -125,5 +134,10 @@ done + + + echo All tests have passed successfully +-grep quality logs_mono.txt | awk '{sum+=$4}END{print "Average mono quality is", sum/NR, "%"}' +-grep quality logs_stereo.txt | awk '{sum+=$4}END{print "Average stereo quality is", sum/NR, "%"}' ++mono1=`grep quality logs_mono.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` ++mono2=`grep quality logs_mono2.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` ++echo $mono1 $mono2 | awk '{if ($2 > $1) $1 = $2; print "Average mono quality is", $1, "%"}' ++ ++stereo1=`grep quality logs_stereo.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` ++stereo2=`grep quality logs_stereo2.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` ++echo $stereo1 $stereo2 | awk '{if ($2 > $1) $1 = $2; print "Average stereo quality is", $1, "%"}' diff --git a/lib/opus/doc/release.txt b/lib/opus/doc/release.txt new file mode 100644 index 00000000..411ab7f4 --- /dev/null +++ b/lib/opus/doc/release.txt @@ -0,0 +1,43 @@ += Release checklist = + +== Source release == + +- Check for uncommitted changes to master. +- Update OPUS_LT_* API versioning in configure.ac. +- Tag the release commit with 'git tag -s vN.M'. + - Include release notes in the tag annotation. +- Verify 'make distcheck' produces a tarball with + the desired name. +- Push tag to public repo. +- Upload source package 'opus-${version}.tar.gz' + - Add to https://svn.xiph.org/releases/opus/ + - Update checksum files + - svn commit + - Copy to archive.mozilla.org/pub/opus/ + - Update checksum files there as well. +- Add release notes to https://gitlab.xiph.org/xiph/opus-website.git +- Update links and checksums on the downloads page. +- Add a copy of the documentation to + and update the links. +- Update /topic in #opus IRC channel. + +Releases are commited to https://svn.xiph.org/releases/opus/ +which propagates to downloads.xiph.org, and copied manually +to https://archive.mozilla.org/pub/opus/ + +Website updates are committed to https://gitlab.xiph.org/xiph/opus-website.git +which propagates to https://opus-codec.org/ + +== Binary release == + +We usually build opus-tools binaries for MacOS and Windows. + +Binary releases are copied manually to +https://archive.mozilla.org/pub/opus/win32/ + +For Mac, submit a pull request to homebrew. + +== Website updates == + +For major releases, recreate the files on https://opus-codec.org/examples/ +with the next encoder. diff --git a/lib/opus/doc/trivial_example.c b/lib/opus/doc/trivial_example.c new file mode 100644 index 00000000..9cf435b4 --- /dev/null +++ b/lib/opus/doc/trivial_example.c @@ -0,0 +1,171 @@ +/* Copyright (c) 2013 Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* This is meant to be a simple example of encoding and decoding audio + using Opus. It should make it easy to understand how the Opus API + works. For more information, see the full API documentation at: + https://www.opus-codec.org/docs/ */ + +#include +#include +#include +#include +#include + +/*The frame size is hardcoded for this sample code but it doesn't have to be*/ +#define FRAME_SIZE 960 +#define SAMPLE_RATE 48000 +#define CHANNELS 2 +#define APPLICATION OPUS_APPLICATION_AUDIO +#define BITRATE 64000 + +#define MAX_FRAME_SIZE 6*960 +#define MAX_PACKET_SIZE (3*1276) + +int main(int argc, char **argv) +{ + char *inFile; + FILE *fin; + char *outFile; + FILE *fout; + opus_int16 in[FRAME_SIZE*CHANNELS]; + opus_int16 out[MAX_FRAME_SIZE*CHANNELS]; + unsigned char cbits[MAX_PACKET_SIZE]; + int nbBytes; + /*Holds the state of the encoder and decoder */ + OpusEncoder *encoder; + OpusDecoder *decoder; + int err; + + if (argc != 3) + { + fprintf(stderr, "usage: trivial_example input.pcm output.pcm\n"); + fprintf(stderr, "input and output are 16-bit little-endian raw files\n"); + return EXIT_FAILURE; + } + + /*Create a new encoder state */ + encoder = opus_encoder_create(SAMPLE_RATE, CHANNELS, APPLICATION, &err); + if (err<0) + { + fprintf(stderr, "failed to create an encoder: %s\n", opus_strerror(err)); + return EXIT_FAILURE; + } + /* Set the desired bit-rate. You can also set other parameters if needed. + The Opus library is designed to have good defaults, so only set + parameters you know you need. Doing otherwise is likely to result + in worse quality, but better. */ + err = opus_encoder_ctl(encoder, OPUS_SET_BITRATE(BITRATE)); + if (err<0) + { + fprintf(stderr, "failed to set bitrate: %s\n", opus_strerror(err)); + return EXIT_FAILURE; + } + inFile = argv[1]; + fin = fopen(inFile, "rb"); + if (fin==NULL) + { + fprintf(stderr, "failed to open input file: %s\n", strerror(errno)); + return EXIT_FAILURE; + } + + + /* Create a new decoder state. */ + decoder = opus_decoder_create(SAMPLE_RATE, CHANNELS, &err); + if (err<0) + { + fprintf(stderr, "failed to create decoder: %s\n", opus_strerror(err)); + return EXIT_FAILURE; + } + outFile = argv[2]; + fout = fopen(outFile, "wb"); + if (fout==NULL) + { + fprintf(stderr, "failed to open output file: %s\n", strerror(errno)); + return EXIT_FAILURE; + } + + while (1) + { + int i; + unsigned char pcm_bytes[MAX_FRAME_SIZE*CHANNELS*2]; + int frame_size; + size_t samples; + + /* Read a 16 bits/sample audio frame. */ + samples = fread(pcm_bytes, sizeof(short)*CHANNELS, FRAME_SIZE, fin); + + /* For simplicity, only read whole frames. In a real application, + * we'd pad the final partial frame with zeroes, record the exact + * duration, and trim the decoded audio to match. + */ + if (samples != FRAME_SIZE) + { + break; + } + + /* Convert from little-endian ordering. */ + for (i=0;i>8)&0xFF; + } + /* Write the decoded audio to file. */ + fwrite(pcm_bytes, sizeof(short), frame_size*CHANNELS, fout); + } + /*Destroy the encoder state*/ + opus_encoder_destroy(encoder); + opus_decoder_destroy(decoder); + fclose(fin); + fclose(fout); + return EXIT_SUCCESS; +} diff --git a/lib/opus/include/meson.build b/lib/opus/include/meson.build new file mode 100644 index 00000000..c1fb0b76 --- /dev/null +++ b/lib/opus/include/meson.build @@ -0,0 +1,13 @@ +opus_headers = [ + 'opus.h', + 'opus_multistream.h', + 'opus_projection.h', + 'opus_types.h', + 'opus_defines.h', +] + +if opt_custom_modes + opus_headers += ['opus_custom.h'] +endif + +install_headers(opus_headers, subdir: 'opus') diff --git a/lib/opus/include/opus.h b/lib/opus/include/opus.h new file mode 100644 index 00000000..0c69c627 --- /dev/null +++ b/lib/opus/include/opus.h @@ -0,0 +1,981 @@ +/* Copyright (c) 2010-2011 Xiph.Org Foundation, Skype Limited + Written by Jean-Marc Valin and Koen Vos */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/** + * @file opus.h + * @brief Opus reference implementation API + */ + +#ifndef OPUS_H +#define OPUS_H + +#include "opus_types.h" +#include "opus_defines.h" + +#ifdef __cplusplus +extern "C" { +#endif + +/** + * @mainpage Opus + * + * The Opus codec is designed for interactive speech and audio transmission over the Internet. + * It is designed by the IETF Codec Working Group and incorporates technology from + * Skype's SILK codec and Xiph.Org's CELT codec. + * + * The Opus codec is designed to handle a wide range of interactive audio applications, + * including Voice over IP, videoconferencing, in-game chat, and even remote live music + * performances. It can scale from low bit-rate narrowband speech to very high quality + * stereo music. Its main features are: + + * @li Sampling rates from 8 to 48 kHz + * @li Bit-rates from 6 kb/s to 510 kb/s + * @li Support for both constant bit-rate (CBR) and variable bit-rate (VBR) + * @li Audio bandwidth from narrowband to full-band + * @li Support for speech and music + * @li Support for mono and stereo + * @li Support for multichannel (up to 255 channels) + * @li Frame sizes from 2.5 ms to 60 ms + * @li Good loss robustness and packet loss concealment (PLC) + * @li Floating point and fixed-point implementation + * + * Documentation sections: + * @li @ref opus_encoder + * @li @ref opus_decoder + * @li @ref opus_repacketizer + * @li @ref opus_multistream + * @li @ref opus_libinfo + * @li @ref opus_custom + */ + +/** @defgroup opus_encoder Opus Encoder + * @{ + * + * @brief This page describes the process and functions used to encode Opus. + * + * Since Opus is a stateful codec, the encoding process starts with creating an encoder + * state. This can be done with: + * + * @code + * int error; + * OpusEncoder *enc; + * enc = opus_encoder_create(Fs, channels, application, &error); + * @endcode + * + * From this point, @c enc can be used for encoding an audio stream. An encoder state + * @b must @b not be used for more than one stream at the same time. Similarly, the encoder + * state @b must @b not be re-initialized for each frame. + * + * While opus_encoder_create() allocates memory for the state, it's also possible + * to initialize pre-allocated memory: + * + * @code + * int size; + * int error; + * OpusEncoder *enc; + * size = opus_encoder_get_size(channels); + * enc = malloc(size); + * error = opus_encoder_init(enc, Fs, channels, application); + * @endcode + * + * where opus_encoder_get_size() returns the required size for the encoder state. Note that + * future versions of this code may change the size, so no assuptions should be made about it. + * + * The encoder state is always continuous in memory and only a shallow copy is sufficient + * to copy it (e.g. memcpy()) + * + * It is possible to change some of the encoder's settings using the opus_encoder_ctl() + * interface. All these settings already default to the recommended value, so they should + * only be changed when necessary. The most common settings one may want to change are: + * + * @code + * opus_encoder_ctl(enc, OPUS_SET_BITRATE(bitrate)); + * opus_encoder_ctl(enc, OPUS_SET_COMPLEXITY(complexity)); + * opus_encoder_ctl(enc, OPUS_SET_SIGNAL(signal_type)); + * @endcode + * + * where + * + * @arg bitrate is in bits per second (b/s) + * @arg complexity is a value from 1 to 10, where 1 is the lowest complexity and 10 is the highest + * @arg signal_type is either OPUS_AUTO (default), OPUS_SIGNAL_VOICE, or OPUS_SIGNAL_MUSIC + * + * See @ref opus_encoderctls and @ref opus_genericctls for a complete list of parameters that can be set or queried. Most parameters can be set or changed at any time during a stream. + * + * To encode a frame, opus_encode() or opus_encode_float() must be called with exactly one frame (2.5, 5, 10, 20, 40 or 60 ms) of audio data: + * @code + * len = opus_encode(enc, audio_frame, frame_size, packet, max_packet); + * @endcode + * + * where + *
    + *
  • audio_frame is the audio data in opus_int16 (or float for opus_encode_float())
    • + *
  • frame_size is the duration of the frame in samples (per channel)
    • + *
  • packet is the byte array to which the compressed data is written
    • + *
  • max_packet is the maximum number of bytes that can be written in the packet (4000 bytes is recommended). + * Do not use max_packet to control VBR target bitrate, instead use the #OPUS_SET_BITRATE CTL.
    • + *
+ * + * opus_encode() and opus_encode_float() return the number of bytes actually written to the packet. + * The return value can be negative, which indicates that an error has occurred. If the return value + * is 2 bytes or less, then the packet does not need to be transmitted (DTX). + * + * Once the encoder state if no longer needed, it can be destroyed with + * + * @code + * opus_encoder_destroy(enc); + * @endcode + * + * If the encoder was created with opus_encoder_init() rather than opus_encoder_create(), + * then no action is required aside from potentially freeing the memory that was manually + * allocated for it (calling free(enc) for the example above) + * + */ + +/** Opus encoder state. + * This contains the complete state of an Opus encoder. + * It is position independent and can be freely copied. + * @see opus_encoder_create,opus_encoder_init + */ +typedef struct OpusEncoder OpusEncoder; + +/** Gets the size of an OpusEncoder structure. + * @param[in] channels int: Number of channels. + * This must be 1 or 2. + * @returns The size in bytes. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_encoder_get_size(int channels); + +/** + */ + +/** Allocates and initializes an encoder state. + * There are three coding modes: + * + * @ref OPUS_APPLICATION_VOIP gives best quality at a given bitrate for voice + * signals. It enhances the input signal by high-pass filtering and + * emphasizing formants and harmonics. Optionally it includes in-band + * forward error correction to protect against packet loss. Use this + * mode for typical VoIP applications. Because of the enhancement, + * even at high bitrates the output may sound different from the input. + * + * @ref OPUS_APPLICATION_AUDIO gives best quality at a given bitrate for most + * non-voice signals like music. Use this mode for music and mixed + * (music/voice) content, broadcast, and applications requiring less + * than 15 ms of coding delay. + * + * @ref OPUS_APPLICATION_RESTRICTED_LOWDELAY configures low-delay mode that + * disables the speech-optimized mode in exchange for slightly reduced delay. + * This mode can only be set on an newly initialized or freshly reset encoder + * because it changes the codec delay. + * + * This is useful when the caller knows that the speech-optimized modes will not be needed (use with caution). + * @param [in] Fs opus_int32: Sampling rate of input signal (Hz) + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param [in] channels int: Number of channels (1 or 2) in input signal + * @param [in] application int: Coding mode (one of @ref OPUS_APPLICATION_VOIP, @ref OPUS_APPLICATION_AUDIO, or @ref OPUS_APPLICATION_RESTRICTED_LOWDELAY) + * @param [out] error int*: @ref opus_errorcodes + * @note Regardless of the sampling rate and number channels selected, the Opus encoder + * can switch to a lower audio bandwidth or number of channels if the bitrate + * selected is too low. This also means that it is safe to always use 48 kHz stereo input + * and let the encoder optimize the encoding. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusEncoder *opus_encoder_create( + opus_int32 Fs, + int channels, + int application, + int *error +); + +/** Initializes a previously allocated encoder state + * The memory pointed to by st must be at least the size returned by opus_encoder_get_size(). + * This is intended for applications which use their own allocator instead of malloc. + * @see opus_encoder_create(),opus_encoder_get_size() + * To reset a previously initialized state, use the #OPUS_RESET_STATE CTL. + * @param [in] st OpusEncoder*: Encoder state + * @param [in] Fs opus_int32: Sampling rate of input signal (Hz) + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param [in] channels int: Number of channels (1 or 2) in input signal + * @param [in] application int: Coding mode (one of OPUS_APPLICATION_VOIP, OPUS_APPLICATION_AUDIO, or OPUS_APPLICATION_RESTRICTED_LOWDELAY) + * @retval #OPUS_OK Success or @ref opus_errorcodes + */ +OPUS_EXPORT int opus_encoder_init( + OpusEncoder *st, + opus_int32 Fs, + int channels, + int application +) OPUS_ARG_NONNULL(1); + +/** Encodes an Opus frame. + * @param [in] st OpusEncoder*: Encoder state + * @param [in] pcm opus_int16*: Input signal (interleaved if 2 channels). length is frame_size*channels*sizeof(opus_int16) + * @param [in] frame_size int: Number of samples per channel in the + * input signal. + * This must be an Opus frame size for + * the encoder's sampling rate. + * For example, at 48 kHz the permitted + * values are 120, 240, 480, 960, 1920, + * and 2880. + * Passing in a duration of less than + * 10 ms (480 samples at 48 kHz) will + * prevent the encoder from using the LPC + * or hybrid modes. + * @param [out] data unsigned char*: Output payload. + * This must contain storage for at + * least \a max_data_bytes. + * @param [in] max_data_bytes opus_int32: Size of the allocated + * memory for the output + * payload. This may be + * used to impose an upper limit on + * the instant bitrate, but should + * not be used as the only bitrate + * control. Use #OPUS_SET_BITRATE to + * control the bitrate. + * @returns The length of the encoded packet (in bytes) on success or a + * negative error code (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_encode( + OpusEncoder *st, + const opus_int16 *pcm, + int frame_size, + unsigned char *data, + opus_int32 max_data_bytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + +/** Encodes an Opus frame from floating point input. + * @param [in] st OpusEncoder*: Encoder state + * @param [in] pcm float*: Input in float format (interleaved if 2 channels), with a normal range of +/-1.0. + * Samples with a range beyond +/-1.0 are supported but will + * be clipped by decoders using the integer API and should + * only be used if it is known that the far end supports + * extended dynamic range. + * length is frame_size*channels*sizeof(float) + * @param [in] frame_size int: Number of samples per channel in the + * input signal. + * This must be an Opus frame size for + * the encoder's sampling rate. + * For example, at 48 kHz the permitted + * values are 120, 240, 480, 960, 1920, + * and 2880. + * Passing in a duration of less than + * 10 ms (480 samples at 48 kHz) will + * prevent the encoder from using the LPC + * or hybrid modes. + * @param [out] data unsigned char*: Output payload. + * This must contain storage for at + * least \a max_data_bytes. + * @param [in] max_data_bytes opus_int32: Size of the allocated + * memory for the output + * payload. This may be + * used to impose an upper limit on + * the instant bitrate, but should + * not be used as the only bitrate + * control. Use #OPUS_SET_BITRATE to + * control the bitrate. + * @returns The length of the encoded packet (in bytes) on success or a + * negative error code (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_encode_float( + OpusEncoder *st, + const float *pcm, + int frame_size, + unsigned char *data, + opus_int32 max_data_bytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + +/** Frees an OpusEncoder allocated by opus_encoder_create(). + * @param[in] st OpusEncoder*: State to be freed. + */ +OPUS_EXPORT void opus_encoder_destroy(OpusEncoder *st); + +/** Perform a CTL function on an Opus encoder. + * + * Generally the request and subsequent arguments are generated + * by a convenience macro. + * @param st OpusEncoder*: Encoder state. + * @param request This and all remaining parameters should be replaced by one + * of the convenience macros in @ref opus_genericctls or + * @ref opus_encoderctls. + * @see opus_genericctls + * @see opus_encoderctls + */ +OPUS_EXPORT int opus_encoder_ctl(OpusEncoder *st, int request, ...) OPUS_ARG_NONNULL(1); +/**@}*/ + +/** @defgroup opus_decoder Opus Decoder + * @{ + * + * @brief This page describes the process and functions used to decode Opus. + * + * The decoding process also starts with creating a decoder + * state. This can be done with: + * @code + * int error; + * OpusDecoder *dec; + * dec = opus_decoder_create(Fs, channels, &error); + * @endcode + * where + * @li Fs is the sampling rate and must be 8000, 12000, 16000, 24000, or 48000 + * @li channels is the number of channels (1 or 2) + * @li error will hold the error code in case of failure (or #OPUS_OK on success) + * @li the return value is a newly created decoder state to be used for decoding + * + * While opus_decoder_create() allocates memory for the state, it's also possible + * to initialize pre-allocated memory: + * @code + * int size; + * int error; + * OpusDecoder *dec; + * size = opus_decoder_get_size(channels); + * dec = malloc(size); + * error = opus_decoder_init(dec, Fs, channels); + * @endcode + * where opus_decoder_get_size() returns the required size for the decoder state. Note that + * future versions of this code may change the size, so no assuptions should be made about it. + * + * The decoder state is always continuous in memory and only a shallow copy is sufficient + * to copy it (e.g. memcpy()) + * + * To decode a frame, opus_decode() or opus_decode_float() must be called with a packet of compressed audio data: + * @code + * frame_size = opus_decode(dec, packet, len, decoded, max_size, 0); + * @endcode + * where + * + * @li packet is the byte array containing the compressed data + * @li len is the exact number of bytes contained in the packet + * @li decoded is the decoded audio data in opus_int16 (or float for opus_decode_float()) + * @li max_size is the max duration of the frame in samples (per channel) that can fit into the decoded_frame array + * + * opus_decode() and opus_decode_float() return the number of samples (per channel) decoded from the packet. + * If that value is negative, then an error has occurred. This can occur if the packet is corrupted or if the audio + * buffer is too small to hold the decoded audio. + * + * Opus is a stateful codec with overlapping blocks and as a result Opus + * packets are not coded independently of each other. Packets must be + * passed into the decoder serially and in the correct order for a correct + * decode. Lost packets can be replaced with loss concealment by calling + * the decoder with a null pointer and zero length for the missing packet. + * + * A single codec state may only be accessed from a single thread at + * a time and any required locking must be performed by the caller. Separate + * streams must be decoded with separate decoder states and can be decoded + * in parallel unless the library was compiled with NONTHREADSAFE_PSEUDOSTACK + * defined. + * + */ + +/** Opus decoder state. + * This contains the complete state of an Opus decoder. + * It is position independent and can be freely copied. + * @see opus_decoder_create,opus_decoder_init + */ +typedef struct OpusDecoder OpusDecoder; + +/** Gets the size of an OpusDecoder structure. + * @param [in] channels int: Number of channels. + * This must be 1 or 2. + * @returns The size in bytes. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_decoder_get_size(int channels); + +/** Allocates and initializes a decoder state. + * @param [in] Fs opus_int32: Sample rate to decode at (Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param [in] channels int: Number of channels (1 or 2) to decode + * @param [out] error int*: #OPUS_OK Success or @ref opus_errorcodes + * + * Internally Opus stores data at 48000 Hz, so that should be the default + * value for Fs. However, the decoder can efficiently decode to buffers + * at 8, 12, 16, and 24 kHz so if for some reason the caller cannot use + * data at the full sample rate, or knows the compressed data doesn't + * use the full frequency range, it can request decoding at a reduced + * rate. Likewise, the decoder is capable of filling in either mono or + * interleaved stereo pcm buffers, at the caller's request. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusDecoder *opus_decoder_create( + opus_int32 Fs, + int channels, + int *error +); + +/** Initializes a previously allocated decoder state. + * The state must be at least the size returned by opus_decoder_get_size(). + * This is intended for applications which use their own allocator instead of malloc. @see opus_decoder_create,opus_decoder_get_size + * To reset a previously initialized state, use the #OPUS_RESET_STATE CTL. + * @param [in] st OpusDecoder*: Decoder state. + * @param [in] Fs opus_int32: Sampling rate to decode to (Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param [in] channels int: Number of channels (1 or 2) to decode + * @retval #OPUS_OK Success or @ref opus_errorcodes + */ +OPUS_EXPORT int opus_decoder_init( + OpusDecoder *st, + opus_int32 Fs, + int channels +) OPUS_ARG_NONNULL(1); + +/** Decode an Opus packet. + * @param [in] st OpusDecoder*: Decoder state + * @param [in] data char*: Input payload. Use a NULL pointer to indicate packet loss + * @param [in] len opus_int32: Number of bytes in payload* + * @param [out] pcm opus_int16*: Output signal (interleaved if 2 channels). length + * is frame_size*channels*sizeof(opus_int16) + * @param [in] frame_size Number of samples per channel of available space in \a pcm. + * If this is less than the maximum packet duration (120ms; 5760 for 48kHz), this function will + * not be capable of decoding some packets. In the case of PLC (data==NULL) or FEC (decode_fec=1), + * then frame_size needs to be exactly the duration of audio that is missing, otherwise the + * decoder will not be in the optimal state to decode the next incoming packet. For the PLC and + * FEC cases, frame_size must be a multiple of 2.5 ms. + * @param [in] decode_fec int: Flag (0 or 1) to request that any in-band forward error correction data be + * decoded. If no such data is available, the frame is decoded as if it were lost. + * @returns Number of decoded samples or @ref opus_errorcodes + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_decode( + OpusDecoder *st, + const unsigned char *data, + opus_int32 len, + opus_int16 *pcm, + int frame_size, + int decode_fec +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + +/** Decode an Opus packet with floating point output. + * @param [in] st OpusDecoder*: Decoder state + * @param [in] data char*: Input payload. Use a NULL pointer to indicate packet loss + * @param [in] len opus_int32: Number of bytes in payload + * @param [out] pcm float*: Output signal (interleaved if 2 channels). length + * is frame_size*channels*sizeof(float) + * @param [in] frame_size Number of samples per channel of available space in \a pcm. + * If this is less than the maximum packet duration (120ms; 5760 for 48kHz), this function will + * not be capable of decoding some packets. In the case of PLC (data==NULL) or FEC (decode_fec=1), + * then frame_size needs to be exactly the duration of audio that is missing, otherwise the + * decoder will not be in the optimal state to decode the next incoming packet. For the PLC and + * FEC cases, frame_size must be a multiple of 2.5 ms. + * @param [in] decode_fec int: Flag (0 or 1) to request that any in-band forward error correction data be + * decoded. If no such data is available the frame is decoded as if it were lost. + * @returns Number of decoded samples or @ref opus_errorcodes + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_decode_float( + OpusDecoder *st, + const unsigned char *data, + opus_int32 len, + float *pcm, + int frame_size, + int decode_fec +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + +/** Perform a CTL function on an Opus decoder. + * + * Generally the request and subsequent arguments are generated + * by a convenience macro. + * @param st OpusDecoder*: Decoder state. + * @param request This and all remaining parameters should be replaced by one + * of the convenience macros in @ref opus_genericctls or + * @ref opus_decoderctls. + * @see opus_genericctls + * @see opus_decoderctls + */ +OPUS_EXPORT int opus_decoder_ctl(OpusDecoder *st, int request, ...) OPUS_ARG_NONNULL(1); + +/** Frees an OpusDecoder allocated by opus_decoder_create(). + * @param[in] st OpusDecoder*: State to be freed. + */ +OPUS_EXPORT void opus_decoder_destroy(OpusDecoder *st); + +/** Parse an opus packet into one or more frames. + * Opus_decode will perform this operation internally so most applications do + * not need to use this function. + * This function does not copy the frames, the returned pointers are pointers into + * the input packet. + * @param [in] data char*: Opus packet to be parsed + * @param [in] len opus_int32: size of data + * @param [out] out_toc char*: TOC pointer + * @param [out] frames char*[48] encapsulated frames + * @param [out] size opus_int16[48] sizes of the encapsulated frames + * @param [out] payload_offset int*: returns the position of the payload within the packet (in bytes) + * @returns number of frames + */ +OPUS_EXPORT int opus_packet_parse( + const unsigned char *data, + opus_int32 len, + unsigned char *out_toc, + const unsigned char *frames[48], + opus_int16 size[48], + int *payload_offset +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(5); + +/** Gets the bandwidth of an Opus packet. + * @param [in] data char*: Opus packet + * @retval OPUS_BANDWIDTH_NARROWBAND Narrowband (4kHz bandpass) + * @retval OPUS_BANDWIDTH_MEDIUMBAND Mediumband (6kHz bandpass) + * @retval OPUS_BANDWIDTH_WIDEBAND Wideband (8kHz bandpass) + * @retval OPUS_BANDWIDTH_SUPERWIDEBAND Superwideband (12kHz bandpass) + * @retval OPUS_BANDWIDTH_FULLBAND Fullband (20kHz bandpass) + * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_bandwidth(const unsigned char *data) OPUS_ARG_NONNULL(1); + +/** Gets the number of samples per frame from an Opus packet. + * @param [in] data char*: Opus packet. + * This must contain at least one byte of + * data. + * @param [in] Fs opus_int32: Sampling rate in Hz. + * This must be a multiple of 400, or + * inaccurate results will be returned. + * @returns Number of samples per frame. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_samples_per_frame(const unsigned char *data, opus_int32 Fs) OPUS_ARG_NONNULL(1); + +/** Gets the number of channels from an Opus packet. + * @param [in] data char*: Opus packet + * @returns Number of channels + * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_channels(const unsigned char *data) OPUS_ARG_NONNULL(1); + +/** Gets the number of frames in an Opus packet. + * @param [in] packet char*: Opus packet + * @param [in] len opus_int32: Length of packet + * @returns Number of frames + * @retval OPUS_BAD_ARG Insufficient data was passed to the function + * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_frames(const unsigned char packet[], opus_int32 len) OPUS_ARG_NONNULL(1); + +/** Gets the number of samples of an Opus packet. + * @param [in] packet char*: Opus packet + * @param [in] len opus_int32: Length of packet + * @param [in] Fs opus_int32: Sampling rate in Hz. + * This must be a multiple of 400, or + * inaccurate results will be returned. + * @returns Number of samples + * @retval OPUS_BAD_ARG Insufficient data was passed to the function + * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_packet_get_nb_samples(const unsigned char packet[], opus_int32 len, opus_int32 Fs) OPUS_ARG_NONNULL(1); + +/** Gets the number of samples of an Opus packet. + * @param [in] dec OpusDecoder*: Decoder state + * @param [in] packet char*: Opus packet + * @param [in] len opus_int32: Length of packet + * @returns Number of samples + * @retval OPUS_BAD_ARG Insufficient data was passed to the function + * @retval OPUS_INVALID_PACKET The compressed data passed is corrupted or of an unsupported type + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_decoder_get_nb_samples(const OpusDecoder *dec, const unsigned char packet[], opus_int32 len) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); + +/** Applies soft-clipping to bring a float signal within the [-1,1] range. If + * the signal is already in that range, nothing is done. If there are values + * outside of [-1,1], then the signal is clipped as smoothly as possible to + * both fit in the range and avoid creating excessive distortion in the + * process. + * @param [in,out] pcm float*: Input PCM and modified PCM + * @param [in] frame_size int Number of samples per channel to process + * @param [in] channels int: Number of channels + * @param [in,out] softclip_mem float*: State memory for the soft clipping process (one float per channel, initialized to zero) + */ +OPUS_EXPORT void opus_pcm_soft_clip(float *pcm, int frame_size, int channels, float *softclip_mem); + + +/**@}*/ + +/** @defgroup opus_repacketizer Repacketizer + * @{ + * + * The repacketizer can be used to merge multiple Opus packets into a single + * packet or alternatively to split Opus packets that have previously been + * merged. Splitting valid Opus packets is always guaranteed to succeed, + * whereas merging valid packets only succeeds if all frames have the same + * mode, bandwidth, and frame size, and when the total duration of the merged + * packet is no more than 120 ms. The 120 ms limit comes from the + * specification and limits decoder memory requirements at a point where + * framing overhead becomes negligible. + * + * The repacketizer currently only operates on elementary Opus + * streams. It will not manipualte multistream packets successfully, except in + * the degenerate case where they consist of data from a single stream. + * + * The repacketizing process starts with creating a repacketizer state, either + * by calling opus_repacketizer_create() or by allocating the memory yourself, + * e.g., + * @code + * OpusRepacketizer *rp; + * rp = (OpusRepacketizer*)malloc(opus_repacketizer_get_size()); + * if (rp != NULL) + * opus_repacketizer_init(rp); + * @endcode + * + * Then the application should submit packets with opus_repacketizer_cat(), + * extract new packets with opus_repacketizer_out() or + * opus_repacketizer_out_range(), and then reset the state for the next set of + * input packets via opus_repacketizer_init(). + * + * For example, to split a sequence of packets into individual frames: + * @code + * unsigned char *data; + * int len; + * while (get_next_packet(&data, &len)) + * { + * unsigned char out[1276]; + * opus_int32 out_len; + * int nb_frames; + * int err; + * int i; + * err = opus_repacketizer_cat(rp, data, len); + * if (err != OPUS_OK) + * { + * release_packet(data); + * return err; + * } + * nb_frames = opus_repacketizer_get_nb_frames(rp); + * for (i = 0; i < nb_frames; i++) + * { + * out_len = opus_repacketizer_out_range(rp, i, i+1, out, sizeof(out)); + * if (out_len < 0) + * { + * release_packet(data); + * return (int)out_len; + * } + * output_next_packet(out, out_len); + * } + * opus_repacketizer_init(rp); + * release_packet(data); + * } + * @endcode + * + * Alternatively, to combine a sequence of frames into packets that each + * contain up to TARGET_DURATION_MS milliseconds of data: + * @code + * // The maximum number of packets with duration TARGET_DURATION_MS occurs + * // when the frame size is 2.5 ms, for a total of (TARGET_DURATION_MS*2/5) + * // packets. + * unsigned char *data[(TARGET_DURATION_MS*2/5)+1]; + * opus_int32 len[(TARGET_DURATION_MS*2/5)+1]; + * int nb_packets; + * unsigned char out[1277*(TARGET_DURATION_MS*2/2)]; + * opus_int32 out_len; + * int prev_toc; + * nb_packets = 0; + * while (get_next_packet(data+nb_packets, len+nb_packets)) + * { + * int nb_frames; + * int err; + * nb_frames = opus_packet_get_nb_frames(data[nb_packets], len[nb_packets]); + * if (nb_frames < 1) + * { + * release_packets(data, nb_packets+1); + * return nb_frames; + * } + * nb_frames += opus_repacketizer_get_nb_frames(rp); + * // If adding the next packet would exceed our target, or it has an + * // incompatible TOC sequence, output the packets we already have before + * // submitting it. + * // N.B., The nb_packets > 0 check ensures we've submitted at least one + * // packet since the last call to opus_repacketizer_init(). Otherwise a + * // single packet longer than TARGET_DURATION_MS would cause us to try to + * // output an (invalid) empty packet. It also ensures that prev_toc has + * // been set to a valid value. Additionally, len[nb_packets] > 0 is + * // guaranteed by the call to opus_packet_get_nb_frames() above, so the + * // reference to data[nb_packets][0] should be valid. + * if (nb_packets > 0 && ( + * ((prev_toc & 0xFC) != (data[nb_packets][0] & 0xFC)) || + * opus_packet_get_samples_per_frame(data[nb_packets], 48000)*nb_frames > + * TARGET_DURATION_MS*48)) + * { + * out_len = opus_repacketizer_out(rp, out, sizeof(out)); + * if (out_len < 0) + * { + * release_packets(data, nb_packets+1); + * return (int)out_len; + * } + * output_next_packet(out, out_len); + * opus_repacketizer_init(rp); + * release_packets(data, nb_packets); + * data[0] = data[nb_packets]; + * len[0] = len[nb_packets]; + * nb_packets = 0; + * } + * err = opus_repacketizer_cat(rp, data[nb_packets], len[nb_packets]); + * if (err != OPUS_OK) + * { + * release_packets(data, nb_packets+1); + * return err; + * } + * prev_toc = data[nb_packets][0]; + * nb_packets++; + * } + * // Output the final, partial packet. + * if (nb_packets > 0) + * { + * out_len = opus_repacketizer_out(rp, out, sizeof(out)); + * release_packets(data, nb_packets); + * if (out_len < 0) + * return (int)out_len; + * output_next_packet(out, out_len); + * } + * @endcode + * + * An alternate way of merging packets is to simply call opus_repacketizer_cat() + * unconditionally until it fails. At that point, the merged packet can be + * obtained with opus_repacketizer_out() and the input packet for which + * opus_repacketizer_cat() needs to be re-added to a newly reinitialized + * repacketizer state. + */ + +typedef struct OpusRepacketizer OpusRepacketizer; + +/** Gets the size of an OpusRepacketizer structure. + * @returns The size in bytes. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_repacketizer_get_size(void); + +/** (Re)initializes a previously allocated repacketizer state. + * The state must be at least the size returned by opus_repacketizer_get_size(). + * This can be used for applications which use their own allocator instead of + * malloc(). + * It must also be called to reset the queue of packets waiting to be + * repacketized, which is necessary if the maximum packet duration of 120 ms + * is reached or if you wish to submit packets with a different Opus + * configuration (coding mode, audio bandwidth, frame size, or channel count). + * Failure to do so will prevent a new packet from being added with + * opus_repacketizer_cat(). + * @see opus_repacketizer_create + * @see opus_repacketizer_get_size + * @see opus_repacketizer_cat + * @param rp OpusRepacketizer*: The repacketizer state to + * (re)initialize. + * @returns A pointer to the same repacketizer state that was passed in. + */ +OPUS_EXPORT OpusRepacketizer *opus_repacketizer_init(OpusRepacketizer *rp) OPUS_ARG_NONNULL(1); + +/** Allocates memory and initializes the new repacketizer with + * opus_repacketizer_init(). + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusRepacketizer *opus_repacketizer_create(void); + +/** Frees an OpusRepacketizer allocated by + * opus_repacketizer_create(). + * @param[in] rp OpusRepacketizer*: State to be freed. + */ +OPUS_EXPORT void opus_repacketizer_destroy(OpusRepacketizer *rp); + +/** Add a packet to the current repacketizer state. + * This packet must match the configuration of any packets already submitted + * for repacketization since the last call to opus_repacketizer_init(). + * This means that it must have the same coding mode, audio bandwidth, frame + * size, and channel count. + * This can be checked in advance by examining the top 6 bits of the first + * byte of the packet, and ensuring they match the top 6 bits of the first + * byte of any previously submitted packet. + * The total duration of audio in the repacketizer state also must not exceed + * 120 ms, the maximum duration of a single packet, after adding this packet. + * + * The contents of the current repacketizer state can be extracted into new + * packets using opus_repacketizer_out() or opus_repacketizer_out_range(). + * + * In order to add a packet with a different configuration or to add more + * audio beyond 120 ms, you must clear the repacketizer state by calling + * opus_repacketizer_init(). + * If a packet is too large to add to the current repacketizer state, no part + * of it is added, even if it contains multiple frames, some of which might + * fit. + * If you wish to be able to add parts of such packets, you should first use + * another repacketizer to split the packet into pieces and add them + * individually. + * @see opus_repacketizer_out_range + * @see opus_repacketizer_out + * @see opus_repacketizer_init + * @param rp OpusRepacketizer*: The repacketizer state to which to + * add the packet. + * @param[in] data const unsigned char*: The packet data. + * The application must ensure + * this pointer remains valid + * until the next call to + * opus_repacketizer_init() or + * opus_repacketizer_destroy(). + * @param len opus_int32: The number of bytes in the packet data. + * @returns An error code indicating whether or not the operation succeeded. + * @retval #OPUS_OK The packet's contents have been added to the repacketizer + * state. + * @retval #OPUS_INVALID_PACKET The packet did not have a valid TOC sequence, + * the packet's TOC sequence was not compatible + * with previously submitted packets (because + * the coding mode, audio bandwidth, frame size, + * or channel count did not match), or adding + * this packet would increase the total amount of + * audio stored in the repacketizer state to more + * than 120 ms. + */ +OPUS_EXPORT int opus_repacketizer_cat(OpusRepacketizer *rp, const unsigned char *data, opus_int32 len) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); + + +/** Construct a new packet from data previously submitted to the repacketizer + * state via opus_repacketizer_cat(). + * @param rp OpusRepacketizer*: The repacketizer state from which to + * construct the new packet. + * @param begin int: The index of the first frame in the current + * repacketizer state to include in the output. + * @param end int: One past the index of the last frame in the + * current repacketizer state to include in the + * output. + * @param[out] data const unsigned char*: The buffer in which to + * store the output packet. + * @param maxlen opus_int32: The maximum number of bytes to store in + * the output buffer. In order to guarantee + * success, this should be at least + * 1276 for a single frame, + * or for multiple frames, + * 1277*(end-begin). + * However, 1*(end-begin) plus + * the size of all packet data submitted to + * the repacketizer since the last call to + * opus_repacketizer_init() or + * opus_repacketizer_create() is also + * sufficient, and possibly much smaller. + * @returns The total size of the output packet on success, or an error code + * on failure. + * @retval #OPUS_BAD_ARG [begin,end) was an invalid range of + * frames (begin < 0, begin >= end, or end > + * opus_repacketizer_get_nb_frames()). + * @retval #OPUS_BUFFER_TOO_SMALL \a maxlen was insufficient to contain the + * complete output packet. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_repacketizer_out_range(OpusRepacketizer *rp, int begin, int end, unsigned char *data, opus_int32 maxlen) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + +/** Return the total number of frames contained in packet data submitted to + * the repacketizer state so far via opus_repacketizer_cat() since the last + * call to opus_repacketizer_init() or opus_repacketizer_create(). + * This defines the valid range of packets that can be extracted with + * opus_repacketizer_out_range() or opus_repacketizer_out(). + * @param rp OpusRepacketizer*: The repacketizer state containing the + * frames. + * @returns The total number of frames contained in the packet data submitted + * to the repacketizer state. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_repacketizer_get_nb_frames(OpusRepacketizer *rp) OPUS_ARG_NONNULL(1); + +/** Construct a new packet from data previously submitted to the repacketizer + * state via opus_repacketizer_cat(). + * This is a convenience routine that returns all the data submitted so far + * in a single packet. + * It is equivalent to calling + * @code + * opus_repacketizer_out_range(rp, 0, opus_repacketizer_get_nb_frames(rp), + * data, maxlen) + * @endcode + * @param rp OpusRepacketizer*: The repacketizer state from which to + * construct the new packet. + * @param[out] data const unsigned char*: The buffer in which to + * store the output packet. + * @param maxlen opus_int32: The maximum number of bytes to store in + * the output buffer. In order to guarantee + * success, this should be at least + * 1277*opus_repacketizer_get_nb_frames(rp). + * However, + * 1*opus_repacketizer_get_nb_frames(rp) + * plus the size of all packet data + * submitted to the repacketizer since the + * last call to opus_repacketizer_init() or + * opus_repacketizer_create() is also + * sufficient, and possibly much smaller. + * @returns The total size of the output packet on success, or an error code + * on failure. + * @retval #OPUS_BUFFER_TOO_SMALL \a maxlen was insufficient to contain the + * complete output packet. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_repacketizer_out(OpusRepacketizer *rp, unsigned char *data, opus_int32 maxlen) OPUS_ARG_NONNULL(1); + +/** Pads a given Opus packet to a larger size (possibly changing the TOC sequence). + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to pad. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @param new_len opus_int32: The desired size of the packet after padding. + * This must be at least as large as len. + * @returns an error code + * @retval #OPUS_OK \a on success. + * @retval #OPUS_BAD_ARG \a len was less than 1 or new_len was less than len. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT int opus_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len); + +/** Remove all padding from a given Opus packet and rewrite the TOC sequence to + * minimize space usage. + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to strip. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @returns The new size of the output packet on success, or an error code + * on failure. + * @retval #OPUS_BAD_ARG \a len was less than 1. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_packet_unpad(unsigned char *data, opus_int32 len); + +/** Pads a given Opus multi-stream packet to a larger size (possibly changing the TOC sequence). + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to pad. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @param new_len opus_int32: The desired size of the packet after padding. + * This must be at least 1. + * @param nb_streams opus_int32: The number of streams (not channels) in the packet. + * This must be at least as large as len. + * @returns an error code + * @retval #OPUS_OK \a on success. + * @retval #OPUS_BAD_ARG \a len was less than 1. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT int opus_multistream_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len, int nb_streams); + +/** Remove all padding from a given Opus multi-stream packet and rewrite the TOC sequence to + * minimize space usage. + * @param[in,out] data const unsigned char*: The buffer containing the + * packet to strip. + * @param len opus_int32: The size of the packet. + * This must be at least 1. + * @param nb_streams opus_int32: The number of streams (not channels) in the packet. + * This must be at least 1. + * @returns The new size of the output packet on success, or an error code + * on failure. + * @retval #OPUS_BAD_ARG \a len was less than 1 or new_len was less than len. + * @retval #OPUS_INVALID_PACKET \a data did not contain a valid Opus packet. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_multistream_packet_unpad(unsigned char *data, opus_int32 len, int nb_streams); + +/**@}*/ + +#ifdef __cplusplus +} +#endif + +#endif /* OPUS_H */ diff --git a/lib/opus/include/opus_custom.h b/lib/opus/include/opus_custom.h new file mode 100644 index 00000000..2f22d4b3 --- /dev/null +++ b/lib/opus/include/opus_custom.h @@ -0,0 +1,343 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Copyright (c) 2008-2012 Gregory Maxwell + Written by Jean-Marc Valin and Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/** + @file opus_custom.h + @brief Opus-Custom reference implementation API + */ + +#ifndef OPUS_CUSTOM_H +#define OPUS_CUSTOM_H + +#include "opus_defines.h" + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef CUSTOM_MODES +# define OPUS_CUSTOM_EXPORT OPUS_EXPORT +# define OPUS_CUSTOM_EXPORT_STATIC OPUS_EXPORT +#else +# define OPUS_CUSTOM_EXPORT +# ifdef OPUS_BUILD +# define OPUS_CUSTOM_EXPORT_STATIC static OPUS_INLINE +# else +# define OPUS_CUSTOM_EXPORT_STATIC +# endif +#endif + +/** @defgroup opus_custom Opus Custom + * @{ + * Opus Custom is an optional part of the Opus specification and + * reference implementation which uses a distinct API from the regular + * API and supports frame sizes that are not normally supported.\ Use + * of Opus Custom is discouraged for all but very special applications + * for which a frame size different from 2.5, 5, 10, or 20 ms is needed + * (for either complexity or latency reasons) and where interoperability + * is less important. + * + * In addition to the interoperability limitations the use of Opus custom + * disables a substantial chunk of the codec and generally lowers the + * quality available at a given bitrate. Normally when an application needs + * a different frame size from the codec it should buffer to match the + * sizes but this adds a small amount of delay which may be important + * in some very low latency applications. Some transports (especially + * constant rate RF transports) may also work best with frames of + * particular durations. + * + * Libopus only supports custom modes if they are enabled at compile time. + * + * The Opus Custom API is similar to the regular API but the + * @ref opus_encoder_create and @ref opus_decoder_create calls take + * an additional mode parameter which is a structure produced by + * a call to @ref opus_custom_mode_create. Both the encoder and decoder + * must create a mode using the same sample rate (fs) and frame size + * (frame size) so these parameters must either be signaled out of band + * or fixed in a particular implementation. + * + * Similar to regular Opus the custom modes support on the fly frame size + * switching, but the sizes available depend on the particular frame size in + * use. For some initial frame sizes on a single on the fly size is available. + */ + +/** Contains the state of an encoder. One encoder state is needed + for each stream. It is initialized once at the beginning of the + stream. Do *not* re-initialize the state for every frame. + @brief Encoder state + */ +typedef struct OpusCustomEncoder OpusCustomEncoder; + +/** State of the decoder. One decoder state is needed for each stream. + It is initialized once at the beginning of the stream. Do *not* + re-initialize the state for every frame. + @brief Decoder state + */ +typedef struct OpusCustomDecoder OpusCustomDecoder; + +/** The mode contains all the information necessary to create an + encoder. Both the encoder and decoder need to be initialized + with exactly the same mode, otherwise the output will be + corrupted. The mode MUST NOT BE DESTROYED until the encoders and + decoders that use it are destroyed as well. + @brief Mode configuration + */ +typedef struct OpusCustomMode OpusCustomMode; + +/** Creates a new mode struct. This will be passed to an encoder or + * decoder. The mode MUST NOT BE DESTROYED until the encoders and + * decoders that use it are destroyed as well. + * @param [in] Fs int: Sampling rate (8000 to 96000 Hz) + * @param [in] frame_size int: Number of samples (per channel) to encode in each + * packet (64 - 1024, prime factorization must contain zero or more 2s, 3s, or 5s and no other primes) + * @param [out] error int*: Returned error code (if NULL, no error will be returned) + * @return A newly created mode + */ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT OpusCustomMode *opus_custom_mode_create(opus_int32 Fs, int frame_size, int *error); + +/** Destroys a mode struct. Only call this after all encoders and + * decoders using this mode are destroyed as well. + * @param [in] mode OpusCustomMode*: Mode to be freed. + */ +OPUS_CUSTOM_EXPORT void opus_custom_mode_destroy(OpusCustomMode *mode); + + +#if !defined(OPUS_BUILD) || defined(CELT_ENCODER_C) + +/* Encoder */ +/** Gets the size of an OpusCustomEncoder structure. + * @param [in] mode OpusCustomMode *: Mode configuration + * @param [in] channels int: Number of channels + * @returns size + */ +OPUS_CUSTOM_EXPORT_STATIC OPUS_WARN_UNUSED_RESULT int opus_custom_encoder_get_size( + const OpusCustomMode *mode, + int channels +) OPUS_ARG_NONNULL(1); + +# ifdef CUSTOM_MODES +/** Initializes a previously allocated encoder state + * The memory pointed to by st must be the size returned by opus_custom_encoder_get_size. + * This is intended for applications which use their own allocator instead of malloc. + * @see opus_custom_encoder_create(),opus_custom_encoder_get_size() + * To reset a previously initialized state use the OPUS_RESET_STATE CTL. + * @param [in] st OpusCustomEncoder*: Encoder state + * @param [in] mode OpusCustomMode *: Contains all the information about the characteristics of + * the stream (must be the same characteristics as used for the + * decoder) + * @param [in] channels int: Number of channels + * @return OPUS_OK Success or @ref opus_errorcodes + */ +OPUS_CUSTOM_EXPORT int opus_custom_encoder_init( + OpusCustomEncoder *st, + const OpusCustomMode *mode, + int channels +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); +# endif +#endif + + +/** Creates a new encoder state. Each stream needs its own encoder + * state (can't be shared across simultaneous streams). + * @param [in] mode OpusCustomMode*: Contains all the information about the characteristics of + * the stream (must be the same characteristics as used for the + * decoder) + * @param [in] channels int: Number of channels + * @param [out] error int*: Returns an error code + * @return Newly created encoder state. +*/ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT OpusCustomEncoder *opus_custom_encoder_create( + const OpusCustomMode *mode, + int channels, + int *error +) OPUS_ARG_NONNULL(1); + + +/** Destroys an encoder state. + * @param[in] st OpusCustomEncoder*: State to be freed. + */ +OPUS_CUSTOM_EXPORT void opus_custom_encoder_destroy(OpusCustomEncoder *st); + +/** Encodes a frame of audio. + * @param [in] st OpusCustomEncoder*: Encoder state + * @param [in] pcm float*: PCM audio in float format, with a normal range of +/-1.0. + * Samples with a range beyond +/-1.0 are supported but will + * be clipped by decoders using the integer API and should + * only be used if it is known that the far end supports + * extended dynamic range. There must be exactly + * frame_size samples per channel. + * @param [in] frame_size int: Number of samples per frame of input signal + * @param [out] compressed char *: The compressed data is written here. This may not alias pcm and must be at least maxCompressedBytes long. + * @param [in] maxCompressedBytes int: Maximum number of bytes to use for compressing the frame + * (can change from one frame to another) + * @return Number of bytes written to "compressed". + * If negative, an error has occurred (see error codes). It is IMPORTANT that + * the length returned be somehow transmitted to the decoder. Otherwise, no + * decoding is possible. + */ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT int opus_custom_encode_float( + OpusCustomEncoder *st, + const float *pcm, + int frame_size, + unsigned char *compressed, + int maxCompressedBytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + +/** Encodes a frame of audio. + * @param [in] st OpusCustomEncoder*: Encoder state + * @param [in] pcm opus_int16*: PCM audio in signed 16-bit format (native endian). + * There must be exactly frame_size samples per channel. + * @param [in] frame_size int: Number of samples per frame of input signal + * @param [out] compressed char *: The compressed data is written here. This may not alias pcm and must be at least maxCompressedBytes long. + * @param [in] maxCompressedBytes int: Maximum number of bytes to use for compressing the frame + * (can change from one frame to another) + * @return Number of bytes written to "compressed". + * If negative, an error has occurred (see error codes). It is IMPORTANT that + * the length returned be somehow transmitted to the decoder. Otherwise, no + * decoding is possible. + */ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT int opus_custom_encode( + OpusCustomEncoder *st, + const opus_int16 *pcm, + int frame_size, + unsigned char *compressed, + int maxCompressedBytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + +/** Perform a CTL function on an Opus custom encoder. + * + * Generally the request and subsequent arguments are generated + * by a convenience macro. + * @see opus_encoderctls + */ +OPUS_CUSTOM_EXPORT int opus_custom_encoder_ctl(OpusCustomEncoder * OPUS_RESTRICT st, int request, ...) OPUS_ARG_NONNULL(1); + + +#if !defined(OPUS_BUILD) || defined(CELT_DECODER_C) +/* Decoder */ + +/** Gets the size of an OpusCustomDecoder structure. + * @param [in] mode OpusCustomMode *: Mode configuration + * @param [in] channels int: Number of channels + * @returns size + */ +OPUS_CUSTOM_EXPORT_STATIC OPUS_WARN_UNUSED_RESULT int opus_custom_decoder_get_size( + const OpusCustomMode *mode, + int channels +) OPUS_ARG_NONNULL(1); + +/** Initializes a previously allocated decoder state + * The memory pointed to by st must be the size returned by opus_custom_decoder_get_size. + * This is intended for applications which use their own allocator instead of malloc. + * @see opus_custom_decoder_create(),opus_custom_decoder_get_size() + * To reset a previously initialized state use the OPUS_RESET_STATE CTL. + * @param [in] st OpusCustomDecoder*: Decoder state + * @param [in] mode OpusCustomMode *: Contains all the information about the characteristics of + * the stream (must be the same characteristics as used for the + * encoder) + * @param [in] channels int: Number of channels + * @return OPUS_OK Success or @ref opus_errorcodes + */ +OPUS_CUSTOM_EXPORT_STATIC int opus_custom_decoder_init( + OpusCustomDecoder *st, + const OpusCustomMode *mode, + int channels +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2); + +#endif + + +/** Creates a new decoder state. Each stream needs its own decoder state (can't + * be shared across simultaneous streams). + * @param [in] mode OpusCustomMode: Contains all the information about the characteristics of the + * stream (must be the same characteristics as used for the encoder) + * @param [in] channels int: Number of channels + * @param [out] error int*: Returns an error code + * @return Newly created decoder state. + */ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT OpusCustomDecoder *opus_custom_decoder_create( + const OpusCustomMode *mode, + int channels, + int *error +) OPUS_ARG_NONNULL(1); + +/** Destroys a decoder state. + * @param[in] st OpusCustomDecoder*: State to be freed. + */ +OPUS_CUSTOM_EXPORT void opus_custom_decoder_destroy(OpusCustomDecoder *st); + +/** Decode an opus custom frame with floating point output + * @param [in] st OpusCustomDecoder*: Decoder state + * @param [in] data char*: Input payload. Use a NULL pointer to indicate packet loss + * @param [in] len int: Number of bytes in payload + * @param [out] pcm float*: Output signal (interleaved if 2 channels). length + * is frame_size*channels*sizeof(float) + * @param [in] frame_size Number of samples per channel of available space in *pcm. + * @returns Number of decoded samples or @ref opus_errorcodes + */ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT int opus_custom_decode_float( + OpusCustomDecoder *st, + const unsigned char *data, + int len, + float *pcm, + int frame_size +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + +/** Decode an opus custom frame + * @param [in] st OpusCustomDecoder*: Decoder state + * @param [in] data char*: Input payload. Use a NULL pointer to indicate packet loss + * @param [in] len int: Number of bytes in payload + * @param [out] pcm opus_int16*: Output signal (interleaved if 2 channels). length + * is frame_size*channels*sizeof(opus_int16) + * @param [in] frame_size Number of samples per channel of available space in *pcm. + * @returns Number of decoded samples or @ref opus_errorcodes + */ +OPUS_CUSTOM_EXPORT OPUS_WARN_UNUSED_RESULT int opus_custom_decode( + OpusCustomDecoder *st, + const unsigned char *data, + int len, + opus_int16 *pcm, + int frame_size +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + +/** Perform a CTL function on an Opus custom decoder. + * + * Generally the request and subsequent arguments are generated + * by a convenience macro. + * @see opus_genericctls + */ +OPUS_CUSTOM_EXPORT int opus_custom_decoder_ctl(OpusCustomDecoder * OPUS_RESTRICT st, int request, ...) OPUS_ARG_NONNULL(1); + +/**@}*/ + +#ifdef __cplusplus +} +#endif + +#endif /* OPUS_CUSTOM_H */ diff --git a/lib/opus/include/opus_defines.h b/lib/opus/include/opus_defines.h new file mode 100644 index 00000000..94b9e0d9 --- /dev/null +++ b/lib/opus/include/opus_defines.h @@ -0,0 +1,801 @@ +/* Copyright (c) 2010-2011 Xiph.Org Foundation, Skype Limited + Written by Jean-Marc Valin and Koen Vos */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/** + * @file opus_defines.h + * @brief Opus reference implementation constants + */ + +#ifndef OPUS_DEFINES_H +#define OPUS_DEFINES_H + +#include "opus_types.h" + +#ifdef __cplusplus +extern "C" { +#endif + +/** @defgroup opus_errorcodes Error codes + * @{ + */ +/** No error @hideinitializer*/ +#define OPUS_OK 0 +/** One or more invalid/out of range arguments @hideinitializer*/ +#define OPUS_BAD_ARG -1 +/** Not enough bytes allocated in the buffer @hideinitializer*/ +#define OPUS_BUFFER_TOO_SMALL -2 +/** An internal error was detected @hideinitializer*/ +#define OPUS_INTERNAL_ERROR -3 +/** The compressed data passed is corrupted @hideinitializer*/ +#define OPUS_INVALID_PACKET -4 +/** Invalid/unsupported request number @hideinitializer*/ +#define OPUS_UNIMPLEMENTED -5 +/** An encoder or decoder structure is invalid or already freed @hideinitializer*/ +#define OPUS_INVALID_STATE -6 +/** Memory allocation has failed @hideinitializer*/ +#define OPUS_ALLOC_FAIL -7 +/**@}*/ + +/** @cond OPUS_INTERNAL_DOC */ +/**Export control for opus functions */ + +#ifndef OPUS_EXPORT +# if defined(_WIN32) +# if defined(OPUS_BUILD) && defined(DLL_EXPORT) +# define OPUS_EXPORT __declspec(dllexport) +# else +# define OPUS_EXPORT +# endif +# elif defined(__GNUC__) && defined(OPUS_BUILD) +# define OPUS_EXPORT __attribute__ ((visibility ("default"))) +# else +# define OPUS_EXPORT +# endif +#endif + +# if !defined(OPUS_GNUC_PREREQ) +# if defined(__GNUC__)&&defined(__GNUC_MINOR__) +# define OPUS_GNUC_PREREQ(_maj,_min) \ + ((__GNUC__<<16)+__GNUC_MINOR__>=((_maj)<<16)+(_min)) +# else +# define OPUS_GNUC_PREREQ(_maj,_min) 0 +# endif +# endif + +#if (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L) ) +# if OPUS_GNUC_PREREQ(3,0) +# define OPUS_RESTRICT __restrict__ +# elif (defined(_MSC_VER) && _MSC_VER >= 1400) +# define OPUS_RESTRICT __restrict +# else +# define OPUS_RESTRICT +# endif +#else +# define OPUS_RESTRICT restrict +#endif + +#if (!defined(__STDC_VERSION__) || (__STDC_VERSION__ < 199901L) ) +# if OPUS_GNUC_PREREQ(2,7) +# define OPUS_INLINE __inline__ +# elif (defined(_MSC_VER)) +# define OPUS_INLINE __inline +# else +# define OPUS_INLINE +# endif +#else +# define OPUS_INLINE inline +#endif + +/**Warning attributes for opus functions + * NONNULL is not used in OPUS_BUILD to avoid the compiler optimizing out + * some paranoid null checks. */ +#if defined(__GNUC__) && OPUS_GNUC_PREREQ(3, 4) +# define OPUS_WARN_UNUSED_RESULT __attribute__ ((__warn_unused_result__)) +#else +# define OPUS_WARN_UNUSED_RESULT +#endif +#if !defined(OPUS_BUILD) && defined(__GNUC__) && OPUS_GNUC_PREREQ(3, 4) +# define OPUS_ARG_NONNULL(_x) __attribute__ ((__nonnull__(_x))) +#else +# define OPUS_ARG_NONNULL(_x) +#endif + +/** These are the actual Encoder CTL ID numbers. + * They should not be used directly by applications. + * In general, SETs should be even and GETs should be odd.*/ +#define OPUS_SET_APPLICATION_REQUEST 4000 +#define OPUS_GET_APPLICATION_REQUEST 4001 +#define OPUS_SET_BITRATE_REQUEST 4002 +#define OPUS_GET_BITRATE_REQUEST 4003 +#define OPUS_SET_MAX_BANDWIDTH_REQUEST 4004 +#define OPUS_GET_MAX_BANDWIDTH_REQUEST 4005 +#define OPUS_SET_VBR_REQUEST 4006 +#define OPUS_GET_VBR_REQUEST 4007 +#define OPUS_SET_BANDWIDTH_REQUEST 4008 +#define OPUS_GET_BANDWIDTH_REQUEST 4009 +#define OPUS_SET_COMPLEXITY_REQUEST 4010 +#define OPUS_GET_COMPLEXITY_REQUEST 4011 +#define OPUS_SET_INBAND_FEC_REQUEST 4012 +#define OPUS_GET_INBAND_FEC_REQUEST 4013 +#define OPUS_SET_PACKET_LOSS_PERC_REQUEST 4014 +#define OPUS_GET_PACKET_LOSS_PERC_REQUEST 4015 +#define OPUS_SET_DTX_REQUEST 4016 +#define OPUS_GET_DTX_REQUEST 4017 +#define OPUS_SET_VBR_CONSTRAINT_REQUEST 4020 +#define OPUS_GET_VBR_CONSTRAINT_REQUEST 4021 +#define OPUS_SET_FORCE_CHANNELS_REQUEST 4022 +#define OPUS_GET_FORCE_CHANNELS_REQUEST 4023 +#define OPUS_SET_SIGNAL_REQUEST 4024 +#define OPUS_GET_SIGNAL_REQUEST 4025 +#define OPUS_GET_LOOKAHEAD_REQUEST 4027 +/* #define OPUS_RESET_STATE 4028 */ +#define OPUS_GET_SAMPLE_RATE_REQUEST 4029 +#define OPUS_GET_FINAL_RANGE_REQUEST 4031 +#define OPUS_GET_PITCH_REQUEST 4033 +#define OPUS_SET_GAIN_REQUEST 4034 +#define OPUS_GET_GAIN_REQUEST 4045 /* Should have been 4035 */ +#define OPUS_SET_LSB_DEPTH_REQUEST 4036 +#define OPUS_GET_LSB_DEPTH_REQUEST 4037 +#define OPUS_GET_LAST_PACKET_DURATION_REQUEST 4039 +#define OPUS_SET_EXPERT_FRAME_DURATION_REQUEST 4040 +#define OPUS_GET_EXPERT_FRAME_DURATION_REQUEST 4041 +#define OPUS_SET_PREDICTION_DISABLED_REQUEST 4042 +#define OPUS_GET_PREDICTION_DISABLED_REQUEST 4043 +/* Don't use 4045, it's already taken by OPUS_GET_GAIN_REQUEST */ +#define OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST 4046 +#define OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST 4047 +#define OPUS_GET_IN_DTX_REQUEST 4049 + +/** Defines for the presence of extended APIs. */ +#define OPUS_HAVE_OPUS_PROJECTION_H + +/* Macros to trigger compilation errors when the wrong types are provided to a CTL */ +#define __opus_check_int(x) (((void)((x) == (opus_int32)0)), (opus_int32)(x)) +#define __opus_check_int_ptr(ptr) ((ptr) + ((ptr) - (opus_int32*)(ptr))) +#define __opus_check_uint_ptr(ptr) ((ptr) + ((ptr) - (opus_uint32*)(ptr))) +#define __opus_check_val16_ptr(ptr) ((ptr) + ((ptr) - (opus_val16*)(ptr))) +/** @endcond */ + +/** @defgroup opus_ctlvalues Pre-defined values for CTL interface + * @see opus_genericctls, opus_encoderctls + * @{ + */ +/* Values for the various encoder CTLs */ +#define OPUS_AUTO -1000 /**opus_int32: Allowed values: 0-10, inclusive. + * + * @hideinitializer */ +#define OPUS_SET_COMPLEXITY(x) OPUS_SET_COMPLEXITY_REQUEST, __opus_check_int(x) +/** Gets the encoder's complexity configuration. + * @see OPUS_SET_COMPLEXITY + * @param[out] x opus_int32 *: Returns a value in the range 0-10, + * inclusive. + * @hideinitializer */ +#define OPUS_GET_COMPLEXITY(x) OPUS_GET_COMPLEXITY_REQUEST, __opus_check_int_ptr(x) + +/** Configures the bitrate in the encoder. + * Rates from 500 to 512000 bits per second are meaningful, as well as the + * special values #OPUS_AUTO and #OPUS_BITRATE_MAX. + * The value #OPUS_BITRATE_MAX can be used to cause the codec to use as much + * rate as it can, which is useful for controlling the rate by adjusting the + * output buffer size. + * @see OPUS_GET_BITRATE + * @param[in] x opus_int32: Bitrate in bits per second. The default + * is determined based on the number of + * channels and the input sampling rate. + * @hideinitializer */ +#define OPUS_SET_BITRATE(x) OPUS_SET_BITRATE_REQUEST, __opus_check_int(x) +/** Gets the encoder's bitrate configuration. + * @see OPUS_SET_BITRATE + * @param[out] x opus_int32 *: Returns the bitrate in bits per second. + * The default is determined based on the + * number of channels and the input + * sampling rate. + * @hideinitializer */ +#define OPUS_GET_BITRATE(x) OPUS_GET_BITRATE_REQUEST, __opus_check_int_ptr(x) + +/** Enables or disables variable bitrate (VBR) in the encoder. + * The configured bitrate may not be met exactly because frames must + * be an integer number of bytes in length. + * @see OPUS_GET_VBR + * @see OPUS_SET_VBR_CONSTRAINT + * @param[in] x opus_int32: Allowed values: + *
+ *
0
Hard CBR. For LPC/hybrid modes at very low bit-rate, this can + * cause noticeable quality degradation.
+ *
1
VBR (default). The exact type of VBR is controlled by + * #OPUS_SET_VBR_CONSTRAINT.
+ *
+ * @hideinitializer */ +#define OPUS_SET_VBR(x) OPUS_SET_VBR_REQUEST, __opus_check_int(x) +/** Determine if variable bitrate (VBR) is enabled in the encoder. + * @see OPUS_SET_VBR + * @see OPUS_GET_VBR_CONSTRAINT + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
0
Hard CBR.
+ *
1
VBR (default). The exact type of VBR may be retrieved via + * #OPUS_GET_VBR_CONSTRAINT.
+ *
+ * @hideinitializer */ +#define OPUS_GET_VBR(x) OPUS_GET_VBR_REQUEST, __opus_check_int_ptr(x) + +/** Enables or disables constrained VBR in the encoder. + * This setting is ignored when the encoder is in CBR mode. + * @warning Only the MDCT mode of Opus currently heeds the constraint. + * Speech mode ignores it completely, hybrid mode may fail to obey it + * if the LPC layer uses more bitrate than the constraint would have + * permitted. + * @see OPUS_GET_VBR_CONSTRAINT + * @see OPUS_SET_VBR + * @param[in] x opus_int32: Allowed values: + *
+ *
0
Unconstrained VBR.
+ *
1
Constrained VBR (default). This creates a maximum of one + * frame of buffering delay assuming a transport with a + * serialization speed of the nominal bitrate.
+ *
+ * @hideinitializer */ +#define OPUS_SET_VBR_CONSTRAINT(x) OPUS_SET_VBR_CONSTRAINT_REQUEST, __opus_check_int(x) +/** Determine if constrained VBR is enabled in the encoder. + * @see OPUS_SET_VBR_CONSTRAINT + * @see OPUS_GET_VBR + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
0
Unconstrained VBR.
+ *
1
Constrained VBR (default).
+ *
+ * @hideinitializer */ +#define OPUS_GET_VBR_CONSTRAINT(x) OPUS_GET_VBR_CONSTRAINT_REQUEST, __opus_check_int_ptr(x) + +/** Configures mono/stereo forcing in the encoder. + * This can force the encoder to produce packets encoded as either mono or + * stereo, regardless of the format of the input audio. This is useful when + * the caller knows that the input signal is currently a mono source embedded + * in a stereo stream. + * @see OPUS_GET_FORCE_CHANNELS + * @param[in] x opus_int32: Allowed values: + *
+ *
#OPUS_AUTO
Not forced (default)
+ *
1
Forced mono
+ *
2
Forced stereo
+ *
+ * @hideinitializer */ +#define OPUS_SET_FORCE_CHANNELS(x) OPUS_SET_FORCE_CHANNELS_REQUEST, __opus_check_int(x) +/** Gets the encoder's forced channel configuration. + * @see OPUS_SET_FORCE_CHANNELS + * @param[out] x opus_int32 *: + *
+ *
#OPUS_AUTO
Not forced (default)
+ *
1
Forced mono
+ *
2
Forced stereo
+ *
+ * @hideinitializer */ +#define OPUS_GET_FORCE_CHANNELS(x) OPUS_GET_FORCE_CHANNELS_REQUEST, __opus_check_int_ptr(x) + +/** Configures the maximum bandpass that the encoder will select automatically. + * Applications should normally use this instead of #OPUS_SET_BANDWIDTH + * (leaving that set to the default, #OPUS_AUTO). This allows the + * application to set an upper bound based on the type of input it is + * providing, but still gives the encoder the freedom to reduce the bandpass + * when the bitrate becomes too low, for better overall quality. + * @see OPUS_GET_MAX_BANDWIDTH + * @param[in] x opus_int32: Allowed values: + *
+ *
OPUS_BANDWIDTH_NARROWBAND
4 kHz passband
+ *
OPUS_BANDWIDTH_MEDIUMBAND
6 kHz passband
+ *
OPUS_BANDWIDTH_WIDEBAND
8 kHz passband
+ *
OPUS_BANDWIDTH_SUPERWIDEBAND
12 kHz passband
+ *
OPUS_BANDWIDTH_FULLBAND
20 kHz passband (default)
+ *
+ * @hideinitializer */ +#define OPUS_SET_MAX_BANDWIDTH(x) OPUS_SET_MAX_BANDWIDTH_REQUEST, __opus_check_int(x) + +/** Gets the encoder's configured maximum allowed bandpass. + * @see OPUS_SET_MAX_BANDWIDTH + * @param[out] x opus_int32 *: Allowed values: + *
+ *
#OPUS_BANDWIDTH_NARROWBAND
4 kHz passband
+ *
#OPUS_BANDWIDTH_MEDIUMBAND
6 kHz passband
+ *
#OPUS_BANDWIDTH_WIDEBAND
8 kHz passband
+ *
#OPUS_BANDWIDTH_SUPERWIDEBAND
12 kHz passband
+ *
#OPUS_BANDWIDTH_FULLBAND
20 kHz passband (default)
+ *
+ * @hideinitializer */ +#define OPUS_GET_MAX_BANDWIDTH(x) OPUS_GET_MAX_BANDWIDTH_REQUEST, __opus_check_int_ptr(x) + +/** Sets the encoder's bandpass to a specific value. + * This prevents the encoder from automatically selecting the bandpass based + * on the available bitrate. If an application knows the bandpass of the input + * audio it is providing, it should normally use #OPUS_SET_MAX_BANDWIDTH + * instead, which still gives the encoder the freedom to reduce the bandpass + * when the bitrate becomes too low, for better overall quality. + * @see OPUS_GET_BANDWIDTH + * @param[in] x opus_int32: Allowed values: + *
+ *
#OPUS_AUTO
(default)
+ *
#OPUS_BANDWIDTH_NARROWBAND
4 kHz passband
+ *
#OPUS_BANDWIDTH_MEDIUMBAND
6 kHz passband
+ *
#OPUS_BANDWIDTH_WIDEBAND
8 kHz passband
+ *
#OPUS_BANDWIDTH_SUPERWIDEBAND
12 kHz passband
+ *
#OPUS_BANDWIDTH_FULLBAND
20 kHz passband
+ *
+ * @hideinitializer */ +#define OPUS_SET_BANDWIDTH(x) OPUS_SET_BANDWIDTH_REQUEST, __opus_check_int(x) + +/** Configures the type of signal being encoded. + * This is a hint which helps the encoder's mode selection. + * @see OPUS_GET_SIGNAL + * @param[in] x opus_int32: Allowed values: + *
+ *
#OPUS_AUTO
(default)
+ *
#OPUS_SIGNAL_VOICE
Bias thresholds towards choosing LPC or Hybrid modes.
+ *
#OPUS_SIGNAL_MUSIC
Bias thresholds towards choosing MDCT modes.
+ *
+ * @hideinitializer */ +#define OPUS_SET_SIGNAL(x) OPUS_SET_SIGNAL_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured signal type. + * @see OPUS_SET_SIGNAL + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
#OPUS_AUTO
(default)
+ *
#OPUS_SIGNAL_VOICE
Bias thresholds towards choosing LPC or Hybrid modes.
+ *
#OPUS_SIGNAL_MUSIC
Bias thresholds towards choosing MDCT modes.
+ *
+ * @hideinitializer */ +#define OPUS_GET_SIGNAL(x) OPUS_GET_SIGNAL_REQUEST, __opus_check_int_ptr(x) + + +/** Configures the encoder's intended application. + * The initial value is a mandatory argument to the encoder_create function. + * @see OPUS_GET_APPLICATION + * @param[in] x opus_int32: Returns one of the following values: + *
+ *
#OPUS_APPLICATION_VOIP
+ *
Process signal for improved speech intelligibility.
+ *
#OPUS_APPLICATION_AUDIO
+ *
Favor faithfulness to the original input.
+ *
#OPUS_APPLICATION_RESTRICTED_LOWDELAY
+ *
Configure the minimum possible coding delay by disabling certain modes + * of operation.
+ *
+ * @hideinitializer */ +#define OPUS_SET_APPLICATION(x) OPUS_SET_APPLICATION_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured application. + * @see OPUS_SET_APPLICATION + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
#OPUS_APPLICATION_VOIP
+ *
Process signal for improved speech intelligibility.
+ *
#OPUS_APPLICATION_AUDIO
+ *
Favor faithfulness to the original input.
+ *
#OPUS_APPLICATION_RESTRICTED_LOWDELAY
+ *
Configure the minimum possible coding delay by disabling certain modes + * of operation.
+ *
+ * @hideinitializer */ +#define OPUS_GET_APPLICATION(x) OPUS_GET_APPLICATION_REQUEST, __opus_check_int_ptr(x) + +/** Gets the total samples of delay added by the entire codec. + * This can be queried by the encoder and then the provided number of samples can be + * skipped on from the start of the decoder's output to provide time aligned input + * and output. From the perspective of a decoding application the real data begins this many + * samples late. + * + * The decoder contribution to this delay is identical for all decoders, but the + * encoder portion of the delay may vary from implementation to implementation, + * version to version, or even depend on the encoder's initial configuration. + * Applications needing delay compensation should call this CTL rather than + * hard-coding a value. + * @param[out] x opus_int32 *: Number of lookahead samples + * @hideinitializer */ +#define OPUS_GET_LOOKAHEAD(x) OPUS_GET_LOOKAHEAD_REQUEST, __opus_check_int_ptr(x) + +/** Configures the encoder's use of inband forward error correction (FEC). + * @note This is only applicable to the LPC layer + * @see OPUS_GET_INBAND_FEC + * @param[in] x opus_int32: Allowed values: + *
+ *
0
Disable inband FEC (default).
+ *
1
Inband FEC enabled. If the packet loss rate is sufficiently high, Opus will automatically switch to SILK even at high rates to enable use of that FEC.
+ *
2
Inband FEC enabled, but does not necessarily switch to SILK if we have music.
+ *
+ * @hideinitializer */ +#define OPUS_SET_INBAND_FEC(x) OPUS_SET_INBAND_FEC_REQUEST, __opus_check_int(x) +/** Gets encoder's configured use of inband forward error correction. + * @see OPUS_SET_INBAND_FEC + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
0
Inband FEC disabled (default).
+ *
1
Inband FEC enabled. If the packet loss rate is sufficiently high, Opus will automatically switch to SILK even at high rates to enable use of that FEC.
+ *
2
Inband FEC enabled, but does not necessarily switch to SILK if we have music.
+ *
+ * @hideinitializer */ +#define OPUS_GET_INBAND_FEC(x) OPUS_GET_INBAND_FEC_REQUEST, __opus_check_int_ptr(x) + +/** Configures the encoder's expected packet loss percentage. + * Higher values trigger progressively more loss resistant behavior in the encoder + * at the expense of quality at a given bitrate in the absence of packet loss, but + * greater quality under loss. + * @see OPUS_GET_PACKET_LOSS_PERC + * @param[in] x opus_int32: Loss percentage in the range 0-100, inclusive (default: 0). + * @hideinitializer */ +#define OPUS_SET_PACKET_LOSS_PERC(x) OPUS_SET_PACKET_LOSS_PERC_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured packet loss percentage. + * @see OPUS_SET_PACKET_LOSS_PERC + * @param[out] x opus_int32 *: Returns the configured loss percentage + * in the range 0-100, inclusive (default: 0). + * @hideinitializer */ +#define OPUS_GET_PACKET_LOSS_PERC(x) OPUS_GET_PACKET_LOSS_PERC_REQUEST, __opus_check_int_ptr(x) + +/** Configures the encoder's use of discontinuous transmission (DTX). + * @note This is only applicable to the LPC layer + * @see OPUS_GET_DTX + * @param[in] x opus_int32: Allowed values: + *
+ *
0
Disable DTX (default).
+ *
1
Enabled DTX.
+ *
+ * @hideinitializer */ +#define OPUS_SET_DTX(x) OPUS_SET_DTX_REQUEST, __opus_check_int(x) +/** Gets encoder's configured use of discontinuous transmission. + * @see OPUS_SET_DTX + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
0
DTX disabled (default).
+ *
1
DTX enabled.
+ *
+ * @hideinitializer */ +#define OPUS_GET_DTX(x) OPUS_GET_DTX_REQUEST, __opus_check_int_ptr(x) +/** Configures the depth of signal being encoded. + * + * This is a hint which helps the encoder identify silence and near-silence. + * It represents the number of significant bits of linear intensity below + * which the signal contains ignorable quantization or other noise. + * + * For example, OPUS_SET_LSB_DEPTH(14) would be an appropriate setting + * for G.711 u-law input. OPUS_SET_LSB_DEPTH(16) would be appropriate + * for 16-bit linear pcm input with opus_encode_float(). + * + * When using opus_encode() instead of opus_encode_float(), or when libopus + * is compiled for fixed-point, the encoder uses the minimum of the value + * set here and the value 16. + * + * @see OPUS_GET_LSB_DEPTH + * @param[in] x opus_int32: Input precision in bits, between 8 and 24 + * (default: 24). + * @hideinitializer */ +#define OPUS_SET_LSB_DEPTH(x) OPUS_SET_LSB_DEPTH_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured signal depth. + * @see OPUS_SET_LSB_DEPTH + * @param[out] x opus_int32 *: Input precision in bits, between 8 and + * 24 (default: 24). + * @hideinitializer */ +#define OPUS_GET_LSB_DEPTH(x) OPUS_GET_LSB_DEPTH_REQUEST, __opus_check_int_ptr(x) + +/** Configures the encoder's use of variable duration frames. + * When variable duration is enabled, the encoder is free to use a shorter frame + * size than the one requested in the opus_encode*() call. + * It is then the user's responsibility + * to verify how much audio was encoded by checking the ToC byte of the encoded + * packet. The part of the audio that was not encoded needs to be resent to the + * encoder for the next call. Do not use this option unless you really + * know what you are doing. + * @see OPUS_GET_EXPERT_FRAME_DURATION + * @param[in] x opus_int32: Allowed values: + *
+ *
OPUS_FRAMESIZE_ARG
Select frame size from the argument (default).
+ *
OPUS_FRAMESIZE_2_5_MS
Use 2.5 ms frames.
+ *
OPUS_FRAMESIZE_5_MS
Use 5 ms frames.
+ *
OPUS_FRAMESIZE_10_MS
Use 10 ms frames.
+ *
OPUS_FRAMESIZE_20_MS
Use 20 ms frames.
+ *
OPUS_FRAMESIZE_40_MS
Use 40 ms frames.
+ *
OPUS_FRAMESIZE_60_MS
Use 60 ms frames.
+ *
OPUS_FRAMESIZE_80_MS
Use 80 ms frames.
+ *
OPUS_FRAMESIZE_100_MS
Use 100 ms frames.
+ *
OPUS_FRAMESIZE_120_MS
Use 120 ms frames.
+ *
+ * @hideinitializer */ +#define OPUS_SET_EXPERT_FRAME_DURATION(x) OPUS_SET_EXPERT_FRAME_DURATION_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured use of variable duration frames. + * @see OPUS_SET_EXPERT_FRAME_DURATION + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
OPUS_FRAMESIZE_ARG
Select frame size from the argument (default).
+ *
OPUS_FRAMESIZE_2_5_MS
Use 2.5 ms frames.
+ *
OPUS_FRAMESIZE_5_MS
Use 5 ms frames.
+ *
OPUS_FRAMESIZE_10_MS
Use 10 ms frames.
+ *
OPUS_FRAMESIZE_20_MS
Use 20 ms frames.
+ *
OPUS_FRAMESIZE_40_MS
Use 40 ms frames.
+ *
OPUS_FRAMESIZE_60_MS
Use 60 ms frames.
+ *
OPUS_FRAMESIZE_80_MS
Use 80 ms frames.
+ *
OPUS_FRAMESIZE_100_MS
Use 100 ms frames.
+ *
OPUS_FRAMESIZE_120_MS
Use 120 ms frames.
+ *
+ * @hideinitializer */ +#define OPUS_GET_EXPERT_FRAME_DURATION(x) OPUS_GET_EXPERT_FRAME_DURATION_REQUEST, __opus_check_int_ptr(x) + +/** If set to 1, disables almost all use of prediction, making frames almost + * completely independent. This reduces quality. + * @see OPUS_GET_PREDICTION_DISABLED + * @param[in] x opus_int32: Allowed values: + *
+ *
0
Enable prediction (default).
+ *
1
Disable prediction.
+ *
+ * @hideinitializer */ +#define OPUS_SET_PREDICTION_DISABLED(x) OPUS_SET_PREDICTION_DISABLED_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured prediction status. + * @see OPUS_SET_PREDICTION_DISABLED + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
0
Prediction enabled (default).
+ *
1
Prediction disabled.
+ *
+ * @hideinitializer */ +#define OPUS_GET_PREDICTION_DISABLED(x) OPUS_GET_PREDICTION_DISABLED_REQUEST, __opus_check_int_ptr(x) + +/**@}*/ + +/** @defgroup opus_genericctls Generic CTLs + * + * These macros are used with the \c opus_decoder_ctl and + * \c opus_encoder_ctl calls to generate a particular + * request. + * + * When called on an \c OpusDecoder they apply to that + * particular decoder instance. When called on an + * \c OpusEncoder they apply to the corresponding setting + * on that encoder instance, if present. + * + * Some usage examples: + * + * @code + * int ret; + * opus_int32 pitch; + * ret = opus_decoder_ctl(dec_ctx, OPUS_GET_PITCH(&pitch)); + * if (ret == OPUS_OK) return ret; + * + * opus_encoder_ctl(enc_ctx, OPUS_RESET_STATE); + * opus_decoder_ctl(dec_ctx, OPUS_RESET_STATE); + * + * opus_int32 enc_bw, dec_bw; + * opus_encoder_ctl(enc_ctx, OPUS_GET_BANDWIDTH(&enc_bw)); + * opus_decoder_ctl(dec_ctx, OPUS_GET_BANDWIDTH(&dec_bw)); + * if (enc_bw != dec_bw) { + * printf("packet bandwidth mismatch!\n"); + * } + * @endcode + * + * @see opus_encoder, opus_decoder_ctl, opus_encoder_ctl, opus_decoderctls, opus_encoderctls + * @{ + */ + +/** Resets the codec state to be equivalent to a freshly initialized state. + * This should be called when switching streams in order to prevent + * the back to back decoding from giving different results from + * one at a time decoding. + * @hideinitializer */ +#define OPUS_RESET_STATE 4028 + +/** Gets the final state of the codec's entropy coder. + * This is used for testing purposes, + * The encoder and decoder state should be identical after coding a payload + * (assuming no data corruption or software bugs) + * + * @param[out] x opus_uint32 *: Entropy coder state + * + * @hideinitializer */ +#define OPUS_GET_FINAL_RANGE(x) OPUS_GET_FINAL_RANGE_REQUEST, __opus_check_uint_ptr(x) + +/** Gets the encoder's configured bandpass or the decoder's last bandpass. + * @see OPUS_SET_BANDWIDTH + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
#OPUS_AUTO
(default)
+ *
#OPUS_BANDWIDTH_NARROWBAND
4 kHz passband
+ *
#OPUS_BANDWIDTH_MEDIUMBAND
6 kHz passband
+ *
#OPUS_BANDWIDTH_WIDEBAND
8 kHz passband
+ *
#OPUS_BANDWIDTH_SUPERWIDEBAND
12 kHz passband
+ *
#OPUS_BANDWIDTH_FULLBAND
20 kHz passband
+ *
+ * @hideinitializer */ +#define OPUS_GET_BANDWIDTH(x) OPUS_GET_BANDWIDTH_REQUEST, __opus_check_int_ptr(x) + +/** Gets the sampling rate the encoder or decoder was initialized with. + * This simply returns the Fs value passed to opus_encoder_init() + * or opus_decoder_init(). + * @param[out] x opus_int32 *: Sampling rate of encoder or decoder. + * @hideinitializer + */ +#define OPUS_GET_SAMPLE_RATE(x) OPUS_GET_SAMPLE_RATE_REQUEST, __opus_check_int_ptr(x) + +/** If set to 1, disables the use of phase inversion for intensity stereo, + * improving the quality of mono downmixes, but slightly reducing normal + * stereo quality. Disabling phase inversion in the decoder does not comply + * with RFC 6716, although it does not cause any interoperability issue and + * is expected to become part of the Opus standard once RFC 6716 is updated + * by draft-ietf-codec-opus-update. + * @see OPUS_GET_PHASE_INVERSION_DISABLED + * @param[in] x opus_int32: Allowed values: + *
+ *
0
Enable phase inversion (default).
+ *
1
Disable phase inversion.
+ *
+ * @hideinitializer */ +#define OPUS_SET_PHASE_INVERSION_DISABLED(x) OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured phase inversion status. + * @see OPUS_SET_PHASE_INVERSION_DISABLED + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
0
Stereo phase inversion enabled (default).
+ *
1
Stereo phase inversion disabled.
+ *
+ * @hideinitializer */ +#define OPUS_GET_PHASE_INVERSION_DISABLED(x) OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST, __opus_check_int_ptr(x) +/** Gets the DTX state of the encoder. + * Returns whether the last encoded frame was either a comfort noise update + * during DTX or not encoded because of DTX. + * @param[out] x opus_int32 *: Returns one of the following values: + *
+ *
0
The encoder is not in DTX.
+ *
1
The encoder is in DTX.
+ *
+ * @hideinitializer */ +#define OPUS_GET_IN_DTX(x) OPUS_GET_IN_DTX_REQUEST, __opus_check_int_ptr(x) + +/**@}*/ + +/** @defgroup opus_decoderctls Decoder related CTLs + * @see opus_genericctls, opus_encoderctls, opus_decoder + * @{ + */ + +/** Configures decoder gain adjustment. + * Scales the decoded output by a factor specified in Q8 dB units. + * This has a maximum range of -32768 to 32767 inclusive, and returns + * OPUS_BAD_ARG otherwise. The default is zero indicating no adjustment. + * This setting survives decoder reset. + * + * gain = pow(10, x/(20.0*256)) + * + * @param[in] x opus_int32: Amount to scale PCM signal by in Q8 dB units. + * @hideinitializer */ +#define OPUS_SET_GAIN(x) OPUS_SET_GAIN_REQUEST, __opus_check_int(x) +/** Gets the decoder's configured gain adjustment. @see OPUS_SET_GAIN + * + * @param[out] x opus_int32 *: Amount to scale PCM signal by in Q8 dB units. + * @hideinitializer */ +#define OPUS_GET_GAIN(x) OPUS_GET_GAIN_REQUEST, __opus_check_int_ptr(x) + +/** Gets the duration (in samples) of the last packet successfully decoded or concealed. + * @param[out] x opus_int32 *: Number of samples (at current sampling rate). + * @hideinitializer */ +#define OPUS_GET_LAST_PACKET_DURATION(x) OPUS_GET_LAST_PACKET_DURATION_REQUEST, __opus_check_int_ptr(x) + +/** Gets the pitch of the last decoded frame, if available. + * This can be used for any post-processing algorithm requiring the use of pitch, + * e.g. time stretching/shortening. If the last frame was not voiced, or if the + * pitch was not coded in the frame, then zero is returned. + * + * This CTL is only implemented for decoder instances. + * + * @param[out] x opus_int32 *: pitch period at 48 kHz (or 0 if not available) + * + * @hideinitializer */ +#define OPUS_GET_PITCH(x) OPUS_GET_PITCH_REQUEST, __opus_check_int_ptr(x) + +/**@}*/ + +/** @defgroup opus_libinfo Opus library information functions + * @{ + */ + +/** Converts an opus error code into a human readable string. + * + * @param[in] error int: Error number + * @returns Error string + */ +OPUS_EXPORT const char *opus_strerror(int error); + +/** Gets the libopus version string. + * + * Applications may look for the substring "-fixed" in the version string to + * determine whether they have a fixed-point or floating-point build at + * runtime. + * + * @returns Version string + */ +OPUS_EXPORT const char *opus_get_version_string(void); +/**@}*/ + +#ifdef __cplusplus +} +#endif + +#endif /* OPUS_DEFINES_H */ diff --git a/lib/opus/include/opus_multistream.h b/lib/opus/include/opus_multistream.h new file mode 100644 index 00000000..babcee69 --- /dev/null +++ b/lib/opus/include/opus_multistream.h @@ -0,0 +1,660 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/** + * @file opus_multistream.h + * @brief Opus reference implementation multistream API + */ + +#ifndef OPUS_MULTISTREAM_H +#define OPUS_MULTISTREAM_H + +#include "opus.h" + +#ifdef __cplusplus +extern "C" { +#endif + +/** @cond OPUS_INTERNAL_DOC */ + +/** Macros to trigger compilation errors when the wrong types are provided to a + * CTL. */ +/**@{*/ +#define __opus_check_encstate_ptr(ptr) ((ptr) + ((ptr) - (OpusEncoder**)(ptr))) +#define __opus_check_decstate_ptr(ptr) ((ptr) + ((ptr) - (OpusDecoder**)(ptr))) +/**@}*/ + +/** These are the actual encoder and decoder CTL ID numbers. + * They should not be used directly by applications. + * In general, SETs should be even and GETs should be odd.*/ +/**@{*/ +#define OPUS_MULTISTREAM_GET_ENCODER_STATE_REQUEST 5120 +#define OPUS_MULTISTREAM_GET_DECODER_STATE_REQUEST 5122 +/**@}*/ + +/** @endcond */ + +/** @defgroup opus_multistream_ctls Multistream specific encoder and decoder CTLs + * + * These are convenience macros that are specific to the + * opus_multistream_encoder_ctl() and opus_multistream_decoder_ctl() + * interface. + * The CTLs from @ref opus_genericctls, @ref opus_encoderctls, and + * @ref opus_decoderctls may be applied to a multistream encoder or decoder as + * well. + * In addition, you may retrieve the encoder or decoder state for an specific + * stream via #OPUS_MULTISTREAM_GET_ENCODER_STATE or + * #OPUS_MULTISTREAM_GET_DECODER_STATE and apply CTLs to it individually. + */ +/**@{*/ + +/** Gets the encoder state for an individual stream of a multistream encoder. + * @param[in] x opus_int32: The index of the stream whose encoder you + * wish to retrieve. + * This must be non-negative and less than + * the streams parameter used + * to initialize the encoder. + * @param[out] y OpusEncoder**: Returns a pointer to the given + * encoder state. + * @retval OPUS_BAD_ARG The index of the requested stream was out of range. + * @hideinitializer + */ +#define OPUS_MULTISTREAM_GET_ENCODER_STATE(x,y) OPUS_MULTISTREAM_GET_ENCODER_STATE_REQUEST, __opus_check_int(x), __opus_check_encstate_ptr(y) + +/** Gets the decoder state for an individual stream of a multistream decoder. + * @param[in] x opus_int32: The index of the stream whose decoder you + * wish to retrieve. + * This must be non-negative and less than + * the streams parameter used + * to initialize the decoder. + * @param[out] y OpusDecoder**: Returns a pointer to the given + * decoder state. + * @retval OPUS_BAD_ARG The index of the requested stream was out of range. + * @hideinitializer + */ +#define OPUS_MULTISTREAM_GET_DECODER_STATE(x,y) OPUS_MULTISTREAM_GET_DECODER_STATE_REQUEST, __opus_check_int(x), __opus_check_decstate_ptr(y) + +/**@}*/ + +/** @defgroup opus_multistream Opus Multistream API + * @{ + * + * The multistream API allows individual Opus streams to be combined into a + * single packet, enabling support for up to 255 channels. Unlike an + * elementary Opus stream, the encoder and decoder must negotiate the channel + * configuration before the decoder can successfully interpret the data in the + * packets produced by the encoder. Some basic information, such as packet + * duration, can be computed without any special negotiation. + * + * The format for multistream Opus packets is defined in + * RFC 7845 + * and is based on the self-delimited Opus framing described in Appendix B of + * RFC 6716. + * Normal Opus packets are just a degenerate case of multistream Opus packets, + * and can be encoded or decoded with the multistream API by setting + * streams to 1 when initializing the encoder or + * decoder. + * + * Multistream Opus streams can contain up to 255 elementary Opus streams. + * These may be either "uncoupled" or "coupled", indicating that the decoder + * is configured to decode them to either 1 or 2 channels, respectively. + * The streams are ordered so that all coupled streams appear at the + * beginning. + * + * A mapping table defines which decoded channel i + * should be used for each input/output (I/O) channel j. This table is + * typically provided as an unsigned char array. + * Let i = mapping[j] be the index for I/O channel j. + * If i < 2*coupled_streams, then I/O channel j is + * encoded as the left channel of stream (i/2) if i + * is even, or as the right channel of stream (i/2) if + * i is odd. Otherwise, I/O channel j is encoded as + * mono in stream (i - coupled_streams), unless it has the special + * value 255, in which case it is omitted from the encoding entirely (the + * decoder will reproduce it as silence). Each value i must either + * be the special value 255 or be less than streams + coupled_streams. + * + * The output channels specified by the encoder + * should use the + * Vorbis + * channel ordering. A decoder may wish to apply an additional permutation + * to the mapping the encoder used to achieve a different output channel + * order (e.g. for outputing in WAV order). + * + * Each multistream packet contains an Opus packet for each stream, and all of + * the Opus packets in a single multistream packet must have the same + * duration. Therefore the duration of a multistream packet can be extracted + * from the TOC sequence of the first stream, which is located at the + * beginning of the packet, just like an elementary Opus stream: + * + * @code + * int nb_samples; + * int nb_frames; + * nb_frames = opus_packet_get_nb_frames(data, len); + * if (nb_frames < 1) + * return nb_frames; + * nb_samples = opus_packet_get_samples_per_frame(data, 48000) * nb_frames; + * @endcode + * + * The general encoding and decoding process proceeds exactly the same as in + * the normal @ref opus_encoder and @ref opus_decoder APIs. + * See their documentation for an overview of how to use the corresponding + * multistream functions. + */ + +/** Opus multistream encoder state. + * This contains the complete state of a multistream Opus encoder. + * It is position independent and can be freely copied. + * @see opus_multistream_encoder_create + * @see opus_multistream_encoder_init + */ +typedef struct OpusMSEncoder OpusMSEncoder; + +/** Opus multistream decoder state. + * This contains the complete state of a multistream Opus decoder. + * It is position independent and can be freely copied. + * @see opus_multistream_decoder_create + * @see opus_multistream_decoder_init + */ +typedef struct OpusMSDecoder OpusMSDecoder; + +/**\name Multistream encoder functions */ +/**@{*/ + +/** Gets the size of an OpusMSEncoder structure. + * @param streams int: The total number of streams to encode from the + * input. + * This must be no more than 255. + * @param coupled_streams int: Number of coupled (2 channel) streams + * to encode. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * encoded channels (streams + + * coupled_streams) must be no + * more than 255. + * @returns The size in bytes on success, or a negative error code + * (see @ref opus_errorcodes) on error. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_multistream_encoder_get_size( + int streams, + int coupled_streams +); + +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_multistream_surround_encoder_get_size( + int channels, + int mapping_family +); + + +/** Allocates and initializes a multistream encoder state. + * Call opus_multistream_encoder_destroy() to release + * this object when finished. + * @param Fs opus_int32: Sampling rate of the input signal (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels in the input signal. + * This must be at most 255. + * It may be greater than the number of + * coded channels (streams + + * coupled_streams). + * @param streams int: The total number of streams to encode from the + * input. + * This must be no more than the number of channels. + * @param coupled_streams int: Number of coupled (2 channel) streams + * to encode. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * encoded channels (streams + + * coupled_streams) must be no + * more than the number of input channels. + * @param[in] mapping const unsigned char[channels]: Mapping from + * encoded channels to input channels, as described in + * @ref opus_multistream. As an extra constraint, the + * multistream encoder does not allow encoding coupled + * streams for which one channel is unused since this + * is never a good idea. + * @param application int: The target encoder application. + * This must be one of the following: + *
+ *
#OPUS_APPLICATION_VOIP
+ *
Process signal for improved speech intelligibility.
+ *
#OPUS_APPLICATION_AUDIO
+ *
Favor faithfulness to the original input.
+ *
#OPUS_APPLICATION_RESTRICTED_LOWDELAY
+ *
Configure the minimum possible coding delay by disabling certain modes + * of operation.
+ *
+ * @param[out] error int *: Returns #OPUS_OK on success, or an error + * code (see @ref opus_errorcodes) on + * failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusMSEncoder *opus_multistream_encoder_create( + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application, + int *error +) OPUS_ARG_NONNULL(5); + +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusMSEncoder *opus_multistream_surround_encoder_create( + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application, + int *error +) OPUS_ARG_NONNULL(4) OPUS_ARG_NONNULL(5) OPUS_ARG_NONNULL(6); + +/** Initialize a previously allocated multistream encoder state. + * The memory pointed to by \a st must be at least the size returned by + * opus_multistream_encoder_get_size(). + * This is intended for applications which use their own allocator instead of + * malloc. + * To reset a previously initialized state, use the #OPUS_RESET_STATE CTL. + * @see opus_multistream_encoder_create + * @see opus_multistream_encoder_get_size + * @param st OpusMSEncoder*: Multistream encoder state to initialize. + * @param Fs opus_int32: Sampling rate of the input signal (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels in the input signal. + * This must be at most 255. + * It may be greater than the number of + * coded channels (streams + + * coupled_streams). + * @param streams int: The total number of streams to encode from the + * input. + * This must be no more than the number of channels. + * @param coupled_streams int: Number of coupled (2 channel) streams + * to encode. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * encoded channels (streams + + * coupled_streams) must be no + * more than the number of input channels. + * @param[in] mapping const unsigned char[channels]: Mapping from + * encoded channels to input channels, as described in + * @ref opus_multistream. As an extra constraint, the + * multistream encoder does not allow encoding coupled + * streams for which one channel is unused since this + * is never a good idea. + * @param application int: The target encoder application. + * This must be one of the following: + *
+ *
#OPUS_APPLICATION_VOIP
+ *
Process signal for improved speech intelligibility.
+ *
#OPUS_APPLICATION_AUDIO
+ *
Favor faithfulness to the original input.
+ *
#OPUS_APPLICATION_RESTRICTED_LOWDELAY
+ *
Configure the minimum possible coding delay by disabling certain modes + * of operation.
+ *
+ * @returns #OPUS_OK on success, or an error code (see @ref opus_errorcodes) + * on failure. + */ +OPUS_EXPORT int opus_multistream_encoder_init( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(6); + +OPUS_EXPORT int opus_multistream_surround_encoder_init( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(5) OPUS_ARG_NONNULL(6) OPUS_ARG_NONNULL(7); + +/** Encodes a multistream Opus frame. + * @param st OpusMSEncoder*: Multistream encoder state. + * @param[in] pcm const opus_int16*: The input signal as interleaved + * samples. + * This must contain + * frame_size*channels + * samples. + * @param frame_size int: Number of samples per channel in the input + * signal. + * This must be an Opus frame size for the + * encoder's sampling rate. + * For example, at 48 kHz the permitted values + * are 120, 240, 480, 960, 1920, and 2880. + * Passing in a duration of less than 10 ms + * (480 samples at 48 kHz) will prevent the + * encoder from using the LPC or hybrid modes. + * @param[out] data unsigned char*: Output payload. + * This must contain storage for at + * least \a max_data_bytes. + * @param [in] max_data_bytes opus_int32: Size of the allocated + * memory for the output + * payload. This may be + * used to impose an upper limit on + * the instant bitrate, but should + * not be used as the only bitrate + * control. Use #OPUS_SET_BITRATE to + * control the bitrate. + * @returns The length of the encoded packet (in bytes) on success or a + * negative error code (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_multistream_encode( + OpusMSEncoder *st, + const opus_int16 *pcm, + int frame_size, + unsigned char *data, + opus_int32 max_data_bytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + +/** Encodes a multistream Opus frame from floating point input. + * @param st OpusMSEncoder*: Multistream encoder state. + * @param[in] pcm const float*: The input signal as interleaved + * samples with a normal range of + * +/-1.0. + * Samples with a range beyond +/-1.0 + * are supported but will be clipped by + * decoders using the integer API and + * should only be used if it is known + * that the far end supports extended + * dynamic range. + * This must contain + * frame_size*channels + * samples. + * @param frame_size int: Number of samples per channel in the input + * signal. + * This must be an Opus frame size for the + * encoder's sampling rate. + * For example, at 48 kHz the permitted values + * are 120, 240, 480, 960, 1920, and 2880. + * Passing in a duration of less than 10 ms + * (480 samples at 48 kHz) will prevent the + * encoder from using the LPC or hybrid modes. + * @param[out] data unsigned char*: Output payload. + * This must contain storage for at + * least \a max_data_bytes. + * @param [in] max_data_bytes opus_int32: Size of the allocated + * memory for the output + * payload. This may be + * used to impose an upper limit on + * the instant bitrate, but should + * not be used as the only bitrate + * control. Use #OPUS_SET_BITRATE to + * control the bitrate. + * @returns The length of the encoded packet (in bytes) on success or a + * negative error code (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_multistream_encode_float( + OpusMSEncoder *st, + const float *pcm, + int frame_size, + unsigned char *data, + opus_int32 max_data_bytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + +/** Frees an OpusMSEncoder allocated by + * opus_multistream_encoder_create(). + * @param st OpusMSEncoder*: Multistream encoder state to be freed. + */ +OPUS_EXPORT void opus_multistream_encoder_destroy(OpusMSEncoder *st); + +/** Perform a CTL function on a multistream Opus encoder. + * + * Generally the request and subsequent arguments are generated by a + * convenience macro. + * @param st OpusMSEncoder*: Multistream encoder state. + * @param request This and all remaining parameters should be replaced by one + * of the convenience macros in @ref opus_genericctls, + * @ref opus_encoderctls, or @ref opus_multistream_ctls. + * @see opus_genericctls + * @see opus_encoderctls + * @see opus_multistream_ctls + */ +OPUS_EXPORT int opus_multistream_encoder_ctl(OpusMSEncoder *st, int request, ...) OPUS_ARG_NONNULL(1); + +/**@}*/ + +/**\name Multistream decoder functions */ +/**@{*/ + +/** Gets the size of an OpusMSDecoder structure. + * @param streams int: The total number of streams coded in the + * input. + * This must be no more than 255. + * @param coupled_streams int: Number streams to decode as coupled + * (2 channel) streams. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * coded channels (streams + + * coupled_streams) must be no + * more than 255. + * @returns The size in bytes on success, or a negative error code + * (see @ref opus_errorcodes) on error. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_multistream_decoder_get_size( + int streams, + int coupled_streams +); + +/** Allocates and initializes a multistream decoder state. + * Call opus_multistream_decoder_destroy() to release + * this object when finished. + * @param Fs opus_int32: Sampling rate to decode at (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels to output. + * This must be at most 255. + * It may be different from the number of coded + * channels (streams + + * coupled_streams). + * @param streams int: The total number of streams coded in the + * input. + * This must be no more than 255. + * @param coupled_streams int: Number of streams to decode as coupled + * (2 channel) streams. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * coded channels (streams + + * coupled_streams) must be no + * more than 255. + * @param[in] mapping const unsigned char[channels]: Mapping from + * coded channels to output channels, as described in + * @ref opus_multistream. + * @param[out] error int *: Returns #OPUS_OK on success, or an error + * code (see @ref opus_errorcodes) on + * failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusMSDecoder *opus_multistream_decoder_create( + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int *error +) OPUS_ARG_NONNULL(5); + +/** Intialize a previously allocated decoder state object. + * The memory pointed to by \a st must be at least the size returned by + * opus_multistream_encoder_get_size(). + * This is intended for applications which use their own allocator instead of + * malloc. + * To reset a previously initialized state, use the #OPUS_RESET_STATE CTL. + * @see opus_multistream_decoder_create + * @see opus_multistream_deocder_get_size + * @param st OpusMSEncoder*: Multistream encoder state to initialize. + * @param Fs opus_int32: Sampling rate to decode at (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels to output. + * This must be at most 255. + * It may be different from the number of coded + * channels (streams + + * coupled_streams). + * @param streams int: The total number of streams coded in the + * input. + * This must be no more than 255. + * @param coupled_streams int: Number of streams to decode as coupled + * (2 channel) streams. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * coded channels (streams + + * coupled_streams) must be no + * more than 255. + * @param[in] mapping const unsigned char[channels]: Mapping from + * coded channels to output channels, as described in + * @ref opus_multistream. + * @returns #OPUS_OK on success, or an error code (see @ref opus_errorcodes) + * on failure. + */ +OPUS_EXPORT int opus_multistream_decoder_init( + OpusMSDecoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(6); + +/** Decode a multistream Opus packet. + * @param st OpusMSDecoder*: Multistream decoder state. + * @param[in] data const unsigned char*: Input payload. + * Use a NULL + * pointer to indicate packet + * loss. + * @param len opus_int32: Number of bytes in payload. + * @param[out] pcm opus_int16*: Output signal, with interleaved + * samples. + * This must contain room for + * frame_size*channels + * samples. + * @param frame_size int: The number of samples per channel of + * available space in \a pcm. + * If this is less than the maximum packet duration + * (120 ms; 5760 for 48kHz), this function will not be capable + * of decoding some packets. In the case of PLC (data==NULL) + * or FEC (decode_fec=1), then frame_size needs to be exactly + * the duration of audio that is missing, otherwise the + * decoder will not be in the optimal state to decode the + * next incoming packet. For the PLC and FEC cases, frame_size + * must be a multiple of 2.5 ms. + * @param decode_fec int: Flag (0 or 1) to request that any in-band + * forward error correction data be decoded. + * If no such data is available, the frame is + * decoded as if it were lost. + * @returns Number of samples decoded on success or a negative error code + * (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_multistream_decode( + OpusMSDecoder *st, + const unsigned char *data, + opus_int32 len, + opus_int16 *pcm, + int frame_size, + int decode_fec +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + +/** Decode a multistream Opus packet with floating point output. + * @param st OpusMSDecoder*: Multistream decoder state. + * @param[in] data const unsigned char*: Input payload. + * Use a NULL + * pointer to indicate packet + * loss. + * @param len opus_int32: Number of bytes in payload. + * @param[out] pcm opus_int16*: Output signal, with interleaved + * samples. + * This must contain room for + * frame_size*channels + * samples. + * @param frame_size int: The number of samples per channel of + * available space in \a pcm. + * If this is less than the maximum packet duration + * (120 ms; 5760 for 48kHz), this function will not be capable + * of decoding some packets. In the case of PLC (data==NULL) + * or FEC (decode_fec=1), then frame_size needs to be exactly + * the duration of audio that is missing, otherwise the + * decoder will not be in the optimal state to decode the + * next incoming packet. For the PLC and FEC cases, frame_size + * must be a multiple of 2.5 ms. + * @param decode_fec int: Flag (0 or 1) to request that any in-band + * forward error correction data be decoded. + * If no such data is available, the frame is + * decoded as if it were lost. + * @returns Number of samples decoded on success or a negative error code + * (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_multistream_decode_float( + OpusMSDecoder *st, + const unsigned char *data, + opus_int32 len, + float *pcm, + int frame_size, + int decode_fec +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + +/** Perform a CTL function on a multistream Opus decoder. + * + * Generally the request and subsequent arguments are generated by a + * convenience macro. + * @param st OpusMSDecoder*: Multistream decoder state. + * @param request This and all remaining parameters should be replaced by one + * of the convenience macros in @ref opus_genericctls, + * @ref opus_decoderctls, or @ref opus_multistream_ctls. + * @see opus_genericctls + * @see opus_decoderctls + * @see opus_multistream_ctls + */ +OPUS_EXPORT int opus_multistream_decoder_ctl(OpusMSDecoder *st, int request, ...) OPUS_ARG_NONNULL(1); + +/** Frees an OpusMSDecoder allocated by + * opus_multistream_decoder_create(). + * @param st OpusMSDecoder: Multistream decoder state to be freed. + */ +OPUS_EXPORT void opus_multistream_decoder_destroy(OpusMSDecoder *st); + +/**@}*/ + +/**@}*/ + +#ifdef __cplusplus +} +#endif + +#endif /* OPUS_MULTISTREAM_H */ diff --git a/lib/opus/include/opus_projection.h b/lib/opus/include/opus_projection.h new file mode 100644 index 00000000..9dabf4e8 --- /dev/null +++ b/lib/opus/include/opus_projection.h @@ -0,0 +1,568 @@ +/* Copyright (c) 2017 Google Inc. + Written by Andrew Allen */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/** + * @file opus_projection.h + * @brief Opus projection reference API + */ + +#ifndef OPUS_PROJECTION_H +#define OPUS_PROJECTION_H + +#include "opus_multistream.h" + +#ifdef __cplusplus +extern "C" { +#endif + +/** @cond OPUS_INTERNAL_DOC */ + +/** These are the actual encoder and decoder CTL ID numbers. + * They should not be used directly by applications.c + * In general, SETs should be even and GETs should be odd.*/ +/**@{*/ +#define OPUS_PROJECTION_GET_DEMIXING_MATRIX_GAIN_REQUEST 6001 +#define OPUS_PROJECTION_GET_DEMIXING_MATRIX_SIZE_REQUEST 6003 +#define OPUS_PROJECTION_GET_DEMIXING_MATRIX_REQUEST 6005 +/**@}*/ + + +/** @endcond */ + +/** @defgroup opus_projection_ctls Projection specific encoder and decoder CTLs + * + * These are convenience macros that are specific to the + * opus_projection_encoder_ctl() and opus_projection_decoder_ctl() + * interface. + * The CTLs from @ref opus_genericctls, @ref opus_encoderctls, + * @ref opus_decoderctls, and @ref opus_multistream_ctls may be applied to a + * projection encoder or decoder as well. + */ +/**@{*/ + +/** Gets the gain (in dB. S7.8-format) of the demixing matrix from the encoder. + * @param[out] x opus_int32 *: Returns the gain (in dB. S7.8-format) + * of the demixing matrix. + * @hideinitializer + */ +#define OPUS_PROJECTION_GET_DEMIXING_MATRIX_GAIN(x) OPUS_PROJECTION_GET_DEMIXING_MATRIX_GAIN_REQUEST, __opus_check_int_ptr(x) + + +/** Gets the size in bytes of the demixing matrix from the encoder. + * @param[out] x opus_int32 *: Returns the size in bytes of the + * demixing matrix. + * @hideinitializer + */ +#define OPUS_PROJECTION_GET_DEMIXING_MATRIX_SIZE(x) OPUS_PROJECTION_GET_DEMIXING_MATRIX_SIZE_REQUEST, __opus_check_int_ptr(x) + + +/** Copies the demixing matrix to the supplied pointer location. + * @param[out] x unsigned char *: Returns the demixing matrix to the + * supplied pointer location. + * @param y opus_int32: The size in bytes of the reserved memory at the + * pointer location. + * @hideinitializer + */ +#define OPUS_PROJECTION_GET_DEMIXING_MATRIX(x,y) OPUS_PROJECTION_GET_DEMIXING_MATRIX_REQUEST, x, __opus_check_int(y) + + +/**@}*/ + +/** Opus projection encoder state. + * This contains the complete state of a projection Opus encoder. + * It is position independent and can be freely copied. + * @see opus_projection_ambisonics_encoder_create + */ +typedef struct OpusProjectionEncoder OpusProjectionEncoder; + + +/** Opus projection decoder state. + * This contains the complete state of a projection Opus decoder. + * It is position independent and can be freely copied. + * @see opus_projection_decoder_create + * @see opus_projection_decoder_init + */ +typedef struct OpusProjectionDecoder OpusProjectionDecoder; + + +/**\name Projection encoder functions */ +/**@{*/ + +/** Gets the size of an OpusProjectionEncoder structure. + * @param channels int: The total number of input channels to encode. + * This must be no more than 255. + * @param mapping_family int: The mapping family to use for selecting + * the appropriate projection. + * @returns The size in bytes on success, or a negative error code + * (see @ref opus_errorcodes) on error. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_projection_ambisonics_encoder_get_size( + int channels, + int mapping_family +); + + +/** Allocates and initializes a projection encoder state. + * Call opus_projection_encoder_destroy() to release + * this object when finished. + * @param Fs opus_int32: Sampling rate of the input signal (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels in the input signal. + * This must be at most 255. + * It may be greater than the number of + * coded channels (streams + + * coupled_streams). + * @param mapping_family int: The mapping family to use for selecting + * the appropriate projection. + * @param[out] streams int *: The total number of streams that will + * be encoded from the input. + * @param[out] coupled_streams int *: Number of coupled (2 channel) + * streams that will be encoded from the input. + * @param application int: The target encoder application. + * This must be one of the following: + *
+ *
#OPUS_APPLICATION_VOIP
+ *
Process signal for improved speech intelligibility.
+ *
#OPUS_APPLICATION_AUDIO
+ *
Favor faithfulness to the original input.
+ *
#OPUS_APPLICATION_RESTRICTED_LOWDELAY
+ *
Configure the minimum possible coding delay by disabling certain modes + * of operation.
+ *
+ * @param[out] error int *: Returns #OPUS_OK on success, or an error + * code (see @ref opus_errorcodes) on + * failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusProjectionEncoder *opus_projection_ambisonics_encoder_create( + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + int application, + int *error +) OPUS_ARG_NONNULL(4) OPUS_ARG_NONNULL(5); + + +/** Initialize a previously allocated projection encoder state. + * The memory pointed to by \a st must be at least the size returned by + * opus_projection_ambisonics_encoder_get_size(). + * This is intended for applications which use their own allocator instead of + * malloc. + * To reset a previously initialized state, use the #OPUS_RESET_STATE CTL. + * @see opus_projection_ambisonics_encoder_create + * @see opus_projection_ambisonics_encoder_get_size + * @param st OpusProjectionEncoder*: Projection encoder state to initialize. + * @param Fs opus_int32: Sampling rate of the input signal (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels in the input signal. + * This must be at most 255. + * It may be greater than the number of + * coded channels (streams + + * coupled_streams). + * @param streams int: The total number of streams to encode from the + * input. + * This must be no more than the number of channels. + * @param coupled_streams int: Number of coupled (2 channel) streams + * to encode. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * encoded channels (streams + + * coupled_streams) must be no + * more than the number of input channels. + * @param application int: The target encoder application. + * This must be one of the following: + *
+ *
#OPUS_APPLICATION_VOIP
+ *
Process signal for improved speech intelligibility.
+ *
#OPUS_APPLICATION_AUDIO
+ *
Favor faithfulness to the original input.
+ *
#OPUS_APPLICATION_RESTRICTED_LOWDELAY
+ *
Configure the minimum possible coding delay by disabling certain modes + * of operation.
+ *
+ * @returns #OPUS_OK on success, or an error code (see @ref opus_errorcodes) + * on failure. + */ +OPUS_EXPORT int opus_projection_ambisonics_encoder_init( + OpusProjectionEncoder *st, + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + int application +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(5) OPUS_ARG_NONNULL(6); + + +/** Encodes a projection Opus frame. + * @param st OpusProjectionEncoder*: Projection encoder state. + * @param[in] pcm const opus_int16*: The input signal as interleaved + * samples. + * This must contain + * frame_size*channels + * samples. + * @param frame_size int: Number of samples per channel in the input + * signal. + * This must be an Opus frame size for the + * encoder's sampling rate. + * For example, at 48 kHz the permitted values + * are 120, 240, 480, 960, 1920, and 2880. + * Passing in a duration of less than 10 ms + * (480 samples at 48 kHz) will prevent the + * encoder from using the LPC or hybrid modes. + * @param[out] data unsigned char*: Output payload. + * This must contain storage for at + * least \a max_data_bytes. + * @param [in] max_data_bytes opus_int32: Size of the allocated + * memory for the output + * payload. This may be + * used to impose an upper limit on + * the instant bitrate, but should + * not be used as the only bitrate + * control. Use #OPUS_SET_BITRATE to + * control the bitrate. + * @returns The length of the encoded packet (in bytes) on success or a + * negative error code (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_projection_encode( + OpusProjectionEncoder *st, + const opus_int16 *pcm, + int frame_size, + unsigned char *data, + opus_int32 max_data_bytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + + +/** Encodes a projection Opus frame from floating point input. + * @param st OpusProjectionEncoder*: Projection encoder state. + * @param[in] pcm const float*: The input signal as interleaved + * samples with a normal range of + * +/-1.0. + * Samples with a range beyond +/-1.0 + * are supported but will be clipped by + * decoders using the integer API and + * should only be used if it is known + * that the far end supports extended + * dynamic range. + * This must contain + * frame_size*channels + * samples. + * @param frame_size int: Number of samples per channel in the input + * signal. + * This must be an Opus frame size for the + * encoder's sampling rate. + * For example, at 48 kHz the permitted values + * are 120, 240, 480, 960, 1920, and 2880. + * Passing in a duration of less than 10 ms + * (480 samples at 48 kHz) will prevent the + * encoder from using the LPC or hybrid modes. + * @param[out] data unsigned char*: Output payload. + * This must contain storage for at + * least \a max_data_bytes. + * @param [in] max_data_bytes opus_int32: Size of the allocated + * memory for the output + * payload. This may be + * used to impose an upper limit on + * the instant bitrate, but should + * not be used as the only bitrate + * control. Use #OPUS_SET_BITRATE to + * control the bitrate. + * @returns The length of the encoded packet (in bytes) on success or a + * negative error code (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_projection_encode_float( + OpusProjectionEncoder *st, + const float *pcm, + int frame_size, + unsigned char *data, + opus_int32 max_data_bytes +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(2) OPUS_ARG_NONNULL(4); + + +/** Frees an OpusProjectionEncoder allocated by + * opus_projection_ambisonics_encoder_create(). + * @param st OpusProjectionEncoder*: Projection encoder state to be freed. + */ +OPUS_EXPORT void opus_projection_encoder_destroy(OpusProjectionEncoder *st); + + +/** Perform a CTL function on a projection Opus encoder. + * + * Generally the request and subsequent arguments are generated by a + * convenience macro. + * @param st OpusProjectionEncoder*: Projection encoder state. + * @param request This and all remaining parameters should be replaced by one + * of the convenience macros in @ref opus_genericctls, + * @ref opus_encoderctls, @ref opus_multistream_ctls, or + * @ref opus_projection_ctls + * @see opus_genericctls + * @see opus_encoderctls + * @see opus_multistream_ctls + * @see opus_projection_ctls + */ +OPUS_EXPORT int opus_projection_encoder_ctl(OpusProjectionEncoder *st, int request, ...) OPUS_ARG_NONNULL(1); + + +/**@}*/ + +/**\name Projection decoder functions */ +/**@{*/ + +/** Gets the size of an OpusProjectionDecoder structure. + * @param channels int: The total number of output channels. + * This must be no more than 255. + * @param streams int: The total number of streams coded in the + * input. + * This must be no more than 255. + * @param coupled_streams int: Number streams to decode as coupled + * (2 channel) streams. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * coded channels (streams + + * coupled_streams) must be no + * more than 255. + * @returns The size in bytes on success, or a negative error code + * (see @ref opus_errorcodes) on error. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT opus_int32 opus_projection_decoder_get_size( + int channels, + int streams, + int coupled_streams +); + + +/** Allocates and initializes a projection decoder state. + * Call opus_projection_decoder_destroy() to release + * this object when finished. + * @param Fs opus_int32: Sampling rate to decode at (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels to output. + * This must be at most 255. + * It may be different from the number of coded + * channels (streams + + * coupled_streams). + * @param streams int: The total number of streams coded in the + * input. + * This must be no more than 255. + * @param coupled_streams int: Number of streams to decode as coupled + * (2 channel) streams. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * coded channels (streams + + * coupled_streams) must be no + * more than 255. + * @param[in] demixing_matrix const unsigned char[demixing_matrix_size]: Demixing matrix + * that mapping from coded channels to output channels, + * as described in @ref opus_projection and + * @ref opus_projection_ctls. + * @param demixing_matrix_size opus_int32: The size in bytes of the + * demixing matrix, as + * described in @ref + * opus_projection_ctls. + * @param[out] error int *: Returns #OPUS_OK on success, or an error + * code (see @ref opus_errorcodes) on + * failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT OpusProjectionDecoder *opus_projection_decoder_create( + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + unsigned char *demixing_matrix, + opus_int32 demixing_matrix_size, + int *error +) OPUS_ARG_NONNULL(5); + + +/** Intialize a previously allocated projection decoder state object. + * The memory pointed to by \a st must be at least the size returned by + * opus_projection_decoder_get_size(). + * This is intended for applications which use their own allocator instead of + * malloc. + * To reset a previously initialized state, use the #OPUS_RESET_STATE CTL. + * @see opus_projection_decoder_create + * @see opus_projection_deocder_get_size + * @param st OpusProjectionDecoder*: Projection encoder state to initialize. + * @param Fs opus_int32: Sampling rate to decode at (in Hz). + * This must be one of 8000, 12000, 16000, + * 24000, or 48000. + * @param channels int: Number of channels to output. + * This must be at most 255. + * It may be different from the number of coded + * channels (streams + + * coupled_streams). + * @param streams int: The total number of streams coded in the + * input. + * This must be no more than 255. + * @param coupled_streams int: Number of streams to decode as coupled + * (2 channel) streams. + * This must be no larger than the total + * number of streams. + * Additionally, The total number of + * coded channels (streams + + * coupled_streams) must be no + * more than 255. + * @param[in] demixing_matrix const unsigned char[demixing_matrix_size]: Demixing matrix + * that mapping from coded channels to output channels, + * as described in @ref opus_projection and + * @ref opus_projection_ctls. + * @param demixing_matrix_size opus_int32: The size in bytes of the + * demixing matrix, as + * described in @ref + * opus_projection_ctls. + * @returns #OPUS_OK on success, or an error code (see @ref opus_errorcodes) + * on failure. + */ +OPUS_EXPORT int opus_projection_decoder_init( + OpusProjectionDecoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + unsigned char *demixing_matrix, + opus_int32 demixing_matrix_size +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(6); + + +/** Decode a projection Opus packet. + * @param st OpusProjectionDecoder*: Projection decoder state. + * @param[in] data const unsigned char*: Input payload. + * Use a NULL + * pointer to indicate packet + * loss. + * @param len opus_int32: Number of bytes in payload. + * @param[out] pcm opus_int16*: Output signal, with interleaved + * samples. + * This must contain room for + * frame_size*channels + * samples. + * @param frame_size int: The number of samples per channel of + * available space in \a pcm. + * If this is less than the maximum packet duration + * (120 ms; 5760 for 48kHz), this function will not be capable + * of decoding some packets. In the case of PLC (data==NULL) + * or FEC (decode_fec=1), then frame_size needs to be exactly + * the duration of audio that is missing, otherwise the + * decoder will not be in the optimal state to decode the + * next incoming packet. For the PLC and FEC cases, frame_size + * must be a multiple of 2.5 ms. + * @param decode_fec int: Flag (0 or 1) to request that any in-band + * forward error correction data be decoded. + * If no such data is available, the frame is + * decoded as if it were lost. + * @returns Number of samples decoded on success or a negative error code + * (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_projection_decode( + OpusProjectionDecoder *st, + const unsigned char *data, + opus_int32 len, + opus_int16 *pcm, + int frame_size, + int decode_fec +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + + +/** Decode a projection Opus packet with floating point output. + * @param st OpusProjectionDecoder*: Projection decoder state. + * @param[in] data const unsigned char*: Input payload. + * Use a NULL + * pointer to indicate packet + * loss. + * @param len opus_int32: Number of bytes in payload. + * @param[out] pcm opus_int16*: Output signal, with interleaved + * samples. + * This must contain room for + * frame_size*channels + * samples. + * @param frame_size int: The number of samples per channel of + * available space in \a pcm. + * If this is less than the maximum packet duration + * (120 ms; 5760 for 48kHz), this function will not be capable + * of decoding some packets. In the case of PLC (data==NULL) + * or FEC (decode_fec=1), then frame_size needs to be exactly + * the duration of audio that is missing, otherwise the + * decoder will not be in the optimal state to decode the + * next incoming packet. For the PLC and FEC cases, frame_size + * must be a multiple of 2.5 ms. + * @param decode_fec int: Flag (0 or 1) to request that any in-band + * forward error correction data be decoded. + * If no such data is available, the frame is + * decoded as if it were lost. + * @returns Number of samples decoded on success or a negative error code + * (see @ref opus_errorcodes) on failure. + */ +OPUS_EXPORT OPUS_WARN_UNUSED_RESULT int opus_projection_decode_float( + OpusProjectionDecoder *st, + const unsigned char *data, + opus_int32 len, + float *pcm, + int frame_size, + int decode_fec +) OPUS_ARG_NONNULL(1) OPUS_ARG_NONNULL(4); + + +/** Perform a CTL function on a projection Opus decoder. + * + * Generally the request and subsequent arguments are generated by a + * convenience macro. + * @param st OpusProjectionDecoder*: Projection decoder state. + * @param request This and all remaining parameters should be replaced by one + * of the convenience macros in @ref opus_genericctls, + * @ref opus_decoderctls, @ref opus_multistream_ctls, or + * @ref opus_projection_ctls. + * @see opus_genericctls + * @see opus_decoderctls + * @see opus_multistream_ctls + * @see opus_projection_ctls + */ +OPUS_EXPORT int opus_projection_decoder_ctl(OpusProjectionDecoder *st, int request, ...) OPUS_ARG_NONNULL(1); + + +/** Frees an OpusProjectionDecoder allocated by + * opus_projection_decoder_create(). + * @param st OpusProjectionDecoder: Projection decoder state to be freed. + */ +OPUS_EXPORT void opus_projection_decoder_destroy(OpusProjectionDecoder *st); + + +/**@}*/ + +/**@}*/ + +#ifdef __cplusplus +} +#endif + +#endif /* OPUS_PROJECTION_H */ diff --git a/lib/opus/include/opus_types.h b/lib/opus/include/opus_types.h new file mode 100644 index 00000000..7cf67558 --- /dev/null +++ b/lib/opus/include/opus_types.h @@ -0,0 +1,166 @@ +/* (C) COPYRIGHT 1994-2002 Xiph.Org Foundation */ +/* Modified by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ +/* opus_types.h based on ogg_types.h from libogg */ + +/** + @file opus_types.h + @brief Opus reference implementation types +*/ +#ifndef OPUS_TYPES_H +#define OPUS_TYPES_H + +#define opus_int int /* used for counters etc; at least 16 bits */ +#define opus_int64 long long +#define opus_int8 signed char + +#define opus_uint unsigned int /* used for counters etc; at least 16 bits */ +#define opus_uint64 unsigned long long +#define opus_uint8 unsigned char + +/* Use the real stdint.h if it's there (taken from Paul Hsieh's pstdint.h) */ +#if (defined(__STDC__) && __STDC__ && defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__GNUC__) && (defined(_STDINT_H) || defined(_STDINT_H_)) || defined (HAVE_STDINT_H)) +#include +# undef opus_int64 +# undef opus_int8 +# undef opus_uint64 +# undef opus_uint8 + typedef int8_t opus_int8; + typedef uint8_t opus_uint8; + typedef int16_t opus_int16; + typedef uint16_t opus_uint16; + typedef int32_t opus_int32; + typedef uint32_t opus_uint32; + typedef int64_t opus_int64; + typedef uint64_t opus_uint64; +#elif defined(_WIN32) + +# if defined(__CYGWIN__) +# include <_G_config.h> + typedef _G_int32_t opus_int32; + typedef _G_uint32_t opus_uint32; + typedef _G_int16 opus_int16; + typedef _G_uint16 opus_uint16; +# elif defined(__MINGW32__) + typedef short opus_int16; + typedef unsigned short opus_uint16; + typedef int opus_int32; + typedef unsigned int opus_uint32; +# elif defined(__MWERKS__) + typedef int opus_int32; + typedef unsigned int opus_uint32; + typedef short opus_int16; + typedef unsigned short opus_uint16; +# else + /* MSVC/Borland */ + typedef __int32 opus_int32; + typedef unsigned __int32 opus_uint32; + typedef __int16 opus_int16; + typedef unsigned __int16 opus_uint16; +# endif + +#elif defined(__MACOS__) + +# include + typedef SInt16 opus_int16; + typedef UInt16 opus_uint16; + typedef SInt32 opus_int32; + typedef UInt32 opus_uint32; + +#elif (defined(__APPLE__) && defined(__MACH__)) /* MacOS X Framework build */ + +# include + typedef int16_t opus_int16; + typedef u_int16_t opus_uint16; + typedef int32_t opus_int32; + typedef u_int32_t opus_uint32; + +#elif defined(__BEOS__) + + /* Be */ +# include + typedef int16 opus_int16; + typedef u_int16 opus_uint16; + typedef int32_t opus_int32; + typedef u_int32_t opus_uint32; + +#elif defined (__EMX__) + + /* OS/2 GCC */ + typedef short opus_int16; + typedef unsigned short opus_uint16; + typedef int opus_int32; + typedef unsigned int opus_uint32; + +#elif defined (DJGPP) + + /* DJGPP */ + typedef short opus_int16; + typedef unsigned short opus_uint16; + typedef int opus_int32; + typedef unsigned int opus_uint32; + +#elif defined(R5900) + + /* PS2 EE */ + typedef int opus_int32; + typedef unsigned opus_uint32; + typedef short opus_int16; + typedef unsigned short opus_uint16; + +#elif defined(__SYMBIAN32__) + + /* Symbian GCC */ + typedef signed short opus_int16; + typedef unsigned short opus_uint16; + typedef signed int opus_int32; + typedef unsigned int opus_uint32; + +#elif defined(CONFIG_TI_C54X) || defined (CONFIG_TI_C55X) + + typedef short opus_int16; + typedef unsigned short opus_uint16; + typedef long opus_int32; + typedef unsigned long opus_uint32; + +#elif defined(CONFIG_TI_C6X) + + typedef short opus_int16; + typedef unsigned short opus_uint16; + typedef int opus_int32; + typedef unsigned int opus_uint32; + +#else + + /* Give up, take a reasonable guess */ + typedef short opus_int16; + typedef unsigned short opus_uint16; + typedef int opus_int32; + typedef unsigned int opus_uint32; + +#endif + +#endif /* OPUS_TYPES_H */ diff --git a/lib/opus/m4/as-gcc-inline-assembly.m4 b/lib/opus/m4/as-gcc-inline-assembly.m4 new file mode 100644 index 00000000..b0d2da41 --- /dev/null +++ b/lib/opus/m4/as-gcc-inline-assembly.m4 @@ -0,0 +1,98 @@ +dnl as-gcc-inline-assembly.m4 0.1.0 + +dnl autostars m4 macro for detection of gcc inline assembly + +dnl David Schleef + +dnl AS_COMPILER_FLAG(ACTION-IF-ACCEPTED, [ACTION-IF-NOT-ACCEPTED]) +dnl Tries to compile with the given CFLAGS. +dnl Runs ACTION-IF-ACCEPTED if the compiler can compile with the flags, +dnl and ACTION-IF-NOT-ACCEPTED otherwise. + +AC_DEFUN([AS_GCC_INLINE_ASSEMBLY], +[ + AC_MSG_CHECKING([if compiler supports gcc-style inline assembly]) + + AC_TRY_COMPILE([], [ +#ifdef __GNUC_MINOR__ +#if (__GNUC__ * 1000 + __GNUC_MINOR__) < 3004 +#error GCC before 3.4 has critical bugs compiling inline assembly +#endif +#endif +__asm__ (""::) ], [flag_ok=yes], [flag_ok=no]) + + if test "X$flag_ok" = Xyes ; then + $1 + true + else + $2 + true + fi + AC_MSG_RESULT([$flag_ok]) +]) + +AC_DEFUN([AS_ASM_ARM_NEON], +[ + AC_MSG_CHECKING([if assembler supports NEON instructions on ARM]) + + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],[__asm__("vorr d0,d0,d0")])], + [AC_MSG_RESULT([yes]) + $1], + [AC_MSG_RESULT([no]) + $2]) +]) + +AC_DEFUN([AS_ASM_ARM_NEON_FORCE], +[ + AC_MSG_CHECKING([if assembler supports NEON instructions on ARM]) + + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],[__asm__(".arch armv7-a\n.fpu neon\n.object_arch armv4t\nvorr d0,d0,d0")])], + [AC_MSG_RESULT([yes]) + $1], + [AC_MSG_RESULT([no]) + $2]) +]) + +AC_DEFUN([AS_ASM_ARM_MEDIA], +[ + AC_MSG_CHECKING([if assembler supports ARMv6 media instructions on ARM]) + + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],[__asm__("shadd8 r3,r3,r3")])], + [AC_MSG_RESULT([yes]) + $1], + [AC_MSG_RESULT([no]) + $2]) +]) + +AC_DEFUN([AS_ASM_ARM_MEDIA_FORCE], +[ + AC_MSG_CHECKING([if assembler supports ARMv6 media instructions on ARM]) + + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],[__asm__(".arch armv6\n.object_arch armv4t\nshadd8 r3,r3,r3")])], + [AC_MSG_RESULT([yes]) + $1], + [AC_MSG_RESULT([no]) + $2]) +]) + +AC_DEFUN([AS_ASM_ARM_EDSP], +[ + AC_MSG_CHECKING([if assembler supports EDSP instructions on ARM]) + + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],[__asm__("qadd r3,r3,r3")])], + [AC_MSG_RESULT([yes]) + $1], + [AC_MSG_RESULT([no]) + $2]) +]) + +AC_DEFUN([AS_ASM_ARM_EDSP_FORCE], +[ + AC_MSG_CHECKING([if assembler supports EDSP instructions on ARM]) + + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([],[__asm__(".arch armv5te\n.object_arch armv4t\nqadd r3,r3,r3")])], + [AC_MSG_RESULT([yes]) + $1], + [AC_MSG_RESULT([no]) + $2]) +]) diff --git a/lib/opus/m4/ax_add_fortify_source.m4 b/lib/opus/m4/ax_add_fortify_source.m4 new file mode 100644 index 00000000..1c89e410 --- /dev/null +++ b/lib/opus/m4/ax_add_fortify_source.m4 @@ -0,0 +1,53 @@ +# =========================================================================== +# Modified from https://www.gnu.org/software/autoconf-archive/ax_add_fortify_source.html +# =========================================================================== +# +# SYNOPSIS +# +# AX_ADD_FORTIFY_SOURCE +# +# DESCRIPTION +# +# Check whether -D_FORTIFY_SOURCE=2 can be added to CFLAGS without macro +# redefinition warnings. Some distributions (such as Gentoo Linux) enable +# _FORTIFY_SOURCE globally in their compilers, leading to unnecessary +# warnings in the form of +# +# :0:0: error: "_FORTIFY_SOURCE" redefined [-Werror] +# : note: this is the location of the previous definition +# +# which is a problem if -Werror is enabled. This macro checks whether +# _FORTIFY_SOURCE is already defined, and if not, adds -D_FORTIFY_SOURCE=2 +# to CFLAGS. +# +# LICENSE +# +# Copyright (c) 2017 David Seifert +# +# Copying and distribution of this file, with or without modification, are +# permitted in any medium without royalty provided the copyright notice +# and this notice are preserved. This file is offered as-is, without any +# warranty. + +#serial 1 + +AC_DEFUN([AX_ADD_FORTIFY_SOURCE],[ + AC_MSG_CHECKING([whether to add -D_FORTIFY_SOURCE=2 to CFLAGS]) + AC_LINK_IFELSE([ + AC_LANG_SOURCE( + [[ + int main() { + #ifndef _FORTIFY_SOURCE + return 0; + #else + this_is_an_error; + #endif + } + ]] + )], [ + AC_MSG_RESULT([yes]) + CFLAGS="$CFLAGS -D_FORTIFY_SOURCE=2" + ], [ + AC_MSG_RESULT([no]) + ]) +]) diff --git a/lib/opus/m4/opus-intrinsics.m4 b/lib/opus/m4/opus-intrinsics.m4 new file mode 100644 index 00000000..a262ca18 --- /dev/null +++ b/lib/opus/m4/opus-intrinsics.m4 @@ -0,0 +1,29 @@ +dnl opus-intrinsics.m4 +dnl macro for testing for support for compiler intrinsics, either by default or with a compiler flag + +dnl OPUS_CHECK_INTRINSICS(NAME-OF-INTRINSICS, COMPILER-FLAG-FOR-INTRINSICS, VAR-IF-PRESENT, VAR-IF-DEFAULT, TEST-PROGRAM-HEADER, TEST-PROGRAM-BODY) +AC_DEFUN([OPUS_CHECK_INTRINSICS], +[ + AC_MSG_CHECKING([if compiler supports $1 intrinsics]) + AC_LINK_IFELSE( + [AC_LANG_PROGRAM($5, $6)], + [ + $3=1 + $4=1 + AC_MSG_RESULT([yes]) + ],[ + $4=0 + AC_MSG_RESULT([no]) + AC_MSG_CHECKING([if compiler supports $1 intrinsics with $2]) + save_CFLAGS="$CFLAGS"; CFLAGS="$CFLAGS $2" + AC_LINK_IFELSE([AC_LANG_PROGRAM($5, $6)], + [ + AC_MSG_RESULT([yes]) + $3=1 + ],[ + AC_MSG_RESULT([no]) + $3=0 + ]) + CFLAGS="$save_CFLAGS" + ]) +]) diff --git a/lib/opus/meson.build b/lib/opus/meson.build new file mode 100644 index 00000000..9f28de33 --- /dev/null +++ b/lib/opus/meson.build @@ -0,0 +1,675 @@ +project('opus', 'c', + version: run_command('meson/get-version.py', '--package-version', check: true).stdout().strip(), + meson_version: '>=0.54.0', + default_options: ['warning_level=2', + 'c_std=gnu99', + 'buildtype=debugoptimized']) + +libversion = run_command('meson/get-version.py', '--libtool-version', check: true).stdout().strip() +macosversion = run_command('meson/get-version.py', '--darwin-version', check: true).stdout().strip() + +cc = meson.get_compiler('c') +host_system = host_machine.system() +host_cpu_family = host_machine.cpu_family() +top_srcdir = meson.current_source_dir() +top_builddir = meson.current_build_dir() + +opus_includes = include_directories('.', 'include', 'celt', 'silk') +opus_public_includes = include_directories('include') + +add_project_arguments('-DOPUS_BUILD', language: 'c') +add_project_arguments('-DHAVE_CONFIG_H', language: 'c') + +if host_system == 'windows' + if cc.get_argument_syntax() == 'msvc' + add_project_arguments('-D_CRT_SECURE_NO_WARNINGS', language: 'c') + endif +endif + +if cc.get_argument_syntax() == 'gnu' + add_project_arguments('-D_FORTIFY_SOURCE=2', language: 'c') +endif + +# Check for extra compiler args +additional_c_args = [] +if cc.get_argument_syntax() != 'msvc' + additional_c_args += [ + '-fvisibility=hidden', + '-Wcast-align', + '-Wnested-externs', + '-Wshadow', + '-Wstrict-prototypes', + ] + + # On Windows, -fstack-protector-strong adds a libssp-0.dll dependency and + # prevents static linking + if host_system != 'windows' + additional_c_args += ['-fstack-protector-strong'] + endif +endif + +foreach arg : additional_c_args + if cc.has_argument(arg) + add_project_arguments(arg, language: 'c') + endif +endforeach + +# Windows MSVC warnings +if cc.get_id() == 'msvc' + # Ignore several spurious warnings. + # If a warning is completely useless and spammy, use '/wdXXXX' to suppress it + # If a warning is harmless but hard to fix, use '/woXXXX' so it's shown once + # NOTE: Only add warnings here if you are sure they're spurious + add_project_arguments('/wd4035', '/wd4715', '/wd4116', '/wd4046', '/wd4068', + '/wd4820', '/wd4244', '/wd4255', '/wd4668', + language : 'c') +endif + +opus_version = meson.project_version() + +opus_conf = configuration_data() +opus_conf.set('PACKAGE_BUGREPORT', '"opus@xiph.org"') +opus_conf.set('PACKAGE_NAME', '"opus"') +opus_conf.set('PACKAGE_STRING', '"opus @0@"'.format(opus_version)) +opus_conf.set('PACKAGE_TARNAME', '"opus"') +opus_conf.set('PACKAGE_URL', '""') +opus_conf.set('PACKAGE_VERSION', '"@0@"'.format(opus_version)) + +# FIXME: optional Ne10 dependency +have_arm_ne10 = false + +libm = cc.find_library('m', required : false) + +opus_conf.set('HAVE_LRINTF', cc.has_function('lrintf', prefix: '#include ', dependencies: libm)) +opus_conf.set('HAVE_LRINT', cc.has_function('lrint', prefix: '#include ', dependencies: libm)) +opus_conf.set('HAVE___MALLOC_HOOK', cc.has_function('__malloc_hook', prefix: '#include ')) +opus_conf.set('HAVE_STDINT_H', cc.check_header('stdint.h')) + +# Check for restrict keyword +restrict_tmpl = ''' +typedef int * int_ptr; +int foo (int_ptr @0@ ip, int * @0@ baz[]) { + return ip[0]; +} +int main (int argc, char ** argv) { + int s[1]; + int * @0@ t = s; + t[0] = 0; + return foo(t, (void *)0); +}''' +# Define restrict to the equivalent of the C99 restrict keyword, or to +# nothing if this is not supported. Do not define if restrict is +# supported directly. +if not cc.compiles(restrict_tmpl.format('restrict'), name : 'restrict keyword') + if cc.compiles(restrict_tmpl.format('__restrict'), name : '__restrict') + opus_conf.set('restrict', '__restrict') + elif cc.compiles(restrict_tmpl.format('__restrict__'), name : '__restrict__') + opus_conf.set('restrict', '__restrict') + elif cc.compiles(restrict_tmpl.format('_Restrict'), name : '_Restrict') + opus_conf.set('restrict', '_Restrict') + else + opus_conf.set('restrict', '/**/') + endif +endif + +# Check for C99 variable-size arrays, or alloca() as fallback +msg_use_alloca = false +if cc.compiles('''static int x; + char some_func (void) { + char a[++x]; + a[sizeof a - 1] = 0; + int N; + return a[0]; + }''', name : 'C99 variable-size arrays') + opus_conf.set('VAR_ARRAYS', 1) + msg_use_alloca = 'NO (using C99 variable-size arrays instead)' +elif cc.compiles('''#include + void some_func (void) { + int foo=10; + int * array = alloca(foo); + }''', name : 'alloca (alloca.h)') + opus_conf.set('USE_ALLOCA', true) + opus_conf.set('HAVE_ALLOCA_H', true) + msg_use_alloca = true +elif cc.compiles('''#include + #include + void some_func (void) { + int foo=10; + int * array = alloca(foo); + }''', name : 'alloca (std)') + opus_conf.set('USE_ALLOCA', true) + msg_use_alloca = true +endif + +opts = [ + [ 'fixed-point', 'FIXED_POINT' ], + [ 'fixed-point-debug', 'FIXED_DEBUG' ], + [ 'custom-modes', 'CUSTOM_MODES' ], + [ 'float-approx', 'FLOAT_APPROX' ], + [ 'assertions', 'ENABLE_ASSERTIONS' ], + [ 'hardening', 'ENABLE_HARDENING' ], + [ 'fuzzing', 'FUZZING' ], + [ 'check-asm', 'OPUS_CHECK_ASM' ], +] + +foreach opt : opts + # we assume these are all boolean options + opt_foo = get_option(opt[0]) + if opt_foo + opus_conf.set(opt[1], 1) + endif + set_variable('opt_' + opt[0].underscorify(), opt_foo) +endforeach + +opt_asm = get_option('asm') +opt_rtcd = get_option('rtcd') +opt_intrinsics = get_option('intrinsics') +extra_programs = get_option('extra-programs') +opt_tests = get_option('tests') + +disable_float_api = not get_option('float-api') +if disable_float_api + opus_conf.set('DISABLE_FLOAT_API', 1) +endif + +# This is for the description in the pkg-config .pc file +if opt_fixed_point + pc_build = 'fixed-point' +else + pc_build = 'floating-point' +endif +if opt_custom_modes + pc_build = pc_build + ', custom modes' +endif + +rtcd_support = [] +# With GCC, Clang, ICC, etc, we differentiate between 'may support this SIMD' +# and 'presume we have this SIMD' by checking whether the SIMD / intrinsics can +# be compiled by the compiler as-is (presume) or with SIMD cflags (may have). +# With MSVC, the compiler will always build SIMD/intrinsics targeting all +# specific instruction sets supported by that version of the compiler. No +# special arguments are ever needed. If runtime CPU detection is not disabled, +# we must always assume that we only 'may have' it. +opus_can_presume_simd = true +if cc.get_argument_syntax() == 'msvc' + if opt_rtcd.disabled() + warning('Building with an MSVC-like compiler and runtime CPU detection is disabled. Outputs may not run on all @0@ CPUs.'.format(host_cpu_family)) + else + opus_can_presume_simd = false + endif +endif + +opus_arm_external_asm = false + +asm_tmpl = ''' +int main (int argc, char ** argv) { + __asm__("@0@"); + return 0; +}''' + +asm_optimization = [] +inline_optimization = [] +if not opt_asm.disabled() + # Currently we only have inline asm for fixed-point + if host_cpu_family == 'arm' and opt_fixed_point + opus_conf.set('OPUS_ARM_ASM', true) + + # Check if compiler supports gcc-style inline assembly + if cc.compiles('''#ifdef __GNUC_MINOR__ + #if (__GNUC__ * 1000 + __GNUC_MINOR__) < 3004 + #error GCC before 3.4 has critical bugs compiling inline assembly + #endif + #endif + __asm__ (""::)''', + name : 'compiler supports gcc-style inline assembly') + + opus_conf.set('OPUS_ARM_INLINE_ASM', 1) + + # AS_ASM_ARM_EDSP + if cc.compiles(asm_tmpl.format('qadd r3,r3,r3'), + name : 'assembler supports EDSP instructions on ARM') + opus_conf.set('OPUS_ARM_INLINE_EDSP', 1) + inline_optimization += ['ESDP'] + endif + + # AS_ASM_ARM_MEDIA + if cc.compiles(asm_tmpl.format('shadd8 r3,r3,r3'), + name : 'assembler supports ARMv6 media instructions on ARM') + opus_conf.set('OPUS_ARM_INLINE_MEDIA', 1) + inline_optimization += ['Media'] + endif + + # AS_ASM_ARM_NEON + if cc.compiles(asm_tmpl.format('vorr d0,d0,d0'), + name : 'assembler supports NEON instructions on ARM') + opus_conf.set('OPUS_ARM_INLINE_NEON', 1) + inline_optimization += ['NEON'] + endif + endif + + # We need Perl to translate RVCT-syntax asm to gas syntax + perl = find_program('perl', required: get_option('asm')) + if perl.found() + opus_arm_external_asm = true + # opus_arm_presume_* mean we can and will use those instructions + # directly without doing runtime CPU detection. + # opus_arm_may_have_* mean we can emit those instructions, but we can + # only use them after runtime detection. + # The same rules apply for x86 assembly and intrinsics. + + opus_arm_may_have_edsp = opus_conf.has('OPUS_ARM_INLINE_EDSP') + opus_arm_presume_edsp = opus_arm_may_have_edsp and opus_can_presume_simd + + opus_arm_may_have_media = opus_conf.has('OPUS_ARM_INLINE_MEDIA') + opus_arm_presume_media = opus_arm_may_have_media and opus_can_presume_simd + + opus_arm_may_have_neon = opus_conf.has('OPUS_ARM_INLINE_NEON') + opus_arm_presume_neon = opus_arm_may_have_neon and opus_can_presume_simd + + if not opt_rtcd.disabled() + if not opus_arm_may_have_edsp + message('Trying to force-enable armv5e EDSP instructions...') + # AS_ASM_ARM_EDSP_FORCE + opus_arm_may_have_edsp = cc.compiles(asm_tmpl.format('.arch armv5te\n.object_arch armv4t\nqadd r3,r3,r3'), + name : 'Assembler supports EDSP instructions on ARM (forced)') + endif + if not opus_arm_may_have_media + message('Trying to force-enable ARMv6 media instructions...') + opus_arm_may_have_media = cc.compiles(asm_tmpl.format('.arch armv6\n.object_arch armv4t\nshadd8 r3,r3,r3'), + name : 'Assembler supports ARMv6 media instructions on ARM (forced)') + endif + if not opus_arm_may_have_neon + message('Trying to force-enable NEON instructions...') + opus_arm_may_have_neon = cc.compiles(asm_tmpl.format('.arch armv7-a\n.fpu neon\n.object_arch armv4t\nvorr d0,d0,d0'), + name : 'Assembler supports NEON instructions on ARM (forced)') + endif + endif + + if opus_arm_may_have_edsp + opus_conf.set('OPUS_ARM_MAY_HAVE_EDSP', 1) + if opus_arm_presume_edsp + opus_conf.set('OPUS_ARM_PRESUME_EDSP', 1) + asm_optimization += ['EDSP'] + else + rtcd_support += ['EDSP'] + endif + endif + if opus_arm_may_have_media + opus_conf.set('OPUS_ARM_MAY_HAVE_MEDIA', 1) + if opus_arm_presume_media + opus_conf.set('OPUS_ARM_PRESUME_MEDIA', 1) + asm_optimization += ['Media'] + else + rtcd_support += ['Media'] + endif + endif + if opus_arm_may_have_neon + opus_conf.set('OPUS_ARM_MAY_HAVE_NEON', 1) + if opus_arm_presume_neon + opus_conf.set('OPUS_ARM_PRESUME_NEON', 1) + asm_optimization += ['NEON'] + else + rtcd_support += ['NEON'] + endif + endif + + if cc.get_define('__APPLE__') + arm2gnu_args = ['--apple'] + else + arm2gnu_args = [] + endif + endif # found perl + else # arm + enable fixed point + if opt_asm.enabled() + error('asm option is enabled, but no assembly support for ' + host_cpu_family) + endif + endif +endif # enable asm + +# Check whether we require assembly and we support assembly on this arch, +# but none were detected. Can happen because of incorrect compiler flags, such +# as missing -mfloat-abi=softfp on ARM32 softfp architectures. +if opt_asm.enabled() and (asm_optimization.length() + inline_optimization.length()) == 0 + error('asm option was enabled, but no assembly support was detected') +endif + +# XXX: NEON has hardfp vs softfp compiler configuration issues +# When targeting ARM32 softfp, we sometimes need to explicitly pass +# -mfloat-abi=softfp to enable NEON. F.ex., on Android. It should +# be set in the cross file. +arm_neon_intr_link_args = ['-mfpu=neon'] + +have_sse = false +have_sse2 = false +have_sse4_1 = false +have_avx = false # no avx opus code yet +have_neon_intr = false + +intrinsics_support = [] +if not opt_intrinsics.disabled() + if host_cpu_family in ['arm', 'aarch64'] + # Check for ARMv7/AArch64 neon intrinsics + intrin_check = ''' + #include + int main (void) { + static float32x4_t A0, A1, SUMM; + SUMM = vmlaq_f32(SUMM, A0, A1); + return (int)vgetq_lane_f32(SUMM, 0); + }''' + intrin_name = 'ARMv7/AArch64 NEON' + if cc.links(intrin_check, + name: 'compiler supports @0@ intrinsics'.format(intrin_name)) + opus_arm_presume_neon_intr = opus_can_presume_simd + opus_arm_may_have_neon_intr = true + else + opus_arm_presume_neon_intr = false + if cc.links(intrin_check, + args: arm_neon_intr_link_args, + name: 'compiler supports @0@ intrinsics with @1@'.format(intrin_name, ' '.join(arm_neon_intr_link_args))) + opus_arm_may_have_neon_intr = true + else + opus_arm_may_have_neon_intr = false + endif + endif + + if opus_arm_may_have_neon_intr + have_neon_intr = true + intrinsics_support += [intrin_name] + opus_conf.set('OPUS_ARM_MAY_HAVE_NEON_INTR', 1) + if opus_arm_presume_neon_intr + opus_conf.set('OPUS_ARM_PRESUME_NEON_INTR', 1) + else + rtcd_support += [intrin_name] + opus_neon_intr_args = arm_neon_intr_link_args + endif + else + message('Compiler does not support @0@ intrinsics'.format(intrin_name)) + endif + + # Check for aarch64 neon intrinsics + intrin_check = ''' + #include + int main (void) { + static int32_t IN; + static int16_t OUT; + OUT = vqmovns_s32(IN); + }''' + intrin_name = 'AArch64 NEON' + if cc.links(intrin_check, + name: 'compiler supports @0@ intrinsics'.format(intrin_name)) + opus_arm_presume_aarch64_neon_intr = opus_can_presume_simd + opus_arm_may_have_aarch64_neon_intr = true + else + opus_arm_presume_aarch64_neon_intr = false + if cc.links(intrin_check, + args: arm_neon_intr_link_args, + name: 'compiler supports @0@ intrinsics with @1@'.format(intrin_name, ' '.join(arm_neon_intr_link_args))) + opus_arm_may_have_aarch64_neon_intr = true + else + opus_arm_may_have_aarch64_neon_intr = false + endif + endif + + if opus_arm_may_have_aarch64_neon_intr + intrinsics_support += [intrin_name] + opus_conf.set('OPUS_X86_MAY_HAVE_AARCH64_NEON_INTR', 1) + if opus_arm_presume_aarch64_neon_intr + opus_conf.set('OPUS_X86_PRESUME_AARCH64_NEON_INTR', 1) + endif + else + message('Compiler does not support @0@ intrinsics'.format(intrin_name)) + endif + elif host_cpu_family in ['x86', 'x86_64'] + # XXX: allow external override/specification of the flags + x86_intrinsics = [ + [ 'SSE', 'xmmintrin.h', '__m128', '_mm_setzero_ps()', ['-msse'] ], + [ 'SSE2', 'emmintrin.h', '__m128i', '_mm_setzero_si128()', ['-msse2'] ], + [ 'SSE4.1', 'smmintrin.h', '__m128i', '_mm_setzero_si128(); mtest = _mm_cmpeq_epi64(mtest, mtest)', ['-msse4.1'] ], + [ 'AVX', 'immintrin.h', '__m256', '_mm256_setzero_ps()', ['-mavx'] ], + ] + + foreach intrin : x86_intrinsics + intrin_check = '''#include <@0@> + int main (int argc, char ** argv) { + static @1@ mtest; + mtest = @2@; + return *((unsigned char *) &mtest) != 0; + }'''.format(intrin[1],intrin[2],intrin[3]) + intrin_name = intrin[0] + # Intrinsics arguments are not available with MSVC-like compilers + intrin_args = cc.get_argument_syntax() == 'msvc' ? [] : intrin[4] + if cc.links(intrin_check, name : 'compiler supports @0@ intrinsics'.format(intrin_name)) + may_have_intrin = true + presume_intrin = opus_can_presume_simd + elif intrin_args.length() > 0 + presume_intrin = false + if cc.links(intrin_check, + args : intrin_args, + name : 'compiler supports @0@ intrinsics with @1@'.format(intrin_name, ' '.join(intrin_args))) + may_have_intrin = true + else + may_have_intrin = false + endif + endif + if may_have_intrin + intrinsics_support += [intrin_name] + intrin_lower_name = intrin_name.to_lower().underscorify() + set_variable('have_' + intrin_lower_name, true) + opus_conf.set('OPUS_X86_MAY_HAVE_' + intrin_name.underscorify(), 1) + if presume_intrin + opus_conf.set('OPUS_X86_PRESUME_' + intrin_name.underscorify(), 1) + else + rtcd_support += [intrin_name] + set_variable('opus_@0@_args'.format(intrin_lower_name), intrin_args) + endif + else + message('Compiler does not support @0@ intrinsics'.format(intrin_name)) + endif + endforeach + + if not opt_rtcd.disabled() + get_cpuid_by_asm = false + cpuid_asm_code = ''' + #include + int main (int argc, char ** argv) { + unsigned int CPUInfo0; + unsigned int CPUInfo1; + unsigned int CPUInfo2; + unsigned int CPUInfo3; + unsigned int InfoType; + #if defined(__i386__) && defined(__PIC__) + __asm__ __volatile__ ( + "xchg %%ebx, %1\n" + "cpuid\n" + "xchg %%ebx, %1\n": + "=a" (CPUInfo0), + "=r" (CPUInfo1), + "=c" (CPUInfo2), + "=d" (CPUInfo3) : + "a" (InfoType), "c" (0) + ); + #else + __asm__ __volatile__ ( + "cpuid": + "=a" (CPUInfo0), + "=b" (CPUInfo1), + "=c" (CPUInfo2), + "=d" (CPUInfo3) : + "a" (InfoType), "c" (0) + ); + #endif + return 0; + }''' + cpuid_c_code = ''' + #include + int main (int argc, char ** argv) { + unsigned int CPUInfo0; + unsigned int CPUInfo1; + unsigned int CPUInfo2; + unsigned int CPUInfo3; + unsigned int InfoType; + __get_cpuid(InfoType, &CPUInfo0, &CPUInfo1, &CPUInfo2, &CPUInfo3); + return 0; + }''' + cpuid_msvc_code = ''' + #include + int main (void) { + int CPUInfo, InfoType; + __cpuid(&CPUInfo, InfoType); + }''' + if cc.links(cpuid_asm_code, name : 'Get X86 CPU info via inline assembly') + opus_conf.set('CPU_INFO_BY_ASM', 1) + elif cc.links(cpuid_c_code, name : 'Get X86 CPU info via C method') + opus_conf.set('CPU_INFO_BY_C', 1) + elif cc.get_define('_MSC_VER') != '' and cc.links(cpuid_msvc_code) + message('Getting X86 CPU info via __cpuid') + else + if opt_intrinsics.enabled() and opt_rtcd.enabled() + error('intrinsics and rtcd options are enabled, but no Get CPU Info method detected') + endif + warning('Get CPU Info method not detected, no rtcd for intrinsics') + endif + endif # opt_rtcd + else + if opt_intrinsics.enabled() + error('intrinsics option enabled, but no intrinsics support for ' + host_cpu_family) + endif + warning('No intrinsics support for ' + host_cpu_family) + endif +endif + +# Check whether we require intrinsics and we support intrinsics on this arch, +# but none were detected. Can happen because of incorrect compiler flags, such +# as missing -mfloat-abi=softfp on ARM32 softfp architectures. +if opt_intrinsics.enabled() and intrinsics_support.length() == 0 + error('intrinsics option was enabled, but none were detected') +endif + +if opt_rtcd.disabled() + rtcd_support = 'disabled' +else + if rtcd_support.length() > 0 + opus_conf.set('OPUS_HAVE_RTCD', 1) + else + if intrinsics_support.length() == 0 + rtcd_support = 'none' + if opt_rtcd.enabled() + error('rtcd option is enabled, but no support for intrinsics or assembly is available') + endif + else + rtcd_support = 'not needed' + endif + endif +endif + +# extract source file lists from .mk files +mk_files = ['silk_sources.mk', 'opus_headers.mk', 'opus_sources.mk', 'silk_headers.mk', 'celt_sources.mk', 'celt_headers.mk'] +lines = run_command('meson/read-sources-list.py', mk_files, check: true).stdout().strip().split('\n') +sources = {} +foreach l : lines + a = l.split(' = ') + var_name = a[0] + file_list = a[1].split() + sources += {var_name: files(file_list)} +endforeach + +subdir('include') +subdir('celt') +subdir('silk') +subdir('src') + +configure_file(output: 'config.h', configuration: opus_conf) + +if not opt_tests.disabled() + subdir('celt/tests') + subdir('silk/tests') + subdir('tests') +endif + +# pkg-config files (not using pkg module so we can use the existing .pc.in file) +pkgconf = configuration_data() + +pkgconf.set('prefix', join_paths(get_option('prefix'))) +pkgconf.set('exec_prefix', '${prefix}') +pkgconf.set('libdir', '${prefix}/@0@'.format(get_option('libdir'))) +pkgconf.set('includedir', '${prefix}/@0@'.format(get_option('includedir'))) +pkgconf.set('VERSION', opus_version) +pkgconf.set('PC_BUILD', pc_build) +pkgconf.set('LIBM', libm.found() ? '-lm' : '') + +pkg_install_dir = '@0@/pkgconfig'.format(get_option('libdir')) + +configure_file(input : 'opus.pc.in', + output : 'opus.pc', + configuration : pkgconf, + install_dir : pkg_install_dir) + +# The uninstalled one has hardcoded libtool + static lib stuff, skip it for now +#configure_file(input : 'opus-uninstalled.pc.in', +# output : 'opus-uninstalled.pc', +# configuration : pkgconf, +# install : false) + +doxygen = find_program('doxygen', required: get_option('docs')) +if doxygen.found() + subdir('doc') +endif + +summary( + { + 'C99 var arrays': opus_conf.has('VAR_ARRAYS'), + 'C99 lrintf': opus_conf.has('HAVE_LRINTF'), + 'Use alloca': msg_use_alloca, + }, + section: 'Compiler support', + bool_yn: true, + list_sep: ', ', +) + +# Parse optimization status +foreach status : [['inline_optimization', opt_asm], + ['asm_optimization', opt_asm], + ['intrinsics_support', opt_intrinsics]] + res = status[0] + opt = status[1] + resval = get_variable(res) + if opt.disabled() + set_variable(res, 'disabled') + elif resval.length() == 0 + if host_cpu_family not in ['arm', 'aarch64', 'x86', 'x86_64'] + set_variable(res, 'No optimizations for your platform, please send patches') + else + set_variable(res, 'none') + endif + endif +endforeach + +summary( + { + 'Floating point support': not opt_fixed_point, + 'Fast float approximations': opt_float_approx, + 'Fixed point debugging': opt_fixed_point_debug, + 'Inline assembly optimizations': inline_optimization, + 'External assembly optimizations': asm_optimization, + 'Intrinsics optimizations': intrinsics_support, + 'Run-time CPU detection': rtcd_support, + }, + section: 'Optimizations', + bool_yn: true, + list_sep: ', ', +) +summary( + { + 'Custom modes': opt_custom_modes, + 'Assertions': opt_assertions, + 'Hardening': opt_hardening, + 'Fuzzing': opt_fuzzing, + 'Check ASM': opt_check_asm, + 'API documentation': doxygen.found(), + 'Extra programs': not extra_programs.disabled(), + 'Tests': not opt_tests.disabled(), + }, + section: 'General configuration', + bool_yn: true, + list_sep: ', ', +) diff --git a/lib/opus/meson/get-version.py b/lib/opus/meson/get-version.py new file mode 100755 index 00000000..d3835f13 --- /dev/null +++ b/lib/opus/meson/get-version.py @@ -0,0 +1,86 @@ +#!/usr/bin/env python3 +# +# Opus get-version.py +# +# Extracts versions for build: +# - Opus package version based on 'git describe' or $srcroot/package_version +# - libtool version based on configure.ac +# - macos lib version based on configure.ac +# +# Usage: +# get-version.py [--package-version | --libtool-version | --darwin-version] +import argparse +import subprocess +import os +import sys +import shutil + +if __name__ == '__main__': + arg_parser = argparse.ArgumentParser(description='Extract Opus package version or libtool version') + group = arg_parser.add_mutually_exclusive_group(required=True) + group.add_argument('--libtool-version', action='store_true') + group.add_argument('--package-version', action='store_true') + group.add_argument('--darwin-version', action='store_true') + args = arg_parser.parse_args() + + srcroot = os.path.normpath(os.path.join(os.path.dirname(__file__), '..')) + + # package version + if args.package_version: + package_version = None + + # check if git checkout + git_dir = os.path.join(srcroot, '.git') + is_git = os.path.isdir(git_dir) or os.path.isfile(git_dir) + have_git = shutil.which('git') is not None + + if is_git and have_git: + git_cmd = subprocess.run(['git', '--git-dir=' + git_dir, 'describe', 'HEAD'], stdout=subprocess.PIPE) + if git_cmd.returncode: + print('ERROR: Could not extract package version via `git describe` in', srcroot, file=sys.stderr) + sys.exit(-1) + package_version = git_cmd.stdout.decode('ascii').strip().lstrip('v') + else: + with open(os.path.join(srcroot, 'package_version'), 'r') as f: + for line in f: + if line.startswith('PACKAGE_VERSION="'): + package_version = line[17:].strip().lstrip('v').rstrip('"') + if package_version: + break + + if not package_version: + print('ERROR: Could not extract package version from package_version file in', srcroot, file=sys.stderr) + sys.exit(-1) + + print(package_version) + sys.exit(0) + + # libtool version + darwin version + elif args.libtool_version or args.darwin_version: + opus_lt_cur = None + opus_lt_rev = None + opus_lt_age = None + + with open(os.path.join(srcroot, 'configure.ac'), 'r') as f: + for line in f: + if line.strip().startswith('OPUS_LT_CURRENT='): + opus_lt_cur = line[16:].strip() + elif line.strip().startswith('OPUS_LT_REVISION='): + opus_lt_rev = line[17:].strip() + elif line.strip().startswith('OPUS_LT_AGE='): + opus_lt_age = line[12:].strip() + + if opus_lt_cur and opus_lt_rev and opus_lt_age: + opus_lt_cur = int(opus_lt_cur) + opus_lt_rev = int(opus_lt_rev) + opus_lt_age = int(opus_lt_age) + if args.libtool_version: + print('{}.{}.{}'.format(opus_lt_cur - opus_lt_age, opus_lt_age, opus_lt_rev)) + elif args.darwin_version: + print('{}.{}.{}'.format(opus_lt_cur + 1, 0, 0)) + sys.exit(0) + else: + print('ERROR: Could not extract libtool version from configure.ac file in', srcroot, file=sys.stderr) + sys.exit(-1) + else: + sys.exit(-1) diff --git a/lib/opus/meson/read-sources-list.py b/lib/opus/meson/read-sources-list.py new file mode 100755 index 00000000..fcbec501 --- /dev/null +++ b/lib/opus/meson/read-sources-list.py @@ -0,0 +1,28 @@ +#!/usr/bin/env python3 +# +# opus/read-sources-list.py +# +# Parses .mk files and extracts list of source files. +# Prints one line per source file list, with filenames space-separated. + +import sys + +if len(sys.argv) < 2: + sys.exit('Usage: {} sources_foo.mk [sources_bar.mk...]'.format(sys.argv[0])) + +for input_fn in sys.argv[1:]: + with open(input_fn, 'r', encoding='utf8') as f: + text = f.read() + text = text.replace('\\\n', '') + + # Remove empty lines + lines = [line for line in text.split('\n') if line.strip()] + + # Print SOURCES_XYZ = file1.c file2.c + for line in lines: + values = line.strip().split('=', maxsplit=2) + if len(values) != 2: + raise RuntimeError('Unable to parse line "{}" from file "{}"'.format(line, input_fn)) + var, files = values + sources_list = [f for f in files.split(' ') if f] + print(var.strip(), '=', ' '.join(sources_list)) diff --git a/lib/opus/meson_options.txt b/lib/opus/meson_options.txt new file mode 100644 index 00000000..360ff263 --- /dev/null +++ b/lib/opus/meson_options.txt @@ -0,0 +1,22 @@ +# Optimizations +option('fixed-point', type : 'boolean', value : false, description : 'Compile without floating point (for machines without a fast enough FPU') +option('fixed-point-debug', type : 'boolean', value : false, description : 'Debug fixed-point implementation') +option('float-api', type : 'boolean', value : true, description : 'Compile with or without the floating point API (for machines with no float library') +option('float-approx', type : 'boolean', value : false, description : 'Enable fast approximations for floating point (not supported on all platforms)') +option('rtcd', type : 'feature', value : 'auto', description : 'Run-time CPU capabilities detection') +option('asm', type : 'feature', value : 'auto', description : 'Assembly optimizations for ARM (fixed-point)') +option('intrinsics', type : 'feature', value : 'auto', description : 'Intrinsics optimizations for ARM NEON or x86') + +option('custom-modes', type : 'boolean', value : false, description : 'Enable non-Opus modes, e.g. 44.1 kHz & 2^n frames') +option('extra-programs', type : 'feature', value : 'auto', description : 'Extra programs (demo and tests)') +option('assertions', type : 'boolean', value : false, description : 'Additional software error checking') +option('hardening', type : 'boolean', value : true, description : 'Run-time checks that are cheap and safe for use in production') +option('fuzzing', type : 'boolean', value : false, description : 'Causes the encoder to make random decisions') +option('check-asm', type : 'boolean', value : false, description : 'Run bit-exactness checks between optimized and c implementations') + +# common feature options +option('tests', type : 'feature', value : 'auto', description : 'Build tests') +option('docs', type: 'feature', value: 'auto', description: 'Build API documentation') + +# other options +option('docdir', type: 'string', value: 'doc/opus', description: 'Directory to install documentation into (default: DATADIR/doc/opus') diff --git a/lib/opus/opus-uninstalled.pc.in b/lib/opus/opus-uninstalled.pc.in new file mode 100644 index 00000000..19f5c933 --- /dev/null +++ b/lib/opus/opus-uninstalled.pc.in @@ -0,0 +1,12 @@ +# Opus codec reference implementation uninstalled pkg-config file + +libdir=${pcfiledir}/.libs +includedir=${pcfiledir} + +Name: opus uninstalled +Description: Opus IETF audio codec (not installed, @PC_BUILD@) +Version: @VERSION@ +Requires: +Conflicts: +Libs: ${libdir}/libopus.la @LIBM@ +Cflags: -I${pcfiledir}/@top_srcdir@/include diff --git a/lib/opus/opus.m4 b/lib/opus/opus.m4 new file mode 100644 index 00000000..263470d4 --- /dev/null +++ b/lib/opus/opus.m4 @@ -0,0 +1,117 @@ +# Configure paths for libopus +# Gregory Maxwell 08-30-2012 +# Shamelessly stolen from Jack Moffitt (libogg) who +# Shamelessly stole from Owen Taylor and Manish Singh + +dnl XIPH_PATH_OPUS([ACTION-IF-FOUND [, ACTION-IF-NOT-FOUND]]) +dnl Test for libopus, and define OPUS_CFLAGS and OPUS_LIBS +dnl +AC_DEFUN([XIPH_PATH_OPUS], +[dnl +dnl Get the cflags and libraries +dnl +AC_ARG_WITH(opus,AC_HELP_STRING([--with-opus=PFX],[Prefix where opus is installed (optional)]), opus_prefix="$withval", opus_prefix="") +AC_ARG_WITH(opus-libraries,AC_HELP_STRING([--with-opus-libraries=DIR],[Directory where the opus library is installed (optional)]), opus_libraries="$withval", opus_libraries="") +AC_ARG_WITH(opus-includes,AC_HELP_STRING([--with-opus-includes=DIR],[Directory where the opus header files are installed (optional)]), opus_includes="$withval", opus_includes="") +AC_ARG_ENABLE(opustest,AC_HELP_STRING([--disable-opustest],[Do not try to compile and run a test opus program]),, enable_opustest=yes) + + if test "x$opus_libraries" != "x" ; then + OPUS_LIBS="-L$opus_libraries" + elif test "x$opus_prefix" = "xno" || test "x$opus_prefix" = "xyes" ; then + OPUS_LIBS="" + elif test "x$opus_prefix" != "x" ; then + OPUS_LIBS="-L$opus_prefix/lib" + elif test "x$prefix" != "xNONE" ; then + OPUS_LIBS="-L$prefix/lib" + fi + + if test "x$opus_prefix" != "xno" ; then + OPUS_LIBS="$OPUS_LIBS -lopus" + fi + + if test "x$opus_includes" != "x" ; then + OPUS_CFLAGS="-I$opus_includes" + elif test "x$opus_prefix" = "xno" || test "x$opus_prefix" = "xyes" ; then + OPUS_CFLAGS="" + elif test "x$opus_prefix" != "x" ; then + OPUS_CFLAGS="-I$opus_prefix/include" + elif test "x$prefix" != "xNONE"; then + OPUS_CFLAGS="-I$prefix/include" + fi + + AC_MSG_CHECKING(for Opus) + if test "x$opus_prefix" = "xno" ; then + no_opus="disabled" + enable_opustest="no" + else + no_opus="" + fi + + + if test "x$enable_opustest" = "xyes" ; then + ac_save_CFLAGS="$CFLAGS" + ac_save_LIBS="$LIBS" + CFLAGS="$CFLAGS $OPUS_CFLAGS" + LIBS="$LIBS $OPUS_LIBS" +dnl +dnl Now check if the installed Opus is sufficiently new. +dnl + rm -f conf.opustest + AC_TRY_RUN([ +#include +#include +#include +#include + +int main (void) +{ + system("touch conf.opustest"); + return 0; +} + +],, no_opus=yes,[echo $ac_n "cross compiling; assumed OK... $ac_c"]) + CFLAGS="$ac_save_CFLAGS" + LIBS="$ac_save_LIBS" + fi + + if test "x$no_opus" = "xdisabled" ; then + AC_MSG_RESULT(no) + ifelse([$2], , :, [$2]) + elif test "x$no_opus" = "x" ; then + AC_MSG_RESULT(yes) + ifelse([$1], , :, [$1]) + else + AC_MSG_RESULT(no) + if test -f conf.opustest ; then + : + else + echo "*** Could not run Opus test program, checking why..." + CFLAGS="$CFLAGS $OPUS_CFLAGS" + LIBS="$LIBS $OPUS_LIBS" + AC_TRY_LINK([ +#include +#include +], [ return 0; ], + [ echo "*** The test program compiled, but did not run. This usually means" + echo "*** that the run-time linker is not finding Opus or finding the wrong" + echo "*** version of Opus. If it is not finding Opus, you'll need to set your" + echo "*** LD_LIBRARY_PATH environment variable, or edit /etc/ld.so.conf to point" + echo "*** to the installed location Also, make sure you have run ldconfig if that" + echo "*** is required on your system" + echo "***" + echo "*** If you have an old version installed, it is best to remove it, although" + echo "*** you may also be able to get things to work by modifying LD_LIBRARY_PATH"], + [ echo "*** The test program failed to compile or link. See the file config.log for the" + echo "*** exact error that occurred. This usually means Opus was incorrectly installed" + echo "*** or that you have moved Opus since it was installed." ]) + CFLAGS="$ac_save_CFLAGS" + LIBS="$ac_save_LIBS" + fi + OPUS_CFLAGS="" + OPUS_LIBS="" + ifelse([$2], , :, [$2]) + fi + AC_SUBST(OPUS_CFLAGS) + AC_SUBST(OPUS_LIBS) + rm -f conf.opustest +]) diff --git a/lib/opus/opus.pc.in b/lib/opus/opus.pc.in new file mode 100644 index 00000000..6946e7de --- /dev/null +++ b/lib/opus/opus.pc.in @@ -0,0 +1,16 @@ +# Opus codec reference implementation pkg-config file + +prefix=@prefix@ +exec_prefix=@exec_prefix@ +libdir=@libdir@ +includedir=@includedir@ + +Name: Opus +Description: Opus IETF audio codec (@PC_BUILD@ build) +URL: https://opus-codec.org/ +Version: @VERSION@ +Requires: +Conflicts: +Libs: -L${libdir} -lopus +Libs.private: @LIBM@ +Cflags: -I${includedir}/opus diff --git a/lib/opus/opus_headers.mk b/lib/opus/opus_headers.mk new file mode 100644 index 00000000..27596f2a --- /dev/null +++ b/lib/opus/opus_headers.mk @@ -0,0 +1,9 @@ +OPUS_HEAD = \ +include/opus.h \ +include/opus_multistream.h \ +include/opus_projection.h \ +src/opus_private.h \ +src/analysis.h \ +src/mapping_matrix.h \ +src/mlp.h \ +src/tansig_table.h diff --git a/lib/opus/opus_sources.mk b/lib/opus/opus_sources.mk new file mode 100644 index 00000000..44153b57 --- /dev/null +++ b/lib/opus/opus_sources.mk @@ -0,0 +1,16 @@ +OPUS_SOURCES = \ +src/opus.c \ +src/opus_decoder.c \ +src/opus_encoder.c \ +src/opus_multistream.c \ +src/opus_multistream_encoder.c \ +src/opus_multistream_decoder.c \ +src/repacketizer.c \ +src/opus_projection_encoder.c \ +src/opus_projection_decoder.c \ +src/mapping_matrix.c + +OPUS_SOURCES_FLOAT = \ +src/analysis.c \ +src/mlp.c \ +src/mlp_data.c diff --git a/lib/opus/releases.sha2 b/lib/opus/releases.sha2 new file mode 100644 index 00000000..334976b1 --- /dev/null +++ b/lib/opus/releases.sha2 @@ -0,0 +1,79 @@ +b2f75c4ac5ab837845eb028413fae2a28754bfb0a6d76416e2af1441ef447649 opus-0.9.0.tar.gz +4e379a98ba95bbbfe9087ef10fdd05c8ac9060b6d695f587ea82a7b43a0df4fe opus-0.9.10.tar.gz +b1cad6846a8f819a141009fe3f8f10c946e8eff7e9c2339cd517bb136cc59eae opus-0.9.14.tar.gz +206221afc47b87496588013bd4523e1e9f556336c0813f4372773fc536dd4293 opus-0.9.1.tar.gz +6e85c1b57e1d7b7dfe2928bf92586b96b73a9067e054ede45bd8e6d24bd30582 opus-0.9.2.tar.gz +d916e34c18a396eb7dffc47af754f441af52a290b761e20db9aedb65928c699e opus-0.9.3.tar.gz +53801066fa97329768e7b871fd1495740269ec46802e1c9051aa7e78c6edee5b opus-0.9.5.tar.gz +3bfaeb25f4b4a625a0bc994d6fc6f6776a05193f60099e0a99f7530c6b256309 opus-0.9.6.tar.gz +1b69772c31c5cbaa43d1dfa5b1c495fc29712e8e0ff69d6f8ad46459e5c6715f opus-0.9.7.tar.gz +4aa30d2e0652ffb4a7a22cc8a29c4ce78267626f560a2d9213b1d2d4e618cf36 opus-0.9.8.tar.gz +2f62359f09151fa3b242040dc9b4c5b6bda15557c5daea59c8420f1a2ff328b7 opus-0.9.9.tar.gz +43bcea51afa531f32a6a5fdd9cba4bd496993e26a141217db3cccce6caa7cd74 opus-1.0.0-rc.tar.gz +9250fcc74472d45c1e14745542ec9c8d09982538aefed56962495614be3e0d2d opus-1.0.0.tar.gz +76bc0a31502a51dae9ab737b4db043b9ecfcd0b5861f0bfda41b662bd5b92227 opus-1.0.1-rc2.tar.gz +3de8d6809dac38971ebb305532d4ea532519d3bed08985f25d6c557f9ce5e8ff opus-1.0.1-rc3.tar.gz +8044397a6365a07117b08cbe8f9818bf7c93746908806ba74a2917187bbdda5f opus-1.0.1-rc.tar.gz +80fa5c3caf2ac0fd68f8a22cce1564fc46b368c773a17554887d0066fe1841ef opus-1.0.1.tar.gz +da615edbee5d019c1833071d69a4782c19f178cf9ca1401375036ecef25cd78a opus-1.0.2.tar.gz +191a089c92dbc403de6980463dd3604b65beb12d283c607e246c8076363cb49c opus-1.0.3.tar.gz +a8d40efe87f6c3e76725391457d46277878c7a816ae1642843261463133fa5c8 opus-1.1-alpha.tar.gz +ec1784287f385aef994b64734aaecae04860e61aa50fc6eef6643fa7e40dd193 opus-1.1-beta.tar.gz +8aa16360f59a94d3e38f38f28d24039f7663179682cbae82aa42f1dd9e52e6ed opus-1.1-rc.tar.gz +ebc87a086d4fe677c5e42d56888b1fd25af858e4179eae4f8656270410dffac3 opus-1.1-rc2.tar.gz +cbfd09c58cc10a4d3fcb727ad5d46d7bb549f8185ac922ee28b4581b52a7bee9 opus-1.1-rc3.tar.gz +b9727015a58affcf3db527322bf8c4d2fcf39f5f6b8f15dbceca20206cbe1d95 opus-1.1.tar.gz +0c668639dcd16b14709fc9dc49e6686606f5a256f2eaa1ebaa2f39a66f8626cd opus-1.1.1-beta.tar.gz +66f2a5877c8803dc9a5a44b4f3d0bdc8f06bd066324222d144eb255612b68152 opus-1.1.1-rc.tar.gz +9b84ff56bd7720d5554103c557664efac2b8b18acc4bbcc234cb881ab9a3371e opus-1.1.1.tar.gz +0e290078e31211baa7b5886bcc8ab6bc048b9fc83882532da4a1a45e58e907fd opus-1.1.2.tar.gz +58b6fe802e7e30182e95d0cde890c0ace40b6f125cffc50635f0ad2eef69b633 opus-1.1.3.tar.gz +9122b6b380081dd2665189f97bfd777f04f92dc3ab6698eea1dbb27ad59d8692 opus-1.1.4.tar.gz +eb84981ca0f40a3e5d5e58d2e8582cb2fee05a022825a6dfe14d14b04eb563e4 opus-1.1.5.tar.gz +654a9bebb73266271a28edcfff431e4cfd9bfcde71f42849a0cdd73bece803a7 opus-1.2-alpha.tar.gz +c0e90507259cf21ce7b2c82fb9ac55367d8543dae91cc3d4d2c59afd37f44023 opus-1.2-alpha2.tar.gz +291e979a8a2fb679ed35a5dff5d761a9d9a5e22586fd07934ed94461e2636c7a opus-1.2-beta.tar.gz +85343fdaed96529d94c1e1f3a210fa51240d04ca62fa01e97ef02f88020c2ce9 opus-1.2-rc1.tar.gz +77db45a87b51578fbc49555ef1b10926179861d854eb2613207dc79d9ec0a9a9 opus-1.2.tar.gz +cfafd339ccd9c5ef8d6ab15d7e1a412c054bf4cb4ecbbbcc78c12ef2def70732 opus-1.2.1.tar.gz +7f56e058c9549d03ae35511ad9e16ef6d1eb257836830d54abff0f495f17e187 opus-1.3-beta.tar.gz +96fa28598e8ccd558b297277ad59a045c551ba0e06d65a9675938e084f837669 opus-1.3-rc.tar.gz +f6bab321fb81db984766f1e4d340a9e71a5ca2c5d4d53f4ee072e84afda271ca opus-1.3-rc2.tar.gz +4f3d69aefdf2dbaf9825408e452a8a414ffc60494c70633560700398820dc550 opus-1.3.tar.gz +65b58e1e25b2a114157014736a3d9dfeaad8d41be1c8179866f144a2fb44ff9d opus-1.3.1.tar.gz +94ac78ca4f74c4e43bc9fe4ec1ad0aa36f38ab90f45b0727c40dd1e96096e767 opus_testvectors-draft11.tar.gz +94ac78ca4f74c4e43bc9fe4ec1ad0aa36f38ab90f45b0727c40dd1e96096e767 opus_testvectors.tar.gz +6b26a22f9ba87b2b836906a9bb7afec5f8e54d49553b1200382520ee6fedfa55 opus_testvectors-rfc8251.tar.gz +5d2b99757bcb628bab2611f3ed27af6f35276ce3abc96c0ed4399d6c6463dda5 opus-tools-0.1.2.tar.gz +008317297d6ce84f84992abf8cc948a048a4fa135e1d1caf429fafde8965a792 opus-tools-0.1.3.tar.gz +de80485c5afa1fd83c0e16a0dd4860470c872997a7dd0a58e99b2ee8a93e5168 opus-tools-0.1.4.tar.gz +76678d0eb7a9b3d793bd0243f9ced9ab0ecdab263f5232ed940c8f5795fb0405 opus-tools-0.1.5.tar.gz +cc86dbc2a4d76da7e1ed9afee85448c8f798c465a5412233f178783220f3a2c1 opus-tools-0.1.6.tar.gz +e0f08d301555dffc417604269b5a85d2bd197f259c7d6c957f370ffd33d6d9cd opus-tools-0.1.7.tar.gz +e4e188579ea1c4e4d5066460d4a7214a7eafe3539e9a4466fdc98af41ba4a2f6 opus-tools-0.1.8.tar.gz +b1873dd78c7fbc98cf65d6e10cfddb5c2c03b3af93f922139a2104baedb4643a opus-tools-0.1.9.tar.gz +a2357532d19471b70666e0e0ec17d514246d8b3cb2eb168f68bb0f6fd372b28c opus-tools-0.1.10.tar.gz +b4e56cb00d3e509acfba9a9b627ffd8273b876b4e2408642259f6da28fa0ff86 opus-tools-0.2.tar.gz +bd6d14e8897a2f80065ef34a516c70e74f8e00060abdbc238e79e5f99bca3e96 libopusenc-0.1.tar.gz +02e6e0b14cbbe0569d948a46420f9c9a81d93bba32dc576a4007cbf96da68ef3 libopusenc-0.1.1.tar.gz +c79e95eeee43a0b965e9b2c59a243763a8f8b0a7e71441df2aa9084f6171c73a libopusenc-0.2.tar.gz +8298db61a8d3d63e41c1a80705baa8ce9ff3f50452ea7ec1c19a564fe106cbb9 libopusenc-0.2.1.tar.gz +8071b968475c1a17f54b6840d6de9d9ee20f930e827b0401abe3c4cf4f3bf30a opusfile-0.1.tar.gz +b4a678b3b6c4adfb6aff1f67ef658becfe146ea7c7ff228e99543762171557f9 opusfile-0.2.tar.gz +4248927f2c4e316ea5b84fb02bd100bfec8fa4624a6910d77f0af7f0c6cb8baa opusfile-0.3.tar.gz +9836ea11706c44f36de92c4c9b1248e03a4c521e7fb2cff18a0cb4f8b0e79140 opusfile-0.4.tar.gz +f187906b1b35f7f0d7de6a759b4aab512a9279d23adb35d8009e7e33bd6a922a opusfile-0.4.zip +2ce52d006aeeec9f10260dbe3073c4636954a1ab19c82b8baafefe0180aa4a39 opusfile-0.5.tar.gz +b940d62beb15b5974764574b9f265481fe5b6ee16902fb705727546caf956261 opusfile-0.5.zip +2428717b356e139f18ed2fdb5ad990b5654a238907a0058200b39c46a7d03ea6 opusfile-0.6.tar.gz +753339225193df605372944889023b9b3c5378d672e8784d69fa241cd465278c opusfile-0.6.zip +9e2bed13bc729058591a0f1cab2505e8cfd8e7ac460bf10a78bcc3b125e7c301 opusfile-0.7.tar.gz +346967d7989bb83b05949483b76bd0f69a12c59bd8b4457e864902b52bb0ac34 opusfile-0.7.zip +2c231ed3cfaa1b3173f52d740e5bbd77d51b9dfecb87014b404917fba4b855a4 opusfile-0.8.tar.gz +89dff4342c3b789574cbea5c57f11b96d4ebe4d28ab90248c1783ea569b1e9e3 opusfile-0.8.zip +f75fb500e40b122775ac1a71ad80c4477698842a8fe9da4a1b4a1a9f16e4e979 opusfile-0.9.tar.gz +e9591da4d4c9e857436c2d46a28a9e470fa5355ea5a76d4d582f137d18755d36 opusfile-0.9.zip +48e03526ba87ef9cf5f1c47b5ebe3aa195bd89b912a57060c36184a6cd19412f opusfile-0.10.tar.gz +9d9e95d01817ecf48bf6daaea8f071f9b45bd1751ca1fc8ce50e5075eb2bc3c8 opusfile-0.10.zip +74ce9b6cf4da103133e7b5c95df810ceb7195471e1162ed57af415fabf5603bf opusfile-0.11.tar.gz +23c5168026c4f1fc34843650135b409d0fc8cf452508163b4ece8077256ac6ff opusfile-0.11.zip diff --git a/lib/opus/scripts/dump_rnn.py b/lib/opus/scripts/dump_rnn.py new file mode 100755 index 00000000..dd66403b --- /dev/null +++ b/lib/opus/scripts/dump_rnn.py @@ -0,0 +1,57 @@ +#!/usr/bin/python + +from __future__ import print_function + +from keras.models import Sequential +from keras.layers import Dense +from keras.layers import LSTM +from keras.layers import GRU +from keras.models import load_model +from keras import backend as K + +import numpy as np + +def printVector(f, vector, name): + v = np.reshape(vector, (-1)); + #print('static const float ', name, '[', len(v), '] = \n', file=f) + f.write('static const opus_int16 {}[{}] = {{\n '.format(name, len(v))) + for i in range(0, len(v)): + f.write('{}'.format(int(round(8192*v[i])))) + if (i!=len(v)-1): + f.write(',') + else: + break; + if (i%8==7): + f.write("\n ") + else: + f.write(" ") + #print(v, file=f) + f.write('\n};\n\n') + return; + +def binary_crossentrop2(y_true, y_pred): + return K.mean(2*K.abs(y_true-0.5) * K.binary_crossentropy(y_pred, y_true), axis=-1) + + +model = load_model("weights.hdf5", custom_objects={'binary_crossentrop2': binary_crossentrop2}) + +weights = model.get_weights() + +f = open('rnn_weights.c', 'w') + +f.write('/*This file is automatically generated from a Keras model*/\n\n') +f.write('#ifdef HAVE_CONFIG_H\n#include "config.h"\n#endif\n\n#include "mlp.h"\n\n') + +printVector(f, weights[0], 'layer0_weights') +printVector(f, weights[1], 'layer0_bias') +printVector(f, weights[2], 'layer1_weights') +printVector(f, weights[3], 'layer1_recur_weights') +printVector(f, weights[4], 'layer1_bias') +printVector(f, weights[5], 'layer2_weights') +printVector(f, weights[6], 'layer2_bias') + +f.write('const DenseLayer layer0 = {\n layer0_bias,\n layer0_weights,\n 25, 16, 0\n};\n\n') +f.write('const GRULayer layer1 = {\n layer1_bias,\n layer1_weights,\n layer1_recur_weights,\n 16, 12\n};\n\n') +f.write('const DenseLayer layer2 = {\n layer2_bias,\n layer2_weights,\n 12, 2, 1\n};\n\n') + +f.close() diff --git a/lib/opus/scripts/rnn_train.py b/lib/opus/scripts/rnn_train.py new file mode 100755 index 00000000..ffdaa1e7 --- /dev/null +++ b/lib/opus/scripts/rnn_train.py @@ -0,0 +1,67 @@ +#!/usr/bin/python + +from __future__ import print_function + +from keras.models import Sequential +from keras.models import Model +from keras.layers import Input +from keras.layers import Dense +from keras.layers import LSTM +from keras.layers import GRU +from keras.layers import SimpleRNN +from keras.layers import Dropout +from keras import losses +import h5py + +from keras import backend as K +import numpy as np + +def binary_crossentrop2(y_true, y_pred): + return K.mean(2*K.abs(y_true-0.5) * K.binary_crossentropy(y_pred, y_true), axis=-1) + +print('Build model...') +#model = Sequential() +#model.add(Dense(16, activation='tanh', input_shape=(None, 25))) +#model.add(GRU(12, dropout=0.0, recurrent_dropout=0.0, activation='tanh', recurrent_activation='sigmoid', return_sequences=True)) +#model.add(Dense(2, activation='sigmoid')) + +main_input = Input(shape=(None, 25), name='main_input') +x = Dense(16, activation='tanh')(main_input) +x = GRU(12, dropout=0.1, recurrent_dropout=0.1, activation='tanh', recurrent_activation='sigmoid', return_sequences=True)(x) +x = Dense(2, activation='sigmoid')(x) +model = Model(inputs=main_input, outputs=x) + +batch_size = 64 + +print('Loading data...') +with h5py.File('features.h5', 'r') as hf: + all_data = hf['features'][:] +print('done.') + +window_size = 1500 + +nb_sequences = len(all_data)/window_size +print(nb_sequences, ' sequences') +x_train = all_data[:nb_sequences*window_size, :-2] +x_train = np.reshape(x_train, (nb_sequences, window_size, 25)) + +y_train = np.copy(all_data[:nb_sequences*window_size, -2:]) +y_train = np.reshape(y_train, (nb_sequences, window_size, 2)) + +all_data = 0; +x_train = x_train.astype('float32') +y_train = y_train.astype('float32') + +print(len(x_train), 'train sequences. x shape =', x_train.shape, 'y shape = ', y_train.shape) + +# try using different optimizers and different optimizer configs +model.compile(loss=binary_crossentrop2, + optimizer='adam', + metrics=['binary_accuracy']) + +print('Train...') +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=200, + validation_data=(x_train, y_train)) +model.save("newweights.hdf5") diff --git a/lib/opus/silk/A2NLSF.c b/lib/opus/silk/A2NLSF.c new file mode 100644 index 00000000..b487686f --- /dev/null +++ b/lib/opus/silk/A2NLSF.c @@ -0,0 +1,267 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +/* Conversion between prediction filter coefficients and NLSFs */ +/* Requires the order to be an even number */ +/* A piecewise linear approximation maps LSF <-> cos(LSF) */ +/* Therefore the result is not accurate NLSFs, but the two */ +/* functions are accurate inverses of each other */ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "tables.h" + +/* Number of binary divisions, when not in low complexity mode */ +#define BIN_DIV_STEPS_A2NLSF_FIX 3 /* must be no higher than 16 - log2( LSF_COS_TAB_SZ_FIX ) */ +#define MAX_ITERATIONS_A2NLSF_FIX 16 + +/* Helper function for A2NLSF(..) */ +/* Transforms polynomials from cos(n*f) to cos(f)^n */ +static OPUS_INLINE void silk_A2NLSF_trans_poly( + opus_int32 *p, /* I/O Polynomial */ + const opus_int dd /* I Polynomial order (= filter order / 2 ) */ +) +{ + opus_int k, n; + + for( k = 2; k <= dd; k++ ) { + for( n = dd; n > k; n-- ) { + p[ n - 2 ] -= p[ n ]; + } + p[ k - 2 ] -= silk_LSHIFT( p[ k ], 1 ); + } +} +/* Helper function for A2NLSF(..) */ +/* Polynomial evaluation */ +static OPUS_INLINE opus_int32 silk_A2NLSF_eval_poly( /* return the polynomial evaluation, in Q16 */ + opus_int32 *p, /* I Polynomial, Q16 */ + const opus_int32 x, /* I Evaluation point, Q12 */ + const opus_int dd /* I Order */ +) +{ + opus_int n; + opus_int32 x_Q16, y32; + + y32 = p[ dd ]; /* Q16 */ + x_Q16 = silk_LSHIFT( x, 4 ); + + if ( opus_likely( 8 == dd ) ) + { + y32 = silk_SMLAWW( p[ 7 ], y32, x_Q16 ); + y32 = silk_SMLAWW( p[ 6 ], y32, x_Q16 ); + y32 = silk_SMLAWW( p[ 5 ], y32, x_Q16 ); + y32 = silk_SMLAWW( p[ 4 ], y32, x_Q16 ); + y32 = silk_SMLAWW( p[ 3 ], y32, x_Q16 ); + y32 = silk_SMLAWW( p[ 2 ], y32, x_Q16 ); + y32 = silk_SMLAWW( p[ 1 ], y32, x_Q16 ); + y32 = silk_SMLAWW( p[ 0 ], y32, x_Q16 ); + } + else + { + for( n = dd - 1; n >= 0; n-- ) { + y32 = silk_SMLAWW( p[ n ], y32, x_Q16 ); /* Q16 */ + } + } + return y32; +} + +static OPUS_INLINE void silk_A2NLSF_init( + const opus_int32 *a_Q16, + opus_int32 *P, + opus_int32 *Q, + const opus_int dd +) +{ + opus_int k; + + /* Convert filter coefs to even and odd polynomials */ + P[dd] = silk_LSHIFT( 1, 16 ); + Q[dd] = silk_LSHIFT( 1, 16 ); + for( k = 0; k < dd; k++ ) { + P[ k ] = -a_Q16[ dd - k - 1 ] - a_Q16[ dd + k ]; /* Q16 */ + Q[ k ] = -a_Q16[ dd - k - 1 ] + a_Q16[ dd + k ]; /* Q16 */ + } + + /* Divide out zeros as we have that for even filter orders, */ + /* z = 1 is always a root in Q, and */ + /* z = -1 is always a root in P */ + for( k = dd; k > 0; k-- ) { + P[ k - 1 ] -= P[ k ]; + Q[ k - 1 ] += Q[ k ]; + } + + /* Transform polynomials from cos(n*f) to cos(f)^n */ + silk_A2NLSF_trans_poly( P, dd ); + silk_A2NLSF_trans_poly( Q, dd ); +} + +/* Compute Normalized Line Spectral Frequencies (NLSFs) from whitening filter coefficients */ +/* If not all roots are found, the a_Q16 coefficients are bandwidth expanded until convergence. */ +void silk_A2NLSF( + opus_int16 *NLSF, /* O Normalized Line Spectral Frequencies in Q15 (0..2^15-1) [d] */ + opus_int32 *a_Q16, /* I/O Monic whitening filter coefficients in Q16 [d] */ + const opus_int d /* I Filter order (must be even) */ +) +{ + opus_int i, k, m, dd, root_ix, ffrac; + opus_int32 xlo, xhi, xmid; + opus_int32 ylo, yhi, ymid, thr; + opus_int32 nom, den; + opus_int32 P[ SILK_MAX_ORDER_LPC / 2 + 1 ]; + opus_int32 Q[ SILK_MAX_ORDER_LPC / 2 + 1 ]; + opus_int32 *PQ[ 2 ]; + opus_int32 *p; + + /* Store pointers to array */ + PQ[ 0 ] = P; + PQ[ 1 ] = Q; + + dd = silk_RSHIFT( d, 1 ); + + silk_A2NLSF_init( a_Q16, P, Q, dd ); + + /* Find roots, alternating between P and Q */ + p = P; /* Pointer to polynomial */ + + xlo = silk_LSFCosTab_FIX_Q12[ 0 ]; /* Q12*/ + ylo = silk_A2NLSF_eval_poly( p, xlo, dd ); + + if( ylo < 0 ) { + /* Set the first NLSF to zero and move on to the next */ + NLSF[ 0 ] = 0; + p = Q; /* Pointer to polynomial */ + ylo = silk_A2NLSF_eval_poly( p, xlo, dd ); + root_ix = 1; /* Index of current root */ + } else { + root_ix = 0; /* Index of current root */ + } + k = 1; /* Loop counter */ + i = 0; /* Counter for bandwidth expansions applied */ + thr = 0; + while( 1 ) { + /* Evaluate polynomial */ + xhi = silk_LSFCosTab_FIX_Q12[ k ]; /* Q12 */ + yhi = silk_A2NLSF_eval_poly( p, xhi, dd ); + + /* Detect zero crossing */ + if( ( ylo <= 0 && yhi >= thr ) || ( ylo >= 0 && yhi <= -thr ) ) { + if( yhi == 0 ) { + /* If the root lies exactly at the end of the current */ + /* interval, look for the next root in the next interval */ + thr = 1; + } else { + thr = 0; + } + /* Binary division */ + ffrac = -256; + for( m = 0; m < BIN_DIV_STEPS_A2NLSF_FIX; m++ ) { + /* Evaluate polynomial */ + xmid = silk_RSHIFT_ROUND( xlo + xhi, 1 ); + ymid = silk_A2NLSF_eval_poly( p, xmid, dd ); + + /* Detect zero crossing */ + if( ( ylo <= 0 && ymid >= 0 ) || ( ylo >= 0 && ymid <= 0 ) ) { + /* Reduce frequency */ + xhi = xmid; + yhi = ymid; + } else { + /* Increase frequency */ + xlo = xmid; + ylo = ymid; + ffrac = silk_ADD_RSHIFT( ffrac, 128, m ); + } + } + + /* Interpolate */ + if( silk_abs( ylo ) < 65536 ) { + /* Avoid dividing by zero */ + den = ylo - yhi; + nom = silk_LSHIFT( ylo, 8 - BIN_DIV_STEPS_A2NLSF_FIX ) + silk_RSHIFT( den, 1 ); + if( den != 0 ) { + ffrac += silk_DIV32( nom, den ); + } + } else { + /* No risk of dividing by zero because abs(ylo - yhi) >= abs(ylo) >= 65536 */ + ffrac += silk_DIV32( ylo, silk_RSHIFT( ylo - yhi, 8 - BIN_DIV_STEPS_A2NLSF_FIX ) ); + } + NLSF[ root_ix ] = (opus_int16)silk_min_32( silk_LSHIFT( (opus_int32)k, 8 ) + ffrac, silk_int16_MAX ); + + silk_assert( NLSF[ root_ix ] >= 0 ); + + root_ix++; /* Next root */ + if( root_ix >= d ) { + /* Found all roots */ + break; + } + /* Alternate pointer to polynomial */ + p = PQ[ root_ix & 1 ]; + + /* Evaluate polynomial */ + xlo = silk_LSFCosTab_FIX_Q12[ k - 1 ]; /* Q12*/ + ylo = silk_LSHIFT( 1 - ( root_ix & 2 ), 12 ); + } else { + /* Increment loop counter */ + k++; + xlo = xhi; + ylo = yhi; + thr = 0; + + if( k > LSF_COS_TAB_SZ_FIX ) { + i++; + if( i > MAX_ITERATIONS_A2NLSF_FIX ) { + /* Set NLSFs to white spectrum and exit */ + NLSF[ 0 ] = (opus_int16)silk_DIV32_16( 1 << 15, d + 1 ); + for( k = 1; k < d; k++ ) { + NLSF[ k ] = (opus_int16)silk_ADD16( NLSF[ k-1 ], NLSF[ 0 ] ); + } + return; + } + + /* Error: Apply progressively more bandwidth expansion and run again */ + silk_bwexpander_32( a_Q16, d, 65536 - silk_LSHIFT( 1, i ) ); + + silk_A2NLSF_init( a_Q16, P, Q, dd ); + p = P; /* Pointer to polynomial */ + xlo = silk_LSFCosTab_FIX_Q12[ 0 ]; /* Q12*/ + ylo = silk_A2NLSF_eval_poly( p, xlo, dd ); + if( ylo < 0 ) { + /* Set the first NLSF to zero and move on to the next */ + NLSF[ 0 ] = 0; + p = Q; /* Pointer to polynomial */ + ylo = silk_A2NLSF_eval_poly( p, xlo, dd ); + root_ix = 1; /* Index of current root */ + } else { + root_ix = 0; /* Index of current root */ + } + k = 1; /* Reset loop counter */ + } + } + } +} diff --git a/lib/opus/silk/API.h b/lib/opus/silk/API.h new file mode 100644 index 00000000..4d90ff9a --- /dev/null +++ b/lib/opus/silk/API.h @@ -0,0 +1,135 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_API_H +#define SILK_API_H + +#include "control.h" +#include "typedef.h" +#include "errors.h" +#include "entenc.h" +#include "entdec.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +#define SILK_MAX_FRAMES_PER_PACKET 3 + +/* Struct for TOC (Table of Contents) */ +typedef struct { + opus_int VADFlag; /* Voice activity for packet */ + opus_int VADFlags[ SILK_MAX_FRAMES_PER_PACKET ]; /* Voice activity for each frame in packet */ + opus_int inbandFECFlag; /* Flag indicating if packet contains in-band FEC */ +} silk_TOC_struct; + +/****************************************/ +/* Encoder functions */ +/****************************************/ + +/***********************************************/ +/* Get size in bytes of the Silk encoder state */ +/***********************************************/ +opus_int silk_Get_Encoder_Size( /* O Returns error code */ + opus_int *encSizeBytes /* O Number of bytes in SILK encoder state */ +); + +/*************************/ +/* Init or reset encoder */ +/*************************/ +opus_int silk_InitEncoder( /* O Returns error code */ + void *encState, /* I/O State */ + int arch, /* I Run-time architecture */ + silk_EncControlStruct *encStatus /* O Encoder Status */ +); + +/**************************/ +/* Encode frame with Silk */ +/**************************/ +/* Note: if prefillFlag is set, the input must contain 10 ms of audio, irrespective of what */ +/* encControl->payloadSize_ms is set to */ +opus_int silk_Encode( /* O Returns error code */ + void *encState, /* I/O State */ + silk_EncControlStruct *encControl, /* I Control status */ + const opus_int16 *samplesIn, /* I Speech sample input vector */ + opus_int nSamplesIn, /* I Number of samples in input vector */ + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int32 *nBytesOut, /* I/O Number of bytes in payload (input: Max bytes) */ + const opus_int prefillFlag, /* I Flag to indicate prefilling buffers no coding */ + int activity /* I Decision of Opus voice activity detector */ +); + +/****************************************/ +/* Decoder functions */ +/****************************************/ + +/***********************************************/ +/* Get size in bytes of the Silk decoder state */ +/***********************************************/ +opus_int silk_Get_Decoder_Size( /* O Returns error code */ + opus_int *decSizeBytes /* O Number of bytes in SILK decoder state */ +); + +/*************************/ +/* Init or Reset decoder */ +/*************************/ +opus_int silk_InitDecoder( /* O Returns error code */ + void *decState /* I/O State */ +); + +/******************/ +/* Decode a frame */ +/******************/ +opus_int silk_Decode( /* O Returns error code */ + void* decState, /* I/O State */ + silk_DecControlStruct* decControl, /* I/O Control Structure */ + opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int newPacketFlag, /* I Indicates first decoder call for this packet */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 *samplesOut, /* O Decoded output speech vector */ + opus_int32 *nSamplesOut, /* O Number of samples decoded */ + int arch /* I Run-time architecture */ +); + +#if 0 +/**************************************/ +/* Get table of contents for a packet */ +/**************************************/ +opus_int silk_get_TOC( + const opus_uint8 *payload, /* I Payload data */ + const opus_int nBytesIn, /* I Number of input bytes */ + const opus_int nFramesPerPayload, /* I Number of SILK frames per payload */ + silk_TOC_struct *Silk_TOC /* O Type of content */ +); +#endif + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/CNG.c b/lib/opus/silk/CNG.c new file mode 100644 index 00000000..2a910099 --- /dev/null +++ b/lib/opus/silk/CNG.c @@ -0,0 +1,188 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" + +/* Generates excitation for CNG LPC synthesis */ +static OPUS_INLINE void silk_CNG_exc( + opus_int32 exc_Q14[], /* O CNG excitation signal Q10 */ + opus_int32 exc_buf_Q14[], /* I Random samples buffer Q10 */ + opus_int length, /* I Length */ + opus_int32 *rand_seed /* I/O Seed to random index generator */ +) +{ + opus_int32 seed; + opus_int i, idx, exc_mask; + + exc_mask = CNG_BUF_MASK_MAX; + while( exc_mask > length ) { + exc_mask = silk_RSHIFT( exc_mask, 1 ); + } + + seed = *rand_seed; + for( i = 0; i < length; i++ ) { + seed = silk_RAND( seed ); + idx = (opus_int)( silk_RSHIFT( seed, 24 ) & exc_mask ); + silk_assert( idx >= 0 ); + silk_assert( idx <= CNG_BUF_MASK_MAX ); + exc_Q14[ i ] = exc_buf_Q14[ idx ]; + } + *rand_seed = seed; +} + +void silk_CNG_Reset( + silk_decoder_state *psDec /* I/O Decoder state */ +) +{ + opus_int i, NLSF_step_Q15, NLSF_acc_Q15; + + NLSF_step_Q15 = silk_DIV32_16( silk_int16_MAX, psDec->LPC_order + 1 ); + NLSF_acc_Q15 = 0; + for( i = 0; i < psDec->LPC_order; i++ ) { + NLSF_acc_Q15 += NLSF_step_Q15; + psDec->sCNG.CNG_smth_NLSF_Q15[ i ] = NLSF_acc_Q15; + } + psDec->sCNG.CNG_smth_Gain_Q16 = 0; + psDec->sCNG.rand_seed = 3176576; +} + +/* Updates CNG estimate, and applies the CNG when packet was lost */ +void silk_CNG( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int16 frame[], /* I/O Signal */ + opus_int length /* I Length of residual */ +) +{ + opus_int i, subfr; + opus_int32 LPC_pred_Q10, max_Gain_Q16, gain_Q16, gain_Q10; + opus_int16 A_Q12[ MAX_LPC_ORDER ]; + silk_CNG_struct *psCNG = &psDec->sCNG; + SAVE_STACK; + + if( psDec->fs_kHz != psCNG->fs_kHz ) { + /* Reset state */ + silk_CNG_Reset( psDec ); + + psCNG->fs_kHz = psDec->fs_kHz; + } + if( psDec->lossCnt == 0 && psDec->prevSignalType == TYPE_NO_VOICE_ACTIVITY ) { + /* Update CNG parameters */ + + /* Smoothing of LSF's */ + for( i = 0; i < psDec->LPC_order; i++ ) { + psCNG->CNG_smth_NLSF_Q15[ i ] += silk_SMULWB( (opus_int32)psDec->prevNLSF_Q15[ i ] - (opus_int32)psCNG->CNG_smth_NLSF_Q15[ i ], CNG_NLSF_SMTH_Q16 ); + } + /* Find the subframe with the highest gain */ + max_Gain_Q16 = 0; + subfr = 0; + for( i = 0; i < psDec->nb_subfr; i++ ) { + if( psDecCtrl->Gains_Q16[ i ] > max_Gain_Q16 ) { + max_Gain_Q16 = psDecCtrl->Gains_Q16[ i ]; + subfr = i; + } + } + /* Update CNG excitation buffer with excitation from this subframe */ + silk_memmove( &psCNG->CNG_exc_buf_Q14[ psDec->subfr_length ], psCNG->CNG_exc_buf_Q14, ( psDec->nb_subfr - 1 ) * psDec->subfr_length * sizeof( opus_int32 ) ); + silk_memcpy( psCNG->CNG_exc_buf_Q14, &psDec->exc_Q14[ subfr * psDec->subfr_length ], psDec->subfr_length * sizeof( opus_int32 ) ); + + /* Smooth gains */ + for( i = 0; i < psDec->nb_subfr; i++ ) { + psCNG->CNG_smth_Gain_Q16 += silk_SMULWB( psDecCtrl->Gains_Q16[ i ] - psCNG->CNG_smth_Gain_Q16, CNG_GAIN_SMTH_Q16 ); + /* If the smoothed gain is 3 dB greater than this subframe's gain, use this subframe's gain to adapt faster. */ + if( silk_SMULWW( psCNG->CNG_smth_Gain_Q16, CNG_GAIN_SMTH_THRESHOLD_Q16 ) > psDecCtrl->Gains_Q16[ i ] ) { + psCNG->CNG_smth_Gain_Q16 = psDecCtrl->Gains_Q16[ i ]; + } + } + } + + /* Add CNG when packet is lost or during DTX */ + if( psDec->lossCnt ) { + VARDECL( opus_int32, CNG_sig_Q14 ); + ALLOC( CNG_sig_Q14, length + MAX_LPC_ORDER, opus_int32 ); + + /* Generate CNG excitation */ + gain_Q16 = silk_SMULWW( psDec->sPLC.randScale_Q14, psDec->sPLC.prevGain_Q16[1] ); + if( gain_Q16 >= (1 << 21) || psCNG->CNG_smth_Gain_Q16 > (1 << 23) ) { + gain_Q16 = silk_SMULTT( gain_Q16, gain_Q16 ); + gain_Q16 = silk_SUB_LSHIFT32(silk_SMULTT( psCNG->CNG_smth_Gain_Q16, psCNG->CNG_smth_Gain_Q16 ), gain_Q16, 5 ); + gain_Q16 = silk_LSHIFT32( silk_SQRT_APPROX( gain_Q16 ), 16 ); + } else { + gain_Q16 = silk_SMULWW( gain_Q16, gain_Q16 ); + gain_Q16 = silk_SUB_LSHIFT32(silk_SMULWW( psCNG->CNG_smth_Gain_Q16, psCNG->CNG_smth_Gain_Q16 ), gain_Q16, 5 ); + gain_Q16 = silk_LSHIFT32( silk_SQRT_APPROX( gain_Q16 ), 8 ); + } + gain_Q10 = silk_RSHIFT( gain_Q16, 6 ); + + silk_CNG_exc( CNG_sig_Q14 + MAX_LPC_ORDER, psCNG->CNG_exc_buf_Q14, length, &psCNG->rand_seed ); + + /* Convert CNG NLSF to filter representation */ + silk_NLSF2A( A_Q12, psCNG->CNG_smth_NLSF_Q15, psDec->LPC_order, psDec->arch ); + + /* Generate CNG signal, by synthesis filtering */ + silk_memcpy( CNG_sig_Q14, psCNG->CNG_synth_state, MAX_LPC_ORDER * sizeof( opus_int32 ) ); + celt_assert( psDec->LPC_order == 10 || psDec->LPC_order == 16 ); + for( i = 0; i < length; i++ ) { + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LPC_pred_Q10 = silk_RSHIFT( psDec->LPC_order, 1 ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 1 ], A_Q12[ 0 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 2 ], A_Q12[ 1 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 3 ], A_Q12[ 2 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 4 ], A_Q12[ 3 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 5 ], A_Q12[ 4 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 6 ], A_Q12[ 5 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 7 ], A_Q12[ 6 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 8 ], A_Q12[ 7 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 9 ], A_Q12[ 8 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 10 ], A_Q12[ 9 ] ); + if( psDec->LPC_order == 16 ) { + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 11 ], A_Q12[ 10 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 12 ], A_Q12[ 11 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 13 ], A_Q12[ 12 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 14 ], A_Q12[ 13 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 15 ], A_Q12[ 14 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, CNG_sig_Q14[ MAX_LPC_ORDER + i - 16 ], A_Q12[ 15 ] ); + } + + /* Update states */ + CNG_sig_Q14[ MAX_LPC_ORDER + i ] = silk_ADD_SAT32( CNG_sig_Q14[ MAX_LPC_ORDER + i ], silk_LSHIFT_SAT32( LPC_pred_Q10, 4 ) ); + + /* Scale with Gain and add to input signal */ + frame[ i ] = (opus_int16)silk_ADD_SAT16( frame[ i ], silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( CNG_sig_Q14[ MAX_LPC_ORDER + i ], gain_Q10 ), 8 ) ) ); + + } + silk_memcpy( psCNG->CNG_synth_state, &CNG_sig_Q14[ length ], MAX_LPC_ORDER * sizeof( opus_int32 ) ); + } else { + silk_memset( psCNG->CNG_synth_state, 0, psDec->LPC_order * sizeof( opus_int32 ) ); + } + RESTORE_STACK; +} diff --git a/lib/opus/silk/HP_variable_cutoff.c b/lib/opus/silk/HP_variable_cutoff.c new file mode 100644 index 00000000..bbe10f04 --- /dev/null +++ b/lib/opus/silk/HP_variable_cutoff.c @@ -0,0 +1,77 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#ifdef FIXED_POINT +#include "main_FIX.h" +#else +#include "main_FLP.h" +#endif +#include "tuning_parameters.h" + +/* High-pass filter with cutoff frequency adaptation based on pitch lag statistics */ +void silk_HP_variable_cutoff( + silk_encoder_state_Fxx state_Fxx[] /* I/O Encoder states */ +) +{ + opus_int quality_Q15; + opus_int32 pitch_freq_Hz_Q16, pitch_freq_log_Q7, delta_freq_Q7; + silk_encoder_state *psEncC1 = &state_Fxx[ 0 ].sCmn; + + /* Adaptive cutoff frequency: estimate low end of pitch frequency range */ + if( psEncC1->prevSignalType == TYPE_VOICED ) { + /* difference, in log domain */ + pitch_freq_Hz_Q16 = silk_DIV32_16( silk_LSHIFT( silk_MUL( psEncC1->fs_kHz, 1000 ), 16 ), psEncC1->prevLag ); + pitch_freq_log_Q7 = silk_lin2log( pitch_freq_Hz_Q16 ) - ( 16 << 7 ); + + /* adjustment based on quality */ + quality_Q15 = psEncC1->input_quality_bands_Q15[ 0 ]; + pitch_freq_log_Q7 = silk_SMLAWB( pitch_freq_log_Q7, silk_SMULWB( silk_LSHIFT( -quality_Q15, 2 ), quality_Q15 ), + pitch_freq_log_Q7 - ( silk_lin2log( SILK_FIX_CONST( VARIABLE_HP_MIN_CUTOFF_HZ, 16 ) ) - ( 16 << 7 ) ) ); + + /* delta_freq = pitch_freq_log - psEnc->variable_HP_smth1; */ + delta_freq_Q7 = pitch_freq_log_Q7 - silk_RSHIFT( psEncC1->variable_HP_smth1_Q15, 8 ); + if( delta_freq_Q7 < 0 ) { + /* less smoothing for decreasing pitch frequency, to track something close to the minimum */ + delta_freq_Q7 = silk_MUL( delta_freq_Q7, 3 ); + } + + /* limit delta, to reduce impact of outliers in pitch estimation */ + delta_freq_Q7 = silk_LIMIT_32( delta_freq_Q7, -SILK_FIX_CONST( VARIABLE_HP_MAX_DELTA_FREQ, 7 ), SILK_FIX_CONST( VARIABLE_HP_MAX_DELTA_FREQ, 7 ) ); + + /* update smoother */ + psEncC1->variable_HP_smth1_Q15 = silk_SMLAWB( psEncC1->variable_HP_smth1_Q15, + silk_SMULBB( psEncC1->speech_activity_Q8, delta_freq_Q7 ), SILK_FIX_CONST( VARIABLE_HP_SMTH_COEF1, 16 ) ); + + /* limit frequency range */ + psEncC1->variable_HP_smth1_Q15 = silk_LIMIT_32( psEncC1->variable_HP_smth1_Q15, + silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 ), + silk_LSHIFT( silk_lin2log( VARIABLE_HP_MAX_CUTOFF_HZ ), 8 ) ); + } +} diff --git a/lib/opus/silk/Inlines.h b/lib/opus/silk/Inlines.h new file mode 100644 index 00000000..ec986cdf --- /dev/null +++ b/lib/opus/silk/Inlines.h @@ -0,0 +1,188 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +/*! \file silk_Inlines.h + * \brief silk_Inlines.h defines OPUS_INLINE signal processing functions. + */ + +#ifndef SILK_FIX_INLINES_H +#define SILK_FIX_INLINES_H + +#ifdef __cplusplus +extern "C" +{ +#endif + +/* count leading zeros of opus_int64 */ +static OPUS_INLINE opus_int32 silk_CLZ64( opus_int64 in ) +{ + opus_int32 in_upper; + + in_upper = (opus_int32)silk_RSHIFT64(in, 32); + if (in_upper == 0) { + /* Search in the lower 32 bits */ + return 32 + silk_CLZ32( (opus_int32) in ); + } else { + /* Search in the upper 32 bits */ + return silk_CLZ32( in_upper ); + } +} + +/* get number of leading zeros and fractional part (the bits right after the leading one */ +static OPUS_INLINE void silk_CLZ_FRAC( + opus_int32 in, /* I input */ + opus_int32 *lz, /* O number of leading zeros */ + opus_int32 *frac_Q7 /* O the 7 bits right after the leading one */ +) +{ + opus_int32 lzeros = silk_CLZ32(in); + + * lz = lzeros; + * frac_Q7 = silk_ROR32(in, 24 - lzeros) & 0x7f; +} + +/* Approximation of square root */ +/* Accuracy: < +/- 10% for output values > 15 */ +/* < +/- 2.5% for output values > 120 */ +static OPUS_INLINE opus_int32 silk_SQRT_APPROX( opus_int32 x ) +{ + opus_int32 y, lz, frac_Q7; + + if( x <= 0 ) { + return 0; + } + + silk_CLZ_FRAC(x, &lz, &frac_Q7); + + if( lz & 1 ) { + y = 32768; + } else { + y = 46214; /* 46214 = sqrt(2) * 32768 */ + } + + /* get scaling right */ + y >>= silk_RSHIFT(lz, 1); + + /* increment using fractional part of input */ + y = silk_SMLAWB(y, y, silk_SMULBB(213, frac_Q7)); + + return y; +} + +/* Divide two int32 values and return result as int32 in a given Q-domain */ +static OPUS_INLINE opus_int32 silk_DIV32_varQ( /* O returns a good approximation of "(a32 << Qres) / b32" */ + const opus_int32 a32, /* I numerator (Q0) */ + const opus_int32 b32, /* I denominator (Q0) */ + const opus_int Qres /* I Q-domain of result (>= 0) */ +) +{ + opus_int a_headrm, b_headrm, lshift; + opus_int32 b32_inv, a32_nrm, b32_nrm, result; + + silk_assert( b32 != 0 ); + silk_assert( Qres >= 0 ); + + /* Compute number of bits head room and normalize inputs */ + a_headrm = silk_CLZ32( silk_abs(a32) ) - 1; + a32_nrm = silk_LSHIFT(a32, a_headrm); /* Q: a_headrm */ + b_headrm = silk_CLZ32( silk_abs(b32) ) - 1; + b32_nrm = silk_LSHIFT(b32, b_headrm); /* Q: b_headrm */ + + /* Inverse of b32, with 14 bits of precision */ + b32_inv = silk_DIV32_16( silk_int32_MAX >> 2, silk_RSHIFT(b32_nrm, 16) ); /* Q: 29 + 16 - b_headrm */ + + /* First approximation */ + result = silk_SMULWB(a32_nrm, b32_inv); /* Q: 29 + a_headrm - b_headrm */ + + /* Compute residual by subtracting product of denominator and first approximation */ + /* It's OK to overflow because the final value of a32_nrm should always be small */ + a32_nrm = silk_SUB32_ovflw(a32_nrm, silk_LSHIFT_ovflw( silk_SMMUL(b32_nrm, result), 3 )); /* Q: a_headrm */ + + /* Refinement */ + result = silk_SMLAWB(result, a32_nrm, b32_inv); /* Q: 29 + a_headrm - b_headrm */ + + /* Convert to Qres domain */ + lshift = 29 + a_headrm - b_headrm - Qres; + if( lshift < 0 ) { + return silk_LSHIFT_SAT32(result, -lshift); + } else { + if( lshift < 32){ + return silk_RSHIFT(result, lshift); + } else { + /* Avoid undefined result */ + return 0; + } + } +} + +/* Invert int32 value and return result as int32 in a given Q-domain */ +static OPUS_INLINE opus_int32 silk_INVERSE32_varQ( /* O returns a good approximation of "(1 << Qres) / b32" */ + const opus_int32 b32, /* I denominator (Q0) */ + const opus_int Qres /* I Q-domain of result (> 0) */ +) +{ + opus_int b_headrm, lshift; + opus_int32 b32_inv, b32_nrm, err_Q32, result; + + silk_assert( b32 != 0 ); + silk_assert( Qres > 0 ); + + /* Compute number of bits head room and normalize input */ + b_headrm = silk_CLZ32( silk_abs(b32) ) - 1; + b32_nrm = silk_LSHIFT(b32, b_headrm); /* Q: b_headrm */ + + /* Inverse of b32, with 14 bits of precision */ + b32_inv = silk_DIV32_16( silk_int32_MAX >> 2, silk_RSHIFT(b32_nrm, 16) ); /* Q: 29 + 16 - b_headrm */ + + /* First approximation */ + result = silk_LSHIFT(b32_inv, 16); /* Q: 61 - b_headrm */ + + /* Compute residual by subtracting product of denominator and first approximation from one */ + err_Q32 = silk_LSHIFT( ((opus_int32)1<<29) - silk_SMULWB(b32_nrm, b32_inv), 3 ); /* Q32 */ + + /* Refinement */ + result = silk_SMLAWW(result, err_Q32, b32_inv); /* Q: 61 - b_headrm */ + + /* Convert to Qres domain */ + lshift = 61 - b_headrm - Qres; + if( lshift <= 0 ) { + return silk_LSHIFT_SAT32(result, -lshift); + } else { + if( lshift < 32){ + return silk_RSHIFT(result, lshift); + }else{ + /* Avoid undefined result */ + return 0; + } + } +} + +#ifdef __cplusplus +} +#endif + +#endif /* SILK_FIX_INLINES_H */ diff --git a/lib/opus/silk/LPC_analysis_filter.c b/lib/opus/silk/LPC_analysis_filter.c new file mode 100644 index 00000000..d34b5eb7 --- /dev/null +++ b/lib/opus/silk/LPC_analysis_filter.c @@ -0,0 +1,111 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "celt_lpc.h" + +/*******************************************/ +/* LPC analysis filter */ +/* NB! State is kept internally and the */ +/* filter always starts with zero state */ +/* first d output samples are set to zero */ +/*******************************************/ + +/* OPT: Using celt_fir() for this function should be faster, but it may cause + integer overflows in intermediate values (not final results), which the + current implementation silences by casting to unsigned. Enabling + this should be safe in pretty much all cases, even though it is not technically + C89-compliant. */ +#define USE_CELT_FIR 0 + +void silk_LPC_analysis_filter( + opus_int16 *out, /* O Output signal */ + const opus_int16 *in, /* I Input signal */ + const opus_int16 *B, /* I MA prediction coefficients, Q12 [order] */ + const opus_int32 len, /* I Signal length */ + const opus_int32 d, /* I Filter order */ + int arch /* I Run-time architecture */ +) +{ + opus_int j; +#if defined(FIXED_POINT) && USE_CELT_FIR + opus_int16 num[SILK_MAX_ORDER_LPC]; +#else + int ix; + opus_int32 out32_Q12, out32; + const opus_int16 *in_ptr; +#endif + + celt_assert( d >= 6 ); + celt_assert( (d & 1) == 0 ); + celt_assert( d <= len ); + +#if defined(FIXED_POINT) && USE_CELT_FIR + celt_assert( d <= SILK_MAX_ORDER_LPC ); + for ( j = 0; j < d; j++ ) { + num[ j ] = -B[ j ]; + } + celt_fir( in + d, num, out + d, len - d, d, arch ); + for ( j = 0; j < d; j++ ) { + out[ j ] = 0; + } +#else + (void)arch; + for( ix = d; ix < len; ix++ ) { + in_ptr = &in[ ix - 1 ]; + + out32_Q12 = silk_SMULBB( in_ptr[ 0 ], B[ 0 ] ); + /* Allowing wrap around so that two wraps can cancel each other. The rare + cases where the result wraps around can only be triggered by invalid streams*/ + out32_Q12 = silk_SMLABB_ovflw( out32_Q12, in_ptr[ -1 ], B[ 1 ] ); + out32_Q12 = silk_SMLABB_ovflw( out32_Q12, in_ptr[ -2 ], B[ 2 ] ); + out32_Q12 = silk_SMLABB_ovflw( out32_Q12, in_ptr[ -3 ], B[ 3 ] ); + out32_Q12 = silk_SMLABB_ovflw( out32_Q12, in_ptr[ -4 ], B[ 4 ] ); + out32_Q12 = silk_SMLABB_ovflw( out32_Q12, in_ptr[ -5 ], B[ 5 ] ); + for( j = 6; j < d; j += 2 ) { + out32_Q12 = silk_SMLABB_ovflw( out32_Q12, in_ptr[ -j ], B[ j ] ); + out32_Q12 = silk_SMLABB_ovflw( out32_Q12, in_ptr[ -j - 1 ], B[ j + 1 ] ); + } + + /* Subtract prediction */ + out32_Q12 = silk_SUB32_ovflw( silk_LSHIFT( (opus_int32)in_ptr[ 1 ], 12 ), out32_Q12 ); + + /* Scale to Q0 */ + out32 = silk_RSHIFT_ROUND( out32_Q12, 12 ); + + /* Saturate output */ + out[ ix ] = (opus_int16)silk_SAT16( out32 ); + } + + /* Set first d output samples to zero */ + silk_memset( out, 0, d * sizeof( opus_int16 ) ); +#endif +} diff --git a/lib/opus/silk/LPC_fit.c b/lib/opus/silk/LPC_fit.c new file mode 100644 index 00000000..c0690a1f --- /dev/null +++ b/lib/opus/silk/LPC_fit.c @@ -0,0 +1,82 @@ +/*********************************************************************** +Copyright (c) 2013, Koen Vos. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Convert int32 coefficients to int16 coefs and make sure there's no wrap-around. + This logic is reused in _celt_lpc(). Any bug fixes should also be applied there. */ +void silk_LPC_fit( + opus_int16 *a_QOUT, /* O Output signal */ + opus_int32 *a_QIN, /* I/O Input signal */ + const opus_int QOUT, /* I Input Q domain */ + const opus_int QIN, /* I Input Q domain */ + const opus_int d /* I Filter order */ +) +{ + opus_int i, k, idx = 0; + opus_int32 maxabs, absval, chirp_Q16; + + /* Limit the maximum absolute value of the prediction coefficients, so that they'll fit in int16 */ + for( i = 0; i < 10; i++ ) { + /* Find maximum absolute value and its index */ + maxabs = 0; + for( k = 0; k < d; k++ ) { + absval = silk_abs( a_QIN[k] ); + if( absval > maxabs ) { + maxabs = absval; + idx = k; + } + } + maxabs = silk_RSHIFT_ROUND( maxabs, QIN - QOUT ); + + if( maxabs > silk_int16_MAX ) { + /* Reduce magnitude of prediction coefficients */ + maxabs = silk_min( maxabs, 163838 ); /* ( silk_int32_MAX >> 14 ) + silk_int16_MAX = 163838 */ + chirp_Q16 = SILK_FIX_CONST( 0.999, 16 ) - silk_DIV32( silk_LSHIFT( maxabs - silk_int16_MAX, 14 ), + silk_RSHIFT32( silk_MUL( maxabs, idx + 1), 2 ) ); + silk_bwexpander_32( a_QIN, d, chirp_Q16 ); + } else { + break; + } + } + + if( i == 10 ) { + /* Reached the last iteration, clip the coefficients */ + for( k = 0; k < d; k++ ) { + a_QOUT[ k ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( a_QIN[ k ], QIN - QOUT ) ); + a_QIN[ k ] = silk_LSHIFT( (opus_int32)a_QOUT[ k ], QIN - QOUT ); + } + } else { + for( k = 0; k < d; k++ ) { + a_QOUT[ k ] = (opus_int16)silk_RSHIFT_ROUND( a_QIN[ k ], QIN - QOUT ); + } + } +} diff --git a/lib/opus/silk/LPC_inv_pred_gain.c b/lib/opus/silk/LPC_inv_pred_gain.c new file mode 100644 index 00000000..a3746a6e --- /dev/null +++ b/lib/opus/silk/LPC_inv_pred_gain.c @@ -0,0 +1,141 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "define.h" + +#define QA 24 +#define A_LIMIT SILK_FIX_CONST( 0.99975, QA ) + +#define MUL32_FRAC_Q(a32, b32, Q) ((opus_int32)(silk_RSHIFT_ROUND64(silk_SMULL(a32, b32), Q))) + +/* Compute inverse of LPC prediction gain, and */ +/* test if LPC coefficients are stable (all poles within unit circle) */ +static opus_int32 LPC_inverse_pred_gain_QA_c( /* O Returns inverse prediction gain in energy domain, Q30 */ + opus_int32 A_QA[ SILK_MAX_ORDER_LPC ], /* I Prediction coefficients */ + const opus_int order /* I Prediction order */ +) +{ + opus_int k, n, mult2Q; + opus_int32 invGain_Q30, rc_Q31, rc_mult1_Q30, rc_mult2, tmp1, tmp2; + + invGain_Q30 = SILK_FIX_CONST( 1, 30 ); + for( k = order - 1; k > 0; k-- ) { + /* Check for stability */ + if( ( A_QA[ k ] > A_LIMIT ) || ( A_QA[ k ] < -A_LIMIT ) ) { + return 0; + } + + /* Set RC equal to negated AR coef */ + rc_Q31 = -silk_LSHIFT( A_QA[ k ], 31 - QA ); + + /* rc_mult1_Q30 range: [ 1 : 2^30 ] */ + rc_mult1_Q30 = silk_SUB32( SILK_FIX_CONST( 1, 30 ), silk_SMMUL( rc_Q31, rc_Q31 ) ); + silk_assert( rc_mult1_Q30 > ( 1 << 15 ) ); /* reduce A_LIMIT if fails */ + silk_assert( rc_mult1_Q30 <= ( 1 << 30 ) ); + + /* Update inverse gain */ + /* invGain_Q30 range: [ 0 : 2^30 ] */ + invGain_Q30 = silk_LSHIFT( silk_SMMUL( invGain_Q30, rc_mult1_Q30 ), 2 ); + silk_assert( invGain_Q30 >= 0 ); + silk_assert( invGain_Q30 <= ( 1 << 30 ) ); + if( invGain_Q30 < SILK_FIX_CONST( 1.0f / MAX_PREDICTION_POWER_GAIN, 30 ) ) { + return 0; + } + + /* rc_mult2 range: [ 2^30 : silk_int32_MAX ] */ + mult2Q = 32 - silk_CLZ32( silk_abs( rc_mult1_Q30 ) ); + rc_mult2 = silk_INVERSE32_varQ( rc_mult1_Q30, mult2Q + 30 ); + + /* Update AR coefficient */ + for( n = 0; n < (k + 1) >> 1; n++ ) { + opus_int64 tmp64; + tmp1 = A_QA[ n ]; + tmp2 = A_QA[ k - n - 1 ]; + tmp64 = silk_RSHIFT_ROUND64( silk_SMULL( silk_SUB_SAT32(tmp1, + MUL32_FRAC_Q( tmp2, rc_Q31, 31 ) ), rc_mult2 ), mult2Q); + if( tmp64 > silk_int32_MAX || tmp64 < silk_int32_MIN ) { + return 0; + } + A_QA[ n ] = ( opus_int32 )tmp64; + tmp64 = silk_RSHIFT_ROUND64( silk_SMULL( silk_SUB_SAT32(tmp2, + MUL32_FRAC_Q( tmp1, rc_Q31, 31 ) ), rc_mult2), mult2Q); + if( tmp64 > silk_int32_MAX || tmp64 < silk_int32_MIN ) { + return 0; + } + A_QA[ k - n - 1 ] = ( opus_int32 )tmp64; + } + } + + /* Check for stability */ + if( ( A_QA[ k ] > A_LIMIT ) || ( A_QA[ k ] < -A_LIMIT ) ) { + return 0; + } + + /* Set RC equal to negated AR coef */ + rc_Q31 = -silk_LSHIFT( A_QA[ 0 ], 31 - QA ); + + /* Range: [ 1 : 2^30 ] */ + rc_mult1_Q30 = silk_SUB32( SILK_FIX_CONST( 1, 30 ), silk_SMMUL( rc_Q31, rc_Q31 ) ); + + /* Update inverse gain */ + /* Range: [ 0 : 2^30 ] */ + invGain_Q30 = silk_LSHIFT( silk_SMMUL( invGain_Q30, rc_mult1_Q30 ), 2 ); + silk_assert( invGain_Q30 >= 0 ); + silk_assert( invGain_Q30 <= ( 1 << 30 ) ); + if( invGain_Q30 < SILK_FIX_CONST( 1.0f / MAX_PREDICTION_POWER_GAIN, 30 ) ) { + return 0; + } + + return invGain_Q30; +} + +/* For input in Q12 domain */ +opus_int32 silk_LPC_inverse_pred_gain_c( /* O Returns inverse prediction gain in energy domain, Q30 */ + const opus_int16 *A_Q12, /* I Prediction coefficients, Q12 [order] */ + const opus_int order /* I Prediction order */ +) +{ + opus_int k; + opus_int32 Atmp_QA[ SILK_MAX_ORDER_LPC ]; + opus_int32 DC_resp = 0; + + /* Increase Q domain of the AR coefficients */ + for( k = 0; k < order; k++ ) { + DC_resp += (opus_int32)A_Q12[ k ]; + Atmp_QA[ k ] = silk_LSHIFT32( (opus_int32)A_Q12[ k ], QA - 12 ); + } + /* If the DC is unstable, we don't even need to do the full calculations */ + if( DC_resp >= 4096 ) { + return 0; + } + return LPC_inverse_pred_gain_QA_c( Atmp_QA, order ); +} diff --git a/lib/opus/silk/LP_variable_cutoff.c b/lib/opus/silk/LP_variable_cutoff.c new file mode 100644 index 00000000..79112ad3 --- /dev/null +++ b/lib/opus/silk/LP_variable_cutoff.c @@ -0,0 +1,135 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* + Elliptic/Cauer filters designed with 0.1 dB passband ripple, + 80 dB minimum stopband attenuation, and + [0.95 : 0.15 : 0.35] normalized cut off frequencies. +*/ + +#include "main.h" + +/* Helper function, interpolates the filter taps */ +static OPUS_INLINE void silk_LP_interpolate_filter_taps( + opus_int32 B_Q28[ TRANSITION_NB ], + opus_int32 A_Q28[ TRANSITION_NA ], + const opus_int ind, + const opus_int32 fac_Q16 +) +{ + opus_int nb, na; + + if( ind < TRANSITION_INT_NUM - 1 ) { + if( fac_Q16 > 0 ) { + if( fac_Q16 < 32768 ) { /* fac_Q16 is in range of a 16-bit int */ + /* Piece-wise linear interpolation of B and A */ + for( nb = 0; nb < TRANSITION_NB; nb++ ) { + B_Q28[ nb ] = silk_SMLAWB( + silk_Transition_LP_B_Q28[ ind ][ nb ], + silk_Transition_LP_B_Q28[ ind + 1 ][ nb ] - + silk_Transition_LP_B_Q28[ ind ][ nb ], + fac_Q16 ); + } + for( na = 0; na < TRANSITION_NA; na++ ) { + A_Q28[ na ] = silk_SMLAWB( + silk_Transition_LP_A_Q28[ ind ][ na ], + silk_Transition_LP_A_Q28[ ind + 1 ][ na ] - + silk_Transition_LP_A_Q28[ ind ][ na ], + fac_Q16 ); + } + } else { /* ( fac_Q16 - ( 1 << 16 ) ) is in range of a 16-bit int */ + silk_assert( fac_Q16 - ( 1 << 16 ) == silk_SAT16( fac_Q16 - ( 1 << 16 ) ) ); + /* Piece-wise linear interpolation of B and A */ + for( nb = 0; nb < TRANSITION_NB; nb++ ) { + B_Q28[ nb ] = silk_SMLAWB( + silk_Transition_LP_B_Q28[ ind + 1 ][ nb ], + silk_Transition_LP_B_Q28[ ind + 1 ][ nb ] - + silk_Transition_LP_B_Q28[ ind ][ nb ], + fac_Q16 - ( (opus_int32)1 << 16 ) ); + } + for( na = 0; na < TRANSITION_NA; na++ ) { + A_Q28[ na ] = silk_SMLAWB( + silk_Transition_LP_A_Q28[ ind + 1 ][ na ], + silk_Transition_LP_A_Q28[ ind + 1 ][ na ] - + silk_Transition_LP_A_Q28[ ind ][ na ], + fac_Q16 - ( (opus_int32)1 << 16 ) ); + } + } + } else { + silk_memcpy( B_Q28, silk_Transition_LP_B_Q28[ ind ], TRANSITION_NB * sizeof( opus_int32 ) ); + silk_memcpy( A_Q28, silk_Transition_LP_A_Q28[ ind ], TRANSITION_NA * sizeof( opus_int32 ) ); + } + } else { + silk_memcpy( B_Q28, silk_Transition_LP_B_Q28[ TRANSITION_INT_NUM - 1 ], TRANSITION_NB * sizeof( opus_int32 ) ); + silk_memcpy( A_Q28, silk_Transition_LP_A_Q28[ TRANSITION_INT_NUM - 1 ], TRANSITION_NA * sizeof( opus_int32 ) ); + } +} + +/* Low-pass filter with variable cutoff frequency based on */ +/* piece-wise linear interpolation between elliptic filters */ +/* Start by setting psEncC->mode <> 0; */ +/* Deactivate by setting psEncC->mode = 0; */ +void silk_LP_variable_cutoff( + silk_LP_state *psLP, /* I/O LP filter state */ + opus_int16 *frame, /* I/O Low-pass filtered output signal */ + const opus_int frame_length /* I Frame length */ +) +{ + opus_int32 B_Q28[ TRANSITION_NB ], A_Q28[ TRANSITION_NA ], fac_Q16 = 0; + opus_int ind = 0; + + silk_assert( psLP->transition_frame_no >= 0 && psLP->transition_frame_no <= TRANSITION_FRAMES ); + + /* Run filter if needed */ + if( psLP->mode != 0 ) { + /* Calculate index and interpolation factor for interpolation */ +#if( TRANSITION_INT_STEPS == 64 ) + fac_Q16 = silk_LSHIFT( TRANSITION_FRAMES - psLP->transition_frame_no, 16 - 6 ); +#else + fac_Q16 = silk_DIV32_16( silk_LSHIFT( TRANSITION_FRAMES - psLP->transition_frame_no, 16 ), TRANSITION_FRAMES ); +#endif + ind = silk_RSHIFT( fac_Q16, 16 ); + fac_Q16 -= silk_LSHIFT( ind, 16 ); + + silk_assert( ind >= 0 ); + silk_assert( ind < TRANSITION_INT_NUM ); + + /* Interpolate filter coefficients */ + silk_LP_interpolate_filter_taps( B_Q28, A_Q28, ind, fac_Q16 ); + + /* Update transition frame number for next frame */ + psLP->transition_frame_no = silk_LIMIT( psLP->transition_frame_no + psLP->mode, 0, TRANSITION_FRAMES ); + + /* ARMA low-pass filtering */ + silk_assert( TRANSITION_NB == 3 && TRANSITION_NA == 2 ); + silk_biquad_alt_stride1( frame, B_Q28, A_Q28, psLP->In_LP_State, frame, frame_length); + } +} diff --git a/lib/opus/silk/MacroCount.h b/lib/opus/silk/MacroCount.h new file mode 100644 index 00000000..dab2f57a --- /dev/null +++ b/lib/opus/silk/MacroCount.h @@ -0,0 +1,710 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SIGPROCFIX_API_MACROCOUNT_H +#define SIGPROCFIX_API_MACROCOUNT_H + +#ifdef silk_MACRO_COUNT +#include +#define varDefine opus_int64 ops_count = 0; + +extern opus_int64 ops_count; + +static OPUS_INLINE opus_int64 silk_SaveCount(){ + return(ops_count); +} + +static OPUS_INLINE opus_int64 silk_SaveResetCount(){ + opus_int64 ret; + + ret = ops_count; + ops_count = 0; + return(ret); +} + +static OPUS_INLINE silk_PrintCount(){ + printf("ops_count = %d \n ", (opus_int32)ops_count); +} + +#undef silk_MUL +static OPUS_INLINE opus_int32 silk_MUL(opus_int32 a32, opus_int32 b32){ + opus_int32 ret; + ops_count += 4; + ret = a32 * b32; + return ret; +} + +#undef silk_MUL_uint +static OPUS_INLINE opus_uint32 silk_MUL_uint(opus_uint32 a32, opus_uint32 b32){ + opus_uint32 ret; + ops_count += 4; + ret = a32 * b32; + return ret; +} +#undef silk_MLA +static OPUS_INLINE opus_int32 silk_MLA(opus_int32 a32, opus_int32 b32, opus_int32 c32){ + opus_int32 ret; + ops_count += 4; + ret = a32 + b32 * c32; + return ret; +} + +#undef silk_MLA_uint +static OPUS_INLINE opus_int32 silk_MLA_uint(opus_uint32 a32, opus_uint32 b32, opus_uint32 c32){ + opus_uint32 ret; + ops_count += 4; + ret = a32 + b32 * c32; + return ret; +} + +#undef silk_SMULWB +static OPUS_INLINE opus_int32 silk_SMULWB(opus_int32 a32, opus_int32 b32){ + opus_int32 ret; + ops_count += 5; + ret = (a32 >> 16) * (opus_int32)((opus_int16)b32) + (((a32 & 0x0000FFFF) * (opus_int32)((opus_int16)b32)) >> 16); + return ret; +} +#undef silk_SMLAWB +static OPUS_INLINE opus_int32 silk_SMLAWB(opus_int32 a32, opus_int32 b32, opus_int32 c32){ + opus_int32 ret; + ops_count += 5; + ret = ((a32) + ((((b32) >> 16) * (opus_int32)((opus_int16)(c32))) + ((((b32) & 0x0000FFFF) * (opus_int32)((opus_int16)(c32))) >> 16))); + return ret; +} + +#undef silk_SMULWT +static OPUS_INLINE opus_int32 silk_SMULWT(opus_int32 a32, opus_int32 b32){ + opus_int32 ret; + ops_count += 4; + ret = (a32 >> 16) * (b32 >> 16) + (((a32 & 0x0000FFFF) * (b32 >> 16)) >> 16); + return ret; +} +#undef silk_SMLAWT +static OPUS_INLINE opus_int32 silk_SMLAWT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ + opus_int32 ret; + ops_count += 4; + ret = a32 + ((b32 >> 16) * (c32 >> 16)) + (((b32 & 0x0000FFFF) * ((c32 >> 16)) >> 16)); + return ret; +} + +#undef silk_SMULBB +static OPUS_INLINE opus_int32 silk_SMULBB(opus_int32 a32, opus_int32 b32){ + opus_int32 ret; + ops_count += 1; + ret = (opus_int32)((opus_int16)a32) * (opus_int32)((opus_int16)b32); + return ret; +} +#undef silk_SMLABB +static OPUS_INLINE opus_int32 silk_SMLABB(opus_int32 a32, opus_int32 b32, opus_int32 c32){ + opus_int32 ret; + ops_count += 1; + ret = a32 + (opus_int32)((opus_int16)b32) * (opus_int32)((opus_int16)c32); + return ret; +} + +#undef silk_SMULBT +static OPUS_INLINE opus_int32 silk_SMULBT(opus_int32 a32, opus_int32 b32 ){ + opus_int32 ret; + ops_count += 4; + ret = ((opus_int32)((opus_int16)a32)) * (b32 >> 16); + return ret; +} + +#undef silk_SMLABT +static OPUS_INLINE opus_int32 silk_SMLABT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ + opus_int32 ret; + ops_count += 1; + ret = a32 + ((opus_int32)((opus_int16)b32)) * (c32 >> 16); + return ret; +} + +#undef silk_SMULTT +static OPUS_INLINE opus_int32 silk_SMULTT(opus_int32 a32, opus_int32 b32){ + opus_int32 ret; + ops_count += 1; + ret = (a32 >> 16) * (b32 >> 16); + return ret; +} + +#undef silk_SMLATT +static OPUS_INLINE opus_int32 silk_SMLATT(opus_int32 a32, opus_int32 b32, opus_int32 c32){ + opus_int32 ret; + ops_count += 1; + ret = a32 + (b32 >> 16) * (c32 >> 16); + return ret; +} + + +/* multiply-accumulate macros that allow overflow in the addition (ie, no asserts in debug mode)*/ +#undef silk_MLA_ovflw +#define silk_MLA_ovflw silk_MLA + +#undef silk_SMLABB_ovflw +#define silk_SMLABB_ovflw silk_SMLABB + +#undef silk_SMLABT_ovflw +#define silk_SMLABT_ovflw silk_SMLABT + +#undef silk_SMLATT_ovflw +#define silk_SMLATT_ovflw silk_SMLATT + +#undef silk_SMLAWB_ovflw +#define silk_SMLAWB_ovflw silk_SMLAWB + +#undef silk_SMLAWT_ovflw +#define silk_SMLAWT_ovflw silk_SMLAWT + +#undef silk_SMULL +static OPUS_INLINE opus_int64 silk_SMULL(opus_int32 a32, opus_int32 b32){ + opus_int64 ret; + ops_count += 8; + ret = ((opus_int64)(a32) * /*(opus_int64)*/(b32)); + return ret; +} + +#undef silk_SMLAL +static OPUS_INLINE opus_int64 silk_SMLAL(opus_int64 a64, opus_int32 b32, opus_int32 c32){ + opus_int64 ret; + ops_count += 8; + ret = a64 + ((opus_int64)(b32) * /*(opus_int64)*/(c32)); + return ret; +} +#undef silk_SMLALBB +static OPUS_INLINE opus_int64 silk_SMLALBB(opus_int64 a64, opus_int16 b16, opus_int16 c16){ + opus_int64 ret; + ops_count += 4; + ret = a64 + ((opus_int64)(b16) * /*(opus_int64)*/(c16)); + return ret; +} + +#undef SigProcFIX_CLZ16 +static OPUS_INLINE opus_int32 SigProcFIX_CLZ16(opus_int16 in16) +{ + opus_int32 out32 = 0; + ops_count += 10; + if( in16 == 0 ) { + return 16; + } + /* test nibbles */ + if( in16 & 0xFF00 ) { + if( in16 & 0xF000 ) { + in16 >>= 12; + } else { + out32 += 4; + in16 >>= 8; + } + } else { + if( in16 & 0xFFF0 ) { + out32 += 8; + in16 >>= 4; + } else { + out32 += 12; + } + } + /* test bits and return */ + if( in16 & 0xC ) { + if( in16 & 0x8 ) + return out32 + 0; + else + return out32 + 1; + } else { + if( in16 & 0xE ) + return out32 + 2; + else + return out32 + 3; + } +} + +#undef SigProcFIX_CLZ32 +static OPUS_INLINE opus_int32 SigProcFIX_CLZ32(opus_int32 in32) +{ + /* test highest 16 bits and convert to opus_int16 */ + ops_count += 2; + if( in32 & 0xFFFF0000 ) { + return SigProcFIX_CLZ16((opus_int16)(in32 >> 16)); + } else { + return SigProcFIX_CLZ16((opus_int16)in32) + 16; + } +} + +#undef silk_DIV32 +static OPUS_INLINE opus_int32 silk_DIV32(opus_int32 a32, opus_int32 b32){ + ops_count += 64; + return a32 / b32; +} + +#undef silk_DIV32_16 +static OPUS_INLINE opus_int32 silk_DIV32_16(opus_int32 a32, opus_int32 b32){ + ops_count += 32; + return a32 / b32; +} + +#undef silk_SAT8 +static OPUS_INLINE opus_int8 silk_SAT8(opus_int64 a){ + opus_int8 tmp; + ops_count += 1; + tmp = (opus_int8)((a) > silk_int8_MAX ? silk_int8_MAX : \ + ((a) < silk_int8_MIN ? silk_int8_MIN : (a))); + return(tmp); +} + +#undef silk_SAT16 +static OPUS_INLINE opus_int16 silk_SAT16(opus_int64 a){ + opus_int16 tmp; + ops_count += 1; + tmp = (opus_int16)((a) > silk_int16_MAX ? silk_int16_MAX : \ + ((a) < silk_int16_MIN ? silk_int16_MIN : (a))); + return(tmp); +} +#undef silk_SAT32 +static OPUS_INLINE opus_int32 silk_SAT32(opus_int64 a){ + opus_int32 tmp; + ops_count += 1; + tmp = (opus_int32)((a) > silk_int32_MAX ? silk_int32_MAX : \ + ((a) < silk_int32_MIN ? silk_int32_MIN : (a))); + return(tmp); +} +#undef silk_POS_SAT32 +static OPUS_INLINE opus_int32 silk_POS_SAT32(opus_int64 a){ + opus_int32 tmp; + ops_count += 1; + tmp = (opus_int32)((a) > silk_int32_MAX ? silk_int32_MAX : (a)); + return(tmp); +} + +#undef silk_ADD_POS_SAT8 +static OPUS_INLINE opus_int8 silk_ADD_POS_SAT8(opus_int64 a, opus_int64 b){ + opus_int8 tmp; + ops_count += 1; + tmp = (opus_int8)((((a)+(b)) & 0x80) ? silk_int8_MAX : ((a)+(b))); + return(tmp); +} +#undef silk_ADD_POS_SAT16 +static OPUS_INLINE opus_int16 silk_ADD_POS_SAT16(opus_int64 a, opus_int64 b){ + opus_int16 tmp; + ops_count += 1; + tmp = (opus_int16)((((a)+(b)) & 0x8000) ? silk_int16_MAX : ((a)+(b))); + return(tmp); +} + +#undef silk_ADD_POS_SAT32 +static OPUS_INLINE opus_int32 silk_ADD_POS_SAT32(opus_int64 a, opus_int64 b){ + opus_int32 tmp; + ops_count += 1; + tmp = (opus_int32)((((a)+(b)) & 0x80000000) ? silk_int32_MAX : ((a)+(b))); + return(tmp); +} + +#undef silk_LSHIFT8 +static OPUS_INLINE opus_int8 silk_LSHIFT8(opus_int8 a, opus_int32 shift){ + opus_int8 ret; + ops_count += 1; + ret = a << shift; + return ret; +} +#undef silk_LSHIFT16 +static OPUS_INLINE opus_int16 silk_LSHIFT16(opus_int16 a, opus_int32 shift){ + opus_int16 ret; + ops_count += 1; + ret = a << shift; + return ret; +} +#undef silk_LSHIFT32 +static OPUS_INLINE opus_int32 silk_LSHIFT32(opus_int32 a, opus_int32 shift){ + opus_int32 ret; + ops_count += 1; + ret = a << shift; + return ret; +} +#undef silk_LSHIFT64 +static OPUS_INLINE opus_int64 silk_LSHIFT64(opus_int64 a, opus_int shift){ + ops_count += 1; + return a << shift; +} + +#undef silk_LSHIFT_ovflw +static OPUS_INLINE opus_int32 silk_LSHIFT_ovflw(opus_int32 a, opus_int32 shift){ + ops_count += 1; + return a << shift; +} + +#undef silk_LSHIFT_uint +static OPUS_INLINE opus_uint32 silk_LSHIFT_uint(opus_uint32 a, opus_int32 shift){ + opus_uint32 ret; + ops_count += 1; + ret = a << shift; + return ret; +} + +#undef silk_RSHIFT8 +static OPUS_INLINE opus_int8 silk_RSHIFT8(opus_int8 a, opus_int32 shift){ + ops_count += 1; + return a >> shift; +} +#undef silk_RSHIFT16 +static OPUS_INLINE opus_int16 silk_RSHIFT16(opus_int16 a, opus_int32 shift){ + ops_count += 1; + return a >> shift; +} +#undef silk_RSHIFT32 +static OPUS_INLINE opus_int32 silk_RSHIFT32(opus_int32 a, opus_int32 shift){ + ops_count += 1; + return a >> shift; +} +#undef silk_RSHIFT64 +static OPUS_INLINE opus_int64 silk_RSHIFT64(opus_int64 a, opus_int64 shift){ + ops_count += 1; + return a >> shift; +} + +#undef silk_RSHIFT_uint +static OPUS_INLINE opus_uint32 silk_RSHIFT_uint(opus_uint32 a, opus_int32 shift){ + ops_count += 1; + return a >> shift; +} + +#undef silk_ADD_LSHIFT +static OPUS_INLINE opus_int32 silk_ADD_LSHIFT(opus_int32 a, opus_int32 b, opus_int32 shift){ + opus_int32 ret; + ops_count += 1; + ret = a + (b << shift); + return ret; /* shift >= 0*/ +} +#undef silk_ADD_LSHIFT32 +static OPUS_INLINE opus_int32 silk_ADD_LSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ + opus_int32 ret; + ops_count += 1; + ret = a + (b << shift); + return ret; /* shift >= 0*/ +} +#undef silk_ADD_LSHIFT_uint +static OPUS_INLINE opus_uint32 silk_ADD_LSHIFT_uint(opus_uint32 a, opus_uint32 b, opus_int32 shift){ + opus_uint32 ret; + ops_count += 1; + ret = a + (b << shift); + return ret; /* shift >= 0*/ +} +#undef silk_ADD_RSHIFT +static OPUS_INLINE opus_int32 silk_ADD_RSHIFT(opus_int32 a, opus_int32 b, opus_int32 shift){ + opus_int32 ret; + ops_count += 1; + ret = a + (b >> shift); + return ret; /* shift > 0*/ +} +#undef silk_ADD_RSHIFT32 +static OPUS_INLINE opus_int32 silk_ADD_RSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ + opus_int32 ret; + ops_count += 1; + ret = a + (b >> shift); + return ret; /* shift > 0*/ +} +#undef silk_ADD_RSHIFT_uint +static OPUS_INLINE opus_uint32 silk_ADD_RSHIFT_uint(opus_uint32 a, opus_uint32 b, opus_int32 shift){ + opus_uint32 ret; + ops_count += 1; + ret = a + (b >> shift); + return ret; /* shift > 0*/ +} +#undef silk_SUB_LSHIFT32 +static OPUS_INLINE opus_int32 silk_SUB_LSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ + opus_int32 ret; + ops_count += 1; + ret = a - (b << shift); + return ret; /* shift >= 0*/ +} +#undef silk_SUB_RSHIFT32 +static OPUS_INLINE opus_int32 silk_SUB_RSHIFT32(opus_int32 a, opus_int32 b, opus_int32 shift){ + opus_int32 ret; + ops_count += 1; + ret = a - (b >> shift); + return ret; /* shift > 0*/ +} + +#undef silk_RSHIFT_ROUND +static OPUS_INLINE opus_int32 silk_RSHIFT_ROUND(opus_int32 a, opus_int32 shift){ + opus_int32 ret; + ops_count += 3; + ret = shift == 1 ? (a >> 1) + (a & 1) : ((a >> (shift - 1)) + 1) >> 1; + return ret; +} + +#undef silk_RSHIFT_ROUND64 +static OPUS_INLINE opus_int64 silk_RSHIFT_ROUND64(opus_int64 a, opus_int32 shift){ + opus_int64 ret; + ops_count += 6; + ret = shift == 1 ? (a >> 1) + (a & 1) : ((a >> (shift - 1)) + 1) >> 1; + return ret; +} + +#undef silk_abs_int64 +static OPUS_INLINE opus_int64 silk_abs_int64(opus_int64 a){ + ops_count += 1; + return (((a) > 0) ? (a) : -(a)); /* Be careful, silk_abs returns wrong when input equals to silk_intXX_MIN*/ +} + +#undef silk_abs_int32 +static OPUS_INLINE opus_int32 silk_abs_int32(opus_int32 a){ + ops_count += 1; + return silk_abs(a); +} + + +#undef silk_min +static silk_min(a, b){ + ops_count += 1; + return (((a) < (b)) ? (a) : (b)); +} +#undef silk_max +static silk_max(a, b){ + ops_count += 1; + return (((a) > (b)) ? (a) : (b)); +} +#undef silk_sign +static silk_sign(a){ + ops_count += 1; + return ((a) > 0 ? 1 : ( (a) < 0 ? -1 : 0 )); +} + +#undef silk_ADD16 +static OPUS_INLINE opus_int16 silk_ADD16(opus_int16 a, opus_int16 b){ + opus_int16 ret; + ops_count += 1; + ret = a + b; + return ret; +} + +#undef silk_ADD32 +static OPUS_INLINE opus_int32 silk_ADD32(opus_int32 a, opus_int32 b){ + opus_int32 ret; + ops_count += 1; + ret = a + b; + return ret; +} + +#undef silk_ADD64 +static OPUS_INLINE opus_int64 silk_ADD64(opus_int64 a, opus_int64 b){ + opus_int64 ret; + ops_count += 2; + ret = a + b; + return ret; +} + +#undef silk_SUB16 +static OPUS_INLINE opus_int16 silk_SUB16(opus_int16 a, opus_int16 b){ + opus_int16 ret; + ops_count += 1; + ret = a - b; + return ret; +} + +#undef silk_SUB32 +static OPUS_INLINE opus_int32 silk_SUB32(opus_int32 a, opus_int32 b){ + opus_int32 ret; + ops_count += 1; + ret = a - b; + return ret; +} + +#undef silk_SUB64 +static OPUS_INLINE opus_int64 silk_SUB64(opus_int64 a, opus_int64 b){ + opus_int64 ret; + ops_count += 2; + ret = a - b; + return ret; +} + +#undef silk_ADD_SAT16 +static OPUS_INLINE opus_int16 silk_ADD_SAT16( opus_int16 a16, opus_int16 b16 ) { + opus_int16 res; + /* Nb will be counted in AKP_add32 and silk_SAT16*/ + res = (opus_int16)silk_SAT16( silk_ADD32( (opus_int32)(a16), (b16) ) ); + return res; +} + +#undef silk_ADD_SAT32 +static OPUS_INLINE opus_int32 silk_ADD_SAT32(opus_int32 a32, opus_int32 b32){ + opus_int32 res; + ops_count += 1; + res = ((((a32) + (b32)) & 0x80000000) == 0 ? \ + ((((a32) & (b32)) & 0x80000000) != 0 ? silk_int32_MIN : (a32)+(b32)) : \ + ((((a32) | (b32)) & 0x80000000) == 0 ? silk_int32_MAX : (a32)+(b32)) ); + return res; +} + +#undef silk_ADD_SAT64 +static OPUS_INLINE opus_int64 silk_ADD_SAT64( opus_int64 a64, opus_int64 b64 ) { + opus_int64 res; + ops_count += 1; + res = ((((a64) + (b64)) & 0x8000000000000000LL) == 0 ? \ + ((((a64) & (b64)) & 0x8000000000000000LL) != 0 ? silk_int64_MIN : (a64)+(b64)) : \ + ((((a64) | (b64)) & 0x8000000000000000LL) == 0 ? silk_int64_MAX : (a64)+(b64)) ); + return res; +} + +#undef silk_SUB_SAT16 +static OPUS_INLINE opus_int16 silk_SUB_SAT16( opus_int16 a16, opus_int16 b16 ) { + opus_int16 res; + silk_assert(0); + /* Nb will be counted in sub-macros*/ + res = (opus_int16)silk_SAT16( silk_SUB32( (opus_int32)(a16), (b16) ) ); + return res; +} + +#undef silk_SUB_SAT32 +static OPUS_INLINE opus_int32 silk_SUB_SAT32( opus_int32 a32, opus_int32 b32 ) { + opus_int32 res; + ops_count += 1; + res = ((((a32)-(b32)) & 0x80000000) == 0 ? \ + (( (a32) & ((b32)^0x80000000) & 0x80000000) ? silk_int32_MIN : (a32)-(b32)) : \ + ((((a32)^0x80000000) & (b32) & 0x80000000) ? silk_int32_MAX : (a32)-(b32)) ); + return res; +} + +#undef silk_SUB_SAT64 +static OPUS_INLINE opus_int64 silk_SUB_SAT64( opus_int64 a64, opus_int64 b64 ) { + opus_int64 res; + ops_count += 1; + res = ((((a64)-(b64)) & 0x8000000000000000LL) == 0 ? \ + (( (a64) & ((b64)^0x8000000000000000LL) & 0x8000000000000000LL) ? silk_int64_MIN : (a64)-(b64)) : \ + ((((a64)^0x8000000000000000LL) & (b64) & 0x8000000000000000LL) ? silk_int64_MAX : (a64)-(b64)) ); + + return res; +} + +#undef silk_SMULWW +static OPUS_INLINE opus_int32 silk_SMULWW(opus_int32 a32, opus_int32 b32){ + opus_int32 ret; + /* Nb will be counted in sub-macros*/ + ret = silk_MLA(silk_SMULWB((a32), (b32)), (a32), silk_RSHIFT_ROUND((b32), 16)); + return ret; +} + +#undef silk_SMLAWW +static OPUS_INLINE opus_int32 silk_SMLAWW(opus_int32 a32, opus_int32 b32, opus_int32 c32){ + opus_int32 ret; + /* Nb will be counted in sub-macros*/ + ret = silk_MLA(silk_SMLAWB((a32), (b32), (c32)), (b32), silk_RSHIFT_ROUND((c32), 16)); + return ret; +} + +#undef silk_min_int +static OPUS_INLINE opus_int silk_min_int(opus_int a, opus_int b) +{ + ops_count += 1; + return (((a) < (b)) ? (a) : (b)); +} + +#undef silk_min_16 +static OPUS_INLINE opus_int16 silk_min_16(opus_int16 a, opus_int16 b) +{ + ops_count += 1; + return (((a) < (b)) ? (a) : (b)); +} +#undef silk_min_32 +static OPUS_INLINE opus_int32 silk_min_32(opus_int32 a, opus_int32 b) +{ + ops_count += 1; + return (((a) < (b)) ? (a) : (b)); +} +#undef silk_min_64 +static OPUS_INLINE opus_int64 silk_min_64(opus_int64 a, opus_int64 b) +{ + ops_count += 1; + return (((a) < (b)) ? (a) : (b)); +} + +/* silk_min() versions with typecast in the function call */ +#undef silk_max_int +static OPUS_INLINE opus_int silk_max_int(opus_int a, opus_int b) +{ + ops_count += 1; + return (((a) > (b)) ? (a) : (b)); +} +#undef silk_max_16 +static OPUS_INLINE opus_int16 silk_max_16(opus_int16 a, opus_int16 b) +{ + ops_count += 1; + return (((a) > (b)) ? (a) : (b)); +} +#undef silk_max_32 +static OPUS_INLINE opus_int32 silk_max_32(opus_int32 a, opus_int32 b) +{ + ops_count += 1; + return (((a) > (b)) ? (a) : (b)); +} + +#undef silk_max_64 +static OPUS_INLINE opus_int64 silk_max_64(opus_int64 a, opus_int64 b) +{ + ops_count += 1; + return (((a) > (b)) ? (a) : (b)); +} + + +#undef silk_LIMIT_int +static OPUS_INLINE opus_int silk_LIMIT_int(opus_int a, opus_int limit1, opus_int limit2) +{ + opus_int ret; + ops_count += 6; + + ret = ((limit1) > (limit2) ? ((a) > (limit1) ? (limit1) : ((a) < (limit2) ? (limit2) : (a))) \ + : ((a) > (limit2) ? (limit2) : ((a) < (limit1) ? (limit1) : (a)))); + + return(ret); +} + +#undef silk_LIMIT_16 +static OPUS_INLINE opus_int16 silk_LIMIT_16(opus_int16 a, opus_int16 limit1, opus_int16 limit2) +{ + opus_int16 ret; + ops_count += 6; + + ret = ((limit1) > (limit2) ? ((a) > (limit1) ? (limit1) : ((a) < (limit2) ? (limit2) : (a))) \ + : ((a) > (limit2) ? (limit2) : ((a) < (limit1) ? (limit1) : (a)))); + +return(ret); +} + + +#undef silk_LIMIT_32 +static OPUS_INLINE opus_int32 silk_LIMIT_32(opus_int32 a, opus_int32 limit1, opus_int32 limit2) +{ + opus_int32 ret; + ops_count += 6; + + ret = ((limit1) > (limit2) ? ((a) > (limit1) ? (limit1) : ((a) < (limit2) ? (limit2) : (a))) \ + : ((a) > (limit2) ? (limit2) : ((a) < (limit1) ? (limit1) : (a)))); + return(ret); +} + +#else +#define varDefine +#define silk_SaveCount() + +#endif +#endif + diff --git a/lib/opus/silk/MacroDebug.h b/lib/opus/silk/MacroDebug.h new file mode 100644 index 00000000..3110da9a --- /dev/null +++ b/lib/opus/silk/MacroDebug.h @@ -0,0 +1,945 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Copyright (C) 2012 Xiph.Org Foundation +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef MACRO_DEBUG_H +#define MACRO_DEBUG_H + +/* Redefine macro functions with extensive assertion in DEBUG mode. + As functions can't be undefined, this file can't work with SigProcFIX_MacroCount.h */ + +#if ( defined (FIXED_DEBUG) || ( 0 && defined (_DEBUG) ) ) && !defined (silk_MACRO_COUNT) + +#undef silk_ADD16 +#define silk_ADD16(a,b) silk_ADD16_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int16 silk_ADD16_(opus_int16 a, opus_int16 b, char *file, int line){ + opus_int16 ret; + + ret = a + b; + if ( ret != silk_ADD_SAT16( a, b ) ) + { + fprintf (stderr, "silk_ADD16(%d, %d) in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_ADD32 +#define silk_ADD32(a,b) silk_ADD32_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_ADD32_(opus_int32 a, opus_int32 b, char *file, int line){ + opus_int32 ret; + + ret = (opus_int32)((opus_uint32)a + (opus_uint32)b); + if ( ret != silk_ADD_SAT32( a, b ) ) + { + fprintf (stderr, "silk_ADD32(%d, %d) in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_ADD64 +#define silk_ADD64(a,b) silk_ADD64_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_ADD64_(opus_int64 a, opus_int64 b, char *file, int line){ + opus_int64 ret; + + ret = a + b; + if ( ret != silk_ADD_SAT64( a, b ) ) + { + fprintf (stderr, "silk_ADD64(%lld, %lld) in %s: line %d\n", (long long)a, (long long)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SUB16 +#define silk_SUB16(a,b) silk_SUB16_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int16 silk_SUB16_(opus_int16 a, opus_int16 b, char *file, int line){ + opus_int16 ret; + + ret = a - b; + if ( ret != silk_SUB_SAT16( a, b ) ) + { + fprintf (stderr, "silk_SUB16(%d, %d) in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SUB32 +#define silk_SUB32(a,b) silk_SUB32_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SUB32_(opus_int32 a, opus_int32 b, char *file, int line){ + opus_int64 ret; + + ret = a - (opus_int64)b; + if ( ret != silk_SUB_SAT32( a, b ) ) + { + fprintf (stderr, "silk_SUB32(%d, %d) in %s: line %d\n", a, b, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SUB64 +#define silk_SUB64(a,b) silk_SUB64_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_SUB64_(opus_int64 a, opus_int64 b, char *file, int line){ + opus_int64 ret; + + ret = a - b; + if ( ret != silk_SUB_SAT64( a, b ) ) + { + fprintf (stderr, "silk_SUB64(%lld, %lld) in %s: line %d\n", (long long)a, (long long)b, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_ADD_SAT16 +#define silk_ADD_SAT16(a,b) silk_ADD_SAT16_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int16 silk_ADD_SAT16_( opus_int16 a16, opus_int16 b16, char *file, int line) { + opus_int16 res; + res = (opus_int16)silk_SAT16( silk_ADD32( (opus_int32)(a16), (b16) ) ); + if ( res != silk_SAT16( (opus_int32)a16 + (opus_int32)b16 ) ) + { + fprintf (stderr, "silk_ADD_SAT16(%d, %d) in %s: line %d\n", a16, b16, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return res; +} + +#undef silk_ADD_SAT32 +#define silk_ADD_SAT32(a,b) silk_ADD_SAT32_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_ADD_SAT32_(opus_int32 a32, opus_int32 b32, char *file, int line){ + opus_int32 res; + res = ((((opus_uint32)(a32) + (opus_uint32)(b32)) & 0x80000000) == 0 ? \ + ((((a32) & (b32)) & 0x80000000) != 0 ? silk_int32_MIN : (a32)+(b32)) : \ + ((((a32) | (b32)) & 0x80000000) == 0 ? silk_int32_MAX : (a32)+(b32)) ); + if ( res != silk_SAT32( (opus_int64)a32 + (opus_int64)b32 ) ) + { + fprintf (stderr, "silk_ADD_SAT32(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return res; +} + +#undef silk_ADD_SAT64 +#define silk_ADD_SAT64(a,b) silk_ADD_SAT64_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_ADD_SAT64_( opus_int64 a64, opus_int64 b64, char *file, int line) { + opus_int64 res; + int fail = 0; + res = ((((a64) + (b64)) & 0x8000000000000000LL) == 0 ? \ + ((((a64) & (b64)) & 0x8000000000000000LL) != 0 ? silk_int64_MIN : (a64)+(b64)) : \ + ((((a64) | (b64)) & 0x8000000000000000LL) == 0 ? silk_int64_MAX : (a64)+(b64)) ); + if( res != a64 + b64 ) { + /* Check that we saturated to the correct extreme value */ + if ( !(( res == silk_int64_MAX && ( ( a64 >> 1 ) + ( b64 >> 1 ) > ( silk_int64_MAX >> 3 ) ) ) || + ( res == silk_int64_MIN && ( ( a64 >> 1 ) + ( b64 >> 1 ) < ( silk_int64_MIN >> 3 ) ) ) ) ) + { + fail = 1; + } + } else { + /* Saturation not necessary */ + fail = res != a64 + b64; + } + if ( fail ) + { + fprintf (stderr, "silk_ADD_SAT64(%lld, %lld) in %s: line %d\n", (long long)a64, (long long)b64, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return res; +} + +#undef silk_SUB_SAT16 +#define silk_SUB_SAT16(a,b) silk_SUB_SAT16_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int16 silk_SUB_SAT16_( opus_int16 a16, opus_int16 b16, char *file, int line ) { + opus_int16 res; + res = (opus_int16)silk_SAT16( silk_SUB32( (opus_int32)(a16), (b16) ) ); + if ( res != silk_SAT16( (opus_int32)a16 - (opus_int32)b16 ) ) + { + fprintf (stderr, "silk_SUB_SAT16(%d, %d) in %s: line %d\n", a16, b16, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return res; +} + +#undef silk_SUB_SAT32 +#define silk_SUB_SAT32(a,b) silk_SUB_SAT32_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SUB_SAT32_( opus_int32 a32, opus_int32 b32, char *file, int line ) { + opus_int32 res; + res = ((((opus_uint32)(a32)-(opus_uint32)(b32)) & 0x80000000) == 0 ? \ + (( (a32) & ((b32)^0x80000000) & 0x80000000) ? silk_int32_MIN : (a32)-(b32)) : \ + ((((a32)^0x80000000) & (b32) & 0x80000000) ? silk_int32_MAX : (a32)-(b32)) ); + if ( res != silk_SAT32( (opus_int64)a32 - (opus_int64)b32 ) ) + { + fprintf (stderr, "silk_SUB_SAT32(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return res; +} + +#undef silk_SUB_SAT64 +#define silk_SUB_SAT64(a,b) silk_SUB_SAT64_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_SUB_SAT64_( opus_int64 a64, opus_int64 b64, char *file, int line ) { + opus_int64 res; + int fail = 0; + res = ((((a64)-(b64)) & 0x8000000000000000LL) == 0 ? \ + (( (a64) & ((b64)^0x8000000000000000LL) & 0x8000000000000000LL) ? silk_int64_MIN : (a64)-(b64)) : \ + ((((a64)^0x8000000000000000LL) & (b64) & 0x8000000000000000LL) ? silk_int64_MAX : (a64)-(b64)) ); + if( res != a64 - b64 ) { + /* Check that we saturated to the correct extreme value */ + if( !(( res == silk_int64_MAX && ( ( a64 >> 1 ) + ( b64 >> 1 ) > ( silk_int64_MAX >> 3 ) ) ) || + ( res == silk_int64_MIN && ( ( a64 >> 1 ) + ( b64 >> 1 ) < ( silk_int64_MIN >> 3 ) ) ) )) + { + fail = 1; + } + } else { + /* Saturation not necessary */ + fail = res != a64 - b64; + } + if ( fail ) + { + fprintf (stderr, "silk_SUB_SAT64(%lld, %lld) in %s: line %d\n", (long long)a64, (long long)b64, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return res; +} + +#undef silk_MUL +#define silk_MUL(a,b) silk_MUL_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_MUL_(opus_int32 a32, opus_int32 b32, char *file, int line){ + opus_int32 ret; + opus_int64 ret64; + ret = (opus_int32)((opus_uint32)a32 * (opus_uint32)b32); + ret64 = (opus_int64)a32 * (opus_int64)b32; + if ( (opus_int64)ret != ret64 ) + { + fprintf (stderr, "silk_MUL(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_MUL_uint +#define silk_MUL_uint(a,b) silk_MUL_uint_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_uint32 silk_MUL_uint_(opus_uint32 a32, opus_uint32 b32, char *file, int line){ + opus_uint32 ret; + ret = a32 * b32; + if ( (opus_uint64)ret != (opus_uint64)a32 * (opus_uint64)b32 ) + { + fprintf (stderr, "silk_MUL_uint(%u, %u) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_MLA +#define silk_MLA(a,b,c) silk_MLA_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_MLA_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ + opus_int32 ret; + ret = a32 + b32 * c32; + if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (opus_int64)c32 ) + { + fprintf (stderr, "silk_MLA(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_MLA_uint +#define silk_MLA_uint(a,b,c) silk_MLA_uint_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_MLA_uint_(opus_uint32 a32, opus_uint32 b32, opus_uint32 c32, char *file, int line){ + opus_uint32 ret; + ret = a32 + b32 * c32; + if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (opus_int64)c32 ) + { + fprintf (stderr, "silk_MLA_uint(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SMULWB +#define silk_SMULWB(a,b) silk_SMULWB_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMULWB_(opus_int32 a32, opus_int32 b32, char *file, int line){ + opus_int32 ret; + ret = (a32 >> 16) * (opus_int32)((opus_int16)b32) + (((a32 & 0x0000FFFF) * (opus_int32)((opus_int16)b32)) >> 16); + if ( (opus_int64)ret != ((opus_int64)a32 * (opus_int16)b32) >> 16 ) + { + fprintf (stderr, "silk_SMULWB(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SMLAWB +#define silk_SMLAWB(a,b,c) silk_SMLAWB_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMLAWB_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ + opus_int32 ret; + ret = silk_ADD32_ovflw( a32, silk_SMULWB( b32, c32 ) ); + if ( ret != silk_ADD_SAT32( a32, silk_SMULWB( b32, c32 ) ) ) + { + fprintf (stderr, "silk_SMLAWB(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SMULWT +#define silk_SMULWT(a,b) silk_SMULWT_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMULWT_(opus_int32 a32, opus_int32 b32, char *file, int line){ + opus_int32 ret; + ret = (a32 >> 16) * (b32 >> 16) + (((a32 & 0x0000FFFF) * (b32 >> 16)) >> 16); + if ( (opus_int64)ret != ((opus_int64)a32 * (b32 >> 16)) >> 16 ) + { + fprintf (stderr, "silk_SMULWT(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SMLAWT +#define silk_SMLAWT(a,b,c) silk_SMLAWT_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMLAWT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ + opus_int32 ret; + ret = a32 + ((b32 >> 16) * (c32 >> 16)) + (((b32 & 0x0000FFFF) * ((c32 >> 16)) >> 16)); + if ( (opus_int64)ret != (opus_int64)a32 + (((opus_int64)b32 * (c32 >> 16)) >> 16) ) + { + fprintf (stderr, "silk_SMLAWT(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SMULL +#define silk_SMULL(a,b) silk_SMULL_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_SMULL_(opus_int64 a64, opus_int64 b64, char *file, int line){ + opus_int64 ret64; + int fail = 0; + ret64 = a64 * b64; + if( b64 != 0 ) { + fail = a64 != (ret64 / b64); + } else if( a64 != 0 ) { + fail = b64 != (ret64 / a64); + } + if ( fail ) + { + fprintf (stderr, "silk_SMULL(%lld, %lld) in %s: line %d\n", (long long)a64, (long long)b64, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret64; +} + +/* no checking needed for silk_SMULBB */ +#undef silk_SMLABB +#define silk_SMLABB(a,b,c) silk_SMLABB_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMLABB_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ + opus_int32 ret; + ret = a32 + (opus_int32)((opus_int16)b32) * (opus_int32)((opus_int16)c32); + if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (opus_int16)c32 ) + { + fprintf (stderr, "silk_SMLABB(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +/* no checking needed for silk_SMULBT */ +#undef silk_SMLABT +#define silk_SMLABT(a,b,c) silk_SMLABT_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMLABT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ + opus_int32 ret; + ret = a32 + ((opus_int32)((opus_int16)b32)) * (c32 >> 16); + if ( (opus_int64)ret != (opus_int64)a32 + (opus_int64)b32 * (c32 >> 16) ) + { + fprintf (stderr, "silk_SMLABT(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +/* no checking needed for silk_SMULTT */ +#undef silk_SMLATT +#define silk_SMLATT(a,b,c) silk_SMLATT_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMLATT_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ + opus_int32 ret; + ret = a32 + (b32 >> 16) * (c32 >> 16); + if ( (opus_int64)ret != (opus_int64)a32 + (b32 >> 16) * (c32 >> 16) ) + { + fprintf (stderr, "silk_SMLATT(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_SMULWW +#define silk_SMULWW(a,b) silk_SMULWW_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMULWW_(opus_int32 a32, opus_int32 b32, char *file, int line){ + opus_int32 ret, tmp1, tmp2; + opus_int64 ret64; + int fail = 0; + + ret = silk_SMULWB( a32, b32 ); + tmp1 = silk_RSHIFT_ROUND( b32, 16 ); + tmp2 = silk_MUL( a32, tmp1 ); + + fail |= (opus_int64)tmp2 != (opus_int64) a32 * (opus_int64) tmp1; + + tmp1 = ret; + ret = silk_ADD32( tmp1, tmp2 ); + fail |= silk_ADD32( tmp1, tmp2 ) != silk_ADD_SAT32( tmp1, tmp2 ); + + ret64 = silk_RSHIFT64( silk_SMULL( a32, b32 ), 16 ); + fail |= (opus_int64)ret != ret64; + + if ( fail ) + { + fprintf (stderr, "silk_SMULWW(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + + return ret; +} + +#undef silk_SMLAWW +#define silk_SMLAWW(a,b,c) silk_SMLAWW_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SMLAWW_(opus_int32 a32, opus_int32 b32, opus_int32 c32, char *file, int line){ + opus_int32 ret, tmp; + + tmp = silk_SMULWW( b32, c32 ); + ret = silk_ADD32( a32, tmp ); + if ( ret != silk_ADD_SAT32( a32, tmp ) ) + { + fprintf (stderr, "silk_SMLAWW(%d, %d, %d) in %s: line %d\n", a32, b32, c32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +/* no checking needed for silk_SMULL + no checking needed for silk_SMLAL + no checking needed for silk_SMLALBB + no checking needed for SigProcFIX_CLZ16 + no checking needed for SigProcFIX_CLZ32*/ + +#undef silk_DIV32 +#define silk_DIV32(a,b) silk_DIV32_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_DIV32_(opus_int32 a32, opus_int32 b32, char *file, int line){ + if ( b32 == 0 ) + { + fprintf (stderr, "silk_DIV32(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a32 / b32; +} + +#undef silk_DIV32_16 +#define silk_DIV32_16(a,b) silk_DIV32_16_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_DIV32_16_(opus_int32 a32, opus_int32 b32, char *file, int line){ + int fail = 0; + fail |= b32 == 0; + fail |= b32 > silk_int16_MAX; + fail |= b32 < silk_int16_MIN; + if ( fail ) + { + fprintf (stderr, "silk_DIV32_16(%d, %d) in %s: line %d\n", a32, b32, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a32 / b32; +} + +/* no checking needed for silk_SAT8 + no checking needed for silk_SAT16 + no checking needed for silk_SAT32 + no checking needed for silk_POS_SAT32 + no checking needed for silk_ADD_POS_SAT8 + no checking needed for silk_ADD_POS_SAT16 + no checking needed for silk_ADD_POS_SAT32 */ + +#undef silk_LSHIFT8 +#define silk_LSHIFT8(a,b) silk_LSHIFT8_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int8 silk_LSHIFT8_(opus_int8 a, opus_int32 shift, char *file, int line){ + opus_int8 ret; + int fail = 0; + ret = (opus_int8)((opus_uint8)a << shift); + fail |= shift < 0; + fail |= shift >= 8; + fail |= (opus_int64)ret != (opus_int64)(((opus_uint64)a) << shift); + if ( fail ) + { + fprintf (stderr, "silk_LSHIFT8(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_LSHIFT16 +#define silk_LSHIFT16(a,b) silk_LSHIFT16_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int16 silk_LSHIFT16_(opus_int16 a, opus_int32 shift, char *file, int line){ + opus_int16 ret; + int fail = 0; + ret = (opus_int16)((opus_uint16)a << shift); + fail |= shift < 0; + fail |= shift >= 16; + fail |= (opus_int64)ret != (opus_int64)(((opus_uint64)a) << shift); + if ( fail ) + { + fprintf (stderr, "silk_LSHIFT16(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_LSHIFT32 +#define silk_LSHIFT32(a,b) silk_LSHIFT32_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_LSHIFT32_(opus_int32 a, opus_int32 shift, char *file, int line){ + opus_int32 ret; + int fail = 0; + ret = (opus_int32)((opus_uint32)a << shift); + fail |= shift < 0; + fail |= shift >= 32; + fail |= (opus_int64)ret != (opus_int64)(((opus_uint64)a) << shift); + if ( fail ) + { + fprintf (stderr, "silk_LSHIFT32(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_LSHIFT64 +#define silk_LSHIFT64(a,b) silk_LSHIFT64_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_LSHIFT64_(opus_int64 a, opus_int shift, char *file, int line){ + opus_int64 ret; + int fail = 0; + ret = (opus_int64)((opus_uint64)a << shift); + fail |= shift < 0; + fail |= shift >= 64; + fail |= (ret>>shift) != ((opus_int64)a); + if ( fail ) + { + fprintf (stderr, "silk_LSHIFT64(%lld, %d) in %s: line %d\n", (long long)a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_LSHIFT_ovflw +#define silk_LSHIFT_ovflw(a,b) silk_LSHIFT_ovflw_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_LSHIFT_ovflw_(opus_int32 a, opus_int32 shift, char *file, int line){ + if ( (shift < 0) || (shift >= 32) ) /* no check for overflow */ + { + fprintf (stderr, "silk_LSHIFT_ovflw(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a << shift; +} + +#undef silk_LSHIFT_uint +#define silk_LSHIFT_uint(a,b) silk_LSHIFT_uint_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_uint32 silk_LSHIFT_uint_(opus_uint32 a, opus_int32 shift, char *file, int line){ + opus_uint32 ret; + ret = a << shift; + if ( (shift < 0) || ((opus_int64)ret != ((opus_int64)a) << shift)) + { + fprintf (stderr, "silk_LSHIFT_uint(%u, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_RSHIFT8 +#define silk_RSHITF8(a,b) silk_RSHIFT8_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int8 silk_RSHIFT8_(opus_int8 a, opus_int32 shift, char *file, int line){ + if ( (shift < 0) || (shift>=8) ) + { + fprintf (stderr, "silk_RSHITF8(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a >> shift; +} + +#undef silk_RSHIFT16 +#define silk_RSHITF16(a,b) silk_RSHIFT16_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int16 silk_RSHIFT16_(opus_int16 a, opus_int32 shift, char *file, int line){ + if ( (shift < 0) || (shift>=16) ) + { + fprintf (stderr, "silk_RSHITF16(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a >> shift; +} + +#undef silk_RSHIFT32 +#define silk_RSHIFT32(a,b) silk_RSHIFT32_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_RSHIFT32_(opus_int32 a, opus_int32 shift, char *file, int line){ + if ( (shift < 0) || (shift>=32) ) + { + fprintf (stderr, "silk_RSHITF32(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a >> shift; +} + +#undef silk_RSHIFT64 +#define silk_RSHIFT64(a,b) silk_RSHIFT64_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_RSHIFT64_(opus_int64 a, opus_int64 shift, char *file, int line){ + if ( (shift < 0) || (shift>=64) ) + { + fprintf (stderr, "silk_RSHITF64(%lld, %lld) in %s: line %d\n", (long long)a, (long long)shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a >> shift; +} + +#undef silk_RSHIFT_uint +#define silk_RSHIFT_uint(a,b) silk_RSHIFT_uint_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_uint32 silk_RSHIFT_uint_(opus_uint32 a, opus_int32 shift, char *file, int line){ + if ( (shift < 0) || (shift>32) ) + { + fprintf (stderr, "silk_RSHIFT_uint(%u, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return a >> shift; +} + +#undef silk_ADD_LSHIFT +#define silk_ADD_LSHIFT(a,b,c) silk_ADD_LSHIFT_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE int silk_ADD_LSHIFT_(int a, int b, int shift, char *file, int line){ + opus_int16 ret; + ret = a + (opus_int16)((opus_uint16)b << shift); + if ( (shift < 0) || (shift>15) || ((opus_int64)ret != (opus_int64)a + (opus_int64)(((opus_uint64)b) << shift)) ) + { + fprintf (stderr, "silk_ADD_LSHIFT(%d, %d, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift >= 0 */ +} + +#undef silk_ADD_LSHIFT32 +#define silk_ADD_LSHIFT32(a,b,c) silk_ADD_LSHIFT32_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_ADD_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ + opus_int32 ret; + ret = silk_ADD32_ovflw(a, (opus_int32)((opus_uint32)b << shift)); + if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a + (opus_int64)(((opus_uint64)b) << shift)) ) + { + fprintf (stderr, "silk_ADD_LSHIFT32(%d, %d, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift >= 0 */ +} + +#undef silk_ADD_LSHIFT_uint +#define silk_ADD_LSHIFT_uint(a,b,c) silk_ADD_LSHIFT_uint_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_uint32 silk_ADD_LSHIFT_uint_(opus_uint32 a, opus_uint32 b, opus_int32 shift, char *file, int line){ + opus_uint32 ret; + ret = a + (b << shift); + if ( (shift < 0) || (shift>32) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) << shift)) ) + { + fprintf (stderr, "silk_ADD_LSHIFT_uint(%u, %u, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift >= 0 */ +} + +#undef silk_ADD_RSHIFT +#define silk_ADD_RSHIFT(a,b,c) silk_ADD_RSHIFT_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE int silk_ADD_RSHIFT_(int a, int b, int shift, char *file, int line){ + opus_int16 ret; + ret = a + (b >> shift); + if ( (shift < 0) || (shift>15) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) >> shift)) ) + { + fprintf (stderr, "silk_ADD_RSHIFT(%d, %d, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift > 0 */ +} + +#undef silk_ADD_RSHIFT32 +#define silk_ADD_RSHIFT32(a,b,c) silk_ADD_RSHIFT32_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_ADD_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ + opus_int32 ret; + ret = silk_ADD32_ovflw(a, (b >> shift)); + if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) >> shift)) ) + { + fprintf (stderr, "silk_ADD_RSHIFT32(%d, %d, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift > 0 */ +} + +#undef silk_ADD_RSHIFT_uint +#define silk_ADD_RSHIFT_uint(a,b,c) silk_ADD_RSHIFT_uint_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_uint32 silk_ADD_RSHIFT_uint_(opus_uint32 a, opus_uint32 b, opus_int32 shift, char *file, int line){ + opus_uint32 ret; + ret = a + (b >> shift); + if ( (shift < 0) || (shift>32) || ((opus_int64)ret != (opus_int64)a + (((opus_int64)b) >> shift)) ) + { + fprintf (stderr, "silk_ADD_RSHIFT_uint(%u, %u, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift > 0 */ +} + +#undef silk_SUB_LSHIFT32 +#define silk_SUB_LSHIFT32(a,b,c) silk_SUB_LSHIFT32_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SUB_LSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ + opus_int32 ret; + ret = silk_SUB32_ovflw(a, (opus_int32)((opus_uint32)b << shift)); + if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a - (opus_int64)(((opus_uint64)b) << shift)) ) + { + fprintf (stderr, "silk_SUB_LSHIFT32(%d, %d, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift >= 0 */ +} + +#undef silk_SUB_RSHIFT32 +#define silk_SUB_RSHIFT32(a,b,c) silk_SUB_RSHIFT32_((a), (b), (c), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_SUB_RSHIFT32_(opus_int32 a, opus_int32 b, opus_int32 shift, char *file, int line){ + opus_int32 ret; + ret = silk_SUB32_ovflw(a, (b >> shift)); + if ( (shift < 0) || (shift>31) || ((opus_int64)ret != (opus_int64)a - (((opus_int64)b) >> shift)) ) + { + fprintf (stderr, "silk_SUB_RSHIFT32(%d, %d, %d) in %s: line %d\n", a, b, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; /* shift > 0 */ +} + +#undef silk_RSHIFT_ROUND +#define silk_RSHIFT_ROUND(a,b) silk_RSHIFT_ROUND_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_RSHIFT_ROUND_(opus_int32 a, opus_int32 shift, char *file, int line){ + opus_int32 ret; + ret = shift == 1 ? (a >> 1) + (a & 1) : ((a >> (shift - 1)) + 1) >> 1; + /* the macro definition can't handle a shift of zero */ + if ( (shift <= 0) || (shift>31) || ((opus_int64)ret != ((opus_int64)a + ((opus_int64)1 << (shift - 1))) >> shift) ) + { + fprintf (stderr, "silk_RSHIFT_ROUND(%d, %d) in %s: line %d\n", a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return ret; +} + +#undef silk_RSHIFT_ROUND64 +#define silk_RSHIFT_ROUND64(a,b) silk_RSHIFT_ROUND64_((a), (b), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_RSHIFT_ROUND64_(opus_int64 a, opus_int32 shift, char *file, int line){ + opus_int64 ret; + /* the macro definition can't handle a shift of zero */ + if ( (shift <= 0) || (shift>=64) ) + { + fprintf (stderr, "silk_RSHIFT_ROUND64(%lld, %d) in %s: line %d\n", (long long)a, shift, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + ret = shift == 1 ? (a >> 1) + (a & 1) : ((a >> (shift - 1)) + 1) >> 1; + return ret; +} + +/* silk_abs is used on floats also, so doesn't work... */ +/*#undef silk_abs +static OPUS_INLINE opus_int32 silk_abs(opus_int32 a){ + silk_assert(a != 0x80000000); + return (((a) > 0) ? (a) : -(a)); // Be careful, silk_abs returns wrong when input equals to silk_intXX_MIN +}*/ + +#undef silk_abs_int64 +#define silk_abs_int64(a) silk_abs_int64_((a), __FILE__, __LINE__) +static OPUS_INLINE opus_int64 silk_abs_int64_(opus_int64 a, char *file, int line){ + if ( a == silk_int64_MIN ) + { + fprintf (stderr, "silk_abs_int64(%lld) in %s: line %d\n", (long long)a, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return (((a) > 0) ? (a) : -(a)); /* Be careful, silk_abs returns wrong when input equals to silk_intXX_MIN */ +} + +#undef silk_abs_int32 +#define silk_abs_int32(a) silk_abs_int32_((a), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_abs_int32_(opus_int32 a, char *file, int line){ + if ( a == silk_int32_MIN ) + { + fprintf (stderr, "silk_abs_int32(%d) in %s: line %d\n", a, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return silk_abs(a); +} + +#undef silk_CHECK_FIT8 +#define silk_CHECK_FIT8(a) silk_CHECK_FIT8_((a), __FILE__, __LINE__) +static OPUS_INLINE opus_int8 silk_CHECK_FIT8_( opus_int64 a, char *file, int line ){ + opus_int8 ret; + ret = (opus_int8)a; + if ( (opus_int64)ret != a ) + { + fprintf (stderr, "silk_CHECK_FIT8(%lld) in %s: line %d\n", (long long)a, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return( ret ); +} + +#undef silk_CHECK_FIT16 +#define silk_CHECK_FIT16(a) silk_CHECK_FIT16_((a), __FILE__, __LINE__) +static OPUS_INLINE opus_int16 silk_CHECK_FIT16_( opus_int64 a, char *file, int line ){ + opus_int16 ret; + ret = (opus_int16)a; + if ( (opus_int64)ret != a ) + { + fprintf (stderr, "silk_CHECK_FIT16(%lld) in %s: line %d\n", (long long)a, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return( ret ); +} + +#undef silk_CHECK_FIT32 +#define silk_CHECK_FIT32(a) silk_CHECK_FIT32_((a), __FILE__, __LINE__) +static OPUS_INLINE opus_int32 silk_CHECK_FIT32_( opus_int64 a, char *file, int line ){ + opus_int32 ret; + ret = (opus_int32)a; + if ( (opus_int64)ret != a ) + { + fprintf (stderr, "silk_CHECK_FIT32(%lld) in %s: line %d\n", (long long)a, file, line); +#ifdef FIXED_DEBUG_ASSERT + silk_assert( 0 ); +#endif + } + return( ret ); +} + +/* no checking for silk_NSHIFT_MUL_32_32 + no checking for silk_NSHIFT_MUL_16_16 + no checking needed for silk_min + no checking needed for silk_max + no checking needed for silk_sign +*/ + +#endif +#endif /* MACRO_DEBUG_H */ diff --git a/lib/opus/silk/NLSF2A.c b/lib/opus/silk/NLSF2A.c new file mode 100644 index 00000000..d5b77306 --- /dev/null +++ b/lib/opus/silk/NLSF2A.c @@ -0,0 +1,141 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* conversion between prediction filter coefficients and LSFs */ +/* order should be even */ +/* a piecewise linear approximation maps LSF <-> cos(LSF) */ +/* therefore the result is not accurate LSFs, but the two */ +/* functions are accurate inverses of each other */ + +#include "SigProc_FIX.h" +#include "tables.h" + +#define QA 16 + +/* helper function for NLSF2A(..) */ +static OPUS_INLINE void silk_NLSF2A_find_poly( + opus_int32 *out, /* O intermediate polynomial, QA [dd+1] */ + const opus_int32 *cLSF, /* I vector of interleaved 2*cos(LSFs), QA [d] */ + opus_int dd /* I polynomial order (= 1/2 * filter order) */ +) +{ + opus_int k, n; + opus_int32 ftmp; + + out[0] = silk_LSHIFT( 1, QA ); + out[1] = -cLSF[0]; + for( k = 1; k < dd; k++ ) { + ftmp = cLSF[2*k]; /* QA*/ + out[k+1] = silk_LSHIFT( out[k-1], 1 ) - (opus_int32)silk_RSHIFT_ROUND64( silk_SMULL( ftmp, out[k] ), QA ); + for( n = k; n > 1; n-- ) { + out[n] += out[n-2] - (opus_int32)silk_RSHIFT_ROUND64( silk_SMULL( ftmp, out[n-1] ), QA ); + } + out[1] -= ftmp; + } +} + +/* compute whitening filter coefficients from normalized line spectral frequencies */ +void silk_NLSF2A( + opus_int16 *a_Q12, /* O monic whitening filter coefficients in Q12, [ d ] */ + const opus_int16 *NLSF, /* I normalized line spectral frequencies in Q15, [ d ] */ + const opus_int d, /* I filter order (should be even) */ + int arch /* I Run-time architecture */ +) +{ + /* This ordering was found to maximize quality. It improves numerical accuracy of + silk_NLSF2A_find_poly() compared to "standard" ordering. */ + static const unsigned char ordering16[16] = { + 0, 15, 8, 7, 4, 11, 12, 3, 2, 13, 10, 5, 6, 9, 14, 1 + }; + static const unsigned char ordering10[10] = { + 0, 9, 6, 3, 4, 5, 8, 1, 2, 7 + }; + const unsigned char *ordering; + opus_int k, i, dd; + opus_int32 cos_LSF_QA[ SILK_MAX_ORDER_LPC ]; + opus_int32 P[ SILK_MAX_ORDER_LPC / 2 + 1 ], Q[ SILK_MAX_ORDER_LPC / 2 + 1 ]; + opus_int32 Ptmp, Qtmp, f_int, f_frac, cos_val, delta; + opus_int32 a32_QA1[ SILK_MAX_ORDER_LPC ]; + + silk_assert( LSF_COS_TAB_SZ_FIX == 128 ); + celt_assert( d==10 || d==16 ); + + /* convert LSFs to 2*cos(LSF), using piecewise linear curve from table */ + ordering = d == 16 ? ordering16 : ordering10; + for( k = 0; k < d; k++ ) { + silk_assert( NLSF[k] >= 0 ); + + /* f_int on a scale 0-127 (rounded down) */ + f_int = silk_RSHIFT( NLSF[k], 15 - 7 ); + + /* f_frac, range: 0..255 */ + f_frac = NLSF[k] - silk_LSHIFT( f_int, 15 - 7 ); + + silk_assert(f_int >= 0); + silk_assert(f_int < LSF_COS_TAB_SZ_FIX ); + + /* Read start and end value from table */ + cos_val = silk_LSFCosTab_FIX_Q12[ f_int ]; /* Q12 */ + delta = silk_LSFCosTab_FIX_Q12[ f_int + 1 ] - cos_val; /* Q12, with a range of 0..200 */ + + /* Linear interpolation */ + cos_LSF_QA[ordering[k]] = silk_RSHIFT_ROUND( silk_LSHIFT( cos_val, 8 ) + silk_MUL( delta, f_frac ), 20 - QA ); /* QA */ + } + + dd = silk_RSHIFT( d, 1 ); + + /* generate even and odd polynomials using convolution */ + silk_NLSF2A_find_poly( P, &cos_LSF_QA[ 0 ], dd ); + silk_NLSF2A_find_poly( Q, &cos_LSF_QA[ 1 ], dd ); + + /* convert even and odd polynomials to opus_int32 Q12 filter coefs */ + for( k = 0; k < dd; k++ ) { + Ptmp = P[ k+1 ] + P[ k ]; + Qtmp = Q[ k+1 ] - Q[ k ]; + + /* the Ptmp and Qtmp values at this stage need to fit in int32 */ + a32_QA1[ k ] = -Qtmp - Ptmp; /* QA+1 */ + a32_QA1[ d-k-1 ] = Qtmp - Ptmp; /* QA+1 */ + } + + /* Convert int32 coefficients to Q12 int16 coefs */ + silk_LPC_fit( a_Q12, a32_QA1, 12, QA + 1, d ); + + for( i = 0; silk_LPC_inverse_pred_gain( a_Q12, d, arch ) == 0 && i < MAX_LPC_STABILIZE_ITERATIONS; i++ ) { + /* Prediction coefficients are (too close to) unstable; apply bandwidth expansion */ + /* on the unscaled coefficients, convert to Q12 and measure again */ + silk_bwexpander_32( a32_QA1, d, 65536 - silk_LSHIFT( 2, i ) ); + for( k = 0; k < d; k++ ) { + a_Q12[ k ] = (opus_int16)silk_RSHIFT_ROUND( a32_QA1[ k ], QA + 1 - 12 ); /* QA+1 -> Q12 */ + } + } +} + diff --git a/lib/opus/silk/NLSF_VQ.c b/lib/opus/silk/NLSF_VQ.c new file mode 100644 index 00000000..b83182a7 --- /dev/null +++ b/lib/opus/silk/NLSF_VQ.c @@ -0,0 +1,76 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Compute quantization errors for an LPC_order element input vector for a VQ codebook */ +void silk_NLSF_VQ( + opus_int32 err_Q24[], /* O Quantization errors [K] */ + const opus_int16 in_Q15[], /* I Input vectors to be quantized [LPC_order] */ + const opus_uint8 pCB_Q8[], /* I Codebook vectors [K*LPC_order] */ + const opus_int16 pWght_Q9[], /* I Codebook weights [K*LPC_order] */ + const opus_int K, /* I Number of codebook vectors */ + const opus_int LPC_order /* I Number of LPCs */ +) +{ + opus_int i, m; + opus_int32 diff_Q15, diffw_Q24, sum_error_Q24, pred_Q24; + const opus_int16 *w_Q9_ptr; + const opus_uint8 *cb_Q8_ptr; + + celt_assert( ( LPC_order & 1 ) == 0 ); + + /* Loop over codebook */ + cb_Q8_ptr = pCB_Q8; + w_Q9_ptr = pWght_Q9; + for( i = 0; i < K; i++ ) { + sum_error_Q24 = 0; + pred_Q24 = 0; + for( m = LPC_order-2; m >= 0; m -= 2 ) { + /* Compute weighted absolute predictive quantization error for index m + 1 */ + diff_Q15 = silk_SUB_LSHIFT32( in_Q15[ m + 1 ], (opus_int32)cb_Q8_ptr[ m + 1 ], 7 ); /* range: [ -32767 : 32767 ]*/ + diffw_Q24 = silk_SMULBB( diff_Q15, w_Q9_ptr[ m + 1 ] ); + sum_error_Q24 = silk_ADD32( sum_error_Q24, silk_abs( silk_SUB_RSHIFT32( diffw_Q24, pred_Q24, 1 ) ) ); + pred_Q24 = diffw_Q24; + + /* Compute weighted absolute predictive quantization error for index m */ + diff_Q15 = silk_SUB_LSHIFT32( in_Q15[ m ], (opus_int32)cb_Q8_ptr[ m ], 7 ); /* range: [ -32767 : 32767 ]*/ + diffw_Q24 = silk_SMULBB( diff_Q15, w_Q9_ptr[ m ] ); + sum_error_Q24 = silk_ADD32( sum_error_Q24, silk_abs( silk_SUB_RSHIFT32( diffw_Q24, pred_Q24, 1 ) ) ); + pred_Q24 = diffw_Q24; + + silk_assert( sum_error_Q24 >= 0 ); + } + err_Q24[ i ] = sum_error_Q24; + cb_Q8_ptr += LPC_order; + w_Q9_ptr += LPC_order; + } +} diff --git a/lib/opus/silk/NLSF_VQ_weights_laroia.c b/lib/opus/silk/NLSF_VQ_weights_laroia.c new file mode 100644 index 00000000..9873bcde --- /dev/null +++ b/lib/opus/silk/NLSF_VQ_weights_laroia.c @@ -0,0 +1,80 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "define.h" +#include "SigProc_FIX.h" + +/* +R. Laroia, N. Phamdo and N. Farvardin, "Robust and Efficient Quantization of Speech LSP +Parameters Using Structured Vector Quantization", Proc. IEEE Int. Conf. Acoust., Speech, +Signal Processing, pp. 641-644, 1991. +*/ + +/* Laroia low complexity NLSF weights */ +void silk_NLSF_VQ_weights_laroia( + opus_int16 *pNLSFW_Q_OUT, /* O Pointer to input vector weights [D] */ + const opus_int16 *pNLSF_Q15, /* I Pointer to input vector [D] */ + const opus_int D /* I Input vector dimension (even) */ +) +{ + opus_int k; + opus_int32 tmp1_int, tmp2_int; + + celt_assert( D > 0 ); + celt_assert( ( D & 1 ) == 0 ); + + /* First value */ + tmp1_int = silk_max_int( pNLSF_Q15[ 0 ], 1 ); + tmp1_int = silk_DIV32_16( (opus_int32)1 << ( 15 + NLSF_W_Q ), tmp1_int ); + tmp2_int = silk_max_int( pNLSF_Q15[ 1 ] - pNLSF_Q15[ 0 ], 1 ); + tmp2_int = silk_DIV32_16( (opus_int32)1 << ( 15 + NLSF_W_Q ), tmp2_int ); + pNLSFW_Q_OUT[ 0 ] = (opus_int16)silk_min_int( tmp1_int + tmp2_int, silk_int16_MAX ); + silk_assert( pNLSFW_Q_OUT[ 0 ] > 0 ); + + /* Main loop */ + for( k = 1; k < D - 1; k += 2 ) { + tmp1_int = silk_max_int( pNLSF_Q15[ k + 1 ] - pNLSF_Q15[ k ], 1 ); + tmp1_int = silk_DIV32_16( (opus_int32)1 << ( 15 + NLSF_W_Q ), tmp1_int ); + pNLSFW_Q_OUT[ k ] = (opus_int16)silk_min_int( tmp1_int + tmp2_int, silk_int16_MAX ); + silk_assert( pNLSFW_Q_OUT[ k ] > 0 ); + + tmp2_int = silk_max_int( pNLSF_Q15[ k + 2 ] - pNLSF_Q15[ k + 1 ], 1 ); + tmp2_int = silk_DIV32_16( (opus_int32)1 << ( 15 + NLSF_W_Q ), tmp2_int ); + pNLSFW_Q_OUT[ k + 1 ] = (opus_int16)silk_min_int( tmp1_int + tmp2_int, silk_int16_MAX ); + silk_assert( pNLSFW_Q_OUT[ k + 1 ] > 0 ); + } + + /* Last value */ + tmp1_int = silk_max_int( ( 1 << 15 ) - pNLSF_Q15[ D - 1 ], 1 ); + tmp1_int = silk_DIV32_16( (opus_int32)1 << ( 15 + NLSF_W_Q ), tmp1_int ); + pNLSFW_Q_OUT[ D - 1 ] = (opus_int16)silk_min_int( tmp1_int + tmp2_int, silk_int16_MAX ); + silk_assert( pNLSFW_Q_OUT[ D - 1 ] > 0 ); +} diff --git a/lib/opus/silk/NLSF_decode.c b/lib/opus/silk/NLSF_decode.c new file mode 100644 index 00000000..eeb0ba8c --- /dev/null +++ b/lib/opus/silk/NLSF_decode.c @@ -0,0 +1,93 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Predictive dequantizer for NLSF residuals */ +static OPUS_INLINE void silk_NLSF_residual_dequant( /* O Returns RD value in Q30 */ + opus_int16 x_Q10[], /* O Output [ order ] */ + const opus_int8 indices[], /* I Quantization indices [ order ] */ + const opus_uint8 pred_coef_Q8[], /* I Backward predictor coefs [ order ] */ + const opus_int quant_step_size_Q16, /* I Quantization step size */ + const opus_int16 order /* I Number of input values */ +) +{ + opus_int i, out_Q10, pred_Q10; + + out_Q10 = 0; + for( i = order-1; i >= 0; i-- ) { + pred_Q10 = silk_RSHIFT( silk_SMULBB( out_Q10, (opus_int16)pred_coef_Q8[ i ] ), 8 ); + out_Q10 = silk_LSHIFT( indices[ i ], 10 ); + if( out_Q10 > 0 ) { + out_Q10 = silk_SUB16( out_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + } else if( out_Q10 < 0 ) { + out_Q10 = silk_ADD16( out_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + } + out_Q10 = silk_SMLAWB( pred_Q10, (opus_int32)out_Q10, quant_step_size_Q16 ); + x_Q10[ i ] = out_Q10; + } +} + + +/***********************/ +/* NLSF vector decoder */ +/***********************/ +void silk_NLSF_decode( + opus_int16 *pNLSF_Q15, /* O Quantized NLSF vector [ LPC_ORDER ] */ + opus_int8 *NLSFIndices, /* I Codebook path vector [ LPC_ORDER + 1 ] */ + const silk_NLSF_CB_struct *psNLSF_CB /* I Codebook object */ +) +{ + opus_int i; + opus_uint8 pred_Q8[ MAX_LPC_ORDER ]; + opus_int16 ec_ix[ MAX_LPC_ORDER ]; + opus_int16 res_Q10[ MAX_LPC_ORDER ]; + opus_int32 NLSF_Q15_tmp; + const opus_uint8 *pCB_element; + const opus_int16 *pCB_Wght_Q9; + + /* Unpack entropy table indices and predictor for current CB1 index */ + silk_NLSF_unpack( ec_ix, pred_Q8, psNLSF_CB, NLSFIndices[ 0 ] ); + + /* Predictive residual dequantizer */ + silk_NLSF_residual_dequant( res_Q10, &NLSFIndices[ 1 ], pred_Q8, psNLSF_CB->quantStepSize_Q16, psNLSF_CB->order ); + + /* Apply inverse square-rooted weights to first stage and add to output */ + pCB_element = &psNLSF_CB->CB1_NLSF_Q8[ NLSFIndices[ 0 ] * psNLSF_CB->order ]; + pCB_Wght_Q9 = &psNLSF_CB->CB1_Wght_Q9[ NLSFIndices[ 0 ] * psNLSF_CB->order ]; + for( i = 0; i < psNLSF_CB->order; i++ ) { + NLSF_Q15_tmp = silk_ADD_LSHIFT32( silk_DIV32_16( silk_LSHIFT( (opus_int32)res_Q10[ i ], 14 ), pCB_Wght_Q9[ i ] ), (opus_int16)pCB_element[ i ], 7 ); + pNLSF_Q15[ i ] = (opus_int16)silk_LIMIT( NLSF_Q15_tmp, 0, 32767 ); + } + + /* NLSF stabilization */ + silk_NLSF_stabilize( pNLSF_Q15, psNLSF_CB->deltaMin_Q15, psNLSF_CB->order ); +} diff --git a/lib/opus/silk/NLSF_del_dec_quant.c b/lib/opus/silk/NLSF_del_dec_quant.c new file mode 100644 index 00000000..44a16acd --- /dev/null +++ b/lib/opus/silk/NLSF_del_dec_quant.c @@ -0,0 +1,215 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Delayed-decision quantizer for NLSF residuals */ +opus_int32 silk_NLSF_del_dec_quant( /* O Returns RD value in Q25 */ + opus_int8 indices[], /* O Quantization indices [ order ] */ + const opus_int16 x_Q10[], /* I Input [ order ] */ + const opus_int16 w_Q5[], /* I Weights [ order ] */ + const opus_uint8 pred_coef_Q8[], /* I Backward predictor coefs [ order ] */ + const opus_int16 ec_ix[], /* I Indices to entropy coding tables [ order ] */ + const opus_uint8 ec_rates_Q5[], /* I Rates [] */ + const opus_int quant_step_size_Q16, /* I Quantization step size */ + const opus_int16 inv_quant_step_size_Q6, /* I Inverse quantization step size */ + const opus_int32 mu_Q20, /* I R/D tradeoff */ + const opus_int16 order /* I Number of input values */ +) +{ + opus_int i, j, nStates, ind_tmp, ind_min_max, ind_max_min, in_Q10, res_Q10; + opus_int pred_Q10, diff_Q10, rate0_Q5, rate1_Q5; + opus_int16 out0_Q10, out1_Q10; + opus_int32 RD_tmp_Q25, min_Q25, min_max_Q25, max_min_Q25; + opus_int ind_sort[ NLSF_QUANT_DEL_DEC_STATES ]; + opus_int8 ind[ NLSF_QUANT_DEL_DEC_STATES ][ MAX_LPC_ORDER ]; + opus_int16 prev_out_Q10[ 2 * NLSF_QUANT_DEL_DEC_STATES ]; + opus_int32 RD_Q25[ 2 * NLSF_QUANT_DEL_DEC_STATES ]; + opus_int32 RD_min_Q25[ NLSF_QUANT_DEL_DEC_STATES ]; + opus_int32 RD_max_Q25[ NLSF_QUANT_DEL_DEC_STATES ]; + const opus_uint8 *rates_Q5; + + opus_int out0_Q10_table[2 * NLSF_QUANT_MAX_AMPLITUDE_EXT]; + opus_int out1_Q10_table[2 * NLSF_QUANT_MAX_AMPLITUDE_EXT]; + + for (i = -NLSF_QUANT_MAX_AMPLITUDE_EXT; i <= NLSF_QUANT_MAX_AMPLITUDE_EXT-1; i++) + { + out0_Q10 = silk_LSHIFT( i, 10 ); + out1_Q10 = silk_ADD16( out0_Q10, 1024 ); + if( i > 0 ) { + out0_Q10 = silk_SUB16( out0_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + out1_Q10 = silk_SUB16( out1_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + } else if( i == 0 ) { + out1_Q10 = silk_SUB16( out1_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + } else if( i == -1 ) { + out0_Q10 = silk_ADD16( out0_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + } else { + out0_Q10 = silk_ADD16( out0_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + out1_Q10 = silk_ADD16( out1_Q10, SILK_FIX_CONST( NLSF_QUANT_LEVEL_ADJ, 10 ) ); + } + out0_Q10_table[ i + NLSF_QUANT_MAX_AMPLITUDE_EXT ] = silk_RSHIFT( silk_SMULBB( out0_Q10, quant_step_size_Q16 ), 16 ); + out1_Q10_table[ i + NLSF_QUANT_MAX_AMPLITUDE_EXT ] = silk_RSHIFT( silk_SMULBB( out1_Q10, quant_step_size_Q16 ), 16 ); + } + + silk_assert( (NLSF_QUANT_DEL_DEC_STATES & (NLSF_QUANT_DEL_DEC_STATES-1)) == 0 ); /* must be power of two */ + + nStates = 1; + RD_Q25[ 0 ] = 0; + prev_out_Q10[ 0 ] = 0; + for( i = order - 1; i >= 0; i-- ) { + rates_Q5 = &ec_rates_Q5[ ec_ix[ i ] ]; + in_Q10 = x_Q10[ i ]; + for( j = 0; j < nStates; j++ ) { + pred_Q10 = silk_RSHIFT( silk_SMULBB( (opus_int16)pred_coef_Q8[ i ], prev_out_Q10[ j ] ), 8 ); + res_Q10 = silk_SUB16( in_Q10, pred_Q10 ); + ind_tmp = silk_RSHIFT( silk_SMULBB( inv_quant_step_size_Q6, res_Q10 ), 16 ); + ind_tmp = silk_LIMIT( ind_tmp, -NLSF_QUANT_MAX_AMPLITUDE_EXT, NLSF_QUANT_MAX_AMPLITUDE_EXT-1 ); + ind[ j ][ i ] = (opus_int8)ind_tmp; + + /* compute outputs for ind_tmp and ind_tmp + 1 */ + out0_Q10 = out0_Q10_table[ ind_tmp + NLSF_QUANT_MAX_AMPLITUDE_EXT ]; + out1_Q10 = out1_Q10_table[ ind_tmp + NLSF_QUANT_MAX_AMPLITUDE_EXT ]; + + out0_Q10 = silk_ADD16( out0_Q10, pred_Q10 ); + out1_Q10 = silk_ADD16( out1_Q10, pred_Q10 ); + prev_out_Q10[ j ] = out0_Q10; + prev_out_Q10[ j + nStates ] = out1_Q10; + + /* compute RD for ind_tmp and ind_tmp + 1 */ + if( ind_tmp + 1 >= NLSF_QUANT_MAX_AMPLITUDE ) { + if( ind_tmp + 1 == NLSF_QUANT_MAX_AMPLITUDE ) { + rate0_Q5 = rates_Q5[ ind_tmp + NLSF_QUANT_MAX_AMPLITUDE ]; + rate1_Q5 = 280; + } else { + rate0_Q5 = silk_SMLABB( 280 - 43 * NLSF_QUANT_MAX_AMPLITUDE, 43, ind_tmp ); + rate1_Q5 = silk_ADD16( rate0_Q5, 43 ); + } + } else if( ind_tmp <= -NLSF_QUANT_MAX_AMPLITUDE ) { + if( ind_tmp == -NLSF_QUANT_MAX_AMPLITUDE ) { + rate0_Q5 = 280; + rate1_Q5 = rates_Q5[ ind_tmp + 1 + NLSF_QUANT_MAX_AMPLITUDE ]; + } else { + rate0_Q5 = silk_SMLABB( 280 - 43 * NLSF_QUANT_MAX_AMPLITUDE, -43, ind_tmp ); + rate1_Q5 = silk_SUB16( rate0_Q5, 43 ); + } + } else { + rate0_Q5 = rates_Q5[ ind_tmp + NLSF_QUANT_MAX_AMPLITUDE ]; + rate1_Q5 = rates_Q5[ ind_tmp + 1 + NLSF_QUANT_MAX_AMPLITUDE ]; + } + RD_tmp_Q25 = RD_Q25[ j ]; + diff_Q10 = silk_SUB16( in_Q10, out0_Q10 ); + RD_Q25[ j ] = silk_SMLABB( silk_MLA( RD_tmp_Q25, silk_SMULBB( diff_Q10, diff_Q10 ), w_Q5[ i ] ), mu_Q20, rate0_Q5 ); + diff_Q10 = silk_SUB16( in_Q10, out1_Q10 ); + RD_Q25[ j + nStates ] = silk_SMLABB( silk_MLA( RD_tmp_Q25, silk_SMULBB( diff_Q10, diff_Q10 ), w_Q5[ i ] ), mu_Q20, rate1_Q5 ); + } + + if( nStates <= NLSF_QUANT_DEL_DEC_STATES/2 ) { + /* double number of states and copy */ + for( j = 0; j < nStates; j++ ) { + ind[ j + nStates ][ i ] = ind[ j ][ i ] + 1; + } + nStates = silk_LSHIFT( nStates, 1 ); + for( j = nStates; j < NLSF_QUANT_DEL_DEC_STATES; j++ ) { + ind[ j ][ i ] = ind[ j - nStates ][ i ]; + } + } else { + /* sort lower and upper half of RD_Q25, pairwise */ + for( j = 0; j < NLSF_QUANT_DEL_DEC_STATES; j++ ) { + if( RD_Q25[ j ] > RD_Q25[ j + NLSF_QUANT_DEL_DEC_STATES ] ) { + RD_max_Q25[ j ] = RD_Q25[ j ]; + RD_min_Q25[ j ] = RD_Q25[ j + NLSF_QUANT_DEL_DEC_STATES ]; + RD_Q25[ j ] = RD_min_Q25[ j ]; + RD_Q25[ j + NLSF_QUANT_DEL_DEC_STATES ] = RD_max_Q25[ j ]; + /* swap prev_out values */ + out0_Q10 = prev_out_Q10[ j ]; + prev_out_Q10[ j ] = prev_out_Q10[ j + NLSF_QUANT_DEL_DEC_STATES ]; + prev_out_Q10[ j + NLSF_QUANT_DEL_DEC_STATES ] = out0_Q10; + ind_sort[ j ] = j + NLSF_QUANT_DEL_DEC_STATES; + } else { + RD_min_Q25[ j ] = RD_Q25[ j ]; + RD_max_Q25[ j ] = RD_Q25[ j + NLSF_QUANT_DEL_DEC_STATES ]; + ind_sort[ j ] = j; + } + } + /* compare the highest RD values of the winning half with the lowest one in the losing half, and copy if necessary */ + /* afterwards ind_sort[] will contain the indices of the NLSF_QUANT_DEL_DEC_STATES winning RD values */ + while( 1 ) { + min_max_Q25 = silk_int32_MAX; + max_min_Q25 = 0; + ind_min_max = 0; + ind_max_min = 0; + for( j = 0; j < NLSF_QUANT_DEL_DEC_STATES; j++ ) { + if( min_max_Q25 > RD_max_Q25[ j ] ) { + min_max_Q25 = RD_max_Q25[ j ]; + ind_min_max = j; + } + if( max_min_Q25 < RD_min_Q25[ j ] ) { + max_min_Q25 = RD_min_Q25[ j ]; + ind_max_min = j; + } + } + if( min_max_Q25 >= max_min_Q25 ) { + break; + } + /* copy ind_min_max to ind_max_min */ + ind_sort[ ind_max_min ] = ind_sort[ ind_min_max ] ^ NLSF_QUANT_DEL_DEC_STATES; + RD_Q25[ ind_max_min ] = RD_Q25[ ind_min_max + NLSF_QUANT_DEL_DEC_STATES ]; + prev_out_Q10[ ind_max_min ] = prev_out_Q10[ ind_min_max + NLSF_QUANT_DEL_DEC_STATES ]; + RD_min_Q25[ ind_max_min ] = 0; + RD_max_Q25[ ind_min_max ] = silk_int32_MAX; + silk_memcpy( ind[ ind_max_min ], ind[ ind_min_max ], MAX_LPC_ORDER * sizeof( opus_int8 ) ); + } + /* increment index if it comes from the upper half */ + for( j = 0; j < NLSF_QUANT_DEL_DEC_STATES; j++ ) { + ind[ j ][ i ] += silk_RSHIFT( ind_sort[ j ], NLSF_QUANT_DEL_DEC_STATES_LOG2 ); + } + } + } + + /* last sample: find winner, copy indices and return RD value */ + ind_tmp = 0; + min_Q25 = silk_int32_MAX; + for( j = 0; j < 2 * NLSF_QUANT_DEL_DEC_STATES; j++ ) { + if( min_Q25 > RD_Q25[ j ] ) { + min_Q25 = RD_Q25[ j ]; + ind_tmp = j; + } + } + for( j = 0; j < order; j++ ) { + indices[ j ] = ind[ ind_tmp & ( NLSF_QUANT_DEL_DEC_STATES - 1 ) ][ j ]; + silk_assert( indices[ j ] >= -NLSF_QUANT_MAX_AMPLITUDE_EXT ); + silk_assert( indices[ j ] <= NLSF_QUANT_MAX_AMPLITUDE_EXT ); + } + indices[ 0 ] += silk_RSHIFT( ind_tmp, NLSF_QUANT_DEL_DEC_STATES_LOG2 ); + silk_assert( indices[ 0 ] <= NLSF_QUANT_MAX_AMPLITUDE_EXT ); + silk_assert( min_Q25 >= 0 ); + return min_Q25; +} diff --git a/lib/opus/silk/NLSF_encode.c b/lib/opus/silk/NLSF_encode.c new file mode 100644 index 00000000..01ac7db7 --- /dev/null +++ b/lib/opus/silk/NLSF_encode.c @@ -0,0 +1,124 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" + +/***********************/ +/* NLSF vector encoder */ +/***********************/ +opus_int32 silk_NLSF_encode( /* O Returns RD value in Q25 */ + opus_int8 *NLSFIndices, /* I Codebook path vector [ LPC_ORDER + 1 ] */ + opus_int16 *pNLSF_Q15, /* I/O (Un)quantized NLSF vector [ LPC_ORDER ] */ + const silk_NLSF_CB_struct *psNLSF_CB, /* I Codebook object */ + const opus_int16 *pW_Q2, /* I NLSF weight vector [ LPC_ORDER ] */ + const opus_int NLSF_mu_Q20, /* I Rate weight for the RD optimization */ + const opus_int nSurvivors, /* I Max survivors after first stage */ + const opus_int signalType /* I Signal type: 0/1/2 */ +) +{ + opus_int i, s, ind1, bestIndex, prob_Q8, bits_q7; + opus_int32 W_tmp_Q9, ret; + VARDECL( opus_int32, err_Q24 ); + VARDECL( opus_int32, RD_Q25 ); + VARDECL( opus_int, tempIndices1 ); + VARDECL( opus_int8, tempIndices2 ); + opus_int16 res_Q10[ MAX_LPC_ORDER ]; + opus_int16 NLSF_tmp_Q15[ MAX_LPC_ORDER ]; + opus_int16 W_adj_Q5[ MAX_LPC_ORDER ]; + opus_uint8 pred_Q8[ MAX_LPC_ORDER ]; + opus_int16 ec_ix[ MAX_LPC_ORDER ]; + const opus_uint8 *pCB_element, *iCDF_ptr; + const opus_int16 *pCB_Wght_Q9; + SAVE_STACK; + + celt_assert( signalType >= 0 && signalType <= 2 ); + silk_assert( NLSF_mu_Q20 <= 32767 && NLSF_mu_Q20 >= 0 ); + + /* NLSF stabilization */ + silk_NLSF_stabilize( pNLSF_Q15, psNLSF_CB->deltaMin_Q15, psNLSF_CB->order ); + + /* First stage: VQ */ + ALLOC( err_Q24, psNLSF_CB->nVectors, opus_int32 ); + silk_NLSF_VQ( err_Q24, pNLSF_Q15, psNLSF_CB->CB1_NLSF_Q8, psNLSF_CB->CB1_Wght_Q9, psNLSF_CB->nVectors, psNLSF_CB->order ); + + /* Sort the quantization errors */ + ALLOC( tempIndices1, nSurvivors, opus_int ); + silk_insertion_sort_increasing( err_Q24, tempIndices1, psNLSF_CB->nVectors, nSurvivors ); + + ALLOC( RD_Q25, nSurvivors, opus_int32 ); + ALLOC( tempIndices2, nSurvivors * MAX_LPC_ORDER, opus_int8 ); + + /* Loop over survivors */ + for( s = 0; s < nSurvivors; s++ ) { + ind1 = tempIndices1[ s ]; + + /* Residual after first stage */ + pCB_element = &psNLSF_CB->CB1_NLSF_Q8[ ind1 * psNLSF_CB->order ]; + pCB_Wght_Q9 = &psNLSF_CB->CB1_Wght_Q9[ ind1 * psNLSF_CB->order ]; + for( i = 0; i < psNLSF_CB->order; i++ ) { + NLSF_tmp_Q15[ i ] = silk_LSHIFT16( (opus_int16)pCB_element[ i ], 7 ); + W_tmp_Q9 = pCB_Wght_Q9[ i ]; + res_Q10[ i ] = (opus_int16)silk_RSHIFT( silk_SMULBB( pNLSF_Q15[ i ] - NLSF_tmp_Q15[ i ], W_tmp_Q9 ), 14 ); + W_adj_Q5[ i ] = silk_DIV32_varQ( (opus_int32)pW_Q2[ i ], silk_SMULBB( W_tmp_Q9, W_tmp_Q9 ), 21 ); + } + + /* Unpack entropy table indices and predictor for current CB1 index */ + silk_NLSF_unpack( ec_ix, pred_Q8, psNLSF_CB, ind1 ); + + /* Trellis quantizer */ + RD_Q25[ s ] = silk_NLSF_del_dec_quant( &tempIndices2[ s * MAX_LPC_ORDER ], res_Q10, W_adj_Q5, pred_Q8, ec_ix, + psNLSF_CB->ec_Rates_Q5, psNLSF_CB->quantStepSize_Q16, psNLSF_CB->invQuantStepSize_Q6, NLSF_mu_Q20, psNLSF_CB->order ); + + /* Add rate for first stage */ + iCDF_ptr = &psNLSF_CB->CB1_iCDF[ ( signalType >> 1 ) * psNLSF_CB->nVectors ]; + if( ind1 == 0 ) { + prob_Q8 = 256 - iCDF_ptr[ ind1 ]; + } else { + prob_Q8 = iCDF_ptr[ ind1 - 1 ] - iCDF_ptr[ ind1 ]; + } + bits_q7 = ( 8 << 7 ) - silk_lin2log( prob_Q8 ); + RD_Q25[ s ] = silk_SMLABB( RD_Q25[ s ], bits_q7, silk_RSHIFT( NLSF_mu_Q20, 2 ) ); + } + + /* Find the lowest rate-distortion error */ + silk_insertion_sort_increasing( RD_Q25, &bestIndex, nSurvivors, 1 ); + + NLSFIndices[ 0 ] = (opus_int8)tempIndices1[ bestIndex ]; + silk_memcpy( &NLSFIndices[ 1 ], &tempIndices2[ bestIndex * MAX_LPC_ORDER ], psNLSF_CB->order * sizeof( opus_int8 ) ); + + /* Decode */ + silk_NLSF_decode( pNLSF_Q15, NLSFIndices, psNLSF_CB ); + + ret = RD_Q25[ 0 ]; + RESTORE_STACK; + return ret; +} diff --git a/lib/opus/silk/NLSF_stabilize.c b/lib/opus/silk/NLSF_stabilize.c new file mode 100644 index 00000000..8f3426b9 --- /dev/null +++ b/lib/opus/silk/NLSF_stabilize.c @@ -0,0 +1,142 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* NLSF stabilizer: */ +/* */ +/* - Moves NLSFs further apart if they are too close */ +/* - Moves NLSFs away from borders if they are too close */ +/* - High effort to achieve a modification with minimum */ +/* Euclidean distance to input vector */ +/* - Output are sorted NLSF coefficients */ +/* */ + +#include "SigProc_FIX.h" + +/* Constant Definitions */ +#define MAX_LOOPS 20 + +/* NLSF stabilizer, for a single input data vector */ +void silk_NLSF_stabilize( + opus_int16 *NLSF_Q15, /* I/O Unstable/stabilized normalized LSF vector in Q15 [L] */ + const opus_int16 *NDeltaMin_Q15, /* I Min distance vector, NDeltaMin_Q15[L] must be >= 1 [L+1] */ + const opus_int L /* I Number of NLSF parameters in the input vector */ +) +{ + opus_int i, I=0, k, loops; + opus_int16 center_freq_Q15; + opus_int32 diff_Q15, min_diff_Q15, min_center_Q15, max_center_Q15; + + /* This is necessary to ensure an output within range of a opus_int16 */ + silk_assert( NDeltaMin_Q15[L] >= 1 ); + + for( loops = 0; loops < MAX_LOOPS; loops++ ) { + /**************************/ + /* Find smallest distance */ + /**************************/ + /* First element */ + min_diff_Q15 = NLSF_Q15[0] - NDeltaMin_Q15[0]; + I = 0; + /* Middle elements */ + for( i = 1; i <= L-1; i++ ) { + diff_Q15 = NLSF_Q15[i] - ( NLSF_Q15[i-1] + NDeltaMin_Q15[i] ); + if( diff_Q15 < min_diff_Q15 ) { + min_diff_Q15 = diff_Q15; + I = i; + } + } + /* Last element */ + diff_Q15 = ( 1 << 15 ) - ( NLSF_Q15[L-1] + NDeltaMin_Q15[L] ); + if( diff_Q15 < min_diff_Q15 ) { + min_diff_Q15 = diff_Q15; + I = L; + } + + /***************************************************/ + /* Now check if the smallest distance non-negative */ + /***************************************************/ + if( min_diff_Q15 >= 0 ) { + return; + } + + if( I == 0 ) { + /* Move away from lower limit */ + NLSF_Q15[0] = NDeltaMin_Q15[0]; + + } else if( I == L) { + /* Move away from higher limit */ + NLSF_Q15[L-1] = ( 1 << 15 ) - NDeltaMin_Q15[L]; + + } else { + /* Find the lower extreme for the location of the current center frequency */ + min_center_Q15 = 0; + for( k = 0; k < I; k++ ) { + min_center_Q15 += NDeltaMin_Q15[k]; + } + min_center_Q15 += silk_RSHIFT( NDeltaMin_Q15[I], 1 ); + + /* Find the upper extreme for the location of the current center frequency */ + max_center_Q15 = 1 << 15; + for( k = L; k > I; k-- ) { + max_center_Q15 -= NDeltaMin_Q15[k]; + } + max_center_Q15 -= silk_RSHIFT( NDeltaMin_Q15[I], 1 ); + + /* Move apart, sorted by value, keeping the same center frequency */ + center_freq_Q15 = (opus_int16)silk_LIMIT_32( silk_RSHIFT_ROUND( (opus_int32)NLSF_Q15[I-1] + (opus_int32)NLSF_Q15[I], 1 ), + min_center_Q15, max_center_Q15 ); + NLSF_Q15[I-1] = center_freq_Q15 - silk_RSHIFT( NDeltaMin_Q15[I], 1 ); + NLSF_Q15[I] = NLSF_Q15[I-1] + NDeltaMin_Q15[I]; + } + } + + /* Safe and simple fall back method, which is less ideal than the above */ + if( loops == MAX_LOOPS ) + { + /* Insertion sort (fast for already almost sorted arrays): */ + /* Best case: O(n) for an already sorted array */ + /* Worst case: O(n^2) for an inversely sorted array */ + silk_insertion_sort_increasing_all_values_int16( &NLSF_Q15[0], L ); + + /* First NLSF should be no less than NDeltaMin[0] */ + NLSF_Q15[0] = silk_max_int( NLSF_Q15[0], NDeltaMin_Q15[0] ); + + /* Keep delta_min distance between the NLSFs */ + for( i = 1; i < L; i++ ) + NLSF_Q15[i] = silk_max_int( NLSF_Q15[i], silk_ADD_SAT16( NLSF_Q15[i-1], NDeltaMin_Q15[i] ) ); + + /* Last NLSF should be no higher than 1 - NDeltaMin[L] */ + NLSF_Q15[L-1] = silk_min_int( NLSF_Q15[L-1], (1<<15) - NDeltaMin_Q15[L] ); + + /* Keep NDeltaMin distance between the NLSFs */ + for( i = L-2; i >= 0; i-- ) + NLSF_Q15[i] = silk_min_int( NLSF_Q15[i], NLSF_Q15[i+1] - NDeltaMin_Q15[i+1] ); + } +} diff --git a/lib/opus/silk/NLSF_unpack.c b/lib/opus/silk/NLSF_unpack.c new file mode 100644 index 00000000..17bd23f7 --- /dev/null +++ b/lib/opus/silk/NLSF_unpack.c @@ -0,0 +1,55 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Unpack predictor values and indices for entropy coding tables */ +void silk_NLSF_unpack( + opus_int16 ec_ix[], /* O Indices to entropy tables [ LPC_ORDER ] */ + opus_uint8 pred_Q8[], /* O LSF predictor [ LPC_ORDER ] */ + const silk_NLSF_CB_struct *psNLSF_CB, /* I Codebook object */ + const opus_int CB1_index /* I Index of vector in first LSF codebook */ +) +{ + opus_int i; + opus_uint8 entry; + const opus_uint8 *ec_sel_ptr; + + ec_sel_ptr = &psNLSF_CB->ec_sel[ CB1_index * psNLSF_CB->order / 2 ]; + for( i = 0; i < psNLSF_CB->order; i += 2 ) { + entry = *ec_sel_ptr++; + ec_ix [ i ] = silk_SMULBB( silk_RSHIFT( entry, 1 ) & 7, 2 * NLSF_QUANT_MAX_AMPLITUDE + 1 ); + pred_Q8[ i ] = psNLSF_CB->pred_Q8[ i + ( entry & 1 ) * ( psNLSF_CB->order - 1 ) ]; + ec_ix [ i + 1 ] = silk_SMULBB( silk_RSHIFT( entry, 5 ) & 7, 2 * NLSF_QUANT_MAX_AMPLITUDE + 1 ); + pred_Q8[ i + 1 ] = psNLSF_CB->pred_Q8[ i + ( silk_RSHIFT( entry, 4 ) & 1 ) * ( psNLSF_CB->order - 1 ) + 1 ]; + } +} + diff --git a/lib/opus/silk/NSQ.c b/lib/opus/silk/NSQ.c new file mode 100644 index 00000000..45dd45ce --- /dev/null +++ b/lib/opus/silk/NSQ.c @@ -0,0 +1,437 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" +#include "NSQ.h" + + +static OPUS_INLINE void silk_nsq_scale_states( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + const opus_int16 x16[], /* I input */ + opus_int32 x_sc_Q10[], /* O input scaled with 1/Gain */ + const opus_int16 sLTP[], /* I re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I subframe number */ + const opus_int LTP_scale_Q14, /* I */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type /* I Signal type */ +); + +#if !defined(OPUS_X86_MAY_HAVE_SSE4_1) +static OPUS_INLINE void silk_noise_shape_quantizer( + silk_nsq_state *NSQ, /* I/O NSQ state */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_sc_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP state */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping AR coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int shapingLPCOrder, /* I Noise shaping AR filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + int arch /* I Architecture */ +); +#endif + +void silk_NSQ_c +( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) +{ + opus_int k, lag, start_idx, LSF_interpolation_flag; + const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; + opus_int16 *pxq; + VARDECL( opus_int32, sLTP_Q15 ); + VARDECL( opus_int16, sLTP ); + opus_int32 HarmShapeFIRPacked_Q14; + opus_int offset_Q10; + VARDECL( opus_int32, x_sc_Q10 ); + SAVE_STACK; + + NSQ->rand_seed = psIndices->Seed; + + /* Set unvoiced lag to the previous one, overwrite later for voiced */ + lag = NSQ->lagPrev; + + silk_assert( NSQ->prev_gain_Q16 != 0 ); + + offset_Q10 = silk_Quantization_Offsets_Q10[ psIndices->signalType >> 1 ][ psIndices->quantOffsetType ]; + + if( psIndices->NLSFInterpCoef_Q2 == 4 ) { + LSF_interpolation_flag = 0; + } else { + LSF_interpolation_flag = 1; + } + + ALLOC( sLTP_Q15, psEncC->ltp_mem_length + psEncC->frame_length, opus_int32 ); + ALLOC( sLTP, psEncC->ltp_mem_length + psEncC->frame_length, opus_int16 ); + ALLOC( x_sc_Q10, psEncC->subfr_length, opus_int32 ); + /* Set up pointers to start of sub frame */ + NSQ->sLTP_shp_buf_idx = psEncC->ltp_mem_length; + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + pxq = &NSQ->xq[ psEncC->ltp_mem_length ]; + for( k = 0; k < psEncC->nb_subfr; k++ ) { + A_Q12 = &PredCoef_Q12[ (( k >> 1 ) | ( 1 - LSF_interpolation_flag )) * MAX_LPC_ORDER ]; + B_Q14 = <PCoef_Q14[ k * LTP_ORDER ]; + AR_shp_Q13 = &AR_Q13[ k * MAX_SHAPE_LPC_ORDER ]; + + /* Noise shape parameters */ + silk_assert( HarmShapeGain_Q14[ k ] >= 0 ); + HarmShapeFIRPacked_Q14 = silk_RSHIFT( HarmShapeGain_Q14[ k ], 2 ); + HarmShapeFIRPacked_Q14 |= silk_LSHIFT( (opus_int32)silk_RSHIFT( HarmShapeGain_Q14[ k ], 1 ), 16 ); + + NSQ->rewhite_flag = 0; + if( psIndices->signalType == TYPE_VOICED ) { + /* Voiced */ + lag = pitchL[ k ]; + + /* Re-whitening */ + if( ( k & ( 3 - silk_LSHIFT( LSF_interpolation_flag, 1 ) ) ) == 0 ) { + /* Rewhiten with new A coefs */ + start_idx = psEncC->ltp_mem_length - lag - psEncC->predictLPCOrder - LTP_ORDER / 2; + celt_assert( start_idx > 0 ); + + silk_LPC_analysis_filter( &sLTP[ start_idx ], &NSQ->xq[ start_idx + k * psEncC->subfr_length ], + A_Q12, psEncC->ltp_mem_length - start_idx, psEncC->predictLPCOrder, psEncC->arch ); + + NSQ->rewhite_flag = 1; + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + } + } + + silk_nsq_scale_states( psEncC, NSQ, x16, x_sc_Q10, sLTP, sLTP_Q15, k, LTP_scale_Q14, Gains_Q16, pitchL, psIndices->signalType ); + + silk_noise_shape_quantizer( NSQ, psIndices->signalType, x_sc_Q10, pulses, pxq, sLTP_Q15, A_Q12, B_Q14, + AR_shp_Q13, lag, HarmShapeFIRPacked_Q14, Tilt_Q14[ k ], LF_shp_Q14[ k ], Gains_Q16[ k ], Lambda_Q10, + offset_Q10, psEncC->subfr_length, psEncC->shapingLPCOrder, psEncC->predictLPCOrder, psEncC->arch ); + + x16 += psEncC->subfr_length; + pulses += psEncC->subfr_length; + pxq += psEncC->subfr_length; + } + + /* Update lagPrev for next frame */ + NSQ->lagPrev = pitchL[ psEncC->nb_subfr - 1 ]; + + /* Save quantized speech and noise shaping signals */ + silk_memmove( NSQ->xq, &NSQ->xq[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int16 ) ); + silk_memmove( NSQ->sLTP_shp_Q14, &NSQ->sLTP_shp_Q14[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int32 ) ); + RESTORE_STACK; +} + +/******************************/ +/* silk_noise_shape_quantizer */ +/******************************/ + +#if !defined(OPUS_X86_MAY_HAVE_SSE4_1) +static OPUS_INLINE +#endif +void silk_noise_shape_quantizer( + silk_nsq_state *NSQ, /* I/O NSQ state */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_sc_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP state */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping AR coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int shapingLPCOrder, /* I Noise shaping AR filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + int arch /* I Architecture */ +) +{ + opus_int i; + opus_int32 LTP_pred_Q13, LPC_pred_Q10, n_AR_Q12, n_LTP_Q13; + opus_int32 n_LF_Q12, r_Q10, rr_Q10, q1_Q0, q1_Q10, q2_Q10, rd1_Q20, rd2_Q20; + opus_int32 exc_Q14, LPC_exc_Q14, xq_Q14, Gain_Q10; + opus_int32 tmp1, tmp2, sLF_AR_shp_Q14; + opus_int32 *psLPC_Q14, *shp_lag_ptr, *pred_lag_ptr; +#ifdef silk_short_prediction_create_arch_coef + opus_int32 a_Q12_arch[MAX_LPC_ORDER]; +#endif + + shp_lag_ptr = &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - lag + HARM_SHAPE_FIR_TAPS / 2 ]; + pred_lag_ptr = &sLTP_Q15[ NSQ->sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + Gain_Q10 = silk_RSHIFT( Gain_Q16, 6 ); + + /* Set up short term AR state */ + psLPC_Q14 = &NSQ->sLPC_Q14[ NSQ_LPC_BUF_LENGTH - 1 ]; + +#ifdef silk_short_prediction_create_arch_coef + silk_short_prediction_create_arch_coef(a_Q12_arch, a_Q12, predictLPCOrder); +#endif + + for( i = 0; i < length; i++ ) { + /* Generate dither */ + NSQ->rand_seed = silk_RAND( NSQ->rand_seed ); + + /* Short-term prediction */ + LPC_pred_Q10 = silk_noise_shape_quantizer_short_prediction(psLPC_Q14, a_Q12, a_Q12_arch, predictLPCOrder, arch); + + /* Long-term prediction */ + if( signalType == TYPE_VOICED ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LTP_pred_Q13 = 2; + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ 0 ], b_Q14[ 0 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -1 ], b_Q14[ 1 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -2 ], b_Q14[ 2 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -3 ], b_Q14[ 3 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -4 ], b_Q14[ 4 ] ); + pred_lag_ptr++; + } else { + LTP_pred_Q13 = 0; + } + + /* Noise shape feedback */ + celt_assert( ( shapingLPCOrder & 1 ) == 0 ); /* check that order is even */ + n_AR_Q12 = silk_NSQ_noise_shape_feedback_loop(&NSQ->sDiff_shp_Q14, NSQ->sAR2_Q14, AR_shp_Q13, shapingLPCOrder, arch); + + n_AR_Q12 = silk_SMLAWB( n_AR_Q12, NSQ->sLF_AR_shp_Q14, Tilt_Q14 ); + + n_LF_Q12 = silk_SMULWB( NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - 1 ], LF_shp_Q14 ); + n_LF_Q12 = silk_SMLAWT( n_LF_Q12, NSQ->sLF_AR_shp_Q14, LF_shp_Q14 ); + + celt_assert( lag > 0 || signalType != TYPE_VOICED ); + + /* Combine prediction and noise shaping signals */ + tmp1 = silk_SUB32( silk_LSHIFT32( LPC_pred_Q10, 2 ), n_AR_Q12 ); /* Q12 */ + tmp1 = silk_SUB32( tmp1, n_LF_Q12 ); /* Q12 */ + if( lag > 0 ) { + /* Symmetric, packed FIR coefficients */ + n_LTP_Q13 = silk_SMULWB( silk_ADD_SAT32( shp_lag_ptr[ 0 ], shp_lag_ptr[ -2 ] ), HarmShapeFIRPacked_Q14 ); + n_LTP_Q13 = silk_SMLAWT( n_LTP_Q13, shp_lag_ptr[ -1 ], HarmShapeFIRPacked_Q14 ); + n_LTP_Q13 = silk_LSHIFT( n_LTP_Q13, 1 ); + shp_lag_ptr++; + + tmp2 = silk_SUB32( LTP_pred_Q13, n_LTP_Q13 ); /* Q13 */ + tmp1 = silk_ADD_LSHIFT32( tmp2, tmp1, 1 ); /* Q13 */ + tmp1 = silk_RSHIFT_ROUND( tmp1, 3 ); /* Q10 */ + } else { + tmp1 = silk_RSHIFT_ROUND( tmp1, 2 ); /* Q10 */ + } + + r_Q10 = silk_SUB32( x_sc_Q10[ i ], tmp1 ); /* residual error Q10 */ + + /* Flip sign depending on dither */ + if( NSQ->rand_seed < 0 ) { + r_Q10 = -r_Q10; + } + r_Q10 = silk_LIMIT_32( r_Q10, -(31 << 10), 30 << 10 ); + + /* Find two quantization level candidates and measure their rate-distortion */ + q1_Q10 = silk_SUB32( r_Q10, offset_Q10 ); + q1_Q0 = silk_RSHIFT( q1_Q10, 10 ); + if (Lambda_Q10 > 2048) { + /* For aggressive RDO, the bias becomes more than one pulse. */ + int rdo_offset = Lambda_Q10/2 - 512; + if (q1_Q10 > rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 - rdo_offset, 10 ); + } else if (q1_Q10 < -rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 + rdo_offset, 10 ); + } else if (q1_Q10 < 0) { + q1_Q0 = -1; + } else { + q1_Q0 = 0; + } + } + if( q1_Q0 > 0 ) { + q1_Q10 = silk_SUB32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q20 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q20 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == 0 ) { + q1_Q10 = offset_Q10; + q2_Q10 = silk_ADD32( q1_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q20 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q20 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == -1 ) { + q2_Q10 = offset_Q10; + q1_Q10 = silk_SUB32( q2_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q20 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q20 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else { /* Q1_Q0 < -1 */ + q1_Q10 = silk_ADD32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q20 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q20 = silk_SMULBB( -q2_Q10, Lambda_Q10 ); + } + rr_Q10 = silk_SUB32( r_Q10, q1_Q10 ); + rd1_Q20 = silk_SMLABB( rd1_Q20, rr_Q10, rr_Q10 ); + rr_Q10 = silk_SUB32( r_Q10, q2_Q10 ); + rd2_Q20 = silk_SMLABB( rd2_Q20, rr_Q10, rr_Q10 ); + + if( rd2_Q20 < rd1_Q20 ) { + q1_Q10 = q2_Q10; + } + + pulses[ i ] = (opus_int8)silk_RSHIFT_ROUND( q1_Q10, 10 ); + + /* Excitation */ + exc_Q14 = silk_LSHIFT( q1_Q10, 4 ); + if ( NSQ->rand_seed < 0 ) { + exc_Q14 = -exc_Q14; + } + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD_LSHIFT32( exc_Q14, LTP_pred_Q13, 1 ); + xq_Q14 = silk_ADD_LSHIFT32( LPC_exc_Q14, LPC_pred_Q10, 4 ); + + /* Scale XQ back to normal level before saving */ + xq[ i ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( xq_Q14, Gain_Q10 ), 8 ) ); + + /* Update states */ + psLPC_Q14++; + *psLPC_Q14 = xq_Q14; + NSQ->sDiff_shp_Q14 = silk_SUB_LSHIFT32( xq_Q14, x_sc_Q10[ i ], 4 ); + sLF_AR_shp_Q14 = silk_SUB_LSHIFT32( NSQ->sDiff_shp_Q14, n_AR_Q12, 2 ); + NSQ->sLF_AR_shp_Q14 = sLF_AR_shp_Q14; + + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx ] = silk_SUB_LSHIFT32( sLF_AR_shp_Q14, n_LF_Q12, 2 ); + sLTP_Q15[ NSQ->sLTP_buf_idx ] = silk_LSHIFT( LPC_exc_Q14, 1 ); + NSQ->sLTP_shp_buf_idx++; + NSQ->sLTP_buf_idx++; + + /* Make dither dependent on quantized signal */ + NSQ->rand_seed = silk_ADD32_ovflw( NSQ->rand_seed, pulses[ i ] ); + } + + /* Update LPC synth buffer */ + silk_memcpy( NSQ->sLPC_Q14, &NSQ->sLPC_Q14[ length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); +} + +static OPUS_INLINE void silk_nsq_scale_states( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + const opus_int16 x16[], /* I input */ + opus_int32 x_sc_Q10[], /* O input scaled with 1/Gain */ + const opus_int16 sLTP[], /* I re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I subframe number */ + const opus_int LTP_scale_Q14, /* I */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type /* I Signal type */ +) +{ + opus_int i, lag; + opus_int32 gain_adj_Q16, inv_gain_Q31, inv_gain_Q26; + + lag = pitchL[ subfr ]; + inv_gain_Q31 = silk_INVERSE32_varQ( silk_max( Gains_Q16[ subfr ], 1 ), 47 ); + silk_assert( inv_gain_Q31 != 0 ); + + /* Scale input */ + inv_gain_Q26 = silk_RSHIFT_ROUND( inv_gain_Q31, 5 ); + for( i = 0; i < psEncC->subfr_length; i++ ) { + x_sc_Q10[ i ] = silk_SMULWW( x16[ i ], inv_gain_Q26 ); + } + + /* After rewhitening the LTP state is un-scaled, so scale with inv_gain_Q16 */ + if( NSQ->rewhite_flag ) { + if( subfr == 0 ) { + /* Do LTP downscaling */ + inv_gain_Q31 = silk_LSHIFT( silk_SMULWB( inv_gain_Q31, LTP_scale_Q14 ), 2 ); + } + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx; i++ ) { + silk_assert( i < MAX_FRAME_LENGTH ); + sLTP_Q15[ i ] = silk_SMULWB( inv_gain_Q31, sLTP[ i ] ); + } + } + + /* Adjust for changing gain */ + if( Gains_Q16[ subfr ] != NSQ->prev_gain_Q16 ) { + gain_adj_Q16 = silk_DIV32_varQ( NSQ->prev_gain_Q16, Gains_Q16[ subfr ], 16 ); + + /* Scale long-term shaping state */ + for( i = NSQ->sLTP_shp_buf_idx - psEncC->ltp_mem_length; i < NSQ->sLTP_shp_buf_idx; i++ ) { + NSQ->sLTP_shp_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLTP_shp_Q14[ i ] ); + } + + /* Scale long-term prediction state */ + if( signal_type == TYPE_VOICED && NSQ->rewhite_flag == 0 ) { + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx; i++ ) { + sLTP_Q15[ i ] = silk_SMULWW( gain_adj_Q16, sLTP_Q15[ i ] ); + } + } + + NSQ->sLF_AR_shp_Q14 = silk_SMULWW( gain_adj_Q16, NSQ->sLF_AR_shp_Q14 ); + NSQ->sDiff_shp_Q14 = silk_SMULWW( gain_adj_Q16, NSQ->sDiff_shp_Q14 ); + + /* Scale short-term prediction and shaping states */ + for( i = 0; i < NSQ_LPC_BUF_LENGTH; i++ ) { + NSQ->sLPC_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLPC_Q14[ i ] ); + } + for( i = 0; i < MAX_SHAPE_LPC_ORDER; i++ ) { + NSQ->sAR2_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sAR2_Q14[ i ] ); + } + + /* Save inverse gain */ + NSQ->prev_gain_Q16 = Gains_Q16[ subfr ]; + } +} diff --git a/lib/opus/silk/NSQ.h b/lib/opus/silk/NSQ.h new file mode 100644 index 00000000..971832f6 --- /dev/null +++ b/lib/opus/silk/NSQ.h @@ -0,0 +1,101 @@ +/*********************************************************************** +Copyright (c) 2014 Vidyo. +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ +#ifndef SILK_NSQ_H +#define SILK_NSQ_H + +#include "SigProc_FIX.h" + +#undef silk_short_prediction_create_arch_coef + +static OPUS_INLINE opus_int32 silk_noise_shape_quantizer_short_prediction_c(const opus_int32 *buf32, const opus_int16 *coef16, opus_int order) +{ + opus_int32 out; + silk_assert( order == 10 || order == 16 ); + + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + out = silk_RSHIFT( order, 1 ); + out = silk_SMLAWB( out, buf32[ 0 ], coef16[ 0 ] ); + out = silk_SMLAWB( out, buf32[ -1 ], coef16[ 1 ] ); + out = silk_SMLAWB( out, buf32[ -2 ], coef16[ 2 ] ); + out = silk_SMLAWB( out, buf32[ -3 ], coef16[ 3 ] ); + out = silk_SMLAWB( out, buf32[ -4 ], coef16[ 4 ] ); + out = silk_SMLAWB( out, buf32[ -5 ], coef16[ 5 ] ); + out = silk_SMLAWB( out, buf32[ -6 ], coef16[ 6 ] ); + out = silk_SMLAWB( out, buf32[ -7 ], coef16[ 7 ] ); + out = silk_SMLAWB( out, buf32[ -8 ], coef16[ 8 ] ); + out = silk_SMLAWB( out, buf32[ -9 ], coef16[ 9 ] ); + + if( order == 16 ) + { + out = silk_SMLAWB( out, buf32[ -10 ], coef16[ 10 ] ); + out = silk_SMLAWB( out, buf32[ -11 ], coef16[ 11 ] ); + out = silk_SMLAWB( out, buf32[ -12 ], coef16[ 12 ] ); + out = silk_SMLAWB( out, buf32[ -13 ], coef16[ 13 ] ); + out = silk_SMLAWB( out, buf32[ -14 ], coef16[ 14 ] ); + out = silk_SMLAWB( out, buf32[ -15 ], coef16[ 15 ] ); + } + return out; +} + +#define silk_noise_shape_quantizer_short_prediction(in, coef, coefRev, order, arch) ((void)arch,silk_noise_shape_quantizer_short_prediction_c(in, coef, order)) + +static OPUS_INLINE opus_int32 silk_NSQ_noise_shape_feedback_loop_c(const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef, opus_int order) +{ + opus_int32 out; + opus_int32 tmp1, tmp2; + opus_int j; + + tmp2 = data0[0]; + tmp1 = data1[0]; + data1[0] = tmp2; + + out = silk_RSHIFT(order, 1); + out = silk_SMLAWB(out, tmp2, coef[0]); + + for (j = 2; j < order; j += 2) { + tmp2 = data1[j - 1]; + data1[j - 1] = tmp1; + out = silk_SMLAWB(out, tmp1, coef[j - 1]); + tmp1 = data1[j + 0]; + data1[j + 0] = tmp2; + out = silk_SMLAWB(out, tmp2, coef[j]); + } + data1[order - 1] = tmp1; + out = silk_SMLAWB(out, tmp1, coef[order - 1]); + /* Q11 -> Q12 */ + out = silk_LSHIFT32( out, 1 ); + return out; +} + +#define silk_NSQ_noise_shape_feedback_loop(data0, data1, coef, order, arch) ((void)arch,silk_NSQ_noise_shape_feedback_loop_c(data0, data1, coef, order)) + +#if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +#include "arm/NSQ_neon.h" +#endif + +#endif /* SILK_NSQ_H */ diff --git a/lib/opus/silk/NSQ_del_dec.c b/lib/opus/silk/NSQ_del_dec.c new file mode 100644 index 00000000..41f3fc93 --- /dev/null +++ b/lib/opus/silk/NSQ_del_dec.c @@ -0,0 +1,733 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" +#include "NSQ.h" + + +typedef struct { + opus_int32 sLPC_Q14[ MAX_SUB_FRAME_LENGTH + NSQ_LPC_BUF_LENGTH ]; + opus_int32 RandState[ DECISION_DELAY ]; + opus_int32 Q_Q10[ DECISION_DELAY ]; + opus_int32 Xq_Q14[ DECISION_DELAY ]; + opus_int32 Pred_Q15[ DECISION_DELAY ]; + opus_int32 Shape_Q14[ DECISION_DELAY ]; + opus_int32 sAR2_Q14[ MAX_SHAPE_LPC_ORDER ]; + opus_int32 LF_AR_Q14; + opus_int32 Diff_Q14; + opus_int32 Seed; + opus_int32 SeedInit; + opus_int32 RD_Q10; +} NSQ_del_dec_struct; + +typedef struct { + opus_int32 Q_Q10; + opus_int32 RD_Q10; + opus_int32 xq_Q14; + opus_int32 LF_AR_Q14; + opus_int32 Diff_Q14; + opus_int32 sLTP_shp_Q14; + opus_int32 LPC_exc_Q14; +} NSQ_sample_struct; + +typedef NSQ_sample_struct NSQ_sample_pair[ 2 ]; + +#if defined(MIPSr1_ASM) +#include "mips/NSQ_del_dec_mipsr1.h" +#endif +static OPUS_INLINE void silk_nsq_del_dec_scale_states( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + const opus_int16 x16[], /* I Input */ + opus_int32 x_sc_Q10[], /* O Input scaled with 1/Gain in Q10 */ + const opus_int16 sLTP[], /* I Re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I Subframe number */ + opus_int nStatesDelayedDecision, /* I Number of del dec states */ + const opus_int LTP_scale_Q14, /* I LTP state scaling */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type, /* I Signal type */ + const opus_int decisionDelay /* I Decision delay */ +); + +/******************************************/ +/* Noise shape quantizer for one subframe */ +/******************************************/ +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec( + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP filter state */ + opus_int32 delayedGain_Q10[], /* I/O Gain delay buffer */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int subfr, /* I Subframe number */ + opus_int shapingLPCOrder, /* I Shaping LPC filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + opus_int warping_Q16, /* I */ + opus_int nStatesDelayedDecision, /* I Number of states in decision tree */ + opus_int *smpl_buf_idx, /* I/O Index to newest samples in buffers */ + opus_int decisionDelay, /* I */ + int arch /* I */ +); + +void silk_NSQ_del_dec_c( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) +{ + opus_int i, k, lag, start_idx, LSF_interpolation_flag, Winner_ind, subfr; + opus_int last_smple_idx, smpl_buf_idx, decisionDelay; + const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; + opus_int16 *pxq; + VARDECL( opus_int32, sLTP_Q15 ); + VARDECL( opus_int16, sLTP ); + opus_int32 HarmShapeFIRPacked_Q14; + opus_int offset_Q10; + opus_int32 RDmin_Q10, Gain_Q10; + VARDECL( opus_int32, x_sc_Q10 ); + VARDECL( opus_int32, delayedGain_Q10 ); + VARDECL( NSQ_del_dec_struct, psDelDec ); + NSQ_del_dec_struct *psDD; + SAVE_STACK; + + /* Set unvoiced lag to the previous one, overwrite later for voiced */ + lag = NSQ->lagPrev; + + silk_assert( NSQ->prev_gain_Q16 != 0 ); + + /* Initialize delayed decision states */ + ALLOC( psDelDec, psEncC->nStatesDelayedDecision, NSQ_del_dec_struct ); + silk_memset( psDelDec, 0, psEncC->nStatesDelayedDecision * sizeof( NSQ_del_dec_struct ) ); + for( k = 0; k < psEncC->nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + psDD->Seed = ( k + psIndices->Seed ) & 3; + psDD->SeedInit = psDD->Seed; + psDD->RD_Q10 = 0; + psDD->LF_AR_Q14 = NSQ->sLF_AR_shp_Q14; + psDD->Diff_Q14 = NSQ->sDiff_shp_Q14; + psDD->Shape_Q14[ 0 ] = NSQ->sLTP_shp_Q14[ psEncC->ltp_mem_length - 1 ]; + silk_memcpy( psDD->sLPC_Q14, NSQ->sLPC_Q14, NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + silk_memcpy( psDD->sAR2_Q14, NSQ->sAR2_Q14, sizeof( NSQ->sAR2_Q14 ) ); + } + + offset_Q10 = silk_Quantization_Offsets_Q10[ psIndices->signalType >> 1 ][ psIndices->quantOffsetType ]; + smpl_buf_idx = 0; /* index of oldest samples */ + + decisionDelay = silk_min_int( DECISION_DELAY, psEncC->subfr_length ); + + /* For voiced frames limit the decision delay to lower than the pitch lag */ + if( psIndices->signalType == TYPE_VOICED ) { + for( k = 0; k < psEncC->nb_subfr; k++ ) { + decisionDelay = silk_min_int( decisionDelay, pitchL[ k ] - LTP_ORDER / 2 - 1 ); + } + } else { + if( lag > 0 ) { + decisionDelay = silk_min_int( decisionDelay, lag - LTP_ORDER / 2 - 1 ); + } + } + + if( psIndices->NLSFInterpCoef_Q2 == 4 ) { + LSF_interpolation_flag = 0; + } else { + LSF_interpolation_flag = 1; + } + + ALLOC( sLTP_Q15, psEncC->ltp_mem_length + psEncC->frame_length, opus_int32 ); + ALLOC( sLTP, psEncC->ltp_mem_length + psEncC->frame_length, opus_int16 ); + ALLOC( x_sc_Q10, psEncC->subfr_length, opus_int32 ); + ALLOC( delayedGain_Q10, DECISION_DELAY, opus_int32 ); + /* Set up pointers to start of sub frame */ + pxq = &NSQ->xq[ psEncC->ltp_mem_length ]; + NSQ->sLTP_shp_buf_idx = psEncC->ltp_mem_length; + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + subfr = 0; + for( k = 0; k < psEncC->nb_subfr; k++ ) { + A_Q12 = &PredCoef_Q12[ ( ( k >> 1 ) | ( 1 - LSF_interpolation_flag ) ) * MAX_LPC_ORDER ]; + B_Q14 = <PCoef_Q14[ k * LTP_ORDER ]; + AR_shp_Q13 = &AR_Q13[ k * MAX_SHAPE_LPC_ORDER ]; + + /* Noise shape parameters */ + silk_assert( HarmShapeGain_Q14[ k ] >= 0 ); + HarmShapeFIRPacked_Q14 = silk_RSHIFT( HarmShapeGain_Q14[ k ], 2 ); + HarmShapeFIRPacked_Q14 |= silk_LSHIFT( (opus_int32)silk_RSHIFT( HarmShapeGain_Q14[ k ], 1 ), 16 ); + + NSQ->rewhite_flag = 0; + if( psIndices->signalType == TYPE_VOICED ) { + /* Voiced */ + lag = pitchL[ k ]; + + /* Re-whitening */ + if( ( k & ( 3 - silk_LSHIFT( LSF_interpolation_flag, 1 ) ) ) == 0 ) { + if( k == 2 ) { + /* RESET DELAYED DECISIONS */ + /* Find winner */ + RDmin_Q10 = psDelDec[ 0 ].RD_Q10; + Winner_ind = 0; + for( i = 1; i < psEncC->nStatesDelayedDecision; i++ ) { + if( psDelDec[ i ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psDelDec[ i ].RD_Q10; + Winner_ind = i; + } + } + for( i = 0; i < psEncC->nStatesDelayedDecision; i++ ) { + if( i != Winner_ind ) { + psDelDec[ i ].RD_Q10 += ( silk_int32_MAX >> 4 ); + silk_assert( psDelDec[ i ].RD_Q10 >= 0 ); + } + } + + /* Copy final part of signals from winner state to output and long-term filter states */ + psDD = &psDelDec[ Winner_ind ]; + last_smple_idx = smpl_buf_idx + decisionDelay; + for( i = 0; i < decisionDelay; i++ ) { + last_smple_idx = ( last_smple_idx - 1 ) % DECISION_DELAY; + if( last_smple_idx < 0 ) last_smple_idx += DECISION_DELAY; + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDD->Q_Q10[ last_smple_idx ], 10 ); + pxq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDD->Xq_Q14[ last_smple_idx ], Gains_Q16[ 1 ] ), 14 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay + i ] = psDD->Shape_Q14[ last_smple_idx ]; + } + + subfr = 0; + } + + /* Rewhiten with new A coefs */ + start_idx = psEncC->ltp_mem_length - lag - psEncC->predictLPCOrder - LTP_ORDER / 2; + celt_assert( start_idx > 0 ); + + silk_LPC_analysis_filter( &sLTP[ start_idx ], &NSQ->xq[ start_idx + k * psEncC->subfr_length ], + A_Q12, psEncC->ltp_mem_length - start_idx, psEncC->predictLPCOrder, psEncC->arch ); + + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + NSQ->rewhite_flag = 1; + } + } + + silk_nsq_del_dec_scale_states( psEncC, NSQ, psDelDec, x16, x_sc_Q10, sLTP, sLTP_Q15, k, + psEncC->nStatesDelayedDecision, LTP_scale_Q14, Gains_Q16, pitchL, psIndices->signalType, decisionDelay ); + + silk_noise_shape_quantizer_del_dec( NSQ, psDelDec, psIndices->signalType, x_sc_Q10, pulses, pxq, sLTP_Q15, + delayedGain_Q10, A_Q12, B_Q14, AR_shp_Q13, lag, HarmShapeFIRPacked_Q14, Tilt_Q14[ k ], LF_shp_Q14[ k ], + Gains_Q16[ k ], Lambda_Q10, offset_Q10, psEncC->subfr_length, subfr++, psEncC->shapingLPCOrder, + psEncC->predictLPCOrder, psEncC->warping_Q16, psEncC->nStatesDelayedDecision, &smpl_buf_idx, decisionDelay, psEncC->arch ); + + x16 += psEncC->subfr_length; + pulses += psEncC->subfr_length; + pxq += psEncC->subfr_length; + } + + /* Find winner */ + RDmin_Q10 = psDelDec[ 0 ].RD_Q10; + Winner_ind = 0; + for( k = 1; k < psEncC->nStatesDelayedDecision; k++ ) { + if( psDelDec[ k ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psDelDec[ k ].RD_Q10; + Winner_ind = k; + } + } + + /* Copy final part of signals from winner state to output and long-term filter states */ + psDD = &psDelDec[ Winner_ind ]; + psIndices->Seed = psDD->SeedInit; + last_smple_idx = smpl_buf_idx + decisionDelay; + Gain_Q10 = silk_RSHIFT32( Gains_Q16[ psEncC->nb_subfr - 1 ], 6 ); + for( i = 0; i < decisionDelay; i++ ) { + last_smple_idx = ( last_smple_idx - 1 ) % DECISION_DELAY; + if( last_smple_idx < 0 ) last_smple_idx += DECISION_DELAY; + + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDD->Q_Q10[ last_smple_idx ], 10 ); + pxq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDD->Xq_Q14[ last_smple_idx ], Gain_Q10 ), 8 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay + i ] = psDD->Shape_Q14[ last_smple_idx ]; + } + silk_memcpy( NSQ->sLPC_Q14, &psDD->sLPC_Q14[ psEncC->subfr_length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + silk_memcpy( NSQ->sAR2_Q14, psDD->sAR2_Q14, sizeof( psDD->sAR2_Q14 ) ); + + /* Update states */ + NSQ->sLF_AR_shp_Q14 = psDD->LF_AR_Q14; + NSQ->sDiff_shp_Q14 = psDD->Diff_Q14; + NSQ->lagPrev = pitchL[ psEncC->nb_subfr - 1 ]; + + /* Save quantized speech signal */ + silk_memmove( NSQ->xq, &NSQ->xq[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int16 ) ); + silk_memmove( NSQ->sLTP_shp_Q14, &NSQ->sLTP_shp_Q14[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int32 ) ); + RESTORE_STACK; +} + +/******************************************/ +/* Noise shape quantizer for one subframe */ +/******************************************/ +#ifndef OVERRIDE_silk_noise_shape_quantizer_del_dec +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec( + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP filter state */ + opus_int32 delayedGain_Q10[], /* I/O Gain delay buffer */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int subfr, /* I Subframe number */ + opus_int shapingLPCOrder, /* I Shaping LPC filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + opus_int warping_Q16, /* I */ + opus_int nStatesDelayedDecision, /* I Number of states in decision tree */ + opus_int *smpl_buf_idx, /* I/O Index to newest samples in buffers */ + opus_int decisionDelay, /* I */ + int arch /* I */ +) +{ + opus_int i, j, k, Winner_ind, RDmin_ind, RDmax_ind, last_smple_idx; + opus_int32 Winner_rand_state; + opus_int32 LTP_pred_Q14, LPC_pred_Q14, n_AR_Q14, n_LTP_Q14; + opus_int32 n_LF_Q14, r_Q10, rr_Q10, rd1_Q10, rd2_Q10, RDmin_Q10, RDmax_Q10; + opus_int32 q1_Q0, q1_Q10, q2_Q10, exc_Q14, LPC_exc_Q14, xq_Q14, Gain_Q10; + opus_int32 tmp1, tmp2, sLF_AR_shp_Q14; + opus_int32 *pred_lag_ptr, *shp_lag_ptr, *psLPC_Q14; +#ifdef silk_short_prediction_create_arch_coef + opus_int32 a_Q12_arch[MAX_LPC_ORDER]; +#endif + + VARDECL( NSQ_sample_pair, psSampleState ); + NSQ_del_dec_struct *psDD; + NSQ_sample_struct *psSS; + SAVE_STACK; + + celt_assert( nStatesDelayedDecision > 0 ); + ALLOC( psSampleState, nStatesDelayedDecision, NSQ_sample_pair ); + + shp_lag_ptr = &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - lag + HARM_SHAPE_FIR_TAPS / 2 ]; + pred_lag_ptr = &sLTP_Q15[ NSQ->sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + Gain_Q10 = silk_RSHIFT( Gain_Q16, 6 ); + +#ifdef silk_short_prediction_create_arch_coef + silk_short_prediction_create_arch_coef(a_Q12_arch, a_Q12, predictLPCOrder); +#endif + + for( i = 0; i < length; i++ ) { + /* Perform common calculations used in all states */ + + /* Long-term prediction */ + if( signalType == TYPE_VOICED ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LTP_pred_Q14 = 2; + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ 0 ], b_Q14[ 0 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -1 ], b_Q14[ 1 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -2 ], b_Q14[ 2 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -3 ], b_Q14[ 3 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -4 ], b_Q14[ 4 ] ); + LTP_pred_Q14 = silk_LSHIFT( LTP_pred_Q14, 1 ); /* Q13 -> Q14 */ + pred_lag_ptr++; + } else { + LTP_pred_Q14 = 0; + } + + /* Long-term shaping */ + if( lag > 0 ) { + /* Symmetric, packed FIR coefficients */ + n_LTP_Q14 = silk_SMULWB( silk_ADD_SAT32( shp_lag_ptr[ 0 ], shp_lag_ptr[ -2 ] ), HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SMLAWT( n_LTP_Q14, shp_lag_ptr[ -1 ], HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SUB_LSHIFT32( LTP_pred_Q14, n_LTP_Q14, 2 ); /* Q12 -> Q14 */ + shp_lag_ptr++; + } else { + n_LTP_Q14 = 0; + } + + for( k = 0; k < nStatesDelayedDecision; k++ ) { + /* Delayed decision state */ + psDD = &psDelDec[ k ]; + + /* Sample state */ + psSS = psSampleState[ k ]; + + /* Generate dither */ + psDD->Seed = silk_RAND( psDD->Seed ); + + /* Pointer used in short term prediction and shaping */ + psLPC_Q14 = &psDD->sLPC_Q14[ NSQ_LPC_BUF_LENGTH - 1 + i ]; + /* Short-term prediction */ + LPC_pred_Q14 = silk_noise_shape_quantizer_short_prediction(psLPC_Q14, a_Q12, a_Q12_arch, predictLPCOrder, arch); + LPC_pred_Q14 = silk_LSHIFT( LPC_pred_Q14, 4 ); /* Q10 -> Q14 */ + + /* Noise shape feedback */ + celt_assert( ( shapingLPCOrder & 1 ) == 0 ); /* check that order is even */ + /* Output of lowpass section */ + tmp2 = silk_SMLAWB( psDD->Diff_Q14, psDD->sAR2_Q14[ 0 ], warping_Q16 ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( psDD->sAR2_Q14[ 0 ], psDD->sAR2_Q14[ 1 ] - tmp2, warping_Q16 ); + psDD->sAR2_Q14[ 0 ] = tmp2; + n_AR_Q14 = silk_RSHIFT( shapingLPCOrder, 1 ); + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp2, AR_shp_Q13[ 0 ] ); + /* Loop over allpass sections */ + for( j = 2; j < shapingLPCOrder; j += 2 ) { + /* Output of allpass section */ + tmp2 = silk_SMLAWB( psDD->sAR2_Q14[ j - 1 ], psDD->sAR2_Q14[ j + 0 ] - tmp1, warping_Q16 ); + psDD->sAR2_Q14[ j - 1 ] = tmp1; + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp1, AR_shp_Q13[ j - 1 ] ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( psDD->sAR2_Q14[ j + 0 ], psDD->sAR2_Q14[ j + 1 ] - tmp2, warping_Q16 ); + psDD->sAR2_Q14[ j + 0 ] = tmp2; + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp2, AR_shp_Q13[ j ] ); + } + psDD->sAR2_Q14[ shapingLPCOrder - 1 ] = tmp1; + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp1, AR_shp_Q13[ shapingLPCOrder - 1 ] ); + + n_AR_Q14 = silk_LSHIFT( n_AR_Q14, 1 ); /* Q11 -> Q12 */ + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, psDD->LF_AR_Q14, Tilt_Q14 ); /* Q12 */ + n_AR_Q14 = silk_LSHIFT( n_AR_Q14, 2 ); /* Q12 -> Q14 */ + + n_LF_Q14 = silk_SMULWB( psDD->Shape_Q14[ *smpl_buf_idx ], LF_shp_Q14 ); /* Q12 */ + n_LF_Q14 = silk_SMLAWT( n_LF_Q14, psDD->LF_AR_Q14, LF_shp_Q14 ); /* Q12 */ + n_LF_Q14 = silk_LSHIFT( n_LF_Q14, 2 ); /* Q12 -> Q14 */ + + /* Input minus prediction plus noise feedback */ + /* r = x[ i ] - LTP_pred - LPC_pred + n_AR + n_Tilt + n_LF + n_LTP */ + tmp1 = silk_ADD_SAT32( n_AR_Q14, n_LF_Q14 ); /* Q14 */ + tmp2 = silk_ADD32( n_LTP_Q14, LPC_pred_Q14 ); /* Q13 */ + tmp1 = silk_SUB_SAT32( tmp2, tmp1 ); /* Q13 */ + tmp1 = silk_RSHIFT_ROUND( tmp1, 4 ); /* Q10 */ + + r_Q10 = silk_SUB32( x_Q10[ i ], tmp1 ); /* residual error Q10 */ + + /* Flip sign depending on dither */ + if ( psDD->Seed < 0 ) { + r_Q10 = -r_Q10; + } + r_Q10 = silk_LIMIT_32( r_Q10, -(31 << 10), 30 << 10 ); + + /* Find two quantization level candidates and measure their rate-distortion */ + q1_Q10 = silk_SUB32( r_Q10, offset_Q10 ); + q1_Q0 = silk_RSHIFT( q1_Q10, 10 ); + if (Lambda_Q10 > 2048) { + /* For aggressive RDO, the bias becomes more than one pulse. */ + int rdo_offset = Lambda_Q10/2 - 512; + if (q1_Q10 > rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 - rdo_offset, 10 ); + } else if (q1_Q10 < -rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 + rdo_offset, 10 ); + } else if (q1_Q10 < 0) { + q1_Q0 = -1; + } else { + q1_Q0 = 0; + } + } + if( q1_Q0 > 0 ) { + q1_Q10 = silk_SUB32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q10 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == 0 ) { + q1_Q10 = offset_Q10; + q2_Q10 = silk_ADD32( q1_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q10 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == -1 ) { + q2_Q10 = offset_Q10; + q1_Q10 = silk_SUB32( q2_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q10 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else { /* q1_Q0 < -1 */ + q1_Q10 = silk_ADD32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q10 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( -q2_Q10, Lambda_Q10 ); + } + rr_Q10 = silk_SUB32( r_Q10, q1_Q10 ); + rd1_Q10 = silk_RSHIFT( silk_SMLABB( rd1_Q10, rr_Q10, rr_Q10 ), 10 ); + rr_Q10 = silk_SUB32( r_Q10, q2_Q10 ); + rd2_Q10 = silk_RSHIFT( silk_SMLABB( rd2_Q10, rr_Q10, rr_Q10 ), 10 ); + + if( rd1_Q10 < rd2_Q10 ) { + psSS[ 0 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd1_Q10 ); + psSS[ 1 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd2_Q10 ); + psSS[ 0 ].Q_Q10 = q1_Q10; + psSS[ 1 ].Q_Q10 = q2_Q10; + } else { + psSS[ 0 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd2_Q10 ); + psSS[ 1 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd1_Q10 ); + psSS[ 0 ].Q_Q10 = q2_Q10; + psSS[ 1 ].Q_Q10 = q1_Q10; + } + + /* Update states for best quantization */ + + /* Quantized excitation */ + exc_Q14 = silk_LSHIFT32( psSS[ 0 ].Q_Q10, 4 ); + if ( psDD->Seed < 0 ) { + exc_Q14 = -exc_Q14; + } + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD32( exc_Q14, LTP_pred_Q14 ); + xq_Q14 = silk_ADD32( LPC_exc_Q14, LPC_pred_Q14 ); + + /* Update states */ + psSS[ 0 ].Diff_Q14 = silk_SUB_LSHIFT32( xq_Q14, x_Q10[ i ], 4 ); + sLF_AR_shp_Q14 = silk_SUB32( psSS[ 0 ].Diff_Q14, n_AR_Q14 ); + psSS[ 0 ].sLTP_shp_Q14 = silk_SUB_SAT32( sLF_AR_shp_Q14, n_LF_Q14 ); + psSS[ 0 ].LF_AR_Q14 = sLF_AR_shp_Q14; + psSS[ 0 ].LPC_exc_Q14 = LPC_exc_Q14; + psSS[ 0 ].xq_Q14 = xq_Q14; + + /* Update states for second best quantization */ + + /* Quantized excitation */ + exc_Q14 = silk_LSHIFT32( psSS[ 1 ].Q_Q10, 4 ); + if ( psDD->Seed < 0 ) { + exc_Q14 = -exc_Q14; + } + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD32( exc_Q14, LTP_pred_Q14 ); + xq_Q14 = silk_ADD32( LPC_exc_Q14, LPC_pred_Q14 ); + + /* Update states */ + psSS[ 1 ].Diff_Q14 = silk_SUB_LSHIFT32( xq_Q14, x_Q10[ i ], 4 ); + sLF_AR_shp_Q14 = silk_SUB32( psSS[ 1 ].Diff_Q14, n_AR_Q14 ); + psSS[ 1 ].sLTP_shp_Q14 = silk_SUB_SAT32( sLF_AR_shp_Q14, n_LF_Q14 ); + psSS[ 1 ].LF_AR_Q14 = sLF_AR_shp_Q14; + psSS[ 1 ].LPC_exc_Q14 = LPC_exc_Q14; + psSS[ 1 ].xq_Q14 = xq_Q14; + } + + *smpl_buf_idx = ( *smpl_buf_idx - 1 ) % DECISION_DELAY; + if( *smpl_buf_idx < 0 ) *smpl_buf_idx += DECISION_DELAY; + last_smple_idx = ( *smpl_buf_idx + decisionDelay ) % DECISION_DELAY; + + /* Find winner */ + RDmin_Q10 = psSampleState[ 0 ][ 0 ].RD_Q10; + Winner_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + if( psSampleState[ k ][ 0 ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ k ][ 0 ].RD_Q10; + Winner_ind = k; + } + } + + /* Increase RD values of expired states */ + Winner_rand_state = psDelDec[ Winner_ind ].RandState[ last_smple_idx ]; + for( k = 0; k < nStatesDelayedDecision; k++ ) { + if( psDelDec[ k ].RandState[ last_smple_idx ] != Winner_rand_state ) { + psSampleState[ k ][ 0 ].RD_Q10 = silk_ADD32( psSampleState[ k ][ 0 ].RD_Q10, silk_int32_MAX >> 4 ); + psSampleState[ k ][ 1 ].RD_Q10 = silk_ADD32( psSampleState[ k ][ 1 ].RD_Q10, silk_int32_MAX >> 4 ); + silk_assert( psSampleState[ k ][ 0 ].RD_Q10 >= 0 ); + } + } + + /* Find worst in first set and best in second set */ + RDmax_Q10 = psSampleState[ 0 ][ 0 ].RD_Q10; + RDmin_Q10 = psSampleState[ 0 ][ 1 ].RD_Q10; + RDmax_ind = 0; + RDmin_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + /* find worst in first set */ + if( psSampleState[ k ][ 0 ].RD_Q10 > RDmax_Q10 ) { + RDmax_Q10 = psSampleState[ k ][ 0 ].RD_Q10; + RDmax_ind = k; + } + /* find best in second set */ + if( psSampleState[ k ][ 1 ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ k ][ 1 ].RD_Q10; + RDmin_ind = k; + } + } + + /* Replace a state if best from second set outperforms worst in first set */ + if( RDmin_Q10 < RDmax_Q10 ) { + silk_memcpy( ( (opus_int32 *)&psDelDec[ RDmax_ind ] ) + i, + ( (opus_int32 *)&psDelDec[ RDmin_ind ] ) + i, sizeof( NSQ_del_dec_struct ) - i * sizeof( opus_int32) ); + silk_memcpy( &psSampleState[ RDmax_ind ][ 0 ], &psSampleState[ RDmin_ind ][ 1 ], sizeof( NSQ_sample_struct ) ); + } + + /* Write samples from winner to output and long-term filter states */ + psDD = &psDelDec[ Winner_ind ]; + if( subfr > 0 || i >= decisionDelay ) { + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDD->Q_Q10[ last_smple_idx ], 10 ); + xq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDD->Xq_Q14[ last_smple_idx ], delayedGain_Q10[ last_smple_idx ] ), 8 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay ] = psDD->Shape_Q14[ last_smple_idx ]; + sLTP_Q15[ NSQ->sLTP_buf_idx - decisionDelay ] = psDD->Pred_Q15[ last_smple_idx ]; + } + NSQ->sLTP_shp_buf_idx++; + NSQ->sLTP_buf_idx++; + + /* Update states */ + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + psSS = &psSampleState[ k ][ 0 ]; + psDD->LF_AR_Q14 = psSS->LF_AR_Q14; + psDD->Diff_Q14 = psSS->Diff_Q14; + psDD->sLPC_Q14[ NSQ_LPC_BUF_LENGTH + i ] = psSS->xq_Q14; + psDD->Xq_Q14[ *smpl_buf_idx ] = psSS->xq_Q14; + psDD->Q_Q10[ *smpl_buf_idx ] = psSS->Q_Q10; + psDD->Pred_Q15[ *smpl_buf_idx ] = silk_LSHIFT32( psSS->LPC_exc_Q14, 1 ); + psDD->Shape_Q14[ *smpl_buf_idx ] = psSS->sLTP_shp_Q14; + psDD->Seed = silk_ADD32_ovflw( psDD->Seed, silk_RSHIFT_ROUND( psSS->Q_Q10, 10 ) ); + psDD->RandState[ *smpl_buf_idx ] = psDD->Seed; + psDD->RD_Q10 = psSS->RD_Q10; + } + delayedGain_Q10[ *smpl_buf_idx ] = Gain_Q10; + } + /* Update LPC states */ + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + silk_memcpy( psDD->sLPC_Q14, &psDD->sLPC_Q14[ length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + } + RESTORE_STACK; +} +#endif /* OVERRIDE_silk_noise_shape_quantizer_del_dec */ + +static OPUS_INLINE void silk_nsq_del_dec_scale_states( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + const opus_int16 x16[], /* I Input */ + opus_int32 x_sc_Q10[], /* O Input scaled with 1/Gain in Q10 */ + const opus_int16 sLTP[], /* I Re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I Subframe number */ + opus_int nStatesDelayedDecision, /* I Number of del dec states */ + const opus_int LTP_scale_Q14, /* I LTP state scaling */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type, /* I Signal type */ + const opus_int decisionDelay /* I Decision delay */ +) +{ + opus_int i, k, lag; + opus_int32 gain_adj_Q16, inv_gain_Q31, inv_gain_Q26; + NSQ_del_dec_struct *psDD; + + lag = pitchL[ subfr ]; + inv_gain_Q31 = silk_INVERSE32_varQ( silk_max( Gains_Q16[ subfr ], 1 ), 47 ); + silk_assert( inv_gain_Q31 != 0 ); + + /* Scale input */ + inv_gain_Q26 = silk_RSHIFT_ROUND( inv_gain_Q31, 5 ); + for( i = 0; i < psEncC->subfr_length; i++ ) { + x_sc_Q10[ i ] = silk_SMULWW( x16[ i ], inv_gain_Q26 ); + } + + /* After rewhitening the LTP state is un-scaled, so scale with inv_gain_Q16 */ + if( NSQ->rewhite_flag ) { + if( subfr == 0 ) { + /* Do LTP downscaling */ + inv_gain_Q31 = silk_LSHIFT( silk_SMULWB( inv_gain_Q31, LTP_scale_Q14 ), 2 ); + } + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx; i++ ) { + silk_assert( i < MAX_FRAME_LENGTH ); + sLTP_Q15[ i ] = silk_SMULWB( inv_gain_Q31, sLTP[ i ] ); + } + } + + /* Adjust for changing gain */ + if( Gains_Q16[ subfr ] != NSQ->prev_gain_Q16 ) { + gain_adj_Q16 = silk_DIV32_varQ( NSQ->prev_gain_Q16, Gains_Q16[ subfr ], 16 ); + + /* Scale long-term shaping state */ + for( i = NSQ->sLTP_shp_buf_idx - psEncC->ltp_mem_length; i < NSQ->sLTP_shp_buf_idx; i++ ) { + NSQ->sLTP_shp_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLTP_shp_Q14[ i ] ); + } + + /* Scale long-term prediction state */ + if( signal_type == TYPE_VOICED && NSQ->rewhite_flag == 0 ) { + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx - decisionDelay; i++ ) { + sLTP_Q15[ i ] = silk_SMULWW( gain_adj_Q16, sLTP_Q15[ i ] ); + } + } + + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + + /* Scale scalar states */ + psDD->LF_AR_Q14 = silk_SMULWW( gain_adj_Q16, psDD->LF_AR_Q14 ); + psDD->Diff_Q14 = silk_SMULWW( gain_adj_Q16, psDD->Diff_Q14 ); + + /* Scale short-term prediction and shaping states */ + for( i = 0; i < NSQ_LPC_BUF_LENGTH; i++ ) { + psDD->sLPC_Q14[ i ] = silk_SMULWW( gain_adj_Q16, psDD->sLPC_Q14[ i ] ); + } + for( i = 0; i < MAX_SHAPE_LPC_ORDER; i++ ) { + psDD->sAR2_Q14[ i ] = silk_SMULWW( gain_adj_Q16, psDD->sAR2_Q14[ i ] ); + } + for( i = 0; i < DECISION_DELAY; i++ ) { + psDD->Pred_Q15[ i ] = silk_SMULWW( gain_adj_Q16, psDD->Pred_Q15[ i ] ); + psDD->Shape_Q14[ i ] = silk_SMULWW( gain_adj_Q16, psDD->Shape_Q14[ i ] ); + } + } + + /* Save inverse gain */ + NSQ->prev_gain_Q16 = Gains_Q16[ subfr ]; + } +} diff --git a/lib/opus/silk/PLC.c b/lib/opus/silk/PLC.c new file mode 100644 index 00000000..4667440d --- /dev/null +++ b/lib/opus/silk/PLC.c @@ -0,0 +1,446 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" +#include "PLC.h" + +#define NB_ATT 2 +static const opus_int16 HARM_ATT_Q15[NB_ATT] = { 32440, 31130 }; /* 0.99, 0.95 */ +static const opus_int16 PLC_RAND_ATTENUATE_V_Q15[NB_ATT] = { 31130, 26214 }; /* 0.95, 0.8 */ +static const opus_int16 PLC_RAND_ATTENUATE_UV_Q15[NB_ATT] = { 32440, 29491 }; /* 0.99, 0.9 */ + +static OPUS_INLINE void silk_PLC_update( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl /* I/O Decoder control */ +); + +static OPUS_INLINE void silk_PLC_conceal( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int16 frame[], /* O LPC residual signal */ + int arch /* I Run-time architecture */ +); + + +void silk_PLC_Reset( + silk_decoder_state *psDec /* I/O Decoder state */ +) +{ + psDec->sPLC.pitchL_Q8 = silk_LSHIFT( psDec->frame_length, 8 - 1 ); + psDec->sPLC.prevGain_Q16[ 0 ] = SILK_FIX_CONST( 1, 16 ); + psDec->sPLC.prevGain_Q16[ 1 ] = SILK_FIX_CONST( 1, 16 ); + psDec->sPLC.subfr_length = 20; + psDec->sPLC.nb_subfr = 2; +} + +void silk_PLC( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int16 frame[], /* I/O signal */ + opus_int lost, /* I Loss flag */ + int arch /* I Run-time architecture */ +) +{ + /* PLC control function */ + if( psDec->fs_kHz != psDec->sPLC.fs_kHz ) { + silk_PLC_Reset( psDec ); + psDec->sPLC.fs_kHz = psDec->fs_kHz; + } + + if( lost ) { + /****************************/ + /* Generate Signal */ + /****************************/ + silk_PLC_conceal( psDec, psDecCtrl, frame, arch ); + + psDec->lossCnt++; + } else { + /****************************/ + /* Update state */ + /****************************/ + silk_PLC_update( psDec, psDecCtrl ); + } +} + +/**************************************************/ +/* Update state of PLC */ +/**************************************************/ +static OPUS_INLINE void silk_PLC_update( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl /* I/O Decoder control */ +) +{ + opus_int32 LTP_Gain_Q14, temp_LTP_Gain_Q14; + opus_int i, j; + silk_PLC_struct *psPLC; + + psPLC = &psDec->sPLC; + + /* Update parameters used in case of packet loss */ + psDec->prevSignalType = psDec->indices.signalType; + LTP_Gain_Q14 = 0; + if( psDec->indices.signalType == TYPE_VOICED ) { + /* Find the parameters for the last subframe which contains a pitch pulse */ + for( j = 0; j * psDec->subfr_length < psDecCtrl->pitchL[ psDec->nb_subfr - 1 ]; j++ ) { + if( j == psDec->nb_subfr ) { + break; + } + temp_LTP_Gain_Q14 = 0; + for( i = 0; i < LTP_ORDER; i++ ) { + temp_LTP_Gain_Q14 += psDecCtrl->LTPCoef_Q14[ ( psDec->nb_subfr - 1 - j ) * LTP_ORDER + i ]; + } + if( temp_LTP_Gain_Q14 > LTP_Gain_Q14 ) { + LTP_Gain_Q14 = temp_LTP_Gain_Q14; + silk_memcpy( psPLC->LTPCoef_Q14, + &psDecCtrl->LTPCoef_Q14[ silk_SMULBB( psDec->nb_subfr - 1 - j, LTP_ORDER ) ], + LTP_ORDER * sizeof( opus_int16 ) ); + + psPLC->pitchL_Q8 = silk_LSHIFT( psDecCtrl->pitchL[ psDec->nb_subfr - 1 - j ], 8 ); + } + } + + silk_memset( psPLC->LTPCoef_Q14, 0, LTP_ORDER * sizeof( opus_int16 ) ); + psPLC->LTPCoef_Q14[ LTP_ORDER / 2 ] = LTP_Gain_Q14; + + /* Limit LT coefs */ + if( LTP_Gain_Q14 < V_PITCH_GAIN_START_MIN_Q14 ) { + opus_int scale_Q10; + opus_int32 tmp; + + tmp = silk_LSHIFT( V_PITCH_GAIN_START_MIN_Q14, 10 ); + scale_Q10 = silk_DIV32( tmp, silk_max( LTP_Gain_Q14, 1 ) ); + for( i = 0; i < LTP_ORDER; i++ ) { + psPLC->LTPCoef_Q14[ i ] = silk_RSHIFT( silk_SMULBB( psPLC->LTPCoef_Q14[ i ], scale_Q10 ), 10 ); + } + } else if( LTP_Gain_Q14 > V_PITCH_GAIN_START_MAX_Q14 ) { + opus_int scale_Q14; + opus_int32 tmp; + + tmp = silk_LSHIFT( V_PITCH_GAIN_START_MAX_Q14, 14 ); + scale_Q14 = silk_DIV32( tmp, silk_max( LTP_Gain_Q14, 1 ) ); + for( i = 0; i < LTP_ORDER; i++ ) { + psPLC->LTPCoef_Q14[ i ] = silk_RSHIFT( silk_SMULBB( psPLC->LTPCoef_Q14[ i ], scale_Q14 ), 14 ); + } + } + } else { + psPLC->pitchL_Q8 = silk_LSHIFT( silk_SMULBB( psDec->fs_kHz, 18 ), 8 ); + silk_memset( psPLC->LTPCoef_Q14, 0, LTP_ORDER * sizeof( opus_int16 )); + } + + /* Save LPC coeficients */ + silk_memcpy( psPLC->prevLPC_Q12, psDecCtrl->PredCoef_Q12[ 1 ], psDec->LPC_order * sizeof( opus_int16 ) ); + psPLC->prevLTP_scale_Q14 = psDecCtrl->LTP_scale_Q14; + + /* Save last two gains */ + silk_memcpy( psPLC->prevGain_Q16, &psDecCtrl->Gains_Q16[ psDec->nb_subfr - 2 ], 2 * sizeof( opus_int32 ) ); + + psPLC->subfr_length = psDec->subfr_length; + psPLC->nb_subfr = psDec->nb_subfr; +} + +static OPUS_INLINE void silk_PLC_energy(opus_int32 *energy1, opus_int *shift1, opus_int32 *energy2, opus_int *shift2, + const opus_int32 *exc_Q14, const opus_int32 *prevGain_Q10, int subfr_length, int nb_subfr) +{ + int i, k; + VARDECL( opus_int16, exc_buf ); + opus_int16 *exc_buf_ptr; + SAVE_STACK; + ALLOC( exc_buf, 2*subfr_length, opus_int16 ); + /* Find random noise component */ + /* Scale previous excitation signal */ + exc_buf_ptr = exc_buf; + for( k = 0; k < 2; k++ ) { + for( i = 0; i < subfr_length; i++ ) { + exc_buf_ptr[ i ] = (opus_int16)silk_SAT16( silk_RSHIFT( + silk_SMULWW( exc_Q14[ i + ( k + nb_subfr - 2 ) * subfr_length ], prevGain_Q10[ k ] ), 8 ) ); + } + exc_buf_ptr += subfr_length; + } + /* Find the subframe with lowest energy of the last two and use that as random noise generator */ + silk_sum_sqr_shift( energy1, shift1, exc_buf, subfr_length ); + silk_sum_sqr_shift( energy2, shift2, &exc_buf[ subfr_length ], subfr_length ); + RESTORE_STACK; +} + +static OPUS_INLINE void silk_PLC_conceal( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int16 frame[], /* O LPC residual signal */ + int arch /* I Run-time architecture */ +) +{ + opus_int i, j, k; + opus_int lag, idx, sLTP_buf_idx, shift1, shift2; + opus_int32 rand_seed, harm_Gain_Q15, rand_Gain_Q15, inv_gain_Q30; + opus_int32 energy1, energy2, *rand_ptr, *pred_lag_ptr; + opus_int32 LPC_pred_Q10, LTP_pred_Q12; + opus_int16 rand_scale_Q14; + opus_int16 *B_Q14; + opus_int32 *sLPC_Q14_ptr; + opus_int16 A_Q12[ MAX_LPC_ORDER ]; +#ifdef SMALL_FOOTPRINT + opus_int16 *sLTP; +#else + VARDECL( opus_int16, sLTP ); +#endif + VARDECL( opus_int32, sLTP_Q14 ); + silk_PLC_struct *psPLC = &psDec->sPLC; + opus_int32 prevGain_Q10[2]; + SAVE_STACK; + + ALLOC( sLTP_Q14, psDec->ltp_mem_length + psDec->frame_length, opus_int32 ); +#ifdef SMALL_FOOTPRINT + /* Ugly hack that breaks aliasing rules to save stack: put sLTP at the very end of sLTP_Q14. */ + sLTP = ((opus_int16*)&sLTP_Q14[psDec->ltp_mem_length + psDec->frame_length])-psDec->ltp_mem_length; +#else + ALLOC( sLTP, psDec->ltp_mem_length, opus_int16 ); +#endif + + prevGain_Q10[0] = silk_RSHIFT( psPLC->prevGain_Q16[ 0 ], 6); + prevGain_Q10[1] = silk_RSHIFT( psPLC->prevGain_Q16[ 1 ], 6); + + if( psDec->first_frame_after_reset ) { + silk_memset( psPLC->prevLPC_Q12, 0, sizeof( psPLC->prevLPC_Q12 ) ); + } + + silk_PLC_energy(&energy1, &shift1, &energy2, &shift2, psDec->exc_Q14, prevGain_Q10, psDec->subfr_length, psDec->nb_subfr); + + if( silk_RSHIFT( energy1, shift2 ) < silk_RSHIFT( energy2, shift1 ) ) { + /* First sub-frame has lowest energy */ + rand_ptr = &psDec->exc_Q14[ silk_max_int( 0, ( psPLC->nb_subfr - 1 ) * psPLC->subfr_length - RAND_BUF_SIZE ) ]; + } else { + /* Second sub-frame has lowest energy */ + rand_ptr = &psDec->exc_Q14[ silk_max_int( 0, psPLC->nb_subfr * psPLC->subfr_length - RAND_BUF_SIZE ) ]; + } + + /* Set up Gain to random noise component */ + B_Q14 = psPLC->LTPCoef_Q14; + rand_scale_Q14 = psPLC->randScale_Q14; + + /* Set up attenuation gains */ + harm_Gain_Q15 = HARM_ATT_Q15[ silk_min_int( NB_ATT - 1, psDec->lossCnt ) ]; + if( psDec->prevSignalType == TYPE_VOICED ) { + rand_Gain_Q15 = PLC_RAND_ATTENUATE_V_Q15[ silk_min_int( NB_ATT - 1, psDec->lossCnt ) ]; + } else { + rand_Gain_Q15 = PLC_RAND_ATTENUATE_UV_Q15[ silk_min_int( NB_ATT - 1, psDec->lossCnt ) ]; + } + + /* LPC concealment. Apply BWE to previous LPC */ + silk_bwexpander( psPLC->prevLPC_Q12, psDec->LPC_order, SILK_FIX_CONST( BWE_COEF, 16 ) ); + + /* Preload LPC coeficients to array on stack. Gives small performance gain */ + silk_memcpy( A_Q12, psPLC->prevLPC_Q12, psDec->LPC_order * sizeof( opus_int16 ) ); + + /* First Lost frame */ + if( psDec->lossCnt == 0 ) { + rand_scale_Q14 = 1 << 14; + + /* Reduce random noise Gain for voiced frames */ + if( psDec->prevSignalType == TYPE_VOICED ) { + for( i = 0; i < LTP_ORDER; i++ ) { + rand_scale_Q14 -= B_Q14[ i ]; + } + rand_scale_Q14 = silk_max_16( 3277, rand_scale_Q14 ); /* 0.2 */ + rand_scale_Q14 = (opus_int16)silk_RSHIFT( silk_SMULBB( rand_scale_Q14, psPLC->prevLTP_scale_Q14 ), 14 ); + } else { + /* Reduce random noise for unvoiced frames with high LPC gain */ + opus_int32 invGain_Q30, down_scale_Q30; + + invGain_Q30 = silk_LPC_inverse_pred_gain( psPLC->prevLPC_Q12, psDec->LPC_order, arch ); + + down_scale_Q30 = silk_min_32( silk_RSHIFT( (opus_int32)1 << 30, LOG2_INV_LPC_GAIN_HIGH_THRES ), invGain_Q30 ); + down_scale_Q30 = silk_max_32( silk_RSHIFT( (opus_int32)1 << 30, LOG2_INV_LPC_GAIN_LOW_THRES ), down_scale_Q30 ); + down_scale_Q30 = silk_LSHIFT( down_scale_Q30, LOG2_INV_LPC_GAIN_HIGH_THRES ); + + rand_Gain_Q15 = silk_RSHIFT( silk_SMULWB( down_scale_Q30, rand_Gain_Q15 ), 14 ); + } + } + + rand_seed = psPLC->rand_seed; + lag = silk_RSHIFT_ROUND( psPLC->pitchL_Q8, 8 ); + sLTP_buf_idx = psDec->ltp_mem_length; + + /* Rewhiten LTP state */ + idx = psDec->ltp_mem_length - lag - psDec->LPC_order - LTP_ORDER / 2; + celt_assert( idx > 0 ); + silk_LPC_analysis_filter( &sLTP[ idx ], &psDec->outBuf[ idx ], A_Q12, psDec->ltp_mem_length - idx, psDec->LPC_order, arch ); + /* Scale LTP state */ + inv_gain_Q30 = silk_INVERSE32_varQ( psPLC->prevGain_Q16[ 1 ], 46 ); + inv_gain_Q30 = silk_min( inv_gain_Q30, silk_int32_MAX >> 1 ); + for( i = idx + psDec->LPC_order; i < psDec->ltp_mem_length; i++ ) { + sLTP_Q14[ i ] = silk_SMULWB( inv_gain_Q30, sLTP[ i ] ); + } + + /***************************/ + /* LTP synthesis filtering */ + /***************************/ + for( k = 0; k < psDec->nb_subfr; k++ ) { + /* Set up pointer */ + pred_lag_ptr = &sLTP_Q14[ sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + for( i = 0; i < psDec->subfr_length; i++ ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LTP_pred_Q12 = 2; + LTP_pred_Q12 = silk_SMLAWB( LTP_pred_Q12, pred_lag_ptr[ 0 ], B_Q14[ 0 ] ); + LTP_pred_Q12 = silk_SMLAWB( LTP_pred_Q12, pred_lag_ptr[ -1 ], B_Q14[ 1 ] ); + LTP_pred_Q12 = silk_SMLAWB( LTP_pred_Q12, pred_lag_ptr[ -2 ], B_Q14[ 2 ] ); + LTP_pred_Q12 = silk_SMLAWB( LTP_pred_Q12, pred_lag_ptr[ -3 ], B_Q14[ 3 ] ); + LTP_pred_Q12 = silk_SMLAWB( LTP_pred_Q12, pred_lag_ptr[ -4 ], B_Q14[ 4 ] ); + pred_lag_ptr++; + + /* Generate LPC excitation */ + rand_seed = silk_RAND( rand_seed ); + idx = silk_RSHIFT( rand_seed, 25 ) & RAND_BUF_MASK; + sLTP_Q14[ sLTP_buf_idx ] = silk_LSHIFT32( silk_SMLAWB( LTP_pred_Q12, rand_ptr[ idx ], rand_scale_Q14 ), 2 ); + sLTP_buf_idx++; + } + + /* Gradually reduce LTP gain */ + for( j = 0; j < LTP_ORDER; j++ ) { + B_Q14[ j ] = silk_RSHIFT( silk_SMULBB( harm_Gain_Q15, B_Q14[ j ] ), 15 ); + } + /* Gradually reduce excitation gain */ + rand_scale_Q14 = silk_RSHIFT( silk_SMULBB( rand_scale_Q14, rand_Gain_Q15 ), 15 ); + + /* Slowly increase pitch lag */ + psPLC->pitchL_Q8 = silk_SMLAWB( psPLC->pitchL_Q8, psPLC->pitchL_Q8, PITCH_DRIFT_FAC_Q16 ); + psPLC->pitchL_Q8 = silk_min_32( psPLC->pitchL_Q8, silk_LSHIFT( silk_SMULBB( MAX_PITCH_LAG_MS, psDec->fs_kHz ), 8 ) ); + lag = silk_RSHIFT_ROUND( psPLC->pitchL_Q8, 8 ); + } + + /***************************/ + /* LPC synthesis filtering */ + /***************************/ + sLPC_Q14_ptr = &sLTP_Q14[ psDec->ltp_mem_length - MAX_LPC_ORDER ]; + + /* Copy LPC state */ + silk_memcpy( sLPC_Q14_ptr, psDec->sLPC_Q14_buf, MAX_LPC_ORDER * sizeof( opus_int32 ) ); + + celt_assert( psDec->LPC_order >= 10 ); /* check that unrolling works */ + for( i = 0; i < psDec->frame_length; i++ ) { + /* partly unrolled */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LPC_pred_Q10 = silk_RSHIFT( psDec->LPC_order, 1 ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 1 ], A_Q12[ 0 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 2 ], A_Q12[ 1 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 3 ], A_Q12[ 2 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 4 ], A_Q12[ 3 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 5 ], A_Q12[ 4 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 6 ], A_Q12[ 5 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 7 ], A_Q12[ 6 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 8 ], A_Q12[ 7 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 9 ], A_Q12[ 8 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - 10 ], A_Q12[ 9 ] ); + for( j = 10; j < psDec->LPC_order; j++ ) { + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14_ptr[ MAX_LPC_ORDER + i - j - 1 ], A_Q12[ j ] ); + } + + /* Add prediction to LPC excitation */ + sLPC_Q14_ptr[ MAX_LPC_ORDER + i ] = silk_ADD_SAT32( sLPC_Q14_ptr[ MAX_LPC_ORDER + i ], + silk_LSHIFT_SAT32( LPC_pred_Q10, 4 )); + + /* Scale with Gain */ + frame[ i ] = (opus_int16)silk_SAT16( silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( sLPC_Q14_ptr[ MAX_LPC_ORDER + i ], prevGain_Q10[ 1 ] ), 8 ) ) ); + } + + /* Save LPC state */ + silk_memcpy( psDec->sLPC_Q14_buf, &sLPC_Q14_ptr[ psDec->frame_length ], MAX_LPC_ORDER * sizeof( opus_int32 ) ); + + /**************************************/ + /* Update states */ + /**************************************/ + psPLC->rand_seed = rand_seed; + psPLC->randScale_Q14 = rand_scale_Q14; + for( i = 0; i < MAX_NB_SUBFR; i++ ) { + psDecCtrl->pitchL[ i ] = lag; + } + RESTORE_STACK; +} + +/* Glues concealed frames with new good received frames */ +void silk_PLC_glue_frames( + silk_decoder_state *psDec, /* I/O decoder state */ + opus_int16 frame[], /* I/O signal */ + opus_int length /* I length of signal */ +) +{ + opus_int i, energy_shift; + opus_int32 energy; + silk_PLC_struct *psPLC; + psPLC = &psDec->sPLC; + + if( psDec->lossCnt ) { + /* Calculate energy in concealed residual */ + silk_sum_sqr_shift( &psPLC->conc_energy, &psPLC->conc_energy_shift, frame, length ); + + psPLC->last_frame_lost = 1; + } else { + if( psDec->sPLC.last_frame_lost ) { + /* Calculate residual in decoded signal if last frame was lost */ + silk_sum_sqr_shift( &energy, &energy_shift, frame, length ); + + /* Normalize energies */ + if( energy_shift > psPLC->conc_energy_shift ) { + psPLC->conc_energy = silk_RSHIFT( psPLC->conc_energy, energy_shift - psPLC->conc_energy_shift ); + } else if( energy_shift < psPLC->conc_energy_shift ) { + energy = silk_RSHIFT( energy, psPLC->conc_energy_shift - energy_shift ); + } + + /* Fade in the energy difference */ + if( energy > psPLC->conc_energy ) { + opus_int32 frac_Q24, LZ; + opus_int32 gain_Q16, slope_Q16; + + LZ = silk_CLZ32( psPLC->conc_energy ); + LZ = LZ - 1; + psPLC->conc_energy = silk_LSHIFT( psPLC->conc_energy, LZ ); + energy = silk_RSHIFT( energy, silk_max_32( 24 - LZ, 0 ) ); + + frac_Q24 = silk_DIV32( psPLC->conc_energy, silk_max( energy, 1 ) ); + + gain_Q16 = silk_LSHIFT( silk_SQRT_APPROX( frac_Q24 ), 4 ); + slope_Q16 = silk_DIV32_16( ( (opus_int32)1 << 16 ) - gain_Q16, length ); + /* Make slope 4x steeper to avoid missing onsets after DTX */ + slope_Q16 = silk_LSHIFT( slope_Q16, 2 ); + + for( i = 0; i < length; i++ ) { + frame[ i ] = silk_SMULWB( gain_Q16, frame[ i ] ); + gain_Q16 += slope_Q16; + if( gain_Q16 > (opus_int32)1 << 16 ) { + break; + } + } + } + } + psPLC->last_frame_lost = 0; + } +} diff --git a/lib/opus/silk/PLC.h b/lib/opus/silk/PLC.h new file mode 100644 index 00000000..6438f516 --- /dev/null +++ b/lib/opus/silk/PLC.h @@ -0,0 +1,62 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_PLC_H +#define SILK_PLC_H + +#include "main.h" + +#define BWE_COEF 0.99 +#define V_PITCH_GAIN_START_MIN_Q14 11469 /* 0.7 in Q14 */ +#define V_PITCH_GAIN_START_MAX_Q14 15565 /* 0.95 in Q14 */ +#define MAX_PITCH_LAG_MS 18 +#define RAND_BUF_SIZE 128 +#define RAND_BUF_MASK ( RAND_BUF_SIZE - 1 ) +#define LOG2_INV_LPC_GAIN_HIGH_THRES 3 /* 2^3 = 8 dB LPC gain */ +#define LOG2_INV_LPC_GAIN_LOW_THRES 8 /* 2^8 = 24 dB LPC gain */ +#define PITCH_DRIFT_FAC_Q16 655 /* 0.01 in Q16 */ + +void silk_PLC_Reset( + silk_decoder_state *psDec /* I/O Decoder state */ +); + +void silk_PLC( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int16 frame[], /* I/O signal */ + opus_int lost, /* I Loss flag */ + int arch /* I Run-time architecture */ +); + +void silk_PLC_glue_frames( + silk_decoder_state *psDec, /* I/O decoder state */ + opus_int16 frame[], /* I/O signal */ + opus_int length /* I length of signal */ +); + +#endif + diff --git a/lib/opus/silk/SigProc_FIX.h b/lib/opus/silk/SigProc_FIX.h new file mode 100644 index 00000000..fbdfa82e --- /dev/null +++ b/lib/opus/silk/SigProc_FIX.h @@ -0,0 +1,643 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_SIGPROC_FIX_H +#define SILK_SIGPROC_FIX_H + +#ifdef __cplusplus +extern "C" +{ +#endif + +/*#define silk_MACRO_COUNT */ /* Used to enable WMOPS counting */ + +#define SILK_MAX_ORDER_LPC 24 /* max order of the LPC analysis in schur() and k2a() */ + +#include /* for memset(), memcpy(), memmove() */ +#include "typedef.h" +#include "resampler_structs.h" +#include "macros.h" +#include "cpu_support.h" + +#if defined(OPUS_X86_MAY_HAVE_SSE4_1) +#include "x86/SigProc_FIX_sse.h" +#endif + +#if (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#include "arm/biquad_alt_arm.h" +#include "arm/LPC_inv_pred_gain_arm.h" +#endif + +/********************************************************************/ +/* SIGNAL PROCESSING FUNCTIONS */ +/********************************************************************/ + +/*! + * Initialize/reset the resampler state for a given pair of input/output sampling rates +*/ +opus_int silk_resampler_init( + silk_resampler_state_struct *S, /* I/O Resampler state */ + opus_int32 Fs_Hz_in, /* I Input sampling rate (Hz) */ + opus_int32 Fs_Hz_out, /* I Output sampling rate (Hz) */ + opus_int forEnc /* I If 1: encoder; if 0: decoder */ +); + +/*! + * Resampler: convert from one sampling rate to another + */ +opus_int silk_resampler( + silk_resampler_state_struct *S, /* I/O Resampler state */ + opus_int16 out[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + opus_int32 inLen /* I Number of input samples */ +); + +/*! +* Downsample 2x, mediocre quality +*/ +void silk_resampler_down2( + opus_int32 *S, /* I/O State vector [ 2 ] */ + opus_int16 *out, /* O Output signal [ len ] */ + const opus_int16 *in, /* I Input signal [ floor(len/2) ] */ + opus_int32 inLen /* I Number of input samples */ +); + +/*! + * Downsample by a factor 2/3, low quality +*/ +void silk_resampler_down2_3( + opus_int32 *S, /* I/O State vector [ 6 ] */ + opus_int16 *out, /* O Output signal [ floor(2*inLen/3) ] */ + const opus_int16 *in, /* I Input signal [ inLen ] */ + opus_int32 inLen /* I Number of input samples */ +); + +/*! + * second order ARMA filter; + * slower than biquad() but uses more precise coefficients + * can handle (slowly) varying coefficients + */ +void silk_biquad_alt_stride1( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [2] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ +); + +void silk_biquad_alt_stride2_c( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [4] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ +); + +/* Variable order MA prediction error filter. */ +void silk_LPC_analysis_filter( + opus_int16 *out, /* O Output signal */ + const opus_int16 *in, /* I Input signal */ + const opus_int16 *B, /* I MA prediction coefficients, Q12 [order] */ + const opus_int32 len, /* I Signal length */ + const opus_int32 d, /* I Filter order */ + int arch /* I Run-time architecture */ +); + +/* Chirp (bandwidth expand) LP AR filter */ +void silk_bwexpander( + opus_int16 *ar, /* I/O AR filter to be expanded (without leading 1) */ + const opus_int d, /* I Length of ar */ + opus_int32 chirp_Q16 /* I Chirp factor (typically in the range 0 to 1) */ +); + +/* Chirp (bandwidth expand) LP AR filter */ +void silk_bwexpander_32( + opus_int32 *ar, /* I/O AR filter to be expanded (without leading 1) */ + const opus_int d, /* I Length of ar */ + opus_int32 chirp_Q16 /* I Chirp factor in Q16 */ +); + +/* Compute inverse of LPC prediction gain, and */ +/* test if LPC coefficients are stable (all poles within unit circle) */ +opus_int32 silk_LPC_inverse_pred_gain_c( /* O Returns inverse prediction gain in energy domain, Q30 */ + const opus_int16 *A_Q12, /* I Prediction coefficients, Q12 [order] */ + const opus_int order /* I Prediction order */ +); + +/* Split signal in two decimated bands using first-order allpass filters */ +void silk_ana_filt_bank_1( + const opus_int16 *in, /* I Input signal [N] */ + opus_int32 *S, /* I/O State vector [2] */ + opus_int16 *outL, /* O Low band [N/2] */ + opus_int16 *outH, /* O High band [N/2] */ + const opus_int32 N /* I Number of input samples */ +); + +#if !defined(OVERRIDE_silk_biquad_alt_stride2) +#define silk_biquad_alt_stride2(in, B_Q28, A_Q28, S, out, len, arch) ((void)(arch), silk_biquad_alt_stride2_c(in, B_Q28, A_Q28, S, out, len)) +#endif + +#if !defined(OVERRIDE_silk_LPC_inverse_pred_gain) +#define silk_LPC_inverse_pred_gain(A_Q12, order, arch) ((void)(arch), silk_LPC_inverse_pred_gain_c(A_Q12, order)) +#endif + +/********************************************************************/ +/* SCALAR FUNCTIONS */ +/********************************************************************/ + +/* Approximation of 128 * log2() (exact inverse of approx 2^() below) */ +/* Convert input to a log scale */ +opus_int32 silk_lin2log( + const opus_int32 inLin /* I input in linear scale */ +); + +/* Approximation of a sigmoid function */ +opus_int silk_sigm_Q15( + opus_int in_Q5 /* I */ +); + +/* Approximation of 2^() (exact inverse of approx log2() above) */ +/* Convert input to a linear scale */ +opus_int32 silk_log2lin( + const opus_int32 inLog_Q7 /* I input on log scale */ +); + +/* Compute number of bits to right shift the sum of squares of a vector */ +/* of int16s to make it fit in an int32 */ +void silk_sum_sqr_shift( + opus_int32 *energy, /* O Energy of x, after shifting to the right */ + opus_int *shift, /* O Number of bits right shift applied to energy */ + const opus_int16 *x, /* I Input vector */ + opus_int len /* I Length of input vector */ +); + +/* Calculates the reflection coefficients from the correlation sequence */ +/* Faster than schur64(), but much less accurate. */ +/* uses SMLAWB(), requiring armv5E and higher. */ +opus_int32 silk_schur( /* O Returns residual energy */ + opus_int16 *rc_Q15, /* O reflection coefficients [order] Q15 */ + const opus_int32 *c, /* I correlations [order+1] */ + const opus_int32 order /* I prediction order */ +); + +/* Calculates the reflection coefficients from the correlation sequence */ +/* Slower than schur(), but more accurate. */ +/* Uses SMULL(), available on armv4 */ +opus_int32 silk_schur64( /* O returns residual energy */ + opus_int32 rc_Q16[], /* O Reflection coefficients [order] Q16 */ + const opus_int32 c[], /* I Correlations [order+1] */ + opus_int32 order /* I Prediction order */ +); + +/* Step up function, converts reflection coefficients to prediction coefficients */ +void silk_k2a( + opus_int32 *A_Q24, /* O Prediction coefficients [order] Q24 */ + const opus_int16 *rc_Q15, /* I Reflection coefficients [order] Q15 */ + const opus_int32 order /* I Prediction order */ +); + +/* Step up function, converts reflection coefficients to prediction coefficients */ +void silk_k2a_Q16( + opus_int32 *A_Q24, /* O Prediction coefficients [order] Q24 */ + const opus_int32 *rc_Q16, /* I Reflection coefficients [order] Q16 */ + const opus_int32 order /* I Prediction order */ +); + +/* Apply sine window to signal vector. */ +/* Window types: */ +/* 1 -> sine window from 0 to pi/2 */ +/* 2 -> sine window from pi/2 to pi */ +/* every other sample of window is linearly interpolated, for speed */ +void silk_apply_sine_window( + opus_int16 px_win[], /* O Pointer to windowed signal */ + const opus_int16 px[], /* I Pointer to input signal */ + const opus_int win_type, /* I Selects a window type */ + const opus_int length /* I Window length, multiple of 4 */ +); + +/* Compute autocorrelation */ +void silk_autocorr( + opus_int32 *results, /* O Result (length correlationCount) */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *inputData, /* I Input data to correlate */ + const opus_int inputDataSize, /* I Length of input */ + const opus_int correlationCount, /* I Number of correlation taps to compute */ + int arch /* I Run-time architecture */ +); + +void silk_decode_pitch( + opus_int16 lagIndex, /* I */ + opus_int8 contourIndex, /* O */ + opus_int pitch_lags[], /* O 4 pitch values */ + const opus_int Fs_kHz, /* I sampling frequency (kHz) */ + const opus_int nb_subfr /* I number of sub frames */ +); + +opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 voiced, 1 unvoiced */ + const opus_int16 *frame, /* I Signal of length PE_FRAME_LENGTH_MS*Fs_kHz */ + opus_int *pitch_out, /* O 4 pitch lag values */ + opus_int16 *lagIndex, /* O Lag Index */ + opus_int8 *contourIndex, /* O Pitch contour Index */ + opus_int *LTPCorr_Q15, /* I/O Normalized correlation; input: value from previous frame */ + opus_int prevLag, /* I Last lag of previous frame; set to zero is unvoiced */ + const opus_int32 search_thres1_Q16, /* I First stage threshold for lag candidates 0 - 1 */ + const opus_int search_thres2_Q13, /* I Final threshold for lag candidates 0 - 1 */ + const opus_int Fs_kHz, /* I Sample frequency (kHz) */ + const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ + const opus_int nb_subfr, /* I number of 5 ms subframes */ + int arch /* I Run-time architecture */ +); + +/* Compute Normalized Line Spectral Frequencies (NLSFs) from whitening filter coefficients */ +/* If not all roots are found, the a_Q16 coefficients are bandwidth expanded until convergence. */ +void silk_A2NLSF( + opus_int16 *NLSF, /* O Normalized Line Spectral Frequencies in Q15 (0..2^15-1) [d] */ + opus_int32 *a_Q16, /* I/O Monic whitening filter coefficients in Q16 [d] */ + const opus_int d /* I Filter order (must be even) */ +); + +/* compute whitening filter coefficients from normalized line spectral frequencies */ +void silk_NLSF2A( + opus_int16 *a_Q12, /* O monic whitening filter coefficients in Q12, [ d ] */ + const opus_int16 *NLSF, /* I normalized line spectral frequencies in Q15, [ d ] */ + const opus_int d, /* I filter order (should be even) */ + int arch /* I Run-time architecture */ +); + +/* Convert int32 coefficients to int16 coefs and make sure there's no wrap-around */ +void silk_LPC_fit( + opus_int16 *a_QOUT, /* O Output signal */ + opus_int32 *a_QIN, /* I/O Input signal */ + const opus_int QOUT, /* I Input Q domain */ + const opus_int QIN, /* I Input Q domain */ + const opus_int d /* I Filter order */ +); + +void silk_insertion_sort_increasing( + opus_int32 *a, /* I/O Unsorted / Sorted vector */ + opus_int *idx, /* O Index vector for the sorted elements */ + const opus_int L, /* I Vector length */ + const opus_int K /* I Number of correctly sorted positions */ +); + +void silk_insertion_sort_decreasing_int16( + opus_int16 *a, /* I/O Unsorted / Sorted vector */ + opus_int *idx, /* O Index vector for the sorted elements */ + const opus_int L, /* I Vector length */ + const opus_int K /* I Number of correctly sorted positions */ +); + +void silk_insertion_sort_increasing_all_values_int16( + opus_int16 *a, /* I/O Unsorted / Sorted vector */ + const opus_int L /* I Vector length */ +); + +/* NLSF stabilizer, for a single input data vector */ +void silk_NLSF_stabilize( + opus_int16 *NLSF_Q15, /* I/O Unstable/stabilized normalized LSF vector in Q15 [L] */ + const opus_int16 *NDeltaMin_Q15, /* I Min distance vector, NDeltaMin_Q15[L] must be >= 1 [L+1] */ + const opus_int L /* I Number of NLSF parameters in the input vector */ +); + +/* Laroia low complexity NLSF weights */ +void silk_NLSF_VQ_weights_laroia( + opus_int16 *pNLSFW_Q_OUT, /* O Pointer to input vector weights [D] */ + const opus_int16 *pNLSF_Q15, /* I Pointer to input vector [D] */ + const opus_int D /* I Input vector dimension (even) */ +); + +/* Compute reflection coefficients from input signal */ +void silk_burg_modified_c( + opus_int32 *res_nrg, /* O Residual energy */ + opus_int *res_nrg_Q, /* O Residual energy Q value */ + opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ + const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ + const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ + const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I Number of subframes stacked in x */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */ +); + +/* Copy and multiply a vector by a constant */ +void silk_scale_copy_vector16( + opus_int16 *data_out, + const opus_int16 *data_in, + opus_int32 gain_Q16, /* I Gain in Q16 */ + const opus_int dataSize /* I Length */ +); + +/* Some for the LTP related function requires Q26 to work.*/ +void silk_scale_vector32_Q26_lshift_18( + opus_int32 *data1, /* I/O Q0/Q18 */ + opus_int32 gain_Q26, /* I Q26 */ + opus_int dataSize /* I length */ +); + +/********************************************************************/ +/* INLINE ARM MATH */ +/********************************************************************/ + +/* return sum( inVec1[i] * inVec2[i] ) */ + +opus_int32 silk_inner_prod_aligned( + const opus_int16 *const inVec1, /* I input vector 1 */ + const opus_int16 *const inVec2, /* I input vector 2 */ + const opus_int len, /* I vector lengths */ + int arch /* I Run-time architecture */ +); + + +opus_int32 silk_inner_prod_aligned_scale( + const opus_int16 *const inVec1, /* I input vector 1 */ + const opus_int16 *const inVec2, /* I input vector 2 */ + const opus_int scale, /* I number of bits to shift */ + const opus_int len /* I vector lengths */ +); + +opus_int64 silk_inner_prod16_c( + const opus_int16 *inVec1, /* I input vector 1 */ + const opus_int16 *inVec2, /* I input vector 2 */ + const opus_int len /* I vector lengths */ +); + +/********************************************************************/ +/* MACROS */ +/********************************************************************/ + +/* Rotate a32 right by 'rot' bits. Negative rot values result in rotating + left. Output is 32bit int. + Note: contemporary compilers recognize the C expression below and + compile it into a 'ror' instruction if available. No need for OPUS_INLINE ASM! */ +static OPUS_INLINE opus_int32 silk_ROR32( opus_int32 a32, opus_int rot ) +{ + opus_uint32 x = (opus_uint32) a32; + opus_uint32 r = (opus_uint32) rot; + opus_uint32 m = (opus_uint32) -rot; + if( rot == 0 ) { + return a32; + } else if( rot < 0 ) { + return (opus_int32) ((x << m) | (x >> (32 - m))); + } else { + return (opus_int32) ((x << (32 - r)) | (x >> r)); + } +} + +/* Allocate opus_int16 aligned to 4-byte memory address */ +#if EMBEDDED_ARM +#define silk_DWORD_ALIGN __attribute__((aligned(4))) +#else +#define silk_DWORD_ALIGN +#endif + +/* Useful Macros that can be adjusted to other platforms */ +#define silk_memcpy(dest, src, size) memcpy((dest), (src), (size)) +#define silk_memset(dest, src, size) memset((dest), (src), (size)) +#define silk_memmove(dest, src, size) memmove((dest), (src), (size)) + +/* Fixed point macros */ + +/* (a32 * b32) output have to be 32bit int */ +#define silk_MUL(a32, b32) ((a32) * (b32)) + +/* (a32 * b32) output have to be 32bit uint */ +#define silk_MUL_uint(a32, b32) silk_MUL(a32, b32) + +/* a32 + (b32 * c32) output have to be 32bit int */ +#define silk_MLA(a32, b32, c32) silk_ADD32((a32),((b32) * (c32))) + +/* a32 + (b32 * c32) output have to be 32bit uint */ +#define silk_MLA_uint(a32, b32, c32) silk_MLA(a32, b32, c32) + +/* ((a32 >> 16) * (b32 >> 16)) output have to be 32bit int */ +#define silk_SMULTT(a32, b32) (((a32) >> 16) * ((b32) >> 16)) + +/* a32 + ((a32 >> 16) * (b32 >> 16)) output have to be 32bit int */ +#define silk_SMLATT(a32, b32, c32) silk_ADD32((a32),((b32) >> 16) * ((c32) >> 16)) + +#define silk_SMLALBB(a64, b16, c16) silk_ADD64((a64),(opus_int64)((opus_int32)(b16) * (opus_int32)(c16))) + +/* (a32 * b32) */ +#define silk_SMULL(a32, b32) ((opus_int64)(a32) * /*(opus_int64)*/(b32)) + +/* Adds two signed 32-bit values in a way that can overflow, while not relying on undefined behaviour + (just standard two's complement implementation-specific behaviour) */ +#define silk_ADD32_ovflw(a, b) ((opus_int32)((opus_uint32)(a) + (opus_uint32)(b))) +/* Subtractss two signed 32-bit values in a way that can overflow, while not relying on undefined behaviour + (just standard two's complement implementation-specific behaviour) */ +#define silk_SUB32_ovflw(a, b) ((opus_int32)((opus_uint32)(a) - (opus_uint32)(b))) + +/* Multiply-accumulate macros that allow overflow in the addition (ie, no asserts in debug mode) */ +#define silk_MLA_ovflw(a32, b32, c32) silk_ADD32_ovflw((a32), (opus_uint32)(b32) * (opus_uint32)(c32)) +#define silk_SMLABB_ovflw(a32, b32, c32) (silk_ADD32_ovflw((a32) , ((opus_int32)((opus_int16)(b32))) * (opus_int32)((opus_int16)(c32)))) + +#define silk_DIV32_16(a32, b16) ((opus_int32)((a32) / (b16))) +#define silk_DIV32(a32, b32) ((opus_int32)((a32) / (b32))) + +/* These macros enables checking for overflow in silk_API_Debug.h*/ +#define silk_ADD16(a, b) ((a) + (b)) +#define silk_ADD32(a, b) ((a) + (b)) +#define silk_ADD64(a, b) ((a) + (b)) + +#define silk_SUB16(a, b) ((a) - (b)) +#define silk_SUB32(a, b) ((a) - (b)) +#define silk_SUB64(a, b) ((a) - (b)) + +#define silk_SAT8(a) ((a) > silk_int8_MAX ? silk_int8_MAX : \ + ((a) < silk_int8_MIN ? silk_int8_MIN : (a))) +#define silk_SAT16(a) ((a) > silk_int16_MAX ? silk_int16_MAX : \ + ((a) < silk_int16_MIN ? silk_int16_MIN : (a))) +#define silk_SAT32(a) ((a) > silk_int32_MAX ? silk_int32_MAX : \ + ((a) < silk_int32_MIN ? silk_int32_MIN : (a))) + +#define silk_CHECK_FIT8(a) (a) +#define silk_CHECK_FIT16(a) (a) +#define silk_CHECK_FIT32(a) (a) + +#define silk_ADD_SAT16(a, b) (opus_int16)silk_SAT16( silk_ADD32( (opus_int32)(a), (b) ) ) +#define silk_ADD_SAT64(a, b) ((((a) + (b)) & 0x8000000000000000LL) == 0 ? \ + ((((a) & (b)) & 0x8000000000000000LL) != 0 ? silk_int64_MIN : (a)+(b)) : \ + ((((a) | (b)) & 0x8000000000000000LL) == 0 ? silk_int64_MAX : (a)+(b)) ) + +#define silk_SUB_SAT16(a, b) (opus_int16)silk_SAT16( silk_SUB32( (opus_int32)(a), (b) ) ) +#define silk_SUB_SAT64(a, b) ((((a)-(b)) & 0x8000000000000000LL) == 0 ? \ + (( (a) & ((b)^0x8000000000000000LL) & 0x8000000000000000LL) ? silk_int64_MIN : (a)-(b)) : \ + ((((a)^0x8000000000000000LL) & (b) & 0x8000000000000000LL) ? silk_int64_MAX : (a)-(b)) ) + +/* Saturation for positive input values */ +#define silk_POS_SAT32(a) ((a) > silk_int32_MAX ? silk_int32_MAX : (a)) + +/* Add with saturation for positive input values */ +#define silk_ADD_POS_SAT8(a, b) ((((a)+(b)) & 0x80) ? silk_int8_MAX : ((a)+(b))) +#define silk_ADD_POS_SAT16(a, b) ((((a)+(b)) & 0x8000) ? silk_int16_MAX : ((a)+(b))) +#define silk_ADD_POS_SAT32(a, b) ((((opus_uint32)(a)+(opus_uint32)(b)) & 0x80000000) ? silk_int32_MAX : ((a)+(b))) + +#define silk_LSHIFT8(a, shift) ((opus_int8)((opus_uint8)(a)<<(shift))) /* shift >= 0, shift < 8 */ +#define silk_LSHIFT16(a, shift) ((opus_int16)((opus_uint16)(a)<<(shift))) /* shift >= 0, shift < 16 */ +#define silk_LSHIFT32(a, shift) ((opus_int32)((opus_uint32)(a)<<(shift))) /* shift >= 0, shift < 32 */ +#define silk_LSHIFT64(a, shift) ((opus_int64)((opus_uint64)(a)<<(shift))) /* shift >= 0, shift < 64 */ +#define silk_LSHIFT(a, shift) silk_LSHIFT32(a, shift) /* shift >= 0, shift < 32 */ + +#define silk_RSHIFT8(a, shift) ((a)>>(shift)) /* shift >= 0, shift < 8 */ +#define silk_RSHIFT16(a, shift) ((a)>>(shift)) /* shift >= 0, shift < 16 */ +#define silk_RSHIFT32(a, shift) ((a)>>(shift)) /* shift >= 0, shift < 32 */ +#define silk_RSHIFT64(a, shift) ((a)>>(shift)) /* shift >= 0, shift < 64 */ +#define silk_RSHIFT(a, shift) silk_RSHIFT32(a, shift) /* shift >= 0, shift < 32 */ + +/* saturates before shifting */ +#define silk_LSHIFT_SAT32(a, shift) (silk_LSHIFT32( silk_LIMIT( (a), silk_RSHIFT32( silk_int32_MIN, (shift) ), \ + silk_RSHIFT32( silk_int32_MAX, (shift) ) ), (shift) )) + +#define silk_LSHIFT_ovflw(a, shift) ((opus_int32)((opus_uint32)(a) << (shift))) /* shift >= 0, allowed to overflow */ +#define silk_LSHIFT_uint(a, shift) ((a) << (shift)) /* shift >= 0 */ +#define silk_RSHIFT_uint(a, shift) ((a) >> (shift)) /* shift >= 0 */ + +#define silk_ADD_LSHIFT(a, b, shift) ((a) + silk_LSHIFT((b), (shift))) /* shift >= 0 */ +#define silk_ADD_LSHIFT32(a, b, shift) silk_ADD32((a), silk_LSHIFT32((b), (shift))) /* shift >= 0 */ +#define silk_ADD_LSHIFT_uint(a, b, shift) ((a) + silk_LSHIFT_uint((b), (shift))) /* shift >= 0 */ +#define silk_ADD_RSHIFT(a, b, shift) ((a) + silk_RSHIFT((b), (shift))) /* shift >= 0 */ +#define silk_ADD_RSHIFT32(a, b, shift) silk_ADD32((a), silk_RSHIFT32((b), (shift))) /* shift >= 0 */ +#define silk_ADD_RSHIFT_uint(a, b, shift) ((a) + silk_RSHIFT_uint((b), (shift))) /* shift >= 0 */ +#define silk_SUB_LSHIFT32(a, b, shift) silk_SUB32((a), silk_LSHIFT32((b), (shift))) /* shift >= 0 */ +#define silk_SUB_RSHIFT32(a, b, shift) silk_SUB32((a), silk_RSHIFT32((b), (shift))) /* shift >= 0 */ + +/* Requires that shift > 0 */ +#define silk_RSHIFT_ROUND(a, shift) ((shift) == 1 ? ((a) >> 1) + ((a) & 1) : (((a) >> ((shift) - 1)) + 1) >> 1) +#define silk_RSHIFT_ROUND64(a, shift) ((shift) == 1 ? ((a) >> 1) + ((a) & 1) : (((a) >> ((shift) - 1)) + 1) >> 1) + +/* Number of rightshift required to fit the multiplication */ +#define silk_NSHIFT_MUL_32_32(a, b) ( -(31- (32-silk_CLZ32(silk_abs(a)) + (32-silk_CLZ32(silk_abs(b))))) ) +#define silk_NSHIFT_MUL_16_16(a, b) ( -(15- (16-silk_CLZ16(silk_abs(a)) + (16-silk_CLZ16(silk_abs(b))))) ) + + +#define silk_min(a, b) (((a) < (b)) ? (a) : (b)) +#define silk_max(a, b) (((a) > (b)) ? (a) : (b)) + +/* Macro to convert floating-point constants to fixed-point */ +#define SILK_FIX_CONST( C, Q ) ((opus_int32)((C) * ((opus_int64)1 << (Q)) + 0.5)) + +/* silk_min() versions with typecast in the function call */ +static OPUS_INLINE opus_int silk_min_int(opus_int a, opus_int b) +{ + return (((a) < (b)) ? (a) : (b)); +} +static OPUS_INLINE opus_int16 silk_min_16(opus_int16 a, opus_int16 b) +{ + return (((a) < (b)) ? (a) : (b)); +} +static OPUS_INLINE opus_int32 silk_min_32(opus_int32 a, opus_int32 b) +{ + return (((a) < (b)) ? (a) : (b)); +} +static OPUS_INLINE opus_int64 silk_min_64(opus_int64 a, opus_int64 b) +{ + return (((a) < (b)) ? (a) : (b)); +} + +/* silk_min() versions with typecast in the function call */ +static OPUS_INLINE opus_int silk_max_int(opus_int a, opus_int b) +{ + return (((a) > (b)) ? (a) : (b)); +} +static OPUS_INLINE opus_int16 silk_max_16(opus_int16 a, opus_int16 b) +{ + return (((a) > (b)) ? (a) : (b)); +} +static OPUS_INLINE opus_int32 silk_max_32(opus_int32 a, opus_int32 b) +{ + return (((a) > (b)) ? (a) : (b)); +} +static OPUS_INLINE opus_int64 silk_max_64(opus_int64 a, opus_int64 b) +{ + return (((a) > (b)) ? (a) : (b)); +} + +#define silk_LIMIT( a, limit1, limit2) ((limit1) > (limit2) ? ((a) > (limit1) ? (limit1) : ((a) < (limit2) ? (limit2) : (a))) \ + : ((a) > (limit2) ? (limit2) : ((a) < (limit1) ? (limit1) : (a)))) + +#define silk_LIMIT_int silk_LIMIT +#define silk_LIMIT_16 silk_LIMIT +#define silk_LIMIT_32 silk_LIMIT + +#define silk_abs(a) (((a) > 0) ? (a) : -(a)) /* Be careful, silk_abs returns wrong when input equals to silk_intXX_MIN */ +#define silk_abs_int(a) (((a) ^ ((a) >> (8 * sizeof(a) - 1))) - ((a) >> (8 * sizeof(a) - 1))) +#define silk_abs_int32(a) (((a) ^ ((a) >> 31)) - ((a) >> 31)) +#define silk_abs_int64(a) (((a) > 0) ? (a) : -(a)) + +#define silk_sign(a) ((a) > 0 ? 1 : ( (a) < 0 ? -1 : 0 )) + +/* PSEUDO-RANDOM GENERATOR */ +/* Make sure to store the result as the seed for the next call (also in between */ +/* frames), otherwise result won't be random at all. When only using some of the */ +/* bits, take the most significant bits by right-shifting. */ +#define RAND_MULTIPLIER 196314165 +#define RAND_INCREMENT 907633515 +#define silk_RAND(seed) (silk_MLA_ovflw((RAND_INCREMENT), (seed), (RAND_MULTIPLIER))) + +/* Add some multiplication functions that can be easily mapped to ARM. */ + +/* silk_SMMUL: Signed top word multiply. + ARMv6 2 instruction cycles. + ARMv3M+ 3 instruction cycles. use SMULL and ignore LSB registers.(except xM)*/ +/*#define silk_SMMUL(a32, b32) (opus_int32)silk_RSHIFT(silk_SMLAL(silk_SMULWB((a32), (b32)), (a32), silk_RSHIFT_ROUND((b32), 16)), 16)*/ +/* the following seems faster on x86 */ +#define silk_SMMUL(a32, b32) (opus_int32)silk_RSHIFT64(silk_SMULL((a32), (b32)), 32) + +#if !defined(OVERRIDE_silk_burg_modified) +#define silk_burg_modified(res_nrg, res_nrg_Q, A_Q16, x, minInvGain_Q30, subfr_length, nb_subfr, D, arch) \ + ((void)(arch), silk_burg_modified_c(res_nrg, res_nrg_Q, A_Q16, x, minInvGain_Q30, subfr_length, nb_subfr, D, arch)) +#endif + +#if !defined(OVERRIDE_silk_inner_prod16) +#define silk_inner_prod16(inVec1, inVec2, len, arch) \ + ((void)(arch),silk_inner_prod16_c(inVec1, inVec2, len)) +#endif + +#include "Inlines.h" +#include "MacroCount.h" +#include "MacroDebug.h" + +#ifdef OPUS_ARM_INLINE_ASM +#include "arm/SigProc_FIX_armv4.h" +#endif + +#ifdef OPUS_ARM_INLINE_EDSP +#include "arm/SigProc_FIX_armv5e.h" +#endif + +#if defined(MIPSr1_ASM) +#include "mips/sigproc_fix_mipsr1.h" +#endif + + +#ifdef __cplusplus +} +#endif + +#endif /* SILK_SIGPROC_FIX_H */ diff --git a/lib/opus/silk/VAD.c b/lib/opus/silk/VAD.c new file mode 100644 index 00000000..d0cda521 --- /dev/null +++ b/lib/opus/silk/VAD.c @@ -0,0 +1,360 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" + +/* Silk VAD noise level estimation */ +# if !defined(OPUS_X86_MAY_HAVE_SSE4_1) +static OPUS_INLINE void silk_VAD_GetNoiseLevels( + const opus_int32 pX[ VAD_N_BANDS ], /* I subband energies */ + silk_VAD_state *psSilk_VAD /* I/O Pointer to Silk VAD state */ +); +#endif + +/**********************************/ +/* Initialization of the Silk VAD */ +/**********************************/ +opus_int silk_VAD_Init( /* O Return value, 0 if success */ + silk_VAD_state *psSilk_VAD /* I/O Pointer to Silk VAD state */ +) +{ + opus_int b, ret = 0; + + /* reset state memory */ + silk_memset( psSilk_VAD, 0, sizeof( silk_VAD_state ) ); + + /* init noise levels */ + /* Initialize array with approx pink noise levels (psd proportional to inverse of frequency) */ + for( b = 0; b < VAD_N_BANDS; b++ ) { + psSilk_VAD->NoiseLevelBias[ b ] = silk_max_32( silk_DIV32_16( VAD_NOISE_LEVELS_BIAS, b + 1 ), 1 ); + } + + /* Initialize state */ + for( b = 0; b < VAD_N_BANDS; b++ ) { + psSilk_VAD->NL[ b ] = silk_MUL( 100, psSilk_VAD->NoiseLevelBias[ b ] ); + psSilk_VAD->inv_NL[ b ] = silk_DIV32( silk_int32_MAX, psSilk_VAD->NL[ b ] ); + } + psSilk_VAD->counter = 15; + + /* init smoothed energy-to-noise ratio*/ + for( b = 0; b < VAD_N_BANDS; b++ ) { + psSilk_VAD->NrgRatioSmth_Q8[ b ] = 100 * 256; /* 100 * 256 --> 20 dB SNR */ + } + + return( ret ); +} + +/* Weighting factors for tilt measure */ +static const opus_int32 tiltWeights[ VAD_N_BANDS ] = { 30000, 6000, -12000, -12000 }; + +/***************************************/ +/* Get the speech activity level in Q8 */ +/***************************************/ +opus_int silk_VAD_GetSA_Q8_c( /* O Return value, 0 if success */ + silk_encoder_state *psEncC, /* I/O Encoder state */ + const opus_int16 pIn[] /* I PCM input */ +) +{ + opus_int SA_Q15, pSNR_dB_Q7, input_tilt; + opus_int decimated_framelength1, decimated_framelength2; + opus_int decimated_framelength; + opus_int dec_subframe_length, dec_subframe_offset, SNR_Q7, i, b, s; + opus_int32 sumSquared, smooth_coef_Q16; + opus_int16 HPstateTmp; + VARDECL( opus_int16, X ); + opus_int32 Xnrg[ VAD_N_BANDS ]; + opus_int32 NrgToNoiseRatio_Q8[ VAD_N_BANDS ]; + opus_int32 speech_nrg, x_tmp; + opus_int X_offset[ VAD_N_BANDS ]; + opus_int ret = 0; + silk_VAD_state *psSilk_VAD = &psEncC->sVAD; + SAVE_STACK; + + /* Safety checks */ + silk_assert( VAD_N_BANDS == 4 ); + celt_assert( MAX_FRAME_LENGTH >= psEncC->frame_length ); + celt_assert( psEncC->frame_length <= 512 ); + celt_assert( psEncC->frame_length == 8 * silk_RSHIFT( psEncC->frame_length, 3 ) ); + + /***********************/ + /* Filter and Decimate */ + /***********************/ + decimated_framelength1 = silk_RSHIFT( psEncC->frame_length, 1 ); + decimated_framelength2 = silk_RSHIFT( psEncC->frame_length, 2 ); + decimated_framelength = silk_RSHIFT( psEncC->frame_length, 3 ); + /* Decimate into 4 bands: + 0 L 3L L 3L 5L + - -- - -- -- + 8 8 2 4 4 + + [0-1 kHz| temp. |1-2 kHz| 2-4 kHz | 4-8 kHz | + + They're arranged to allow the minimal ( frame_length / 4 ) extra + scratch space during the downsampling process */ + X_offset[ 0 ] = 0; + X_offset[ 1 ] = decimated_framelength + decimated_framelength2; + X_offset[ 2 ] = X_offset[ 1 ] + decimated_framelength; + X_offset[ 3 ] = X_offset[ 2 ] + decimated_framelength2; + ALLOC( X, X_offset[ 3 ] + decimated_framelength1, opus_int16 ); + + /* 0-8 kHz to 0-4 kHz and 4-8 kHz */ + silk_ana_filt_bank_1( pIn, &psSilk_VAD->AnaState[ 0 ], + X, &X[ X_offset[ 3 ] ], psEncC->frame_length ); + + /* 0-4 kHz to 0-2 kHz and 2-4 kHz */ + silk_ana_filt_bank_1( X, &psSilk_VAD->AnaState1[ 0 ], + X, &X[ X_offset[ 2 ] ], decimated_framelength1 ); + + /* 0-2 kHz to 0-1 kHz and 1-2 kHz */ + silk_ana_filt_bank_1( X, &psSilk_VAD->AnaState2[ 0 ], + X, &X[ X_offset[ 1 ] ], decimated_framelength2 ); + + /*********************************************/ + /* HP filter on lowest band (differentiator) */ + /*********************************************/ + X[ decimated_framelength - 1 ] = silk_RSHIFT( X[ decimated_framelength - 1 ], 1 ); + HPstateTmp = X[ decimated_framelength - 1 ]; + for( i = decimated_framelength - 1; i > 0; i-- ) { + X[ i - 1 ] = silk_RSHIFT( X[ i - 1 ], 1 ); + X[ i ] -= X[ i - 1 ]; + } + X[ 0 ] -= psSilk_VAD->HPstate; + psSilk_VAD->HPstate = HPstateTmp; + + /*************************************/ + /* Calculate the energy in each band */ + /*************************************/ + for( b = 0; b < VAD_N_BANDS; b++ ) { + /* Find the decimated framelength in the non-uniformly divided bands */ + decimated_framelength = silk_RSHIFT( psEncC->frame_length, silk_min_int( VAD_N_BANDS - b, VAD_N_BANDS - 1 ) ); + + /* Split length into subframe lengths */ + dec_subframe_length = silk_RSHIFT( decimated_framelength, VAD_INTERNAL_SUBFRAMES_LOG2 ); + dec_subframe_offset = 0; + + /* Compute energy per sub-frame */ + /* initialize with summed energy of last subframe */ + Xnrg[ b ] = psSilk_VAD->XnrgSubfr[ b ]; + for( s = 0; s < VAD_INTERNAL_SUBFRAMES; s++ ) { + sumSquared = 0; + for( i = 0; i < dec_subframe_length; i++ ) { + /* The energy will be less than dec_subframe_length * ( silk_int16_MIN / 8 ) ^ 2. */ + /* Therefore we can accumulate with no risk of overflow (unless dec_subframe_length > 128) */ + x_tmp = silk_RSHIFT( + X[ X_offset[ b ] + i + dec_subframe_offset ], 3 ); + sumSquared = silk_SMLABB( sumSquared, x_tmp, x_tmp ); + + /* Safety check */ + silk_assert( sumSquared >= 0 ); + } + + /* Add/saturate summed energy of current subframe */ + if( s < VAD_INTERNAL_SUBFRAMES - 1 ) { + Xnrg[ b ] = silk_ADD_POS_SAT32( Xnrg[ b ], sumSquared ); + } else { + /* Look-ahead subframe */ + Xnrg[ b ] = silk_ADD_POS_SAT32( Xnrg[ b ], silk_RSHIFT( sumSquared, 1 ) ); + } + + dec_subframe_offset += dec_subframe_length; + } + psSilk_VAD->XnrgSubfr[ b ] = sumSquared; + } + + /********************/ + /* Noise estimation */ + /********************/ + silk_VAD_GetNoiseLevels( &Xnrg[ 0 ], psSilk_VAD ); + + /***********************************************/ + /* Signal-plus-noise to noise ratio estimation */ + /***********************************************/ + sumSquared = 0; + input_tilt = 0; + for( b = 0; b < VAD_N_BANDS; b++ ) { + speech_nrg = Xnrg[ b ] - psSilk_VAD->NL[ b ]; + if( speech_nrg > 0 ) { + /* Divide, with sufficient resolution */ + if( ( Xnrg[ b ] & 0xFF800000 ) == 0 ) { + NrgToNoiseRatio_Q8[ b ] = silk_DIV32( silk_LSHIFT( Xnrg[ b ], 8 ), psSilk_VAD->NL[ b ] + 1 ); + } else { + NrgToNoiseRatio_Q8[ b ] = silk_DIV32( Xnrg[ b ], silk_RSHIFT( psSilk_VAD->NL[ b ], 8 ) + 1 ); + } + + /* Convert to log domain */ + SNR_Q7 = silk_lin2log( NrgToNoiseRatio_Q8[ b ] ) - 8 * 128; + + /* Sum-of-squares */ + sumSquared = silk_SMLABB( sumSquared, SNR_Q7, SNR_Q7 ); /* Q14 */ + + /* Tilt measure */ + if( speech_nrg < ( (opus_int32)1 << 20 ) ) { + /* Scale down SNR value for small subband speech energies */ + SNR_Q7 = silk_SMULWB( silk_LSHIFT( silk_SQRT_APPROX( speech_nrg ), 6 ), SNR_Q7 ); + } + input_tilt = silk_SMLAWB( input_tilt, tiltWeights[ b ], SNR_Q7 ); + } else { + NrgToNoiseRatio_Q8[ b ] = 256; + } + } + + /* Mean-of-squares */ + sumSquared = silk_DIV32_16( sumSquared, VAD_N_BANDS ); /* Q14 */ + + /* Root-mean-square approximation, scale to dBs, and write to output pointer */ + pSNR_dB_Q7 = (opus_int16)( 3 * silk_SQRT_APPROX( sumSquared ) ); /* Q7 */ + + /*********************************/ + /* Speech Probability Estimation */ + /*********************************/ + SA_Q15 = silk_sigm_Q15( silk_SMULWB( VAD_SNR_FACTOR_Q16, pSNR_dB_Q7 ) - VAD_NEGATIVE_OFFSET_Q5 ); + + /**************************/ + /* Frequency Tilt Measure */ + /**************************/ + psEncC->input_tilt_Q15 = silk_LSHIFT( silk_sigm_Q15( input_tilt ) - 16384, 1 ); + + /**************************************************/ + /* Scale the sigmoid output based on power levels */ + /**************************************************/ + speech_nrg = 0; + for( b = 0; b < VAD_N_BANDS; b++ ) { + /* Accumulate signal-without-noise energies, higher frequency bands have more weight */ + speech_nrg += ( b + 1 ) * silk_RSHIFT( Xnrg[ b ] - psSilk_VAD->NL[ b ], 4 ); + } + + if( psEncC->frame_length == 20 * psEncC->fs_kHz ) { + speech_nrg = silk_RSHIFT32( speech_nrg, 1 ); + } + /* Power scaling */ + if( speech_nrg <= 0 ) { + SA_Q15 = silk_RSHIFT( SA_Q15, 1 ); + } else if( speech_nrg < 16384 ) { + speech_nrg = silk_LSHIFT32( speech_nrg, 16 ); + + /* square-root */ + speech_nrg = silk_SQRT_APPROX( speech_nrg ); + SA_Q15 = silk_SMULWB( 32768 + speech_nrg, SA_Q15 ); + } + + /* Copy the resulting speech activity in Q8 */ + psEncC->speech_activity_Q8 = silk_min_int( silk_RSHIFT( SA_Q15, 7 ), silk_uint8_MAX ); + + /***********************************/ + /* Energy Level and SNR estimation */ + /***********************************/ + /* Smoothing coefficient */ + smooth_coef_Q16 = silk_SMULWB( VAD_SNR_SMOOTH_COEF_Q18, silk_SMULWB( (opus_int32)SA_Q15, SA_Q15 ) ); + + if( psEncC->frame_length == 10 * psEncC->fs_kHz ) { + smooth_coef_Q16 >>= 1; + } + + for( b = 0; b < VAD_N_BANDS; b++ ) { + /* compute smoothed energy-to-noise ratio per band */ + psSilk_VAD->NrgRatioSmth_Q8[ b ] = silk_SMLAWB( psSilk_VAD->NrgRatioSmth_Q8[ b ], + NrgToNoiseRatio_Q8[ b ] - psSilk_VAD->NrgRatioSmth_Q8[ b ], smooth_coef_Q16 ); + + /* signal to noise ratio in dB per band */ + SNR_Q7 = 3 * ( silk_lin2log( psSilk_VAD->NrgRatioSmth_Q8[b] ) - 8 * 128 ); + /* quality = sigmoid( 0.25 * ( SNR_dB - 16 ) ); */ + psEncC->input_quality_bands_Q15[ b ] = silk_sigm_Q15( silk_RSHIFT( SNR_Q7 - 16 * 128, 4 ) ); + } + + RESTORE_STACK; + return( ret ); +} + +/**************************/ +/* Noise level estimation */ +/**************************/ +# if !defined(OPUS_X86_MAY_HAVE_SSE4_1) +static OPUS_INLINE +#endif +void silk_VAD_GetNoiseLevels( + const opus_int32 pX[ VAD_N_BANDS ], /* I subband energies */ + silk_VAD_state *psSilk_VAD /* I/O Pointer to Silk VAD state */ +) +{ + opus_int k; + opus_int32 nl, nrg, inv_nrg; + opus_int coef, min_coef; + + /* Initially faster smoothing */ + if( psSilk_VAD->counter < 1000 ) { /* 1000 = 20 sec */ + min_coef = silk_DIV32_16( silk_int16_MAX, silk_RSHIFT( psSilk_VAD->counter, 4 ) + 1 ); + /* Increment frame counter */ + psSilk_VAD->counter++; + } else { + min_coef = 0; + } + + for( k = 0; k < VAD_N_BANDS; k++ ) { + /* Get old noise level estimate for current band */ + nl = psSilk_VAD->NL[ k ]; + silk_assert( nl >= 0 ); + + /* Add bias */ + nrg = silk_ADD_POS_SAT32( pX[ k ], psSilk_VAD->NoiseLevelBias[ k ] ); + silk_assert( nrg > 0 ); + + /* Invert energies */ + inv_nrg = silk_DIV32( silk_int32_MAX, nrg ); + silk_assert( inv_nrg >= 0 ); + + /* Less update when subband energy is high */ + if( nrg > silk_LSHIFT( nl, 3 ) ) { + coef = VAD_NOISE_LEVEL_SMOOTH_COEF_Q16 >> 3; + } else if( nrg < nl ) { + coef = VAD_NOISE_LEVEL_SMOOTH_COEF_Q16; + } else { + coef = silk_SMULWB( silk_SMULWW( inv_nrg, nl ), VAD_NOISE_LEVEL_SMOOTH_COEF_Q16 << 1 ); + } + + /* Initially faster smoothing */ + coef = silk_max_int( coef, min_coef ); + + /* Smooth inverse energies */ + psSilk_VAD->inv_NL[ k ] = silk_SMLAWB( psSilk_VAD->inv_NL[ k ], inv_nrg - psSilk_VAD->inv_NL[ k ], coef ); + silk_assert( psSilk_VAD->inv_NL[ k ] >= 0 ); + + /* Compute noise level by inverting again */ + nl = silk_DIV32( silk_int32_MAX, psSilk_VAD->inv_NL[ k ] ); + silk_assert( nl >= 0 ); + + /* Limit noise levels (guarantee 7 bits of head room) */ + nl = silk_min( nl, 0x00FFFFFF ); + + /* Store as part of state */ + psSilk_VAD->NL[ k ] = nl; + } +} diff --git a/lib/opus/silk/VQ_WMat_EC.c b/lib/opus/silk/VQ_WMat_EC.c new file mode 100644 index 00000000..245a7e4b --- /dev/null +++ b/lib/opus/silk/VQ_WMat_EC.c @@ -0,0 +1,131 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Entropy constrained matrix-weighted VQ, hard-coded to 5-element vectors, for a single input data vector */ +void silk_VQ_WMat_EC_c( + opus_int8 *ind, /* O index of best codebook vector */ + opus_int32 *res_nrg_Q15, /* O best residual energy */ + opus_int32 *rate_dist_Q8, /* O best total bitrate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ + const opus_int32 *XX_Q17, /* I correlation matrix */ + const opus_int32 *xX_Q17, /* I correlation vector */ + const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ + const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ + const opus_int subfr_len, /* I number of samples per subframe */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ + const opus_int L /* I number of vectors in codebook */ +) +{ + opus_int k, gain_tmp_Q7; + const opus_int8 *cb_row_Q7; + opus_int32 neg_xX_Q24[ 5 ]; + opus_int32 sum1_Q15, sum2_Q24; + opus_int32 bits_res_Q8, bits_tot_Q8; + + /* Negate and convert to new Q domain */ + neg_xX_Q24[ 0 ] = -silk_LSHIFT32( xX_Q17[ 0 ], 7 ); + neg_xX_Q24[ 1 ] = -silk_LSHIFT32( xX_Q17[ 1 ], 7 ); + neg_xX_Q24[ 2 ] = -silk_LSHIFT32( xX_Q17[ 2 ], 7 ); + neg_xX_Q24[ 3 ] = -silk_LSHIFT32( xX_Q17[ 3 ], 7 ); + neg_xX_Q24[ 4 ] = -silk_LSHIFT32( xX_Q17[ 4 ], 7 ); + + /* Loop over codebook */ + *rate_dist_Q8 = silk_int32_MAX; + *res_nrg_Q15 = silk_int32_MAX; + cb_row_Q7 = cb_Q7; + /* If things go really bad, at least *ind is set to something safe. */ + *ind = 0; + for( k = 0; k < L; k++ ) { + opus_int32 penalty; + gain_tmp_Q7 = cb_gain_Q7[k]; + /* Weighted rate */ + /* Quantization error: 1 - 2 * xX * cb + cb' * XX * cb */ + sum1_Q15 = SILK_FIX_CONST( 1.001, 15 ); + + /* Penalty for too large gain */ + penalty = silk_LSHIFT32( silk_max( silk_SUB32( gain_tmp_Q7, max_gain_Q7 ), 0 ), 11 ); + + /* first row of XX_Q17 */ + sum2_Q24 = silk_MLA( neg_xX_Q24[ 0 ], XX_Q17[ 1 ], cb_row_Q7[ 1 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 2 ], cb_row_Q7[ 2 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 3 ], cb_row_Q7[ 3 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 4 ], cb_row_Q7[ 4 ] ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 0 ], cb_row_Q7[ 0 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 0 ] ); + + /* second row of XX_Q17 */ + sum2_Q24 = silk_MLA( neg_xX_Q24[ 1 ], XX_Q17[ 7 ], cb_row_Q7[ 2 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 8 ], cb_row_Q7[ 3 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 9 ], cb_row_Q7[ 4 ] ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 6 ], cb_row_Q7[ 1 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 1 ] ); + + /* third row of XX_Q17 */ + sum2_Q24 = silk_MLA( neg_xX_Q24[ 2 ], XX_Q17[ 13 ], cb_row_Q7[ 3 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 14 ], cb_row_Q7[ 4 ] ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 12 ], cb_row_Q7[ 2 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 2 ] ); + + /* fourth row of XX_Q17 */ + sum2_Q24 = silk_MLA( neg_xX_Q24[ 3 ], XX_Q17[ 19 ], cb_row_Q7[ 4 ] ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 18 ], cb_row_Q7[ 3 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 3 ] ); + + /* last row of XX_Q17 */ + sum2_Q24 = silk_LSHIFT32( neg_xX_Q24[ 4 ], 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 24 ], cb_row_Q7[ 4 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 4 ] ); + + /* find best */ + if( sum1_Q15 >= 0 ) { + /* Translate residual energy to bits using high-rate assumption (6 dB ==> 1 bit/sample) */ + bits_res_Q8 = silk_SMULBB( subfr_len, silk_lin2log( sum1_Q15 + penalty) - (15 << 7) ); + /* In the following line we reduce the codelength component by half ("-1"); seems to slightly improve quality */ + bits_tot_Q8 = silk_ADD_LSHIFT32( bits_res_Q8, cl_Q5[ k ], 3-1 ); + if( bits_tot_Q8 <= *rate_dist_Q8 ) { + *rate_dist_Q8 = bits_tot_Q8; + *res_nrg_Q15 = sum1_Q15 + penalty; + *ind = (opus_int8)k; + *gain_Q7 = gain_tmp_Q7; + } + } + + /* Go to next cbk vector */ + cb_row_Q7 += LTP_ORDER; + } +} diff --git a/lib/opus/silk/ana_filt_bank_1.c b/lib/opus/silk/ana_filt_bank_1.c new file mode 100644 index 00000000..24cfb03f --- /dev/null +++ b/lib/opus/silk/ana_filt_bank_1.c @@ -0,0 +1,74 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Coefficients for 2-band filter bank based on first-order allpass filters */ +static opus_int16 A_fb1_20 = 5394 << 1; +static opus_int16 A_fb1_21 = -24290; /* (opus_int16)(20623 << 1) */ + +/* Split signal into two decimated bands using first-order allpass filters */ +void silk_ana_filt_bank_1( + const opus_int16 *in, /* I Input signal [N] */ + opus_int32 *S, /* I/O State vector [2] */ + opus_int16 *outL, /* O Low band [N/2] */ + opus_int16 *outH, /* O High band [N/2] */ + const opus_int32 N /* I Number of input samples */ +) +{ + opus_int k, N2 = silk_RSHIFT( N, 1 ); + opus_int32 in32, X, Y, out_1, out_2; + + /* Internal variables and state are in Q10 format */ + for( k = 0; k < N2; k++ ) { + /* Convert to Q10 */ + in32 = silk_LSHIFT( (opus_int32)in[ 2 * k ], 10 ); + + /* All-pass section for even input sample */ + Y = silk_SUB32( in32, S[ 0 ] ); + X = silk_SMLAWB( Y, Y, A_fb1_21 ); + out_1 = silk_ADD32( S[ 0 ], X ); + S[ 0 ] = silk_ADD32( in32, X ); + + /* Convert to Q10 */ + in32 = silk_LSHIFT( (opus_int32)in[ 2 * k + 1 ], 10 ); + + /* All-pass section for odd input sample, and add to output of previous section */ + Y = silk_SUB32( in32, S[ 1 ] ); + X = silk_SMULWB( Y, A_fb1_20 ); + out_2 = silk_ADD32( S[ 1 ], X ); + S[ 1 ] = silk_ADD32( in32, X ); + + /* Add/subtract, convert back to int16 and store to output */ + outL[ k ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_ADD32( out_2, out_1 ), 11 ) ); + outH[ k ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_SUB32( out_2, out_1 ), 11 ) ); + } +} diff --git a/lib/opus/silk/arm/LPC_inv_pred_gain_arm.h b/lib/opus/silk/arm/LPC_inv_pred_gain_arm.h new file mode 100644 index 00000000..9895b555 --- /dev/null +++ b/lib/opus/silk/arm/LPC_inv_pred_gain_arm.h @@ -0,0 +1,57 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_LPC_INV_PRED_GAIN_ARM_H +# define SILK_LPC_INV_PRED_GAIN_ARM_H + +# include "celt/arm/armcpu.h" + +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +opus_int32 silk_LPC_inverse_pred_gain_neon( /* O Returns inverse prediction gain in energy domain, Q30 */ + const opus_int16 *A_Q12, /* I Prediction coefficients, Q12 [order] */ + const opus_int order /* I Prediction order */ +); + +# if !defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_PRESUME_NEON) +# define OVERRIDE_silk_LPC_inverse_pred_gain (1) +# define silk_LPC_inverse_pred_gain(A_Q12, order, arch) ((void)(arch), PRESUME_NEON(silk_LPC_inverse_pred_gain)(A_Q12, order)) +# endif +# endif + +# if !defined(OVERRIDE_silk_LPC_inverse_pred_gain) +/*Is run-time CPU detection enabled on this platform?*/ +# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)) +extern opus_int32 (*const SILK_LPC_INVERSE_PRED_GAIN_IMPL[OPUS_ARCHMASK+1])(const opus_int16 *A_Q12, const opus_int order); +# define OVERRIDE_silk_LPC_inverse_pred_gain (1) +# define silk_LPC_inverse_pred_gain(A_Q12, order, arch) ((*SILK_LPC_INVERSE_PRED_GAIN_IMPL[(arch)&OPUS_ARCHMASK])(A_Q12, order)) +# elif defined(OPUS_ARM_PRESUME_NEON_INTR) +# define OVERRIDE_silk_LPC_inverse_pred_gain (1) +# define silk_LPC_inverse_pred_gain(A_Q12, order, arch) ((void)(arch), silk_LPC_inverse_pred_gain_neon(A_Q12, order)) +# endif +# endif + +#endif /* end SILK_LPC_INV_PRED_GAIN_ARM_H */ diff --git a/lib/opus/silk/arm/LPC_inv_pred_gain_neon_intr.c b/lib/opus/silk/arm/LPC_inv_pred_gain_neon_intr.c new file mode 100644 index 00000000..726e6667 --- /dev/null +++ b/lib/opus/silk/arm/LPC_inv_pred_gain_neon_intr.c @@ -0,0 +1,288 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "SigProc_FIX.h" +#include "define.h" + +#define QA 24 +#define A_LIMIT SILK_FIX_CONST( 0.99975, QA ) + +#define MUL32_FRAC_Q(a32, b32, Q) ((opus_int32)(silk_RSHIFT_ROUND64(silk_SMULL(a32, b32), Q))) + +/* The difficulty is how to judge a 64-bit signed integer tmp64 is 32-bit overflowed, + * since NEON has no 64-bit min, max or comparison instructions. + * A failed idea is to compare the results of vmovn(tmp64) and vqmovn(tmp64) whether they are equal or not. + * However, this idea fails when the tmp64 is something like 0xFFFFFFF980000000. + * Here we know that mult2Q >= 1, so the highest bit (bit 63, sign bit) of tmp64 must equal to bit 62. + * tmp64 was shifted left by 1 and we got tmp64'. If high_half(tmp64') != 0 and high_half(tmp64') != -1, + * then we know that bit 31 to bit 63 of tmp64 can not all be the sign bit, and therefore tmp64 is 32-bit overflowed. + * That is, we judge if tmp64' > 0x00000000FFFFFFFF, or tmp64' <= 0xFFFFFFFF00000000. + * We use narrowing shift right 31 bits to tmp32' to save data bandwidth and instructions. + * That is, we judge if tmp32' > 0x00000000, or tmp32' <= 0xFFFFFFFF. + */ + +/* Compute inverse of LPC prediction gain, and */ +/* test if LPC coefficients are stable (all poles within unit circle) */ +static OPUS_INLINE opus_int32 LPC_inverse_pred_gain_QA_neon( /* O Returns inverse prediction gain in energy domain, Q30 */ + opus_int32 A_QA[ SILK_MAX_ORDER_LPC ], /* I Prediction coefficients */ + const opus_int order /* I Prediction order */ +) +{ + opus_int k, n, mult2Q; + opus_int32 invGain_Q30, rc_Q31, rc_mult1_Q30, rc_mult2, tmp1, tmp2; + opus_int32 max, min; + int32x4_t max_s32x4, min_s32x4; + int32x2_t max_s32x2, min_s32x2; + + max_s32x4 = vdupq_n_s32( silk_int32_MIN ); + min_s32x4 = vdupq_n_s32( silk_int32_MAX ); + invGain_Q30 = SILK_FIX_CONST( 1, 30 ); + for( k = order - 1; k > 0; k-- ) { + int32x2_t rc_Q31_s32x2, rc_mult2_s32x2; + int64x2_t mult2Q_s64x2; + + /* Check for stability */ + if( ( A_QA[ k ] > A_LIMIT ) || ( A_QA[ k ] < -A_LIMIT ) ) { + return 0; + } + + /* Set RC equal to negated AR coef */ + rc_Q31 = -silk_LSHIFT( A_QA[ k ], 31 - QA ); + + /* rc_mult1_Q30 range: [ 1 : 2^30 ] */ + rc_mult1_Q30 = silk_SUB32( SILK_FIX_CONST( 1, 30 ), silk_SMMUL( rc_Q31, rc_Q31 ) ); + silk_assert( rc_mult1_Q30 > ( 1 << 15 ) ); /* reduce A_LIMIT if fails */ + silk_assert( rc_mult1_Q30 <= ( 1 << 30 ) ); + + /* Update inverse gain */ + /* invGain_Q30 range: [ 0 : 2^30 ] */ + invGain_Q30 = silk_LSHIFT( silk_SMMUL( invGain_Q30, rc_mult1_Q30 ), 2 ); + silk_assert( invGain_Q30 >= 0 ); + silk_assert( invGain_Q30 <= ( 1 << 30 ) ); + if( invGain_Q30 < SILK_FIX_CONST( 1.0f / MAX_PREDICTION_POWER_GAIN, 30 ) ) { + return 0; + } + + /* rc_mult2 range: [ 2^30 : silk_int32_MAX ] */ + mult2Q = 32 - silk_CLZ32( silk_abs( rc_mult1_Q30 ) ); + rc_mult2 = silk_INVERSE32_varQ( rc_mult1_Q30, mult2Q + 30 ); + + /* Update AR coefficient */ + rc_Q31_s32x2 = vdup_n_s32( rc_Q31 ); + mult2Q_s64x2 = vdupq_n_s64( -mult2Q ); + rc_mult2_s32x2 = vdup_n_s32( rc_mult2 ); + + for( n = 0; n < ( ( k + 1 ) >> 1 ) - 3; n += 4 ) { + /* We always calculate extra elements of A_QA buffer when ( k % 4 ) != 0, to take the advantage of SIMD parallelization. */ + int32x4_t tmp1_s32x4, tmp2_s32x4, t0_s32x4, t1_s32x4, s0_s32x4, s1_s32x4, t_QA0_s32x4, t_QA1_s32x4; + int64x2_t t0_s64x2, t1_s64x2, t2_s64x2, t3_s64x2; + tmp1_s32x4 = vld1q_s32( A_QA + n ); + tmp2_s32x4 = vld1q_s32( A_QA + k - n - 4 ); + tmp2_s32x4 = vrev64q_s32( tmp2_s32x4 ); + tmp2_s32x4 = vcombine_s32( vget_high_s32( tmp2_s32x4 ), vget_low_s32( tmp2_s32x4 ) ); + t0_s32x4 = vqrdmulhq_lane_s32( tmp2_s32x4, rc_Q31_s32x2, 0 ); + t1_s32x4 = vqrdmulhq_lane_s32( tmp1_s32x4, rc_Q31_s32x2, 0 ); + t_QA0_s32x4 = vqsubq_s32( tmp1_s32x4, t0_s32x4 ); + t_QA1_s32x4 = vqsubq_s32( tmp2_s32x4, t1_s32x4 ); + t0_s64x2 = vmull_s32( vget_low_s32 ( t_QA0_s32x4 ), rc_mult2_s32x2 ); + t1_s64x2 = vmull_s32( vget_high_s32( t_QA0_s32x4 ), rc_mult2_s32x2 ); + t2_s64x2 = vmull_s32( vget_low_s32 ( t_QA1_s32x4 ), rc_mult2_s32x2 ); + t3_s64x2 = vmull_s32( vget_high_s32( t_QA1_s32x4 ), rc_mult2_s32x2 ); + t0_s64x2 = vrshlq_s64( t0_s64x2, mult2Q_s64x2 ); + t1_s64x2 = vrshlq_s64( t1_s64x2, mult2Q_s64x2 ); + t2_s64x2 = vrshlq_s64( t2_s64x2, mult2Q_s64x2 ); + t3_s64x2 = vrshlq_s64( t3_s64x2, mult2Q_s64x2 ); + t0_s32x4 = vcombine_s32( vmovn_s64( t0_s64x2 ), vmovn_s64( t1_s64x2 ) ); + t1_s32x4 = vcombine_s32( vmovn_s64( t2_s64x2 ), vmovn_s64( t3_s64x2 ) ); + s0_s32x4 = vcombine_s32( vshrn_n_s64( t0_s64x2, 31 ), vshrn_n_s64( t1_s64x2, 31 ) ); + s1_s32x4 = vcombine_s32( vshrn_n_s64( t2_s64x2, 31 ), vshrn_n_s64( t3_s64x2, 31 ) ); + max_s32x4 = vmaxq_s32( max_s32x4, s0_s32x4 ); + min_s32x4 = vminq_s32( min_s32x4, s0_s32x4 ); + max_s32x4 = vmaxq_s32( max_s32x4, s1_s32x4 ); + min_s32x4 = vminq_s32( min_s32x4, s1_s32x4 ); + t1_s32x4 = vrev64q_s32( t1_s32x4 ); + t1_s32x4 = vcombine_s32( vget_high_s32( t1_s32x4 ), vget_low_s32( t1_s32x4 ) ); + vst1q_s32( A_QA + n, t0_s32x4 ); + vst1q_s32( A_QA + k - n - 4, t1_s32x4 ); + } + for( ; n < (k + 1) >> 1; n++ ) { + opus_int64 tmp64; + tmp1 = A_QA[ n ]; + tmp2 = A_QA[ k - n - 1 ]; + tmp64 = silk_RSHIFT_ROUND64( silk_SMULL( silk_SUB_SAT32(tmp1, + MUL32_FRAC_Q( tmp2, rc_Q31, 31 ) ), rc_mult2 ), mult2Q); + if( tmp64 > silk_int32_MAX || tmp64 < silk_int32_MIN ) { + return 0; + } + A_QA[ n ] = ( opus_int32 )tmp64; + tmp64 = silk_RSHIFT_ROUND64( silk_SMULL( silk_SUB_SAT32(tmp2, + MUL32_FRAC_Q( tmp1, rc_Q31, 31 ) ), rc_mult2), mult2Q); + if( tmp64 > silk_int32_MAX || tmp64 < silk_int32_MIN ) { + return 0; + } + A_QA[ k - n - 1 ] = ( opus_int32 )tmp64; + } + } + + /* Check for stability */ + if( ( A_QA[ k ] > A_LIMIT ) || ( A_QA[ k ] < -A_LIMIT ) ) { + return 0; + } + + max_s32x2 = vmax_s32( vget_low_s32( max_s32x4 ), vget_high_s32( max_s32x4 ) ); + min_s32x2 = vmin_s32( vget_low_s32( min_s32x4 ), vget_high_s32( min_s32x4 ) ); + max_s32x2 = vmax_s32( max_s32x2, vreinterpret_s32_s64( vshr_n_s64( vreinterpret_s64_s32( max_s32x2 ), 32 ) ) ); + min_s32x2 = vmin_s32( min_s32x2, vreinterpret_s32_s64( vshr_n_s64( vreinterpret_s64_s32( min_s32x2 ), 32 ) ) ); + max = vget_lane_s32( max_s32x2, 0 ); + min = vget_lane_s32( min_s32x2, 0 ); + if( ( max > 0 ) || ( min < -1 ) ) { + return 0; + } + + /* Set RC equal to negated AR coef */ + rc_Q31 = -silk_LSHIFT( A_QA[ 0 ], 31 - QA ); + + /* Range: [ 1 : 2^30 ] */ + rc_mult1_Q30 = silk_SUB32( SILK_FIX_CONST( 1, 30 ), silk_SMMUL( rc_Q31, rc_Q31 ) ); + + /* Update inverse gain */ + /* Range: [ 0 : 2^30 ] */ + invGain_Q30 = silk_LSHIFT( silk_SMMUL( invGain_Q30, rc_mult1_Q30 ), 2 ); + silk_assert( invGain_Q30 >= 0 ); + silk_assert( invGain_Q30 <= ( 1 << 30 ) ); + if( invGain_Q30 < SILK_FIX_CONST( 1.0f / MAX_PREDICTION_POWER_GAIN, 30 ) ) { + return 0; + } + + return invGain_Q30; +} + +/* For input in Q12 domain */ +opus_int32 silk_LPC_inverse_pred_gain_neon( /* O Returns inverse prediction gain in energy domain, Q30 */ + const opus_int16 *A_Q12, /* I Prediction coefficients, Q12 [order] */ + const opus_int order /* I Prediction order */ +) +{ +#ifdef OPUS_CHECK_ASM + const opus_int32 invGain_Q30_c = silk_LPC_inverse_pred_gain_c( A_Q12, order ); +#endif + + opus_int32 invGain_Q30; + if( ( SILK_MAX_ORDER_LPC != 24 ) || ( order & 1 )) { + invGain_Q30 = silk_LPC_inverse_pred_gain_c( A_Q12, order ); + } + else { + opus_int32 Atmp_QA[ SILK_MAX_ORDER_LPC ]; + opus_int32 DC_resp; + int16x8_t t0_s16x8, t1_s16x8, t2_s16x8; + int32x4_t t0_s32x4; + const opus_int leftover = order & 7; + + /* Increase Q domain of the AR coefficients */ + t0_s16x8 = vld1q_s16( A_Q12 + 0 ); + t1_s16x8 = vld1q_s16( A_Q12 + 8 ); + if ( order > 16 ) { + t2_s16x8 = vld1q_s16( A_Q12 + 16 ); + } + t0_s32x4 = vpaddlq_s16( t0_s16x8 ); + + switch( order - leftover ) + { + case 24: + t0_s32x4 = vpadalq_s16( t0_s32x4, t2_s16x8 ); + vst1q_s32( Atmp_QA + 16, vshll_n_s16( vget_low_s16 ( t2_s16x8 ), QA - 12 ) ); + vst1q_s32( Atmp_QA + 20, vshll_n_s16( vget_high_s16( t2_s16x8 ), QA - 12 ) ); + /* FALLTHROUGH */ + + case 16: + t0_s32x4 = vpadalq_s16( t0_s32x4, t1_s16x8 ); + vst1q_s32( Atmp_QA + 8, vshll_n_s16( vget_low_s16 ( t1_s16x8 ), QA - 12 ) ); + vst1q_s32( Atmp_QA + 12, vshll_n_s16( vget_high_s16( t1_s16x8 ), QA - 12 ) ); + /* FALLTHROUGH */ + + case 8: + { + const int32x2_t t_s32x2 = vpadd_s32( vget_low_s32( t0_s32x4 ), vget_high_s32( t0_s32x4 ) ); + const int64x1_t t_s64x1 = vpaddl_s32( t_s32x2 ); + DC_resp = vget_lane_s32( vreinterpret_s32_s64( t_s64x1 ), 0 ); + vst1q_s32( Atmp_QA + 0, vshll_n_s16( vget_low_s16 ( t0_s16x8 ), QA - 12 ) ); + vst1q_s32( Atmp_QA + 4, vshll_n_s16( vget_high_s16( t0_s16x8 ), QA - 12 ) ); + } + break; + + default: + DC_resp = 0; + break; + } + A_Q12 += order - leftover; + + switch( leftover ) + { + case 6: + DC_resp += (opus_int32)A_Q12[ 5 ]; + DC_resp += (opus_int32)A_Q12[ 4 ]; + Atmp_QA[ order - leftover + 5 ] = silk_LSHIFT32( (opus_int32)A_Q12[ 5 ], QA - 12 ); + Atmp_QA[ order - leftover + 4 ] = silk_LSHIFT32( (opus_int32)A_Q12[ 4 ], QA - 12 ); + /* FALLTHROUGH */ + + case 4: + DC_resp += (opus_int32)A_Q12[ 3 ]; + DC_resp += (opus_int32)A_Q12[ 2 ]; + Atmp_QA[ order - leftover + 3 ] = silk_LSHIFT32( (opus_int32)A_Q12[ 3 ], QA - 12 ); + Atmp_QA[ order - leftover + 2 ] = silk_LSHIFT32( (opus_int32)A_Q12[ 2 ], QA - 12 ); + /* FALLTHROUGH */ + + case 2: + DC_resp += (opus_int32)A_Q12[ 1 ]; + DC_resp += (opus_int32)A_Q12[ 0 ]; + Atmp_QA[ order - leftover + 1 ] = silk_LSHIFT32( (opus_int32)A_Q12[ 1 ], QA - 12 ); + Atmp_QA[ order - leftover + 0 ] = silk_LSHIFT32( (opus_int32)A_Q12[ 0 ], QA - 12 ); + /* FALLTHROUGH */ + + default: + break; + } + + /* If the DC is unstable, we don't even need to do the full calculations */ + if( DC_resp >= 4096 ) { + invGain_Q30 = 0; + } else { + invGain_Q30 = LPC_inverse_pred_gain_QA_neon( Atmp_QA, order ); + } + } + +#ifdef OPUS_CHECK_ASM + silk_assert( invGain_Q30_c == invGain_Q30 ); +#endif + + return invGain_Q30; +} diff --git a/lib/opus/silk/arm/NSQ_del_dec_arm.h b/lib/opus/silk/arm/NSQ_del_dec_arm.h new file mode 100644 index 00000000..9e76e169 --- /dev/null +++ b/lib/opus/silk/arm/NSQ_del_dec_arm.h @@ -0,0 +1,100 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_NSQ_DEL_DEC_ARM_H +#define SILK_NSQ_DEL_DEC_ARM_H + +#include "celt/arm/armcpu.h" + +#if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +void silk_NSQ_del_dec_neon( + const silk_encoder_state *psEncC, silk_nsq_state *NSQ, + SideInfoIndices *psIndices, const opus_int16 x16[], opus_int8 pulses[], + const opus_int16 PredCoef_Q12[2 * MAX_LPC_ORDER], + const opus_int16 LTPCoef_Q14[LTP_ORDER * MAX_NB_SUBFR], + const opus_int16 AR_Q13[MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER], + const opus_int HarmShapeGain_Q14[MAX_NB_SUBFR], + const opus_int Tilt_Q14[MAX_NB_SUBFR], + const opus_int32 LF_shp_Q14[MAX_NB_SUBFR], + const opus_int32 Gains_Q16[MAX_NB_SUBFR], + const opus_int pitchL[MAX_NB_SUBFR], const opus_int Lambda_Q10, + const opus_int LTP_scale_Q14); + +#if !defined(OPUS_HAVE_RTCD) +#define OVERRIDE_silk_NSQ_del_dec (1) +#define silk_NSQ_del_dec(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, \ + LTPCoef_Q14, AR_Q13, HarmShapeGain_Q14, Tilt_Q14, \ + LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, \ + LTP_scale_Q14, arch) \ + ((void)(arch), \ + PRESUME_NEON(silk_NSQ_del_dec)( \ + psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, \ + AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, \ + Lambda_Q10, LTP_scale_Q14)) +#endif +#endif + +#if !defined(OVERRIDE_silk_NSQ_del_dec) +/*Is run-time CPU detection enabled on this platform?*/ +#if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && \ + !defined(OPUS_ARM_PRESUME_NEON_INTR)) +extern void (*const SILK_NSQ_DEL_DEC_IMPL[OPUS_ARCHMASK + 1])( + const silk_encoder_state *psEncC, silk_nsq_state *NSQ, + SideInfoIndices *psIndices, const opus_int16 x16[], opus_int8 pulses[], + const opus_int16 PredCoef_Q12[2 * MAX_LPC_ORDER], + const opus_int16 LTPCoef_Q14[LTP_ORDER * MAX_NB_SUBFR], + const opus_int16 AR_Q13[MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER], + const opus_int HarmShapeGain_Q14[MAX_NB_SUBFR], + const opus_int Tilt_Q14[MAX_NB_SUBFR], + const opus_int32 LF_shp_Q14[MAX_NB_SUBFR], + const opus_int32 Gains_Q16[MAX_NB_SUBFR], + const opus_int pitchL[MAX_NB_SUBFR], const opus_int Lambda_Q10, + const opus_int LTP_scale_Q14); +#define OVERRIDE_silk_NSQ_del_dec (1) +#define silk_NSQ_del_dec(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, \ + LTPCoef_Q14, AR_Q13, HarmShapeGain_Q14, Tilt_Q14, \ + LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, \ + LTP_scale_Q14, arch) \ + ((*SILK_NSQ_DEL_DEC_IMPL[(arch)&OPUS_ARCHMASK])( \ + psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, \ + AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, \ + Lambda_Q10, LTP_scale_Q14)) +#elif defined(OPUS_ARM_PRESUME_NEON_INTR) +#define OVERRIDE_silk_NSQ_del_dec (1) +#define silk_NSQ_del_dec(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, \ + LTPCoef_Q14, AR_Q13, HarmShapeGain_Q14, Tilt_Q14, \ + LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, \ + LTP_scale_Q14, arch) \ + ((void)(arch), \ + silk_NSQ_del_dec_neon(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, \ + LTPCoef_Q14, AR_Q13, HarmShapeGain_Q14, Tilt_Q14, \ + LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, \ + LTP_scale_Q14)) +#endif +#endif + +#endif /* end SILK_NSQ_DEL_DEC_ARM_H */ diff --git a/lib/opus/silk/arm/NSQ_del_dec_neon_intr.c b/lib/opus/silk/arm/NSQ_del_dec_neon_intr.c new file mode 100644 index 00000000..212410f3 --- /dev/null +++ b/lib/opus/silk/arm/NSQ_del_dec_neon_intr.c @@ -0,0 +1,1124 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#ifdef OPUS_CHECK_ASM +# include +#endif +#include "main.h" +#include "stack_alloc.h" + +/* NEON intrinsics optimization now can only parallelize up to 4 delay decision states. */ +/* If there are more states, C function is called, and this optimization must be expanded. */ +#define NEON_MAX_DEL_DEC_STATES 4 + +typedef struct { + opus_int32 sLPC_Q14[ MAX_SUB_FRAME_LENGTH + NSQ_LPC_BUF_LENGTH ][ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 RandState[ DECISION_DELAY ][ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 Q_Q10[ DECISION_DELAY ][ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 Xq_Q14[ DECISION_DELAY ][ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 Pred_Q15[ DECISION_DELAY ][ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 Shape_Q14[ DECISION_DELAY ][ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 sAR2_Q14[ MAX_SHAPE_LPC_ORDER ][ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 LF_AR_Q14[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 Diff_Q14[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 Seed[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 SeedInit[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 RD_Q10[ NEON_MAX_DEL_DEC_STATES ]; +} NSQ_del_decs_struct; + +typedef struct { + opus_int32 Q_Q10[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 RD_Q10[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 xq_Q14[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 LF_AR_Q14[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 Diff_Q14[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 sLTP_shp_Q14[ NEON_MAX_DEL_DEC_STATES ]; + opus_int32 LPC_exc_Q14[ NEON_MAX_DEL_DEC_STATES ]; +} NSQ_samples_struct; + +static OPUS_INLINE void silk_nsq_del_dec_scale_states_neon( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_decs_struct psDelDec[], /* I/O Delayed decision states */ + const opus_int16 x16[], /* I Input */ + opus_int32 x_sc_Q10[], /* O Input scaled with 1/Gain in Q10 */ + const opus_int16 sLTP[], /* I Re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I Subframe number */ + const opus_int LTP_scale_Q14, /* I LTP state scaling */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type, /* I Signal type */ + const opus_int decisionDelay /* I Decision delay */ +); + +/******************************************/ +/* Noise shape quantizer for one subframe */ +/******************************************/ +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec_neon( + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_decs_struct psDelDec[], /* I/O Delayed decision states */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP filter state */ + opus_int32 delayedGain_Q10[], /* I/O Gain delay buffer */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int subfr, /* I Subframe number */ + opus_int shapingLPCOrder, /* I Shaping LPC filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + opus_int warping_Q16, /* I */ + opus_int nStatesDelayedDecision, /* I Number of states in decision tree */ + opus_int *smpl_buf_idx, /* I/O Index to newest samples in buffers */ + opus_int decisionDelay /* I */ +); + +static OPUS_INLINE void copy_winner_state_kernel( + const NSQ_del_decs_struct *psDelDec, + const opus_int offset, + const opus_int last_smple_idx, + const opus_int Winner_ind, + const int32x2_t gain_lo_s32x2, + const int32x2_t gain_hi_s32x2, + const int32x4_t shift_s32x4, + int32x4_t t0_s32x4, + int32x4_t t1_s32x4, + opus_int8 *const pulses, + opus_int16 *pxq, + silk_nsq_state *NSQ +) +{ + int16x8_t t_s16x8; + int32x4_t o0_s32x4, o1_s32x4; + + t0_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 0 ][ Winner_ind ], t0_s32x4, 0 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 1 ][ Winner_ind ], t0_s32x4, 1 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 2 ][ Winner_ind ], t0_s32x4, 2 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 3 ][ Winner_ind ], t0_s32x4, 3 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 4 ][ Winner_ind ], t1_s32x4, 0 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 5 ][ Winner_ind ], t1_s32x4, 1 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 6 ][ Winner_ind ], t1_s32x4, 2 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Q_Q10[ last_smple_idx - 7 ][ Winner_ind ], t1_s32x4, 3 ); + t_s16x8 = vcombine_s16( vrshrn_n_s32( t0_s32x4, 10 ), vrshrn_n_s32( t1_s32x4, 10 ) ); + vst1_s8( &pulses[ offset ], vmovn_s16( t_s16x8 ) ); + + t0_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 0 ][ Winner_ind ], t0_s32x4, 0 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 1 ][ Winner_ind ], t0_s32x4, 1 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 2 ][ Winner_ind ], t0_s32x4, 2 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 3 ][ Winner_ind ], t0_s32x4, 3 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 4 ][ Winner_ind ], t1_s32x4, 0 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 5 ][ Winner_ind ], t1_s32x4, 1 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 6 ][ Winner_ind ], t1_s32x4, 2 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Xq_Q14[ last_smple_idx - 7 ][ Winner_ind ], t1_s32x4, 3 ); + o0_s32x4 = vqdmulhq_lane_s32( t0_s32x4, gain_lo_s32x2, 0 ); + o1_s32x4 = vqdmulhq_lane_s32( t1_s32x4, gain_lo_s32x2, 0 ); + o0_s32x4 = vmlaq_lane_s32( o0_s32x4, t0_s32x4, gain_hi_s32x2, 0 ); + o1_s32x4 = vmlaq_lane_s32( o1_s32x4, t1_s32x4, gain_hi_s32x2, 0 ); + o0_s32x4 = vrshlq_s32( o0_s32x4, shift_s32x4 ); + o1_s32x4 = vrshlq_s32( o1_s32x4, shift_s32x4 ); + vst1_s16( &pxq[ offset + 0 ], vqmovn_s32( o0_s32x4 ) ); + vst1_s16( &pxq[ offset + 4 ], vqmovn_s32( o1_s32x4 ) ); + + t0_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 0 ][ Winner_ind ], t0_s32x4, 0 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 1 ][ Winner_ind ], t0_s32x4, 1 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 2 ][ Winner_ind ], t0_s32x4, 2 ); + t0_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 3 ][ Winner_ind ], t0_s32x4, 3 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 4 ][ Winner_ind ], t1_s32x4, 0 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 5 ][ Winner_ind ], t1_s32x4, 1 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 6 ][ Winner_ind ], t1_s32x4, 2 ); + t1_s32x4 = vld1q_lane_s32( &psDelDec->Shape_Q14[ last_smple_idx - 7 ][ Winner_ind ], t1_s32x4, 3 ); + vst1q_s32( &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx + offset + 0 ], t0_s32x4 ); + vst1q_s32( &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx + offset + 4 ], t1_s32x4 ); +} + +static OPUS_INLINE void copy_winner_state( + const NSQ_del_decs_struct *psDelDec, + const opus_int decisionDelay, + const opus_int smpl_buf_idx, + const opus_int Winner_ind, + const opus_int32 gain, + const opus_int32 shift, + opus_int8 *const pulses, + opus_int16 *pxq, + silk_nsq_state *NSQ +) +{ + opus_int i, last_smple_idx; + const int32x2_t gain_lo_s32x2 = vdup_n_s32( silk_LSHIFT32( gain & 0x0000FFFF, 15 ) ); + const int32x2_t gain_hi_s32x2 = vdup_n_s32( gain >> 16 ); + const int32x4_t shift_s32x4 = vdupq_n_s32( -shift ); + int32x4_t t0_s32x4, t1_s32x4; + + t0_s32x4 = t1_s32x4 = vdupq_n_s32( 0 ); /* initialization */ + last_smple_idx = smpl_buf_idx + decisionDelay - 1 + DECISION_DELAY; + if( last_smple_idx >= DECISION_DELAY ) last_smple_idx -= DECISION_DELAY; + if( last_smple_idx >= DECISION_DELAY ) last_smple_idx -= DECISION_DELAY; + + for( i = 0; ( i < ( decisionDelay - 7 ) ) && ( last_smple_idx >= 7 ); i += 8, last_smple_idx -= 8 ) { + copy_winner_state_kernel( psDelDec, i - decisionDelay, last_smple_idx, Winner_ind, gain_lo_s32x2, gain_hi_s32x2, shift_s32x4, t0_s32x4, t1_s32x4, pulses, pxq, NSQ ); + } + for( ; ( i < decisionDelay ) && ( last_smple_idx >= 0 ); i++, last_smple_idx-- ) { + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDelDec->Q_Q10[ last_smple_idx ][ Winner_ind ], 10 ); + pxq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( psDelDec->Xq_Q14[ last_smple_idx ][ Winner_ind ], gain ), shift ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay + i ] = psDelDec->Shape_Q14[ last_smple_idx ][ Winner_ind ]; + } + + last_smple_idx += DECISION_DELAY; + for( ; i < ( decisionDelay - 7 ); i++, last_smple_idx-- ) { + copy_winner_state_kernel( psDelDec, i - decisionDelay, last_smple_idx, Winner_ind, gain_lo_s32x2, gain_hi_s32x2, shift_s32x4, t0_s32x4, t1_s32x4, pulses, pxq, NSQ ); + } + for( ; i < decisionDelay; i++, last_smple_idx-- ) { + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDelDec->Q_Q10[ last_smple_idx ][ Winner_ind ], 10 ); + pxq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( psDelDec->Xq_Q14[ last_smple_idx ][ Winner_ind ], gain ), shift ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay + i ] = psDelDec->Shape_Q14[ last_smple_idx ][ Winner_ind ]; + } +} + +void silk_NSQ_del_dec_neon( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) +{ +#ifdef OPUS_CHECK_ASM + silk_nsq_state NSQ_c; + SideInfoIndices psIndices_c; + opus_int8 pulses_c[ MAX_FRAME_LENGTH ]; + const opus_int8 *const pulses_a = pulses; + + ( void )pulses_a; + silk_memcpy( &NSQ_c, NSQ, sizeof( NSQ_c ) ); + silk_memcpy( &psIndices_c, psIndices, sizeof( psIndices_c ) ); + silk_memcpy( pulses_c, pulses, sizeof( pulses_c ) ); + silk_NSQ_del_dec_c( psEncC, &NSQ_c, &psIndices_c, x16, pulses_c, PredCoef_Q12, LTPCoef_Q14, AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, + pitchL, Lambda_Q10, LTP_scale_Q14 ); +#endif + + /* The optimization parallelizes the different delay decision states. */ + if(( psEncC->nStatesDelayedDecision > NEON_MAX_DEL_DEC_STATES ) || ( psEncC->nStatesDelayedDecision <= 2 )) { + /* NEON intrinsics optimization now can only parallelize up to 4 delay decision states. */ + /* If there are more states, C function is called, and this optimization must be expanded. */ + /* When the number of delay decision states is less than 3, there are penalties using this */ + /* optimization, and C function is called. */ + /* When the number of delay decision states is 2, it's better to specialize another */ + /* structure NSQ_del_dec2_struct and optimize with shorter NEON registers. (Low priority) */ + silk_NSQ_del_dec_c( psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, HarmShapeGain_Q14, + Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14 ); + } else { + opus_int i, k, lag, start_idx, LSF_interpolation_flag, Winner_ind, subfr; + opus_int smpl_buf_idx, decisionDelay; + const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; + opus_int16 *pxq; + VARDECL( opus_int32, sLTP_Q15 ); + VARDECL( opus_int16, sLTP ); + opus_int32 HarmShapeFIRPacked_Q14; + opus_int offset_Q10; + opus_int32 RDmin_Q10, Gain_Q10; + VARDECL( opus_int32, x_sc_Q10 ); + VARDECL( opus_int32, delayedGain_Q10 ); + VARDECL( NSQ_del_decs_struct, psDelDec ); + int32x4_t t_s32x4; + SAVE_STACK; + + /* Set unvoiced lag to the previous one, overwrite later for voiced */ + lag = NSQ->lagPrev; + + silk_assert( NSQ->prev_gain_Q16 != 0 ); + + /* Initialize delayed decision states */ + ALLOC( psDelDec, 1, NSQ_del_decs_struct ); + /* Only RandState and RD_Q10 need to be initialized to 0. */ + silk_memset( psDelDec->RandState, 0, sizeof( psDelDec->RandState ) ); + vst1q_s32( psDelDec->RD_Q10, vdupq_n_s32( 0 ) ); + + for( k = 0; k < psEncC->nStatesDelayedDecision; k++ ) { + psDelDec->SeedInit[ k ] = psDelDec->Seed[ k ] = ( k + psIndices->Seed ) & 3; + } + vst1q_s32( psDelDec->LF_AR_Q14, vld1q_dup_s32( &NSQ->sLF_AR_shp_Q14 ) ); + vst1q_s32( psDelDec->Diff_Q14, vld1q_dup_s32( &NSQ->sDiff_shp_Q14 ) ); + vst1q_s32( psDelDec->Shape_Q14[ 0 ], vld1q_dup_s32( &NSQ->sLTP_shp_Q14[ psEncC->ltp_mem_length - 1 ] ) ); + for( i = 0; i < NSQ_LPC_BUF_LENGTH; i++ ) { + vst1q_s32( psDelDec->sLPC_Q14[ i ], vld1q_dup_s32( &NSQ->sLPC_Q14[ i ] ) ); + } + for( i = 0; i < (opus_int)( sizeof( NSQ->sAR2_Q14 ) / sizeof( NSQ->sAR2_Q14[ 0 ] ) ); i++ ) { + vst1q_s32( psDelDec->sAR2_Q14[ i ], vld1q_dup_s32( &NSQ->sAR2_Q14[ i ] ) ); + } + + offset_Q10 = silk_Quantization_Offsets_Q10[ psIndices->signalType >> 1 ][ psIndices->quantOffsetType ]; + smpl_buf_idx = 0; /* index of oldest samples */ + + decisionDelay = silk_min_int( DECISION_DELAY, psEncC->subfr_length ); + + /* For voiced frames limit the decision delay to lower than the pitch lag */ + if( psIndices->signalType == TYPE_VOICED ) { + opus_int pitch_min = pitchL[ 0 ]; + for( k = 1; k < psEncC->nb_subfr; k++ ) { + pitch_min = silk_min_int( pitch_min, pitchL[ k ] ); + } + decisionDelay = silk_min_int( decisionDelay, pitch_min - LTP_ORDER / 2 - 1 ); + } else { + if( lag > 0 ) { + decisionDelay = silk_min_int( decisionDelay, lag - LTP_ORDER / 2 - 1 ); + } + } + + if( psIndices->NLSFInterpCoef_Q2 == 4 ) { + LSF_interpolation_flag = 0; + } else { + LSF_interpolation_flag = 1; + } + + ALLOC( sLTP_Q15, psEncC->ltp_mem_length + psEncC->frame_length, opus_int32 ); + ALLOC( sLTP, psEncC->ltp_mem_length + psEncC->frame_length, opus_int16 ); + ALLOC( x_sc_Q10, psEncC->subfr_length, opus_int32 ); + ALLOC( delayedGain_Q10, DECISION_DELAY, opus_int32 ); + /* Set up pointers to start of sub frame */ + pxq = &NSQ->xq[ psEncC->ltp_mem_length ]; + NSQ->sLTP_shp_buf_idx = psEncC->ltp_mem_length; + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + subfr = 0; + for( k = 0; k < psEncC->nb_subfr; k++ ) { + A_Q12 = &PredCoef_Q12[ ( ( k >> 1 ) | ( 1 - LSF_interpolation_flag ) ) * MAX_LPC_ORDER ]; + B_Q14 = <PCoef_Q14[ k * LTP_ORDER ]; + AR_shp_Q13 = &AR_Q13[ k * MAX_SHAPE_LPC_ORDER ]; + + /* Noise shape parameters */ + silk_assert( HarmShapeGain_Q14[ k ] >= 0 ); + HarmShapeFIRPacked_Q14 = silk_RSHIFT( HarmShapeGain_Q14[ k ], 2 ); + HarmShapeFIRPacked_Q14 |= silk_LSHIFT( (opus_int32)silk_RSHIFT( HarmShapeGain_Q14[ k ], 1 ), 16 ); + + NSQ->rewhite_flag = 0; + if( psIndices->signalType == TYPE_VOICED ) { + /* Voiced */ + lag = pitchL[ k ]; + + /* Re-whitening */ + if( ( k & ( 3 - silk_LSHIFT( LSF_interpolation_flag, 1 ) ) ) == 0 ) { + if( k == 2 ) { + /* RESET DELAYED DECISIONS */ + /* Find winner */ + int32x4_t RD_Q10_s32x4; + RDmin_Q10 = psDelDec->RD_Q10[ 0 ]; + Winner_ind = 0; + for( i = 1; i < psEncC->nStatesDelayedDecision; i++ ) { + if( psDelDec->RD_Q10[ i ] < RDmin_Q10 ) { + RDmin_Q10 = psDelDec->RD_Q10[ i ]; + Winner_ind = i; + } + } + psDelDec->RD_Q10[ Winner_ind ] -= ( silk_int32_MAX >> 4 ); + RD_Q10_s32x4 = vld1q_s32( psDelDec->RD_Q10 ); + RD_Q10_s32x4 = vaddq_s32( RD_Q10_s32x4, vdupq_n_s32( silk_int32_MAX >> 4 ) ); + vst1q_s32( psDelDec->RD_Q10, RD_Q10_s32x4 ); + + /* Copy final part of signals from winner state to output and long-term filter states */ + copy_winner_state( psDelDec, decisionDelay, smpl_buf_idx, Winner_ind, Gains_Q16[ 1 ], 14, pulses, pxq, NSQ ); + + subfr = 0; + } + + /* Rewhiten with new A coefs */ + start_idx = psEncC->ltp_mem_length - lag - psEncC->predictLPCOrder - LTP_ORDER / 2; + silk_assert( start_idx > 0 ); + + silk_LPC_analysis_filter( &sLTP[ start_idx ], &NSQ->xq[ start_idx + k * psEncC->subfr_length ], + A_Q12, psEncC->ltp_mem_length - start_idx, psEncC->predictLPCOrder, psEncC->arch ); + + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + NSQ->rewhite_flag = 1; + } + } + + silk_nsq_del_dec_scale_states_neon( psEncC, NSQ, psDelDec, x16, x_sc_Q10, sLTP, sLTP_Q15, k, + LTP_scale_Q14, Gains_Q16, pitchL, psIndices->signalType, decisionDelay ); + + silk_noise_shape_quantizer_del_dec_neon( NSQ, psDelDec, psIndices->signalType, x_sc_Q10, pulses, pxq, sLTP_Q15, + delayedGain_Q10, A_Q12, B_Q14, AR_shp_Q13, lag, HarmShapeFIRPacked_Q14, Tilt_Q14[ k ], LF_shp_Q14[ k ], + Gains_Q16[ k ], Lambda_Q10, offset_Q10, psEncC->subfr_length, subfr++, psEncC->shapingLPCOrder, + psEncC->predictLPCOrder, psEncC->warping_Q16, psEncC->nStatesDelayedDecision, &smpl_buf_idx, decisionDelay ); + + x16 += psEncC->subfr_length; + pulses += psEncC->subfr_length; + pxq += psEncC->subfr_length; + } + + /* Find winner */ + RDmin_Q10 = psDelDec->RD_Q10[ 0 ]; + Winner_ind = 0; + for( k = 1; k < psEncC->nStatesDelayedDecision; k++ ) { + if( psDelDec->RD_Q10[ k ] < RDmin_Q10 ) { + RDmin_Q10 = psDelDec->RD_Q10[ k ]; + Winner_ind = k; + } + } + + /* Copy final part of signals from winner state to output and long-term filter states */ + psIndices->Seed = psDelDec->SeedInit[ Winner_ind ]; + Gain_Q10 = silk_RSHIFT32( Gains_Q16[ psEncC->nb_subfr - 1 ], 6 ); + copy_winner_state( psDelDec, decisionDelay, smpl_buf_idx, Winner_ind, Gain_Q10, 8, pulses, pxq, NSQ ); + + t_s32x4 = vdupq_n_s32( 0 ); /* initialization */ + for( i = 0; i < ( NSQ_LPC_BUF_LENGTH - 3 ); i += 4 ) { + t_s32x4 = vld1q_lane_s32( &psDelDec->sLPC_Q14[ i + 0 ][ Winner_ind ], t_s32x4, 0 ); + t_s32x4 = vld1q_lane_s32( &psDelDec->sLPC_Q14[ i + 1 ][ Winner_ind ], t_s32x4, 1 ); + t_s32x4 = vld1q_lane_s32( &psDelDec->sLPC_Q14[ i + 2 ][ Winner_ind ], t_s32x4, 2 ); + t_s32x4 = vld1q_lane_s32( &psDelDec->sLPC_Q14[ i + 3 ][ Winner_ind ], t_s32x4, 3 ); + vst1q_s32( &NSQ->sLPC_Q14[ i ], t_s32x4 ); + } + + for( ; i < NSQ_LPC_BUF_LENGTH; i++ ) { + NSQ->sLPC_Q14[ i ] = psDelDec->sLPC_Q14[ i ][ Winner_ind ]; + } + + for( i = 0; i < (opus_int)( sizeof( NSQ->sAR2_Q14 ) / sizeof( NSQ->sAR2_Q14[ 0 ] ) - 3 ); i += 4 ) { + t_s32x4 = vld1q_lane_s32( &psDelDec->sAR2_Q14[ i + 0 ][ Winner_ind ], t_s32x4, 0 ); + t_s32x4 = vld1q_lane_s32( &psDelDec->sAR2_Q14[ i + 1 ][ Winner_ind ], t_s32x4, 1 ); + t_s32x4 = vld1q_lane_s32( &psDelDec->sAR2_Q14[ i + 2 ][ Winner_ind ], t_s32x4, 2 ); + t_s32x4 = vld1q_lane_s32( &psDelDec->sAR2_Q14[ i + 3 ][ Winner_ind ], t_s32x4, 3 ); + vst1q_s32( &NSQ->sAR2_Q14[ i ], t_s32x4 ); + } + + for( ; i < (opus_int)( sizeof( NSQ->sAR2_Q14 ) / sizeof( NSQ->sAR2_Q14[ 0 ] ) ); i++ ) { + NSQ->sAR2_Q14[ i ] = psDelDec->sAR2_Q14[ i ][ Winner_ind ]; + } + + /* Update states */ + NSQ->sLF_AR_shp_Q14 = psDelDec->LF_AR_Q14[ Winner_ind ]; + NSQ->sDiff_shp_Q14 = psDelDec->Diff_Q14[ Winner_ind ]; + NSQ->lagPrev = pitchL[ psEncC->nb_subfr - 1 ]; + + /* Save quantized speech signal */ + silk_memmove( NSQ->xq, &NSQ->xq[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int16 ) ); + silk_memmove( NSQ->sLTP_shp_Q14, &NSQ->sLTP_shp_Q14[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int32 ) ); + RESTORE_STACK; + } + +#ifdef OPUS_CHECK_ASM + silk_assert( !memcmp( &NSQ_c, NSQ, sizeof( NSQ_c ) ) ); + silk_assert( !memcmp( &psIndices_c, psIndices, sizeof( psIndices_c ) ) ); + silk_assert( !memcmp( pulses_c, pulses_a, sizeof( pulses_c ) ) ); +#endif +} + +/******************************************/ +/* Noise shape quantizer for one subframe */ +/******************************************/ +/* Note: Function silk_short_prediction_create_arch_coef_neon() defined in NSQ_neon.h is actually a hacking C function. */ +/* Therefore here we append "_local" to the NEON function name to avoid confusion. */ +static OPUS_INLINE void silk_short_prediction_create_arch_coef_neon_local(opus_int32 *out, const opus_int16 *in, opus_int order) +{ + int16x8_t t_s16x8; + int32x4_t t0_s32x4, t1_s32x4, t2_s32x4, t3_s32x4; + silk_assert( order == 10 || order == 16 ); + + t_s16x8 = vld1q_s16( in + 0 ); /* 7 6 5 4 3 2 1 0 */ + t_s16x8 = vrev64q_s16( t_s16x8 ); /* 4 5 6 7 0 1 2 3 */ + t2_s32x4 = vshll_n_s16( vget_high_s16( t_s16x8 ), 15 ); /* 4 5 6 7 */ + t3_s32x4 = vshll_n_s16( vget_low_s16( t_s16x8 ), 15 ); /* 0 1 2 3 */ + + if( order == 16 ) { + t_s16x8 = vld1q_s16( in + 8 ); /* F E D C B A 9 8 */ + t_s16x8 = vrev64q_s16( t_s16x8 ); /* C D E F 8 9 A B */ + t0_s32x4 = vshll_n_s16( vget_high_s16( t_s16x8 ), 15 ); /* C D E F */ + t1_s32x4 = vshll_n_s16( vget_low_s16( t_s16x8 ), 15 ); /* 8 9 A B */ + } else { + int16x4_t t_s16x4; + + t0_s32x4 = vdupq_n_s32( 0 ); /* zero zero zero zero */ + t_s16x4 = vld1_s16( in + 6 ); /* 9 8 7 6 */ + t_s16x4 = vrev64_s16( t_s16x4 ); /* 6 7 8 9 */ + t1_s32x4 = vshll_n_s16( t_s16x4, 15 ); + t1_s32x4 = vcombine_s32( vget_low_s32(t0_s32x4), vget_low_s32( t1_s32x4 ) ); /* 8 9 zero zero */ + } + vst1q_s32( out + 0, t0_s32x4 ); + vst1q_s32( out + 4, t1_s32x4 ); + vst1q_s32( out + 8, t2_s32x4 ); + vst1q_s32( out + 12, t3_s32x4 ); +} + +static OPUS_INLINE int32x4_t silk_SMLAWB_lane0_neon( + const int32x4_t out_s32x4, + const int32x4_t in_s32x4, + const int32x2_t coef_s32x2 +) +{ + return vaddq_s32( out_s32x4, vqdmulhq_lane_s32( in_s32x4, coef_s32x2, 0 ) ); +} + +static OPUS_INLINE int32x4_t silk_SMLAWB_lane1_neon( + const int32x4_t out_s32x4, + const int32x4_t in_s32x4, + const int32x2_t coef_s32x2 +) +{ + return vaddq_s32( out_s32x4, vqdmulhq_lane_s32( in_s32x4, coef_s32x2, 1 ) ); +} + +/* Note: This function has different return value than silk_noise_shape_quantizer_short_prediction_neon(). */ +/* Therefore here we append "_local" to the function name to avoid confusion. */ +static OPUS_INLINE int32x4_t silk_noise_shape_quantizer_short_prediction_neon_local(const opus_int32 *buf32, const opus_int32 *a_Q12_arch, opus_int order) +{ + const int32x4_t a_Q12_arch0_s32x4 = vld1q_s32( a_Q12_arch + 0 ); + const int32x4_t a_Q12_arch1_s32x4 = vld1q_s32( a_Q12_arch + 4 ); + const int32x4_t a_Q12_arch2_s32x4 = vld1q_s32( a_Q12_arch + 8 ); + const int32x4_t a_Q12_arch3_s32x4 = vld1q_s32( a_Q12_arch + 12 ); + int32x4_t LPC_pred_Q14_s32x4; + + silk_assert( order == 10 || order == 16 ); + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LPC_pred_Q14_s32x4 = vdupq_n_s32( silk_RSHIFT( order, 1 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 0 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch0_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 1 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch0_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 2 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch0_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 3 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch0_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 4 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch1_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 5 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch1_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 6 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch1_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 7 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch1_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 8 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch2_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 9 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch2_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 10 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch2_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 11 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch2_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 12 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch3_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 13 * NEON_MAX_DEL_DEC_STATES ), vget_low_s32( a_Q12_arch3_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane0_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 14 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch3_s32x4 ) ); + LPC_pred_Q14_s32x4 = silk_SMLAWB_lane1_neon( LPC_pred_Q14_s32x4, vld1q_s32( buf32 + 15 * NEON_MAX_DEL_DEC_STATES ), vget_high_s32( a_Q12_arch3_s32x4 ) ); + + return LPC_pred_Q14_s32x4; +} + +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec_neon( + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_decs_struct psDelDec[], /* I/O Delayed decision states */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP filter state */ + opus_int32 delayedGain_Q10[], /* I/O Gain delay buffer */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int subfr, /* I Subframe number */ + opus_int shapingLPCOrder, /* I Shaping LPC filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + opus_int warping_Q16, /* I */ + opus_int nStatesDelayedDecision, /* I Number of states in decision tree */ + opus_int *smpl_buf_idx, /* I/O Index to newest samples in buffers */ + opus_int decisionDelay /* I */ +) +{ + opus_int i, j, k, Winner_ind, RDmin_ind, RDmax_ind, last_smple_idx; + opus_int32 Winner_rand_state; + opus_int32 LTP_pred_Q14, n_LTP_Q14; + opus_int32 RDmin_Q10, RDmax_Q10; + opus_int32 Gain_Q10; + opus_int32 *pred_lag_ptr, *shp_lag_ptr; + opus_int32 a_Q12_arch[MAX_LPC_ORDER]; + const int32x2_t warping_Q16_s32x2 = vdup_n_s32( silk_LSHIFT32( warping_Q16, 16 ) >> 1 ); + const opus_int32 LF_shp_Q29 = silk_LSHIFT32( LF_shp_Q14, 16 ) >> 1; + opus_int32 AR_shp_Q28[ MAX_SHAPE_LPC_ORDER ]; + const uint32x4_t rand_multiplier_u32x4 = vdupq_n_u32( RAND_MULTIPLIER ); + const uint32x4_t rand_increment_u32x4 = vdupq_n_u32( RAND_INCREMENT ); + + VARDECL( NSQ_samples_struct, psSampleState ); + SAVE_STACK; + + silk_assert( nStatesDelayedDecision > 0 ); + silk_assert( ( shapingLPCOrder & 1 ) == 0 ); /* check that order is even */ + ALLOC( psSampleState, 2, NSQ_samples_struct ); + + shp_lag_ptr = &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - lag + HARM_SHAPE_FIR_TAPS / 2 ]; + pred_lag_ptr = &sLTP_Q15[ NSQ->sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + Gain_Q10 = silk_RSHIFT( Gain_Q16, 6 ); + + for( i = 0; i < ( MAX_SHAPE_LPC_ORDER - 7 ); i += 8 ) { + const int16x8_t t_s16x8 = vld1q_s16( AR_shp_Q13 + i ); + vst1q_s32( AR_shp_Q28 + i + 0, vshll_n_s16( vget_low_s16( t_s16x8 ), 15 ) ); + vst1q_s32( AR_shp_Q28 + i + 4, vshll_n_s16( vget_high_s16( t_s16x8 ), 15 ) ); + } + + for( ; i < MAX_SHAPE_LPC_ORDER; i++ ) { + AR_shp_Q28[i] = silk_LSHIFT32( AR_shp_Q13[i], 15 ); + } + + silk_short_prediction_create_arch_coef_neon_local( a_Q12_arch, a_Q12, predictLPCOrder ); + + for( i = 0; i < length; i++ ) { + int32x4_t Seed_s32x4, LPC_pred_Q14_s32x4; + int32x4_t sign_s32x4, tmp1_s32x4, tmp2_s32x4; + int32x4_t n_AR_Q14_s32x4, n_LF_Q14_s32x4; + int32x2_t AR_shp_Q28_s32x2; + int16x4_t r_Q10_s16x4, rr_Q10_s16x4; + + /* Perform common calculations used in all states */ + + /* Long-term prediction */ + if( signalType == TYPE_VOICED ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LTP_pred_Q14 = 2; + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ 0 ], b_Q14[ 0 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -1 ], b_Q14[ 1 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -2 ], b_Q14[ 2 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -3 ], b_Q14[ 3 ] ); + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -4 ], b_Q14[ 4 ] ); + LTP_pred_Q14 = silk_LSHIFT( LTP_pred_Q14, 1 ); /* Q13 -> Q14 */ + pred_lag_ptr++; + } else { + LTP_pred_Q14 = 0; + } + + /* Long-term shaping */ + if( lag > 0 ) { + /* Symmetric, packed FIR coefficients */ + n_LTP_Q14 = silk_SMULWB( silk_ADD32( shp_lag_ptr[ 0 ], shp_lag_ptr[ -2 ] ), HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SMLAWT( n_LTP_Q14, shp_lag_ptr[ -1 ], HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SUB_LSHIFT32( LTP_pred_Q14, n_LTP_Q14, 2 ); /* Q12 -> Q14 */ + shp_lag_ptr++; + } else { + n_LTP_Q14 = 0; + } + + /* Generate dither */ + Seed_s32x4 = vld1q_s32( psDelDec->Seed ); + Seed_s32x4 = vreinterpretq_s32_u32( vmlaq_u32( rand_increment_u32x4, vreinterpretq_u32_s32( Seed_s32x4 ), rand_multiplier_u32x4 ) ); + vst1q_s32( psDelDec->Seed, Seed_s32x4 ); + + /* Short-term prediction */ + LPC_pred_Q14_s32x4 = silk_noise_shape_quantizer_short_prediction_neon_local(psDelDec->sLPC_Q14[ NSQ_LPC_BUF_LENGTH - 16 + i ], a_Q12_arch, predictLPCOrder); + LPC_pred_Q14_s32x4 = vshlq_n_s32( LPC_pred_Q14_s32x4, 4 ); /* Q10 -> Q14 */ + + /* Noise shape feedback */ + /* Output of lowpass section */ + tmp2_s32x4 = silk_SMLAWB_lane0_neon( vld1q_s32( psDelDec->Diff_Q14 ), vld1q_s32( psDelDec->sAR2_Q14[ 0 ] ), warping_Q16_s32x2 ); + /* Output of allpass section */ + tmp1_s32x4 = vsubq_s32( vld1q_s32( psDelDec->sAR2_Q14[ 1 ] ), tmp2_s32x4 ); + tmp1_s32x4 = silk_SMLAWB_lane0_neon( vld1q_s32( psDelDec->sAR2_Q14[ 0 ] ), tmp1_s32x4, warping_Q16_s32x2 ); + vst1q_s32( psDelDec->sAR2_Q14[ 0 ], tmp2_s32x4 ); + AR_shp_Q28_s32x2 = vld1_s32( AR_shp_Q28 ); + n_AR_Q14_s32x4 = vaddq_s32( vdupq_n_s32( silk_RSHIFT( shapingLPCOrder, 1 ) ), vqdmulhq_lane_s32( tmp2_s32x4, AR_shp_Q28_s32x2, 0 ) ); + + /* Loop over allpass sections */ + for( j = 2; j < shapingLPCOrder; j += 2 ) { + /* Output of allpass section */ + tmp2_s32x4 = vsubq_s32( vld1q_s32( psDelDec->sAR2_Q14[ j + 0 ] ), tmp1_s32x4 ); + tmp2_s32x4 = silk_SMLAWB_lane0_neon( vld1q_s32( psDelDec->sAR2_Q14[ j - 1 ] ), tmp2_s32x4, warping_Q16_s32x2 ); + vst1q_s32( psDelDec->sAR2_Q14[ j - 1 ], tmp1_s32x4 ); + n_AR_Q14_s32x4 = vaddq_s32( n_AR_Q14_s32x4, vqdmulhq_lane_s32( tmp1_s32x4, AR_shp_Q28_s32x2, 1 ) ); + /* Output of allpass section */ + tmp1_s32x4 = vsubq_s32( vld1q_s32( psDelDec->sAR2_Q14[ j + 1 ] ), tmp2_s32x4 ); + tmp1_s32x4 = silk_SMLAWB_lane0_neon( vld1q_s32( psDelDec->sAR2_Q14[ j + 0 ] ), tmp1_s32x4, warping_Q16_s32x2 ); + vst1q_s32( psDelDec->sAR2_Q14[ j + 0 ], tmp2_s32x4 ); + AR_shp_Q28_s32x2 = vld1_s32( &AR_shp_Q28[ j ] ); + n_AR_Q14_s32x4 = vaddq_s32( n_AR_Q14_s32x4, vqdmulhq_lane_s32( tmp2_s32x4, AR_shp_Q28_s32x2, 0 ) ); + } + vst1q_s32( psDelDec->sAR2_Q14[ shapingLPCOrder - 1 ], tmp1_s32x4 ); + n_AR_Q14_s32x4 = vaddq_s32( n_AR_Q14_s32x4, vqdmulhq_lane_s32( tmp1_s32x4, AR_shp_Q28_s32x2, 1 ) ); + n_AR_Q14_s32x4 = vshlq_n_s32( n_AR_Q14_s32x4, 1 ); /* Q11 -> Q12 */ + n_AR_Q14_s32x4 = vaddq_s32( n_AR_Q14_s32x4, vqdmulhq_n_s32( vld1q_s32( psDelDec->LF_AR_Q14 ), silk_LSHIFT32( Tilt_Q14, 16 ) >> 1 ) ); /* Q12 */ + n_AR_Q14_s32x4 = vshlq_n_s32( n_AR_Q14_s32x4, 2 ); /* Q12 -> Q14 */ + n_LF_Q14_s32x4 = vqdmulhq_n_s32( vld1q_s32( psDelDec->Shape_Q14[ *smpl_buf_idx ] ), LF_shp_Q29 ); /* Q12 */ + n_LF_Q14_s32x4 = vaddq_s32( n_LF_Q14_s32x4, vqdmulhq_n_s32( vld1q_s32( psDelDec->LF_AR_Q14 ), silk_LSHIFT32( LF_shp_Q14 >> 16 , 15 ) ) ); /* Q12 */ + n_LF_Q14_s32x4 = vshlq_n_s32( n_LF_Q14_s32x4, 2 ); /* Q12 -> Q14 */ + + /* Input minus prediction plus noise feedback */ + /* r = x[ i ] - LTP_pred - LPC_pred + n_AR + n_Tilt + n_LF + n_LTP */ + tmp1_s32x4 = vaddq_s32( n_AR_Q14_s32x4, n_LF_Q14_s32x4 ); /* Q14 */ + tmp2_s32x4 = vaddq_s32( vdupq_n_s32( n_LTP_Q14 ), LPC_pred_Q14_s32x4 ); /* Q13 */ + tmp1_s32x4 = vsubq_s32( tmp2_s32x4, tmp1_s32x4 ); /* Q13 */ + tmp1_s32x4 = vrshrq_n_s32( tmp1_s32x4, 4 ); /* Q10 */ + tmp1_s32x4 = vsubq_s32( vdupq_n_s32( x_Q10[ i ] ), tmp1_s32x4 ); /* residual error Q10 */ + + /* Flip sign depending on dither */ + sign_s32x4 = vreinterpretq_s32_u32( vcltq_s32( Seed_s32x4, vdupq_n_s32( 0 ) ) ); + tmp1_s32x4 = veorq_s32( tmp1_s32x4, sign_s32x4 ); + tmp1_s32x4 = vsubq_s32( tmp1_s32x4, sign_s32x4 ); + tmp1_s32x4 = vmaxq_s32( tmp1_s32x4, vdupq_n_s32( -( 31 << 10 ) ) ); + tmp1_s32x4 = vminq_s32( tmp1_s32x4, vdupq_n_s32( 30 << 10 ) ); + r_Q10_s16x4 = vmovn_s32( tmp1_s32x4 ); + + /* Find two quantization level candidates and measure their rate-distortion */ + { + int16x4_t q1_Q10_s16x4 = vsub_s16( r_Q10_s16x4, vdup_n_s16( offset_Q10 ) ); + int16x4_t q1_Q0_s16x4 = vshr_n_s16( q1_Q10_s16x4, 10 ); + int16x4_t q2_Q10_s16x4; + int32x4_t rd1_Q10_s32x4, rd2_Q10_s32x4; + uint32x4_t t_u32x4; + + if( Lambda_Q10 > 2048 ) { + /* For aggressive RDO, the bias becomes more than one pulse. */ + const int rdo_offset = Lambda_Q10/2 - 512; + const uint16x4_t greaterThanRdo = vcgt_s16( q1_Q10_s16x4, vdup_n_s16( rdo_offset ) ); + const uint16x4_t lessThanMinusRdo = vclt_s16( q1_Q10_s16x4, vdup_n_s16( -rdo_offset ) ); + /* If Lambda_Q10 > 32767, then q1_Q0, q1_Q10 and q2_Q10 must change to 32-bit. */ + silk_assert( Lambda_Q10 <= 32767 ); + + q1_Q0_s16x4 = vreinterpret_s16_u16( vclt_s16( q1_Q10_s16x4, vdup_n_s16( 0 ) ) ); + q1_Q0_s16x4 = vbsl_s16( greaterThanRdo, vsub_s16( q1_Q10_s16x4, vdup_n_s16( rdo_offset ) ), q1_Q0_s16x4 ); + q1_Q0_s16x4 = vbsl_s16( lessThanMinusRdo, vadd_s16( q1_Q10_s16x4, vdup_n_s16( rdo_offset ) ), q1_Q0_s16x4 ); + q1_Q0_s16x4 = vshr_n_s16( q1_Q0_s16x4, 10 ); + } + { + const uint16x4_t equal0_u16x4 = vceq_s16( q1_Q0_s16x4, vdup_n_s16( 0 ) ); + const uint16x4_t equalMinus1_u16x4 = vceq_s16( q1_Q0_s16x4, vdup_n_s16( -1 ) ); + const uint16x4_t lessThanMinus1_u16x4 = vclt_s16( q1_Q0_s16x4, vdup_n_s16( -1 ) ); + int16x4_t tmp1_s16x4, tmp2_s16x4; + + q1_Q10_s16x4 = vshl_n_s16( q1_Q0_s16x4, 10 ); + tmp1_s16x4 = vadd_s16( q1_Q10_s16x4, vdup_n_s16( offset_Q10 - QUANT_LEVEL_ADJUST_Q10 ) ); + q1_Q10_s16x4 = vadd_s16( q1_Q10_s16x4, vdup_n_s16( offset_Q10 + QUANT_LEVEL_ADJUST_Q10 ) ); + q1_Q10_s16x4 = vbsl_s16( lessThanMinus1_u16x4, q1_Q10_s16x4, tmp1_s16x4 ); + q1_Q10_s16x4 = vbsl_s16( equal0_u16x4, vdup_n_s16( offset_Q10 ), q1_Q10_s16x4 ); + q1_Q10_s16x4 = vbsl_s16( equalMinus1_u16x4, vdup_n_s16( offset_Q10 - ( 1024 - QUANT_LEVEL_ADJUST_Q10 ) ), q1_Q10_s16x4 ); + q2_Q10_s16x4 = vadd_s16( q1_Q10_s16x4, vdup_n_s16( 1024 ) ); + q2_Q10_s16x4 = vbsl_s16( equal0_u16x4, vdup_n_s16( offset_Q10 + 1024 - QUANT_LEVEL_ADJUST_Q10 ), q2_Q10_s16x4 ); + q2_Q10_s16x4 = vbsl_s16( equalMinus1_u16x4, vdup_n_s16( offset_Q10 ), q2_Q10_s16x4 ); + tmp1_s16x4 = q1_Q10_s16x4; + tmp2_s16x4 = q2_Q10_s16x4; + tmp1_s16x4 = vbsl_s16( vorr_u16( equalMinus1_u16x4, lessThanMinus1_u16x4 ), vneg_s16( tmp1_s16x4 ), tmp1_s16x4 ); + tmp2_s16x4 = vbsl_s16( lessThanMinus1_u16x4, vneg_s16( tmp2_s16x4 ), tmp2_s16x4 ); + rd1_Q10_s32x4 = vmull_s16( tmp1_s16x4, vdup_n_s16( Lambda_Q10 ) ); + rd2_Q10_s32x4 = vmull_s16( tmp2_s16x4, vdup_n_s16( Lambda_Q10 ) ); + } + + rr_Q10_s16x4 = vsub_s16( r_Q10_s16x4, q1_Q10_s16x4 ); + rd1_Q10_s32x4 = vmlal_s16( rd1_Q10_s32x4, rr_Q10_s16x4, rr_Q10_s16x4 ); + rd1_Q10_s32x4 = vshrq_n_s32( rd1_Q10_s32x4, 10 ); + + rr_Q10_s16x4 = vsub_s16( r_Q10_s16x4, q2_Q10_s16x4 ); + rd2_Q10_s32x4 = vmlal_s16( rd2_Q10_s32x4, rr_Q10_s16x4, rr_Q10_s16x4 ); + rd2_Q10_s32x4 = vshrq_n_s32( rd2_Q10_s32x4, 10 ); + + tmp2_s32x4 = vld1q_s32( psDelDec->RD_Q10 ); + tmp1_s32x4 = vaddq_s32( tmp2_s32x4, vminq_s32( rd1_Q10_s32x4, rd2_Q10_s32x4 ) ); + tmp2_s32x4 = vaddq_s32( tmp2_s32x4, vmaxq_s32( rd1_Q10_s32x4, rd2_Q10_s32x4 ) ); + vst1q_s32( psSampleState[ 0 ].RD_Q10, tmp1_s32x4 ); + vst1q_s32( psSampleState[ 1 ].RD_Q10, tmp2_s32x4 ); + t_u32x4 = vcltq_s32( rd1_Q10_s32x4, rd2_Q10_s32x4 ); + tmp1_s32x4 = vbslq_s32( t_u32x4, vmovl_s16( q1_Q10_s16x4 ), vmovl_s16( q2_Q10_s16x4 ) ); + tmp2_s32x4 = vbslq_s32( t_u32x4, vmovl_s16( q2_Q10_s16x4 ), vmovl_s16( q1_Q10_s16x4 ) ); + vst1q_s32( psSampleState[ 0 ].Q_Q10, tmp1_s32x4 ); + vst1q_s32( psSampleState[ 1 ].Q_Q10, tmp2_s32x4 ); + } + + { + /* Update states for best quantization */ + int32x4_t exc_Q14_s32x4, LPC_exc_Q14_s32x4, xq_Q14_s32x4, sLF_AR_shp_Q14_s32x4; + + /* Quantized excitation */ + exc_Q14_s32x4 = vshlq_n_s32( tmp1_s32x4, 4 ); + exc_Q14_s32x4 = veorq_s32( exc_Q14_s32x4, sign_s32x4 ); + exc_Q14_s32x4 = vsubq_s32( exc_Q14_s32x4, sign_s32x4 ); + + /* Add predictions */ + LPC_exc_Q14_s32x4 = vaddq_s32( exc_Q14_s32x4, vdupq_n_s32( LTP_pred_Q14 ) ); + xq_Q14_s32x4 = vaddq_s32( LPC_exc_Q14_s32x4, LPC_pred_Q14_s32x4 ); + + /* Update states */ + tmp1_s32x4 = vsubq_s32( xq_Q14_s32x4, vshlq_n_s32( vdupq_n_s32( x_Q10[ i ] ), 4 ) ); + vst1q_s32( psSampleState[ 0 ].Diff_Q14, tmp1_s32x4 ); + sLF_AR_shp_Q14_s32x4 = vsubq_s32( tmp1_s32x4, n_AR_Q14_s32x4 ); + vst1q_s32( psSampleState[ 0 ].sLTP_shp_Q14, vsubq_s32( sLF_AR_shp_Q14_s32x4, n_LF_Q14_s32x4 ) ); + vst1q_s32( psSampleState[ 0 ].LF_AR_Q14, sLF_AR_shp_Q14_s32x4 ); + vst1q_s32( psSampleState[ 0 ].LPC_exc_Q14, LPC_exc_Q14_s32x4 ); + vst1q_s32( psSampleState[ 0 ].xq_Q14, xq_Q14_s32x4 ); + + /* Quantized excitation */ + exc_Q14_s32x4 = vshlq_n_s32( tmp2_s32x4, 4 ); + exc_Q14_s32x4 = veorq_s32( exc_Q14_s32x4, sign_s32x4 ); + exc_Q14_s32x4 = vsubq_s32( exc_Q14_s32x4, sign_s32x4 ); + + /* Add predictions */ + LPC_exc_Q14_s32x4 = vaddq_s32( exc_Q14_s32x4, vdupq_n_s32( LTP_pred_Q14 ) ); + xq_Q14_s32x4 = vaddq_s32( LPC_exc_Q14_s32x4, LPC_pred_Q14_s32x4 ); + + /* Update states */ + tmp1_s32x4 = vsubq_s32( xq_Q14_s32x4, vshlq_n_s32( vdupq_n_s32( x_Q10[ i ] ), 4 ) ); + vst1q_s32( psSampleState[ 1 ].Diff_Q14, tmp1_s32x4 ); + sLF_AR_shp_Q14_s32x4 = vsubq_s32( tmp1_s32x4, n_AR_Q14_s32x4 ); + vst1q_s32( psSampleState[ 1 ].sLTP_shp_Q14, vsubq_s32( sLF_AR_shp_Q14_s32x4, n_LF_Q14_s32x4 ) ); + vst1q_s32( psSampleState[ 1 ].LF_AR_Q14, sLF_AR_shp_Q14_s32x4 ); + vst1q_s32( psSampleState[ 1 ].LPC_exc_Q14, LPC_exc_Q14_s32x4 ); + vst1q_s32( psSampleState[ 1 ].xq_Q14, xq_Q14_s32x4 ); + } + + *smpl_buf_idx = *smpl_buf_idx ? ( *smpl_buf_idx - 1 ) : ( DECISION_DELAY - 1); + last_smple_idx = *smpl_buf_idx + decisionDelay + DECISION_DELAY; + if( last_smple_idx >= DECISION_DELAY ) last_smple_idx -= DECISION_DELAY; + if( last_smple_idx >= DECISION_DELAY ) last_smple_idx -= DECISION_DELAY; + + /* Find winner */ + RDmin_Q10 = psSampleState[ 0 ].RD_Q10[ 0 ]; + Winner_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + if( psSampleState[ 0 ].RD_Q10[ k ] < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ 0 ].RD_Q10[ k ]; + Winner_ind = k; + } + } + + /* Increase RD values of expired states */ + { + uint32x4_t t_u32x4; + Winner_rand_state = psDelDec->RandState[ last_smple_idx ][ Winner_ind ]; + t_u32x4 = vceqq_s32( vld1q_s32( psDelDec->RandState[ last_smple_idx ] ), vdupq_n_s32( Winner_rand_state ) ); + t_u32x4 = vmvnq_u32( t_u32x4 ); + t_u32x4 = vshrq_n_u32( t_u32x4, 5 ); + tmp1_s32x4 = vld1q_s32( psSampleState[ 0 ].RD_Q10 ); + tmp2_s32x4 = vld1q_s32( psSampleState[ 1 ].RD_Q10 ); + tmp1_s32x4 = vaddq_s32( tmp1_s32x4, vreinterpretq_s32_u32( t_u32x4 ) ); + tmp2_s32x4 = vaddq_s32( tmp2_s32x4, vreinterpretq_s32_u32( t_u32x4 ) ); + vst1q_s32( psSampleState[ 0 ].RD_Q10, tmp1_s32x4 ); + vst1q_s32( psSampleState[ 1 ].RD_Q10, tmp2_s32x4 ); + + /* Find worst in first set and best in second set */ + RDmax_Q10 = psSampleState[ 0 ].RD_Q10[ 0 ]; + RDmin_Q10 = psSampleState[ 1 ].RD_Q10[ 0 ]; + RDmax_ind = 0; + RDmin_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + /* find worst in first set */ + if( psSampleState[ 0 ].RD_Q10[ k ] > RDmax_Q10 ) { + RDmax_Q10 = psSampleState[ 0 ].RD_Q10[ k ]; + RDmax_ind = k; + } + /* find best in second set */ + if( psSampleState[ 1 ].RD_Q10[ k ] < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ 1 ].RD_Q10[ k ]; + RDmin_ind = k; + } + } + } + + /* Replace a state if best from second set outperforms worst in first set */ + if( RDmin_Q10 < RDmax_Q10 ) { + opus_int32 (*ptr)[NEON_MAX_DEL_DEC_STATES] = psDelDec->RandState; + const int numOthers = (int)( ( sizeof( NSQ_del_decs_struct ) - sizeof( ( (NSQ_del_decs_struct *)0 )->sLPC_Q14 ) ) + / ( NEON_MAX_DEL_DEC_STATES * sizeof( opus_int32 ) ) ); + /* Only ( predictLPCOrder - 1 ) of sLPC_Q14 buffer need to be updated, though the first several */ + /* useless sLPC_Q14[] will be different comparing with C when predictLPCOrder < NSQ_LPC_BUF_LENGTH. */ + /* Here just update constant ( NSQ_LPC_BUF_LENGTH - 1 ) for simplicity. */ + for( j = i + 1; j < i + NSQ_LPC_BUF_LENGTH; j++ ) { + psDelDec->sLPC_Q14[ j ][ RDmax_ind ] = psDelDec->sLPC_Q14[ j ][ RDmin_ind ]; + } + for( j = 0; j < numOthers; j++ ) { + ptr[ j ][ RDmax_ind ] = ptr[ j ][ RDmin_ind ]; + } + + psSampleState[ 0 ].Q_Q10[ RDmax_ind ] = psSampleState[ 1 ].Q_Q10[ RDmin_ind ]; + psSampleState[ 0 ].RD_Q10[ RDmax_ind ] = psSampleState[ 1 ].RD_Q10[ RDmin_ind ]; + psSampleState[ 0 ].xq_Q14[ RDmax_ind ] = psSampleState[ 1 ].xq_Q14[ RDmin_ind ]; + psSampleState[ 0 ].LF_AR_Q14[ RDmax_ind ] = psSampleState[ 1 ].LF_AR_Q14[ RDmin_ind ]; + psSampleState[ 0 ].Diff_Q14[ RDmax_ind ] = psSampleState[ 1 ].Diff_Q14[ RDmin_ind ]; + psSampleState[ 0 ].sLTP_shp_Q14[ RDmax_ind ] = psSampleState[ 1 ].sLTP_shp_Q14[ RDmin_ind ]; + psSampleState[ 0 ].LPC_exc_Q14[ RDmax_ind ] = psSampleState[ 1 ].LPC_exc_Q14[ RDmin_ind ]; + } + + /* Write samples from winner to output and long-term filter states */ + if( subfr > 0 || i >= decisionDelay ) { + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDelDec->Q_Q10[ last_smple_idx ][ Winner_ind ], 10 ); + xq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDelDec->Xq_Q14[ last_smple_idx ][ Winner_ind ], delayedGain_Q10[ last_smple_idx ] ), 8 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay ] = psDelDec->Shape_Q14[ last_smple_idx ][ Winner_ind ]; + sLTP_Q15[ NSQ->sLTP_buf_idx - decisionDelay ] = psDelDec->Pred_Q15[ last_smple_idx ][ Winner_ind ]; + } + NSQ->sLTP_shp_buf_idx++; + NSQ->sLTP_buf_idx++; + + /* Update states */ + vst1q_s32( psDelDec->LF_AR_Q14, vld1q_s32( psSampleState[ 0 ].LF_AR_Q14 ) ); + vst1q_s32( psDelDec->Diff_Q14, vld1q_s32( psSampleState[ 0 ].Diff_Q14 ) ); + vst1q_s32( psDelDec->sLPC_Q14[ NSQ_LPC_BUF_LENGTH + i ], vld1q_s32( psSampleState[ 0 ].xq_Q14 ) ); + vst1q_s32( psDelDec->Xq_Q14[ *smpl_buf_idx ], vld1q_s32( psSampleState[ 0 ].xq_Q14 ) ); + tmp1_s32x4 = vld1q_s32( psSampleState[ 0 ].Q_Q10 ); + vst1q_s32( psDelDec->Q_Q10[ *smpl_buf_idx ], tmp1_s32x4 ); + vst1q_s32( psDelDec->Pred_Q15[ *smpl_buf_idx ], vshlq_n_s32( vld1q_s32( psSampleState[ 0 ].LPC_exc_Q14 ), 1 ) ); + vst1q_s32( psDelDec->Shape_Q14[ *smpl_buf_idx ], vld1q_s32( psSampleState[ 0 ].sLTP_shp_Q14 ) ); + tmp1_s32x4 = vrshrq_n_s32( tmp1_s32x4, 10 ); + tmp1_s32x4 = vaddq_s32( vld1q_s32( psDelDec->Seed ), tmp1_s32x4 ); + vst1q_s32( psDelDec->Seed, tmp1_s32x4 ); + vst1q_s32( psDelDec->RandState[ *smpl_buf_idx ], tmp1_s32x4 ); + vst1q_s32( psDelDec->RD_Q10, vld1q_s32( psSampleState[ 0 ].RD_Q10 ) ); + delayedGain_Q10[ *smpl_buf_idx ] = Gain_Q10; + } + /* Update LPC states */ + silk_memcpy( psDelDec->sLPC_Q14[ 0 ], psDelDec->sLPC_Q14[ length ], NEON_MAX_DEL_DEC_STATES * NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + + RESTORE_STACK; +} + +static OPUS_INLINE void silk_SMULWB_8_neon( + const opus_int16 *a, + const int32x2_t b, + opus_int32 *o +) +{ + const int16x8_t a_s16x8 = vld1q_s16( a ); + int32x4_t o0_s32x4, o1_s32x4; + + o0_s32x4 = vshll_n_s16( vget_low_s16( a_s16x8 ), 15 ); + o1_s32x4 = vshll_n_s16( vget_high_s16( a_s16x8 ), 15 ); + o0_s32x4 = vqdmulhq_lane_s32( o0_s32x4, b, 0 ); + o1_s32x4 = vqdmulhq_lane_s32( o1_s32x4, b, 0 ); + vst1q_s32( o, o0_s32x4 ); + vst1q_s32( o + 4, o1_s32x4 ); +} + +/* Only works when ( b >= -65536 ) && ( b < 65536 ). */ +static OPUS_INLINE void silk_SMULWW_small_b_4_neon( + opus_int32 *a, + const int32x2_t b_s32x2) +{ + int32x4_t o_s32x4; + + o_s32x4 = vld1q_s32( a ); + o_s32x4 = vqdmulhq_lane_s32( o_s32x4, b_s32x2, 0 ); + vst1q_s32( a, o_s32x4 ); +} + +/* Only works when ( b >= -65536 ) && ( b < 65536 ). */ +static OPUS_INLINE void silk_SMULWW_small_b_8_neon( + opus_int32 *a, + const int32x2_t b_s32x2 +) +{ + int32x4_t o0_s32x4, o1_s32x4; + + o0_s32x4 = vld1q_s32( a ); + o1_s32x4 = vld1q_s32( a + 4 ); + o0_s32x4 = vqdmulhq_lane_s32( o0_s32x4, b_s32x2, 0 ); + o1_s32x4 = vqdmulhq_lane_s32( o1_s32x4, b_s32x2, 0 ); + vst1q_s32( a, o0_s32x4 ); + vst1q_s32( a + 4, o1_s32x4 ); +} + +static OPUS_INLINE void silk_SMULWW_4_neon( + opus_int32 *a, + const int32x2_t b_s32x2) +{ + int32x4_t a_s32x4, o_s32x4; + + a_s32x4 = vld1q_s32( a ); + o_s32x4 = vqdmulhq_lane_s32( a_s32x4, b_s32x2, 0 ); + o_s32x4 = vmlaq_lane_s32( o_s32x4, a_s32x4, b_s32x2, 1 ); + vst1q_s32( a, o_s32x4 ); +} + +static OPUS_INLINE void silk_SMULWW_8_neon( + opus_int32 *a, + const int32x2_t b_s32x2 +) +{ + int32x4_t a0_s32x4, a1_s32x4, o0_s32x4, o1_s32x4; + + a0_s32x4 = vld1q_s32( a ); + a1_s32x4 = vld1q_s32( a + 4 ); + o0_s32x4 = vqdmulhq_lane_s32( a0_s32x4, b_s32x2, 0 ); + o1_s32x4 = vqdmulhq_lane_s32( a1_s32x4, b_s32x2, 0 ); + o0_s32x4 = vmlaq_lane_s32( o0_s32x4, a0_s32x4, b_s32x2, 1 ); + o1_s32x4 = vmlaq_lane_s32( o1_s32x4, a1_s32x4, b_s32x2, 1 ); + vst1q_s32( a, o0_s32x4 ); + vst1q_s32( a + 4, o1_s32x4 ); +} + +static OPUS_INLINE void silk_SMULWW_loop_neon( + const opus_int16 *a, + const opus_int32 b, + opus_int32 *o, + const opus_int loop_num +) +{ + opus_int i; + int32x2_t b_s32x2; + + b_s32x2 = vdup_n_s32( b ); + for( i = 0; i < loop_num - 7; i += 8 ) { + silk_SMULWB_8_neon( a + i, b_s32x2, o + i ); + } + for( ; i < loop_num; i++ ) { + o[ i ] = silk_SMULWW( a[ i ], b ); + } +} + +static OPUS_INLINE void silk_nsq_del_dec_scale_states_neon( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_decs_struct psDelDec[], /* I/O Delayed decision states */ + const opus_int16 x16[], /* I Input */ + opus_int32 x_sc_Q10[], /* O Input scaled with 1/Gain in Q10 */ + const opus_int16 sLTP[], /* I Re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I Subframe number */ + const opus_int LTP_scale_Q14, /* I LTP state scaling */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type, /* I Signal type */ + const opus_int decisionDelay /* I Decision delay */ +) +{ + opus_int i, lag; + opus_int32 gain_adj_Q16, inv_gain_Q31, inv_gain_Q26; + + lag = pitchL[ subfr ]; + inv_gain_Q31 = silk_INVERSE32_varQ( silk_max( Gains_Q16[ subfr ], 1 ), 47 ); + silk_assert( inv_gain_Q31 != 0 ); + + /* Scale input */ + inv_gain_Q26 = silk_RSHIFT_ROUND( inv_gain_Q31, 5 ); + silk_SMULWW_loop_neon( x16, inv_gain_Q26, x_sc_Q10, psEncC->subfr_length ); + + /* After rewhitening the LTP state is un-scaled, so scale with inv_gain_Q16 */ + if( NSQ->rewhite_flag ) { + if( subfr == 0 ) { + /* Do LTP downscaling */ + inv_gain_Q31 = silk_LSHIFT( silk_SMULWB( inv_gain_Q31, LTP_scale_Q14 ), 2 ); + } + silk_SMULWW_loop_neon( sLTP + NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2, inv_gain_Q31, sLTP_Q15 + NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2, lag + LTP_ORDER / 2 ); + } + + /* Adjust for changing gain */ + if( Gains_Q16[ subfr ] != NSQ->prev_gain_Q16 ) { + int32x2_t gain_adj_Q16_s32x2; + gain_adj_Q16 = silk_DIV32_varQ( NSQ->prev_gain_Q16, Gains_Q16[ subfr ], 16 ); + + /* Scale long-term shaping state */ + if( ( gain_adj_Q16 >= -65536 ) && ( gain_adj_Q16 < 65536 ) ) { + gain_adj_Q16_s32x2 = vdup_n_s32( silk_LSHIFT32( gain_adj_Q16, 15 ) ); + for( i = NSQ->sLTP_shp_buf_idx - psEncC->ltp_mem_length; i < NSQ->sLTP_shp_buf_idx - 7; i += 8 ) { + silk_SMULWW_small_b_8_neon( NSQ->sLTP_shp_Q14 + i, gain_adj_Q16_s32x2 ); + } + for( ; i < NSQ->sLTP_shp_buf_idx; i++ ) { + NSQ->sLTP_shp_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLTP_shp_Q14[ i ] ); + } + + /* Scale long-term prediction state */ + if( signal_type == TYPE_VOICED && NSQ->rewhite_flag == 0 ) { + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx - decisionDelay - 7; i += 8 ) { + silk_SMULWW_small_b_8_neon( sLTP_Q15 + i, gain_adj_Q16_s32x2 ); + } + for( ; i < NSQ->sLTP_buf_idx - decisionDelay; i++ ) { + sLTP_Q15[ i ] = silk_SMULWW( gain_adj_Q16, sLTP_Q15[ i ] ); + } + } + + /* Scale scalar states */ + silk_SMULWW_small_b_4_neon( psDelDec->LF_AR_Q14, gain_adj_Q16_s32x2 ); + silk_SMULWW_small_b_4_neon( psDelDec->Diff_Q14, gain_adj_Q16_s32x2 ); + + /* Scale short-term prediction and shaping states */ + for( i = 0; i < NSQ_LPC_BUF_LENGTH; i++ ) { + silk_SMULWW_small_b_4_neon( psDelDec->sLPC_Q14[ i ], gain_adj_Q16_s32x2 ); + } + + for( i = 0; i < MAX_SHAPE_LPC_ORDER; i++ ) { + silk_SMULWW_small_b_4_neon( psDelDec->sAR2_Q14[ i ], gain_adj_Q16_s32x2 ); + } + + for( i = 0; i < DECISION_DELAY; i++ ) { + silk_SMULWW_small_b_4_neon( psDelDec->Pred_Q15[ i ], gain_adj_Q16_s32x2 ); + silk_SMULWW_small_b_4_neon( psDelDec->Shape_Q14[ i ], gain_adj_Q16_s32x2 ); + } + } else { + gain_adj_Q16_s32x2 = vdup_n_s32( silk_LSHIFT32( gain_adj_Q16 & 0x0000FFFF, 15 ) ); + gain_adj_Q16_s32x2 = vset_lane_s32( gain_adj_Q16 >> 16, gain_adj_Q16_s32x2, 1 ); + for( i = NSQ->sLTP_shp_buf_idx - psEncC->ltp_mem_length; i < NSQ->sLTP_shp_buf_idx - 7; i += 8 ) { + silk_SMULWW_8_neon( NSQ->sLTP_shp_Q14 + i, gain_adj_Q16_s32x2 ); + } + for( ; i < NSQ->sLTP_shp_buf_idx; i++ ) { + NSQ->sLTP_shp_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLTP_shp_Q14[ i ] ); + } + + /* Scale long-term prediction state */ + if( signal_type == TYPE_VOICED && NSQ->rewhite_flag == 0 ) { + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx - decisionDelay - 7; i += 8 ) { + silk_SMULWW_8_neon( sLTP_Q15 + i, gain_adj_Q16_s32x2 ); + } + for( ; i < NSQ->sLTP_buf_idx - decisionDelay; i++ ) { + sLTP_Q15[ i ] = silk_SMULWW( gain_adj_Q16, sLTP_Q15[ i ] ); + } + } + + /* Scale scalar states */ + silk_SMULWW_4_neon( psDelDec->LF_AR_Q14, gain_adj_Q16_s32x2 ); + silk_SMULWW_4_neon( psDelDec->Diff_Q14, gain_adj_Q16_s32x2 ); + + /* Scale short-term prediction and shaping states */ + for( i = 0; i < NSQ_LPC_BUF_LENGTH; i++ ) { + silk_SMULWW_4_neon( psDelDec->sLPC_Q14[ i ], gain_adj_Q16_s32x2 ); + } + + for( i = 0; i < MAX_SHAPE_LPC_ORDER; i++ ) { + silk_SMULWW_4_neon( psDelDec->sAR2_Q14[ i ], gain_adj_Q16_s32x2 ); + } + + for( i = 0; i < DECISION_DELAY; i++ ) { + silk_SMULWW_4_neon( psDelDec->Pred_Q15[ i ], gain_adj_Q16_s32x2 ); + silk_SMULWW_4_neon( psDelDec->Shape_Q14[ i ], gain_adj_Q16_s32x2 ); + } + } + + /* Save inverse gain */ + NSQ->prev_gain_Q16 = Gains_Q16[ subfr ]; + } +} diff --git a/lib/opus/silk/arm/NSQ_neon.c b/lib/opus/silk/arm/NSQ_neon.c new file mode 100644 index 00000000..96425299 --- /dev/null +++ b/lib/opus/silk/arm/NSQ_neon.c @@ -0,0 +1,112 @@ +/*********************************************************************** +Copyright (C) 2014 Vidyo +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "main.h" +#include "stack_alloc.h" +#include "NSQ.h" +#include "celt/cpu_support.h" +#include "celt/arm/armcpu.h" + +opus_int32 silk_noise_shape_quantizer_short_prediction_neon(const opus_int32 *buf32, const opus_int32 *coef32, opus_int order) +{ + int32x4_t coef0 = vld1q_s32(coef32); + int32x4_t coef1 = vld1q_s32(coef32 + 4); + int32x4_t coef2 = vld1q_s32(coef32 + 8); + int32x4_t coef3 = vld1q_s32(coef32 + 12); + + int32x4_t a0 = vld1q_s32(buf32 - 15); + int32x4_t a1 = vld1q_s32(buf32 - 11); + int32x4_t a2 = vld1q_s32(buf32 - 7); + int32x4_t a3 = vld1q_s32(buf32 - 3); + + int32x4_t b0 = vqdmulhq_s32(coef0, a0); + int32x4_t b1 = vqdmulhq_s32(coef1, a1); + int32x4_t b2 = vqdmulhq_s32(coef2, a2); + int32x4_t b3 = vqdmulhq_s32(coef3, a3); + + int32x4_t c0 = vaddq_s32(b0, b1); + int32x4_t c1 = vaddq_s32(b2, b3); + + int32x4_t d = vaddq_s32(c0, c1); + + int64x2_t e = vpaddlq_s32(d); + + int64x1_t f = vadd_s64(vget_low_s64(e), vget_high_s64(e)); + + opus_int32 out = vget_lane_s32(vreinterpret_s32_s64(f), 0); + + out += silk_RSHIFT( order, 1 ); + + return out; +} + + +opus_int32 silk_NSQ_noise_shape_feedback_loop_neon(const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef, opus_int order) +{ + opus_int32 out; + if (order == 8) + { + int32x4_t a00 = vdupq_n_s32(data0[0]); + int32x4_t a01 = vld1q_s32(data1); /* data1[0] ... [3] */ + + int32x4_t a0 = vextq_s32 (a00, a01, 3); /* data0[0] data1[0] ...[2] */ + int32x4_t a1 = vld1q_s32(data1 + 3); /* data1[3] ... [6] */ + + /*TODO: Convert these once in advance instead of once per sample, like + silk_noise_shape_quantizer_short_prediction_neon() does.*/ + int16x8_t coef16 = vld1q_s16(coef); + int32x4_t coef0 = vmovl_s16(vget_low_s16(coef16)); + int32x4_t coef1 = vmovl_s16(vget_high_s16(coef16)); + + /*This is not bit-exact with the C version, since we do not drop the + lower 16 bits of each multiply, but wait until the end to truncate + precision. This is an encoder-specific calculation (and unlike + silk_noise_shape_quantizer_short_prediction_neon(), is not meant to + simulate what the decoder will do). We still could use vqdmulhq_s32() + like silk_noise_shape_quantizer_short_prediction_neon() and save + half the multiplies, but the speed difference is not large, since we + then need two extra adds.*/ + int64x2_t b0 = vmull_s32(vget_low_s32(a0), vget_low_s32(coef0)); + int64x2_t b1 = vmlal_s32(b0, vget_high_s32(a0), vget_high_s32(coef0)); + int64x2_t b2 = vmlal_s32(b1, vget_low_s32(a1), vget_low_s32(coef1)); + int64x2_t b3 = vmlal_s32(b2, vget_high_s32(a1), vget_high_s32(coef1)); + + int64x1_t c = vadd_s64(vget_low_s64(b3), vget_high_s64(b3)); + int64x1_t cS = vrshr_n_s64(c, 15); + int32x2_t d = vreinterpret_s32_s64(cS); + + out = vget_lane_s32(d, 0); + vst1q_s32(data1, a0); + vst1q_s32(data1 + 4, a1); + return out; + } + return silk_NSQ_noise_shape_feedback_loop_c(data0, data1, coef, order); +} diff --git a/lib/opus/silk/arm/NSQ_neon.h b/lib/opus/silk/arm/NSQ_neon.h new file mode 100644 index 00000000..b31d9442 --- /dev/null +++ b/lib/opus/silk/arm/NSQ_neon.h @@ -0,0 +1,114 @@ +/*********************************************************************** +Copyright (C) 2014 Vidyo +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ +#ifndef SILK_NSQ_NEON_H +#define SILK_NSQ_NEON_H + +#include "cpu_support.h" +#include "SigProc_FIX.h" + +#undef silk_short_prediction_create_arch_coef +/* For vectorized calc, reverse a_Q12 coefs, convert to 32-bit, and shift for vqdmulhq_s32. */ +static OPUS_INLINE void silk_short_prediction_create_arch_coef_neon(opus_int32 *out, const opus_int16 *in, opus_int order) +{ + out[15] = silk_LSHIFT32(in[0], 15); + out[14] = silk_LSHIFT32(in[1], 15); + out[13] = silk_LSHIFT32(in[2], 15); + out[12] = silk_LSHIFT32(in[3], 15); + out[11] = silk_LSHIFT32(in[4], 15); + out[10] = silk_LSHIFT32(in[5], 15); + out[9] = silk_LSHIFT32(in[6], 15); + out[8] = silk_LSHIFT32(in[7], 15); + out[7] = silk_LSHIFT32(in[8], 15); + out[6] = silk_LSHIFT32(in[9], 15); + + if (order == 16) + { + out[5] = silk_LSHIFT32(in[10], 15); + out[4] = silk_LSHIFT32(in[11], 15); + out[3] = silk_LSHIFT32(in[12], 15); + out[2] = silk_LSHIFT32(in[13], 15); + out[1] = silk_LSHIFT32(in[14], 15); + out[0] = silk_LSHIFT32(in[15], 15); + } + else + { + out[5] = 0; + out[4] = 0; + out[3] = 0; + out[2] = 0; + out[1] = 0; + out[0] = 0; + } +} + +#if defined(OPUS_ARM_PRESUME_NEON_INTR) + +#define silk_short_prediction_create_arch_coef(out, in, order) \ + (silk_short_prediction_create_arch_coef_neon(out, in, order)) + +#elif defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + +#define silk_short_prediction_create_arch_coef(out, in, order) \ + do { if (arch == OPUS_ARCH_ARM_NEON) { silk_short_prediction_create_arch_coef_neon(out, in, order); } } while (0) + +#endif + +opus_int32 silk_noise_shape_quantizer_short_prediction_neon(const opus_int32 *buf32, const opus_int32 *coef32, opus_int order); + +opus_int32 silk_NSQ_noise_shape_feedback_loop_neon(const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef, opus_int order); + +#if defined(OPUS_ARM_PRESUME_NEON_INTR) +#undef silk_noise_shape_quantizer_short_prediction +#define silk_noise_shape_quantizer_short_prediction(in, coef, coefRev, order, arch) \ + ((void)arch,silk_noise_shape_quantizer_short_prediction_neon(in, coefRev, order)) + +#undef silk_NSQ_noise_shape_feedback_loop +#define silk_NSQ_noise_shape_feedback_loop(data0, data1, coef, order, arch) ((void)arch,silk_NSQ_noise_shape_feedback_loop_neon(data0, data1, coef, order)) + +#elif defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_MAY_HAVE_NEON_INTR) + +/* silk_noise_shape_quantizer_short_prediction implementations take different parameters based on arch + (coef vs. coefRev) so can't use the usual IMPL table implementation */ +#undef silk_noise_shape_quantizer_short_prediction +#define silk_noise_shape_quantizer_short_prediction(in, coef, coefRev, order, arch) \ + (arch == OPUS_ARCH_ARM_NEON ? \ + silk_noise_shape_quantizer_short_prediction_neon(in, coefRev, order) : \ + silk_noise_shape_quantizer_short_prediction_c(in, coef, order)) + +extern opus_int32 + (*const SILK_NSQ_NOISE_SHAPE_FEEDBACK_LOOP_IMPL[OPUS_ARCHMASK+1])( + const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef, + opus_int order); + +#undef silk_NSQ_noise_shape_feedback_loop +#define silk_NSQ_noise_shape_feedback_loop(data0, data1, coef, order, arch) \ + (SILK_NSQ_NOISE_SHAPE_FEEDBACK_LOOP_IMPL[(arch)&OPUS_ARCHMASK](data0, data1, \ + coef, order)) + +#endif + +#endif /* SILK_NSQ_NEON_H */ diff --git a/lib/opus/silk/arm/SigProc_FIX_armv4.h b/lib/opus/silk/arm/SigProc_FIX_armv4.h new file mode 100644 index 00000000..ff62b1e5 --- /dev/null +++ b/lib/opus/silk/arm/SigProc_FIX_armv4.h @@ -0,0 +1,47 @@ +/*********************************************************************** +Copyright (C) 2013 Xiph.Org Foundation and contributors +Copyright (c) 2013 Parrot +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_SIGPROC_FIX_ARMv4_H +#define SILK_SIGPROC_FIX_ARMv4_H + +#undef silk_MLA +static OPUS_INLINE opus_int32 silk_MLA_armv4(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + opus_int32 res; + __asm__( + "#silk_MLA\n\t" + "mla %0, %1, %2, %3\n\t" + : "=&r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_MLA(a, b, c) (silk_MLA_armv4(a, b, c)) + +#endif diff --git a/lib/opus/silk/arm/SigProc_FIX_armv5e.h b/lib/opus/silk/arm/SigProc_FIX_armv5e.h new file mode 100644 index 00000000..617a09ca --- /dev/null +++ b/lib/opus/silk/arm/SigProc_FIX_armv5e.h @@ -0,0 +1,61 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Copyright (c) 2013 Parrot +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_SIGPROC_FIX_ARMv5E_H +#define SILK_SIGPROC_FIX_ARMv5E_H + +#undef silk_SMULTT +static OPUS_INLINE opus_int32 silk_SMULTT_armv5e(opus_int32 a, opus_int32 b) +{ + opus_int32 res; + __asm__( + "#silk_SMULTT\n\t" + "smultt %0, %1, %2\n\t" + : "=r"(res) + : "%r"(a), "r"(b) + ); + return res; +} +#define silk_SMULTT(a, b) (silk_SMULTT_armv5e(a, b)) + +#undef silk_SMLATT +static OPUS_INLINE opus_int32 silk_SMLATT_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + opus_int32 res; + __asm__( + "#silk_SMLATT\n\t" + "smlatt %0, %1, %2, %3\n\t" + : "=r"(res) + : "%r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLATT(a, b, c) (silk_SMLATT_armv5e(a, b, c)) + +#endif diff --git a/lib/opus/silk/arm/arm_silk_map.c b/lib/opus/silk/arm/arm_silk_map.c new file mode 100644 index 00000000..0b9bfec2 --- /dev/null +++ b/lib/opus/silk/arm/arm_silk_map.c @@ -0,0 +1,123 @@ +/*********************************************************************** +Copyright (C) 2014 Vidyo +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ +#ifdef HAVE_CONFIG_H +# include "config.h" +#endif + +#include "main_FIX.h" +#include "NSQ.h" +#include "SigProc_FIX.h" + +#if defined(OPUS_HAVE_RTCD) + +# if (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && \ + !defined(OPUS_ARM_PRESUME_NEON_INTR)) + +void (*const SILK_BIQUAD_ALT_STRIDE2_IMPL[OPUS_ARCHMASK + 1])( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [4] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ +) = { + silk_biquad_alt_stride2_c, /* ARMv4 */ + silk_biquad_alt_stride2_c, /* EDSP */ + silk_biquad_alt_stride2_c, /* Media */ + silk_biquad_alt_stride2_neon, /* Neon */ +}; + +opus_int32 (*const SILK_LPC_INVERSE_PRED_GAIN_IMPL[OPUS_ARCHMASK + 1])( /* O Returns inverse prediction gain in energy domain, Q30 */ + const opus_int16 *A_Q12, /* I Prediction coefficients, Q12 [order] */ + const opus_int order /* I Prediction order */ +) = { + silk_LPC_inverse_pred_gain_c, /* ARMv4 */ + silk_LPC_inverse_pred_gain_c, /* EDSP */ + silk_LPC_inverse_pred_gain_c, /* Media */ + silk_LPC_inverse_pred_gain_neon, /* Neon */ +}; + +void (*const SILK_NSQ_DEL_DEC_IMPL[OPUS_ARCHMASK + 1])( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) = { + silk_NSQ_del_dec_c, /* ARMv4 */ + silk_NSQ_del_dec_c, /* EDSP */ + silk_NSQ_del_dec_c, /* Media */ + silk_NSQ_del_dec_neon, /* Neon */ +}; + +/*There is no table for silk_noise_shape_quantizer_short_prediction because the + NEON version takes different parameters than the C version. + Instead RTCD is done via if statements at the call sites. + See NSQ_neon.h for details.*/ + +opus_int32 + (*const SILK_NSQ_NOISE_SHAPE_FEEDBACK_LOOP_IMPL[OPUS_ARCHMASK+1])( + const opus_int32 *data0, opus_int32 *data1, const opus_int16 *coef, + opus_int order) = { + silk_NSQ_noise_shape_feedback_loop_c, /* ARMv4 */ + silk_NSQ_noise_shape_feedback_loop_c, /* EDSP */ + silk_NSQ_noise_shape_feedback_loop_c, /* Media */ + silk_NSQ_noise_shape_feedback_loop_neon, /* NEON */ +}; + +# endif + +# if defined(FIXED_POINT) && \ + defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR) + +void (*const SILK_WARPED_AUTOCORRELATION_FIX_IMPL[OPUS_ARCHMASK + 1])( + opus_int32 *corr, /* O Result [order + 1] */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *input, /* I Input data to correlate */ + const opus_int warping_Q16, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +) = { + silk_warped_autocorrelation_FIX_c, /* ARMv4 */ + silk_warped_autocorrelation_FIX_c, /* EDSP */ + silk_warped_autocorrelation_FIX_c, /* Media */ + silk_warped_autocorrelation_FIX_neon, /* Neon */ +}; + +# endif + +#endif /* OPUS_HAVE_RTCD */ diff --git a/lib/opus/silk/arm/biquad_alt_arm.h b/lib/opus/silk/arm/biquad_alt_arm.h new file mode 100644 index 00000000..66ea9f43 --- /dev/null +++ b/lib/opus/silk/arm/biquad_alt_arm.h @@ -0,0 +1,68 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_BIQUAD_ALT_ARM_H +# define SILK_BIQUAD_ALT_ARM_H + +# include "celt/arm/armcpu.h" + +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +void silk_biquad_alt_stride2_neon( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [4] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ +); + +# if !defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_PRESUME_NEON) +# define OVERRIDE_silk_biquad_alt_stride2 (1) +# define silk_biquad_alt_stride2(in, B_Q28, A_Q28, S, out, len, arch) ((void)(arch), PRESUME_NEON(silk_biquad_alt_stride2)(in, B_Q28, A_Q28, S, out, len)) +# endif +# endif + +# if !defined(OVERRIDE_silk_biquad_alt_stride2) +/*Is run-time CPU detection enabled on this platform?*/ +# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)) +extern void (*const SILK_BIQUAD_ALT_STRIDE2_IMPL[OPUS_ARCHMASK+1])( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [4] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ + ); +# define OVERRIDE_silk_biquad_alt_stride2 (1) +# define silk_biquad_alt_stride2(in, B_Q28, A_Q28, S, out, len, arch) ((*SILK_BIQUAD_ALT_STRIDE2_IMPL[(arch)&OPUS_ARCHMASK])(in, B_Q28, A_Q28, S, out, len)) +# elif defined(OPUS_ARM_PRESUME_NEON_INTR) +# define OVERRIDE_silk_biquad_alt_stride2 (1) +# define silk_biquad_alt_stride2(in, B_Q28, A_Q28, S, out, len, arch) ((void)(arch), silk_biquad_alt_stride2_neon(in, B_Q28, A_Q28, S, out, len)) +# endif +# endif + +#endif /* end SILK_BIQUAD_ALT_ARM_H */ diff --git a/lib/opus/silk/arm/biquad_alt_neon_intr.c b/lib/opus/silk/arm/biquad_alt_neon_intr.c new file mode 100644 index 00000000..97157331 --- /dev/null +++ b/lib/opus/silk/arm/biquad_alt_neon_intr.c @@ -0,0 +1,156 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#ifdef OPUS_CHECK_ASM +# include +# include "stack_alloc.h" +#endif +#include "SigProc_FIX.h" + +static inline void silk_biquad_alt_stride2_kernel( const int32x4_t A_L_s32x4, const int32x4_t A_U_s32x4, const int32x4_t B_Q28_s32x4, const int32x2_t t_s32x2, const int32x4_t in_s32x4, int32x4_t *S_s32x4, int32x2_t *out32_Q14_s32x2 ) +{ + int32x4_t t_s32x4, out32_Q14_s32x4; + + *out32_Q14_s32x2 = vadd_s32( vget_low_s32( *S_s32x4 ), t_s32x2 ); /* silk_SMLAWB( S{0,1}, B_Q28[ 0 ], in{0,1} ) */ + *S_s32x4 = vcombine_s32( vget_high_s32( *S_s32x4 ), vdup_n_s32( 0 ) ); /* S{0,1} = S{2,3}; S{2,3} = 0; */ + *out32_Q14_s32x2 = vshl_n_s32( *out32_Q14_s32x2, 2 ); /* out32_Q14_{0,1} = silk_LSHIFT( silk_SMLAWB( S{0,1}, B_Q28[ 0 ], in{0,1} ), 2 ); */ + out32_Q14_s32x4 = vcombine_s32( *out32_Q14_s32x2, *out32_Q14_s32x2 ); /* out32_Q14_{0,1,0,1} */ + t_s32x4 = vqdmulhq_s32( out32_Q14_s32x4, A_L_s32x4 ); /* silk_SMULWB( out32_Q14_{0,1,0,1}, A{0,0,1,1}_L_Q28 ) */ + *S_s32x4 = vrsraq_n_s32( *S_s32x4, t_s32x4, 14 ); /* S{0,1} = S{2,3} + silk_RSHIFT_ROUND(); S{2,3} = silk_RSHIFT_ROUND(); */ + t_s32x4 = vqdmulhq_s32( out32_Q14_s32x4, A_U_s32x4 ); /* silk_SMULWB( out32_Q14_{0,1,0,1}, A{0,0,1,1}_U_Q28 ) */ + *S_s32x4 = vaddq_s32( *S_s32x4, t_s32x4 ); /* S0 = silk_SMLAWB( S{0,1,2,3}, out32_Q14_{0,1,0,1}, A{0,0,1,1}_U_Q28 ); */ + t_s32x4 = vqdmulhq_s32( in_s32x4, B_Q28_s32x4 ); /* silk_SMULWB( B_Q28[ {1,1,2,2} ], in{0,1,0,1} ) */ + *S_s32x4 = vaddq_s32( *S_s32x4, t_s32x4 ); /* S0 = silk_SMLAWB( S0, B_Q28[ {1,1,2,2} ], in{0,1,0,1} ); */ +} + +void silk_biquad_alt_stride2_neon( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [4] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ +) +{ + /* DIRECT FORM II TRANSPOSED (uses 2 element state vector) */ + opus_int k = 0; + const int32x2_t offset_s32x2 = vdup_n_s32( (1<<14) - 1 ); + const int32x4_t offset_s32x4 = vcombine_s32( offset_s32x2, offset_s32x2 ); + int16x4_t in_s16x4 = vdup_n_s16( 0 ); + int16x4_t out_s16x4; + int32x2_t A_Q28_s32x2, A_L_s32x2, A_U_s32x2, B_Q28_s32x2, t_s32x2; + int32x4_t A_L_s32x4, A_U_s32x4, B_Q28_s32x4, S_s32x4, out32_Q14_s32x4; + int32x2x2_t t0_s32x2x2, t1_s32x2x2, t2_s32x2x2, S_s32x2x2; + +#ifdef OPUS_CHECK_ASM + opus_int32 S_c[ 4 ]; + VARDECL( opus_int16, out_c ); + SAVE_STACK; + ALLOC( out_c, 2 * len, opus_int16 ); + + silk_memcpy( &S_c, S, sizeof( S_c ) ); + silk_biquad_alt_stride2_c( in, B_Q28, A_Q28, S_c, out_c, len ); +#endif + + /* Negate A_Q28 values and split in two parts */ + A_Q28_s32x2 = vld1_s32( A_Q28 ); + A_Q28_s32x2 = vneg_s32( A_Q28_s32x2 ); + A_L_s32x2 = vshl_n_s32( A_Q28_s32x2, 18 ); /* ( -A_Q28[] & 0x00003FFF ) << 18 */ + A_L_s32x2 = vreinterpret_s32_u32( vshr_n_u32( vreinterpret_u32_s32( A_L_s32x2 ), 3 ) ); /* ( -A_Q28[] & 0x00003FFF ) << 15 */ + A_U_s32x2 = vshr_n_s32( A_Q28_s32x2, 14 ); /* silk_RSHIFT( -A_Q28[], 14 ) */ + A_U_s32x2 = vshl_n_s32( A_U_s32x2, 16 ); /* silk_RSHIFT( -A_Q28[], 14 ) << 16 (Clip two leading bits to conform to C function.) */ + A_U_s32x2 = vshr_n_s32( A_U_s32x2, 1 ); /* silk_RSHIFT( -A_Q28[], 14 ) << 15 */ + + B_Q28_s32x2 = vld1_s32( B_Q28 ); + t_s32x2 = vld1_s32( B_Q28 + 1 ); + t0_s32x2x2 = vzip_s32( A_L_s32x2, A_L_s32x2 ); + t1_s32x2x2 = vzip_s32( A_U_s32x2, A_U_s32x2 ); + t2_s32x2x2 = vzip_s32( t_s32x2, t_s32x2 ); + A_L_s32x4 = vcombine_s32( t0_s32x2x2.val[ 0 ], t0_s32x2x2.val[ 1 ] ); /* A{0,0,1,1}_L_Q28 */ + A_U_s32x4 = vcombine_s32( t1_s32x2x2.val[ 0 ], t1_s32x2x2.val[ 1 ] ); /* A{0,0,1,1}_U_Q28 */ + B_Q28_s32x4 = vcombine_s32( t2_s32x2x2.val[ 0 ], t2_s32x2x2.val[ 1 ] ); /* B_Q28[ {1,1,2,2} ] */ + S_s32x4 = vld1q_s32( S ); /* S0 = S[ 0 ]; S3 = S[ 3 ]; */ + S_s32x2x2 = vtrn_s32( vget_low_s32( S_s32x4 ), vget_high_s32( S_s32x4 ) ); /* S2 = S[ 1 ]; S1 = S[ 2 ]; */ + S_s32x4 = vcombine_s32( S_s32x2x2.val[ 0 ], S_s32x2x2.val[ 1 ] ); + + for( ; k < len - 1; k += 2 ) { + int32x4_t in_s32x4[ 2 ], t_s32x4; + int32x2_t out32_Q14_s32x2[ 2 ]; + + /* S[ 2 * i + 0 ], S[ 2 * i + 1 ], S[ 2 * i + 2 ], S[ 2 * i + 3 ]: Q12 */ + in_s16x4 = vld1_s16( &in[ 2 * k ] ); /* in{0,1,2,3} = in[ 2 * k + {0,1,2,3} ]; */ + in_s32x4[ 0 ] = vshll_n_s16( in_s16x4, 15 ); /* in{0,1,2,3} << 15 */ + t_s32x4 = vqdmulhq_lane_s32( in_s32x4[ 0 ], B_Q28_s32x2, 0 ); /* silk_SMULWB( B_Q28[ 0 ], in{0,1,2,3} ) */ + in_s32x4[ 1 ] = vcombine_s32( vget_high_s32( in_s32x4[ 0 ] ), vget_high_s32( in_s32x4[ 0 ] ) ); /* in{2,3,2,3} << 15 */ + in_s32x4[ 0 ] = vcombine_s32( vget_low_s32 ( in_s32x4[ 0 ] ), vget_low_s32 ( in_s32x4[ 0 ] ) ); /* in{0,1,0,1} << 15 */ + silk_biquad_alt_stride2_kernel( A_L_s32x4, A_U_s32x4, B_Q28_s32x4, vget_low_s32 ( t_s32x4 ), in_s32x4[ 0 ], &S_s32x4, &out32_Q14_s32x2[ 0 ] ); + silk_biquad_alt_stride2_kernel( A_L_s32x4, A_U_s32x4, B_Q28_s32x4, vget_high_s32( t_s32x4 ), in_s32x4[ 1 ], &S_s32x4, &out32_Q14_s32x2[ 1 ] ); + + /* Scale back to Q0 and saturate */ + out32_Q14_s32x4 = vcombine_s32( out32_Q14_s32x2[ 0 ], out32_Q14_s32x2[ 1 ] ); /* out32_Q14_{0,1,2,3} */ + out32_Q14_s32x4 = vaddq_s32( out32_Q14_s32x4, offset_s32x4 ); /* out32_Q14_{0,1,2,3} + (1<<14) - 1 */ + out_s16x4 = vqshrn_n_s32( out32_Q14_s32x4, 14 ); /* (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14_{0,1,2,3} + (1<<14) - 1, 14 ) ) */ + vst1_s16( &out[ 2 * k ], out_s16x4 ); /* out[ 2 * k + {0,1,2,3} ] = (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14_{0,1,2,3} + (1<<14) - 1, 14 ) ); */ + } + + /* Process leftover. */ + if( k < len ) { + int32x4_t in_s32x4; + int32x2_t out32_Q14_s32x2; + + /* S[ 2 * i + 0 ], S[ 2 * i + 1 ]: Q12 */ + in_s16x4 = vld1_lane_s16( &in[ 2 * k + 0 ], in_s16x4, 0 ); /* in{0,1} = in[ 2 * k + {0,1} ]; */ + in_s16x4 = vld1_lane_s16( &in[ 2 * k + 1 ], in_s16x4, 1 ); /* in{0,1} = in[ 2 * k + {0,1} ]; */ + in_s32x4 = vshll_n_s16( in_s16x4, 15 ); /* in{0,1} << 15 */ + t_s32x2 = vqdmulh_lane_s32( vget_low_s32( in_s32x4 ), B_Q28_s32x2, 0 ); /* silk_SMULWB( B_Q28[ 0 ], in{0,1} ) */ + in_s32x4 = vcombine_s32( vget_low_s32( in_s32x4 ), vget_low_s32( in_s32x4 ) ); /* in{0,1,0,1} << 15 */ + silk_biquad_alt_stride2_kernel( A_L_s32x4, A_U_s32x4, B_Q28_s32x4, t_s32x2, in_s32x4, &S_s32x4, &out32_Q14_s32x2 ); + + /* Scale back to Q0 and saturate */ + out32_Q14_s32x2 = vadd_s32( out32_Q14_s32x2, offset_s32x2 ); /* out32_Q14_{0,1} + (1<<14) - 1 */ + out32_Q14_s32x4 = vcombine_s32( out32_Q14_s32x2, out32_Q14_s32x2 ); /* out32_Q14_{0,1,0,1} + (1<<14) - 1 */ + out_s16x4 = vqshrn_n_s32( out32_Q14_s32x4, 14 ); /* (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14_{0,1,0,1} + (1<<14) - 1, 14 ) ) */ + vst1_lane_s16( &out[ 2 * k + 0 ], out_s16x4, 0 ); /* out[ 2 * k + 0 ] = (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14_0 + (1<<14) - 1, 14 ) ); */ + vst1_lane_s16( &out[ 2 * k + 1 ], out_s16x4, 1 ); /* out[ 2 * k + 1 ] = (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14_1 + (1<<14) - 1, 14 ) ); */ + } + + vst1q_lane_s32( &S[ 0 ], S_s32x4, 0 ); /* S[ 0 ] = S0; */ + vst1q_lane_s32( &S[ 1 ], S_s32x4, 2 ); /* S[ 1 ] = S2; */ + vst1q_lane_s32( &S[ 2 ], S_s32x4, 1 ); /* S[ 2 ] = S1; */ + vst1q_lane_s32( &S[ 3 ], S_s32x4, 3 ); /* S[ 3 ] = S3; */ + +#ifdef OPUS_CHECK_ASM + silk_assert( !memcmp( S_c, S, sizeof( S_c ) ) ); + silk_assert( !memcmp( out_c, out, 2 * len * sizeof( opus_int16 ) ) ); + RESTORE_STACK; +#endif +} diff --git a/lib/opus/silk/arm/macros_arm64.h b/lib/opus/silk/arm/macros_arm64.h new file mode 100644 index 00000000..ed030413 --- /dev/null +++ b/lib/opus/silk/arm/macros_arm64.h @@ -0,0 +1,39 @@ +/*********************************************************************** +Copyright (C) 2015 Vidyo +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MACROS_ARM64_H +#define SILK_MACROS_ARM64_H + +#include + +#undef silk_ADD_SAT32 +#define silk_ADD_SAT32(a, b) (vqadds_s32((a), (b))) + +#undef silk_SUB_SAT32 +#define silk_SUB_SAT32(a, b) (vqsubs_s32((a), (b))) + +#endif /* SILK_MACROS_ARM64_H */ diff --git a/lib/opus/silk/arm/macros_armv4.h b/lib/opus/silk/arm/macros_armv4.h new file mode 100644 index 00000000..877eb18d --- /dev/null +++ b/lib/opus/silk/arm/macros_armv4.h @@ -0,0 +1,110 @@ +/*********************************************************************** +Copyright (C) 2013 Xiph.Org Foundation and contributors. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MACROS_ARMv4_H +#define SILK_MACROS_ARMv4_H + +/* This macro only avoids the undefined behaviour from a left shift of + a negative value. It should only be used in macros that can't include + SigProc_FIX.h. In other cases, use silk_LSHIFT32(). */ +#define SAFE_SHL(a,b) ((opus_int32)((opus_uint32)(a) << (b))) + +/* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */ +#undef silk_SMULWB +static OPUS_INLINE opus_int32 silk_SMULWB_armv4(opus_int32 a, opus_int16 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMULWB\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(a), "r"(SAFE_SHL(b,16)) + ); + return rd_hi; +} +#define silk_SMULWB(a, b) (silk_SMULWB_armv4(a, b)) + +/* a32 + (b32 * (opus_int32)((opus_int16)(c32))) >> 16 output have to be 32bit int */ +#undef silk_SMLAWB +#define silk_SMLAWB(a, b, c) ((a) + silk_SMULWB(b, c)) + +/* (a32 * (b32 >> 16)) >> 16 */ +#undef silk_SMULWT +static OPUS_INLINE opus_int32 silk_SMULWT_armv4(opus_int32 a, opus_int32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMULWT\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(a), "r"(b&~0xFFFF) + ); + return rd_hi; +} +#define silk_SMULWT(a, b) (silk_SMULWT_armv4(a, b)) + +/* a32 + (b32 * (c32 >> 16)) >> 16 */ +#undef silk_SMLAWT +#define silk_SMLAWT(a, b, c) ((a) + silk_SMULWT(b, c)) + +/* (a32 * b32) >> 16 */ +#undef silk_SMULWW +static OPUS_INLINE opus_int32 silk_SMULWW_armv4(opus_int32 a, opus_int32 b) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMULWW\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(a), "r"(b) + ); + return SAFE_SHL(rd_hi,16)+(rd_lo>>16); +} +#define silk_SMULWW(a, b) (silk_SMULWW_armv4(a, b)) + +#undef silk_SMLAWW +static OPUS_INLINE opus_int32 silk_SMLAWW_armv4(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + unsigned rd_lo; + int rd_hi; + __asm__( + "#silk_SMLAWW\n\t" + "smull %0, %1, %2, %3\n\t" + : "=&r"(rd_lo), "=&r"(rd_hi) + : "%r"(b), "r"(c) + ); + return a+SAFE_SHL(rd_hi,16)+(rd_lo>>16); +} +#define silk_SMLAWW(a, b, c) (silk_SMLAWW_armv4(a, b, c)) + +#undef SAFE_SHL + +#endif /* SILK_MACROS_ARMv4_H */ diff --git a/lib/opus/silk/arm/macros_armv5e.h b/lib/opus/silk/arm/macros_armv5e.h new file mode 100644 index 00000000..b14ec65d --- /dev/null +++ b/lib/opus/silk/arm/macros_armv5e.h @@ -0,0 +1,220 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Copyright (c) 2013 Parrot +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MACROS_ARMv5E_H +#define SILK_MACROS_ARMv5E_H + +/* This macro only avoids the undefined behaviour from a left shift of + a negative value. It should only be used in macros that can't include + SigProc_FIX.h. In other cases, use silk_LSHIFT32(). */ +#define SAFE_SHL(a,b) ((opus_int32)((opus_uint32)(a) << (b))) + +/* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */ +#undef silk_SMULWB +static OPUS_INLINE opus_int32 silk_SMULWB_armv5e(opus_int32 a, opus_int16 b) +{ + int res; + __asm__( + "#silk_SMULWB\n\t" + "smulwb %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SMULWB(a, b) (silk_SMULWB_armv5e(a, b)) + +/* a32 + (b32 * (opus_int32)((opus_int16)(c32))) >> 16 output have to be 32bit int */ +#undef silk_SMLAWB +static OPUS_INLINE opus_int32 silk_SMLAWB_armv5e(opus_int32 a, opus_int32 b, + opus_int16 c) +{ + int res; + __asm__( + "#silk_SMLAWB\n\t" + "smlawb %0, %1, %2, %3\n\t" + : "=r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLAWB(a, b, c) (silk_SMLAWB_armv5e(a, b, c)) + +/* (a32 * (b32 >> 16)) >> 16 */ +#undef silk_SMULWT +static OPUS_INLINE opus_int32 silk_SMULWT_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SMULWT\n\t" + "smulwt %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SMULWT(a, b) (silk_SMULWT_armv5e(a, b)) + +/* a32 + (b32 * (c32 >> 16)) >> 16 */ +#undef silk_SMLAWT +static OPUS_INLINE opus_int32 silk_SMLAWT_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + int res; + __asm__( + "#silk_SMLAWT\n\t" + "smlawt %0, %1, %2, %3\n\t" + : "=r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLAWT(a, b, c) (silk_SMLAWT_armv5e(a, b, c)) + +/* (opus_int32)((opus_int16)(a3))) * (opus_int32)((opus_int16)(b32)) output have to be 32bit int */ +#undef silk_SMULBB +static OPUS_INLINE opus_int32 silk_SMULBB_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SMULBB\n\t" + "smulbb %0, %1, %2\n\t" + : "=r"(res) + : "%r"(a), "r"(b) + ); + return res; +} +#define silk_SMULBB(a, b) (silk_SMULBB_armv5e(a, b)) + +/* a32 + (opus_int32)((opus_int16)(b32)) * (opus_int32)((opus_int16)(c32)) output have to be 32bit int */ +#undef silk_SMLABB +static OPUS_INLINE opus_int32 silk_SMLABB_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + int res; + __asm__( + "#silk_SMLABB\n\t" + "smlabb %0, %1, %2, %3\n\t" + : "=r"(res) + : "%r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLABB(a, b, c) (silk_SMLABB_armv5e(a, b, c)) + +/* (opus_int32)((opus_int16)(a32)) * (b32 >> 16) */ +#undef silk_SMULBT +static OPUS_INLINE opus_int32 silk_SMULBT_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SMULBT\n\t" + "smulbt %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SMULBT(a, b) (silk_SMULBT_armv5e(a, b)) + +/* a32 + (opus_int32)((opus_int16)(b32)) * (c32 >> 16) */ +#undef silk_SMLABT +static OPUS_INLINE opus_int32 silk_SMLABT_armv5e(opus_int32 a, opus_int32 b, + opus_int32 c) +{ + int res; + __asm__( + "#silk_SMLABT\n\t" + "smlabt %0, %1, %2, %3\n\t" + : "=r"(res) + : "r"(b), "r"(c), "r"(a) + ); + return res; +} +#define silk_SMLABT(a, b, c) (silk_SMLABT_armv5e(a, b, c)) + +/* add/subtract with output saturated */ +#undef silk_ADD_SAT32 +static OPUS_INLINE opus_int32 silk_ADD_SAT32_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_ADD_SAT32\n\t" + "qadd %0, %1, %2\n\t" + : "=r"(res) + : "%r"(a), "r"(b) + ); + return res; +} +#define silk_ADD_SAT32(a, b) (silk_ADD_SAT32_armv5e(a, b)) + +#undef silk_SUB_SAT32 +static OPUS_INLINE opus_int32 silk_SUB_SAT32_armv5e(opus_int32 a, opus_int32 b) +{ + int res; + __asm__( + "#silk_SUB_SAT32\n\t" + "qsub %0, %1, %2\n\t" + : "=r"(res) + : "r"(a), "r"(b) + ); + return res; +} +#define silk_SUB_SAT32(a, b) (silk_SUB_SAT32_armv5e(a, b)) + +#undef silk_CLZ16 +static OPUS_INLINE opus_int32 silk_CLZ16_armv5(opus_int16 in16) +{ + int res; + __asm__( + "#silk_CLZ16\n\t" + "clz %0, %1;\n" + : "=r"(res) + : "r"(SAFE_SHL(in16,16)|0x8000) + ); + return res; +} +#define silk_CLZ16(in16) (silk_CLZ16_armv5(in16)) + +#undef silk_CLZ32 +static OPUS_INLINE opus_int32 silk_CLZ32_armv5(opus_int32 in32) +{ + int res; + __asm__( + "#silk_CLZ32\n\t" + "clz %0, %1\n\t" + : "=r"(res) + : "r"(in32) + ); + return res; +} +#define silk_CLZ32(in32) (silk_CLZ32_armv5(in32)) + +#undef SAFE_SHL + +#endif /* SILK_MACROS_ARMv5E_H */ diff --git a/lib/opus/silk/biquad_alt.c b/lib/opus/silk/biquad_alt.c new file mode 100644 index 00000000..54566a43 --- /dev/null +++ b/lib/opus/silk/biquad_alt.c @@ -0,0 +1,121 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +/* * + * silk_biquad_alt.c * + * * + * Second order ARMA filter * + * Can handle slowly varying filter coefficients * + * */ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Second order ARMA filter, alternative implementation */ +void silk_biquad_alt_stride1( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [2] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ +) +{ + /* DIRECT FORM II TRANSPOSED (uses 2 element state vector) */ + opus_int k; + opus_int32 inval, A0_U_Q28, A0_L_Q28, A1_U_Q28, A1_L_Q28, out32_Q14; + + /* Negate A_Q28 values and split in two parts */ + A0_L_Q28 = ( -A_Q28[ 0 ] ) & 0x00003FFF; /* lower part */ + A0_U_Q28 = silk_RSHIFT( -A_Q28[ 0 ], 14 ); /* upper part */ + A1_L_Q28 = ( -A_Q28[ 1 ] ) & 0x00003FFF; /* lower part */ + A1_U_Q28 = silk_RSHIFT( -A_Q28[ 1 ], 14 ); /* upper part */ + + for( k = 0; k < len; k++ ) { + /* S[ 0 ], S[ 1 ]: Q12 */ + inval = in[ k ]; + out32_Q14 = silk_LSHIFT( silk_SMLAWB( S[ 0 ], B_Q28[ 0 ], inval ), 2 ); + + S[ 0 ] = S[1] + silk_RSHIFT_ROUND( silk_SMULWB( out32_Q14, A0_L_Q28 ), 14 ); + S[ 0 ] = silk_SMLAWB( S[ 0 ], out32_Q14, A0_U_Q28 ); + S[ 0 ] = silk_SMLAWB( S[ 0 ], B_Q28[ 1 ], inval); + + S[ 1 ] = silk_RSHIFT_ROUND( silk_SMULWB( out32_Q14, A1_L_Q28 ), 14 ); + S[ 1 ] = silk_SMLAWB( S[ 1 ], out32_Q14, A1_U_Q28 ); + S[ 1 ] = silk_SMLAWB( S[ 1 ], B_Q28[ 2 ], inval ); + + /* Scale back to Q0 and saturate */ + out[ k ] = (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14 + (1<<14) - 1, 14 ) ); + } +} + +void silk_biquad_alt_stride2_c( + const opus_int16 *in, /* I input signal */ + const opus_int32 *B_Q28, /* I MA coefficients [3] */ + const opus_int32 *A_Q28, /* I AR coefficients [2] */ + opus_int32 *S, /* I/O State vector [4] */ + opus_int16 *out, /* O output signal */ + const opus_int32 len /* I signal length (must be even) */ +) +{ + /* DIRECT FORM II TRANSPOSED (uses 2 element state vector) */ + opus_int k; + opus_int32 A0_U_Q28, A0_L_Q28, A1_U_Q28, A1_L_Q28, out32_Q14[ 2 ]; + + /* Negate A_Q28 values and split in two parts */ + A0_L_Q28 = ( -A_Q28[ 0 ] ) & 0x00003FFF; /* lower part */ + A0_U_Q28 = silk_RSHIFT( -A_Q28[ 0 ], 14 ); /* upper part */ + A1_L_Q28 = ( -A_Q28[ 1 ] ) & 0x00003FFF; /* lower part */ + A1_U_Q28 = silk_RSHIFT( -A_Q28[ 1 ], 14 ); /* upper part */ + + for( k = 0; k < len; k++ ) { + /* S[ 0 ], S[ 1 ], S[ 2 ], S[ 3 ]: Q12 */ + out32_Q14[ 0 ] = silk_LSHIFT( silk_SMLAWB( S[ 0 ], B_Q28[ 0 ], in[ 2 * k + 0 ] ), 2 ); + out32_Q14[ 1 ] = silk_LSHIFT( silk_SMLAWB( S[ 2 ], B_Q28[ 0 ], in[ 2 * k + 1 ] ), 2 ); + + S[ 0 ] = S[ 1 ] + silk_RSHIFT_ROUND( silk_SMULWB( out32_Q14[ 0 ], A0_L_Q28 ), 14 ); + S[ 2 ] = S[ 3 ] + silk_RSHIFT_ROUND( silk_SMULWB( out32_Q14[ 1 ], A0_L_Q28 ), 14 ); + S[ 0 ] = silk_SMLAWB( S[ 0 ], out32_Q14[ 0 ], A0_U_Q28 ); + S[ 2 ] = silk_SMLAWB( S[ 2 ], out32_Q14[ 1 ], A0_U_Q28 ); + S[ 0 ] = silk_SMLAWB( S[ 0 ], B_Q28[ 1 ], in[ 2 * k + 0 ] ); + S[ 2 ] = silk_SMLAWB( S[ 2 ], B_Q28[ 1 ], in[ 2 * k + 1 ] ); + + S[ 1 ] = silk_RSHIFT_ROUND( silk_SMULWB( out32_Q14[ 0 ], A1_L_Q28 ), 14 ); + S[ 3 ] = silk_RSHIFT_ROUND( silk_SMULWB( out32_Q14[ 1 ], A1_L_Q28 ), 14 ); + S[ 1 ] = silk_SMLAWB( S[ 1 ], out32_Q14[ 0 ], A1_U_Q28 ); + S[ 3 ] = silk_SMLAWB( S[ 3 ], out32_Q14[ 1 ], A1_U_Q28 ); + S[ 1 ] = silk_SMLAWB( S[ 1 ], B_Q28[ 2 ], in[ 2 * k + 0 ] ); + S[ 3 ] = silk_SMLAWB( S[ 3 ], B_Q28[ 2 ], in[ 2 * k + 1 ] ); + + /* Scale back to Q0 and saturate */ + out[ 2 * k + 0 ] = (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14[ 0 ] + (1<<14) - 1, 14 ) ); + out[ 2 * k + 1 ] = (opus_int16)silk_SAT16( silk_RSHIFT( out32_Q14[ 1 ] + (1<<14) - 1, 14 ) ); + } +} diff --git a/lib/opus/silk/bwexpander.c b/lib/opus/silk/bwexpander.c new file mode 100644 index 00000000..afa97907 --- /dev/null +++ b/lib/opus/silk/bwexpander.c @@ -0,0 +1,51 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Chirp (bandwidth expand) LP AR filter */ +void silk_bwexpander( + opus_int16 *ar, /* I/O AR filter to be expanded (without leading 1) */ + const opus_int d, /* I Length of ar */ + opus_int32 chirp_Q16 /* I Chirp factor (typically in the range 0 to 1) */ +) +{ + opus_int i; + opus_int32 chirp_minus_one_Q16 = chirp_Q16 - 65536; + + /* NB: Dont use silk_SMULWB, instead of silk_RSHIFT_ROUND( silk_MUL(), 16 ), below. */ + /* Bias in silk_SMULWB can lead to unstable filters */ + for( i = 0; i < d - 1; i++ ) { + ar[ i ] = (opus_int16)silk_RSHIFT_ROUND( silk_MUL( chirp_Q16, ar[ i ] ), 16 ); + chirp_Q16 += silk_RSHIFT_ROUND( silk_MUL( chirp_Q16, chirp_minus_one_Q16 ), 16 ); + } + ar[ d - 1 ] = (opus_int16)silk_RSHIFT_ROUND( silk_MUL( chirp_Q16, ar[ d - 1 ] ), 16 ); +} diff --git a/lib/opus/silk/bwexpander_32.c b/lib/opus/silk/bwexpander_32.c new file mode 100644 index 00000000..0f32b9df --- /dev/null +++ b/lib/opus/silk/bwexpander_32.c @@ -0,0 +1,51 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Chirp (bandwidth expand) LP AR filter. + This logic is reused in _celt_lpc(). Any bug fixes should also be applied there. */ +void silk_bwexpander_32( + opus_int32 *ar, /* I/O AR filter to be expanded (without leading 1) */ + const opus_int d, /* I Length of ar */ + opus_int32 chirp_Q16 /* I Chirp factor in Q16 */ +) +{ + opus_int i; + opus_int32 chirp_minus_one_Q16 = chirp_Q16 - 65536; + + for( i = 0; i < d - 1; i++ ) { + ar[ i ] = silk_SMULWW( chirp_Q16, ar[ i ] ); + chirp_Q16 += silk_RSHIFT_ROUND( silk_MUL( chirp_Q16, chirp_minus_one_Q16 ), 16 ); + } + ar[ d - 1 ] = silk_SMULWW( chirp_Q16, ar[ d - 1 ] ); +} + diff --git a/lib/opus/silk/check_control_input.c b/lib/opus/silk/check_control_input.c new file mode 100644 index 00000000..739fb01f --- /dev/null +++ b/lib/opus/silk/check_control_input.c @@ -0,0 +1,106 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "control.h" +#include "errors.h" + +/* Check encoder control struct */ +opus_int check_control_input( + silk_EncControlStruct *encControl /* I Control structure */ +) +{ + celt_assert( encControl != NULL ); + + if( ( ( encControl->API_sampleRate != 8000 ) && + ( encControl->API_sampleRate != 12000 ) && + ( encControl->API_sampleRate != 16000 ) && + ( encControl->API_sampleRate != 24000 ) && + ( encControl->API_sampleRate != 32000 ) && + ( encControl->API_sampleRate != 44100 ) && + ( encControl->API_sampleRate != 48000 ) ) || + ( ( encControl->desiredInternalSampleRate != 8000 ) && + ( encControl->desiredInternalSampleRate != 12000 ) && + ( encControl->desiredInternalSampleRate != 16000 ) ) || + ( ( encControl->maxInternalSampleRate != 8000 ) && + ( encControl->maxInternalSampleRate != 12000 ) && + ( encControl->maxInternalSampleRate != 16000 ) ) || + ( ( encControl->minInternalSampleRate != 8000 ) && + ( encControl->minInternalSampleRate != 12000 ) && + ( encControl->minInternalSampleRate != 16000 ) ) || + ( encControl->minInternalSampleRate > encControl->desiredInternalSampleRate ) || + ( encControl->maxInternalSampleRate < encControl->desiredInternalSampleRate ) || + ( encControl->minInternalSampleRate > encControl->maxInternalSampleRate ) ) { + celt_assert( 0 ); + return SILK_ENC_FS_NOT_SUPPORTED; + } + if( encControl->payloadSize_ms != 10 && + encControl->payloadSize_ms != 20 && + encControl->payloadSize_ms != 40 && + encControl->payloadSize_ms != 60 ) { + celt_assert( 0 ); + return SILK_ENC_PACKET_SIZE_NOT_SUPPORTED; + } + if( encControl->packetLossPercentage < 0 || encControl->packetLossPercentage > 100 ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_LOSS_RATE; + } + if( encControl->useDTX < 0 || encControl->useDTX > 1 ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_DTX_SETTING; + } + if( encControl->useCBR < 0 || encControl->useCBR > 1 ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_CBR_SETTING; + } + if( encControl->useInBandFEC < 0 || encControl->useInBandFEC > 1 ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_INBAND_FEC_SETTING; + } + if( encControl->nChannelsAPI < 1 || encControl->nChannelsAPI > ENCODER_NUM_CHANNELS ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_NUMBER_OF_CHANNELS_ERROR; + } + if( encControl->nChannelsInternal < 1 || encControl->nChannelsInternal > ENCODER_NUM_CHANNELS ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_NUMBER_OF_CHANNELS_ERROR; + } + if( encControl->nChannelsInternal > encControl->nChannelsAPI ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_NUMBER_OF_CHANNELS_ERROR; + } + if( encControl->complexity < 0 || encControl->complexity > 10 ) { + celt_assert( 0 ); + return SILK_ENC_INVALID_COMPLEXITY_SETTING; + } + + return SILK_NO_ERROR; +} diff --git a/lib/opus/silk/code_signs.c b/lib/opus/silk/code_signs.c new file mode 100644 index 00000000..dfd1dca9 --- /dev/null +++ b/lib/opus/silk/code_signs.c @@ -0,0 +1,115 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/*#define silk_enc_map(a) ((a) > 0 ? 1 : 0)*/ +/*#define silk_dec_map(a) ((a) > 0 ? 1 : -1)*/ +/* shifting avoids if-statement */ +#define silk_enc_map(a) ( silk_RSHIFT( (a), 15 ) + 1 ) +#define silk_dec_map(a) ( silk_LSHIFT( (a), 1 ) - 1 ) + +/* Encodes signs of excitation */ +void silk_encode_signs( + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + const opus_int8 pulses[], /* I pulse signal */ + opus_int length, /* I length of input */ + const opus_int signalType, /* I Signal type */ + const opus_int quantOffsetType, /* I Quantization offset type */ + const opus_int sum_pulses[ MAX_NB_SHELL_BLOCKS ] /* I Sum of absolute pulses per block */ +) +{ + opus_int i, j, p; + opus_uint8 icdf[ 2 ]; + const opus_int8 *q_ptr; + const opus_uint8 *icdf_ptr; + + icdf[ 1 ] = 0; + q_ptr = pulses; + i = silk_SMULBB( 7, silk_ADD_LSHIFT( quantOffsetType, signalType, 1 ) ); + icdf_ptr = &silk_sign_iCDF[ i ]; + length = silk_RSHIFT( length + SHELL_CODEC_FRAME_LENGTH/2, LOG2_SHELL_CODEC_FRAME_LENGTH ); + for( i = 0; i < length; i++ ) { + p = sum_pulses[ i ]; + if( p > 0 ) { + icdf[ 0 ] = icdf_ptr[ silk_min( p & 0x1F, 6 ) ]; + for( j = 0; j < SHELL_CODEC_FRAME_LENGTH; j++ ) { + if( q_ptr[ j ] != 0 ) { + ec_enc_icdf( psRangeEnc, silk_enc_map( q_ptr[ j ]), icdf, 8 ); + } + } + } + q_ptr += SHELL_CODEC_FRAME_LENGTH; + } +} + +/* Decodes signs of excitation */ +void silk_decode_signs( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 pulses[], /* I/O pulse signal */ + opus_int length, /* I length of input */ + const opus_int signalType, /* I Signal type */ + const opus_int quantOffsetType, /* I Quantization offset type */ + const opus_int sum_pulses[ MAX_NB_SHELL_BLOCKS ] /* I Sum of absolute pulses per block */ +) +{ + opus_int i, j, p; + opus_uint8 icdf[ 2 ]; + opus_int16 *q_ptr; + const opus_uint8 *icdf_ptr; + + icdf[ 1 ] = 0; + q_ptr = pulses; + i = silk_SMULBB( 7, silk_ADD_LSHIFT( quantOffsetType, signalType, 1 ) ); + icdf_ptr = &silk_sign_iCDF[ i ]; + length = silk_RSHIFT( length + SHELL_CODEC_FRAME_LENGTH/2, LOG2_SHELL_CODEC_FRAME_LENGTH ); + for( i = 0; i < length; i++ ) { + p = sum_pulses[ i ]; + if( p > 0 ) { + icdf[ 0 ] = icdf_ptr[ silk_min( p & 0x1F, 6 ) ]; + for( j = 0; j < SHELL_CODEC_FRAME_LENGTH; j++ ) { + if( q_ptr[ j ] > 0 ) { + /* attach sign */ +#if 0 + /* conditional implementation */ + if( ec_dec_icdf( psRangeDec, icdf, 8 ) == 0 ) { + q_ptr[ j ] = -q_ptr[ j ]; + } +#else + /* implementation with shift, subtraction, multiplication */ + q_ptr[ j ] *= silk_dec_map( ec_dec_icdf( psRangeDec, icdf, 8 ) ); +#endif + } + } + } + q_ptr += SHELL_CODEC_FRAME_LENGTH; + } +} diff --git a/lib/opus/silk/control.h b/lib/opus/silk/control.h new file mode 100644 index 00000000..b76ec33c --- /dev/null +++ b/lib/opus/silk/control.h @@ -0,0 +1,150 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_CONTROL_H +#define SILK_CONTROL_H + +#include "typedef.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +/* Decoder API flags */ +#define FLAG_DECODE_NORMAL 0 +#define FLAG_PACKET_LOST 1 +#define FLAG_DECODE_LBRR 2 + +/***********************************************/ +/* Structure for controlling encoder operation */ +/***********************************************/ +typedef struct { + /* I: Number of channels; 1/2 */ + opus_int32 nChannelsAPI; + + /* I: Number of channels; 1/2 */ + opus_int32 nChannelsInternal; + + /* I: Input signal sampling rate in Hertz; 8000/12000/16000/24000/32000/44100/48000 */ + opus_int32 API_sampleRate; + + /* I: Maximum internal sampling rate in Hertz; 8000/12000/16000 */ + opus_int32 maxInternalSampleRate; + + /* I: Minimum internal sampling rate in Hertz; 8000/12000/16000 */ + opus_int32 minInternalSampleRate; + + /* I: Soft request for internal sampling rate in Hertz; 8000/12000/16000 */ + opus_int32 desiredInternalSampleRate; + + /* I: Number of samples per packet in milliseconds; 10/20/40/60 */ + opus_int payloadSize_ms; + + /* I: Bitrate during active speech in bits/second; internally limited */ + opus_int32 bitRate; + + /* I: Uplink packet loss in percent (0-100) */ + opus_int packetLossPercentage; + + /* I: Complexity mode; 0 is lowest, 10 is highest complexity */ + opus_int complexity; + + /* I: Flag to enable in-band Forward Error Correction (FEC); 0/1 */ + opus_int useInBandFEC; + + /* I: Flag to actually code in-band Forward Error Correction (FEC) in the current packet; 0/1 */ + opus_int LBRR_coded; + + /* I: Flag to enable discontinuous transmission (DTX); 0/1 */ + opus_int useDTX; + + /* I: Flag to use constant bitrate */ + opus_int useCBR; + + /* I: Maximum number of bits allowed for the frame */ + opus_int maxBits; + + /* I: Causes a smooth downmix to mono */ + opus_int toMono; + + /* I: Opus encoder is allowing us to switch bandwidth */ + opus_int opusCanSwitch; + + /* I: Make frames as independent as possible (but still use LPC) */ + opus_int reducedDependency; + + /* O: Internal sampling rate used, in Hertz; 8000/12000/16000 */ + opus_int32 internalSampleRate; + + /* O: Flag that bandwidth switching is allowed (because low voice activity) */ + opus_int allowBandwidthSwitch; + + /* O: Flag that SILK runs in WB mode without variable LP filter (use for switching between WB/SWB/FB) */ + opus_int inWBmodeWithoutVariableLP; + + /* O: Stereo width */ + opus_int stereoWidth_Q14; + + /* O: Tells the Opus encoder we're ready to switch */ + opus_int switchReady; + + /* O: SILK Signal type */ + opus_int signalType; + + /* O: SILK offset (dithering) */ + opus_int offset; +} silk_EncControlStruct; + +/**************************************************************************/ +/* Structure for controlling decoder operation and reading decoder status */ +/**************************************************************************/ +typedef struct { + /* I: Number of channels; 1/2 */ + opus_int32 nChannelsAPI; + + /* I: Number of channels; 1/2 */ + opus_int32 nChannelsInternal; + + /* I: Output signal sampling rate in Hertz; 8000/12000/16000/24000/32000/44100/48000 */ + opus_int32 API_sampleRate; + + /* I: Internal sampling rate used, in Hertz; 8000/12000/16000 */ + opus_int32 internalSampleRate; + + /* I: Number of samples per packet in milliseconds; 10/20/40/60 */ + opus_int payloadSize_ms; + + /* O: Pitch lag of previous frame (0 if unvoiced), measured in samples at 48 kHz */ + opus_int prevPitchLag; +} silk_DecControlStruct; + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/control_SNR.c b/lib/opus/silk/control_SNR.c new file mode 100644 index 00000000..9a6db275 --- /dev/null +++ b/lib/opus/silk/control_SNR.c @@ -0,0 +1,113 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "tuning_parameters.h" + +/* These tables hold SNR values divided by 21 (so they fit in 8 bits) + for different target bitrates spaced at 400 bps interval. The first + 10 values are omitted (0-4 kb/s) because they're all zeros. + These tables were obtained by running different SNRs through the + encoder and measuring the active bitrate. */ +static const unsigned char silk_TargetRate_NB_21[117 - 10] = { + 0, 15, 39, 52, 61, 68, + 74, 79, 84, 88, 92, 95, 99,102,105,108,111,114,117,119,122,124, + 126,129,131,133,135,137,139,142,143,145,147,149,151,153,155,157, + 158,160,162,163,165,167,168,170,171,173,174,176,177,179,180,182, + 183,185,186,187,189,190,192,193,194,196,197,199,200,201,203,204, + 205,207,208,209,211,212,213,215,216,217,219,220,221,223,224,225, + 227,228,230,231,232,234,235,236,238,239,241,242,243,245,246,248, + 249,250,252,253,255 +}; + +static const unsigned char silk_TargetRate_MB_21[165 - 10] = { + 0, 0, 28, 43, 52, 59, + 65, 70, 74, 78, 81, 85, 87, 90, 93, 95, 98,100,102,105,107,109, + 111,113,115,116,118,120,122,123,125,127,128,130,131,133,134,136, + 137,138,140,141,143,144,145,147,148,149,151,152,153,154,156,157, + 158,159,160,162,163,164,165,166,167,168,169,171,172,173,174,175, + 176,177,178,179,180,181,182,183,184,185,186,187,188,188,189,190, + 191,192,193,194,195,196,197,198,199,200,201,202,203,203,204,205, + 206,207,208,209,210,211,212,213,214,214,215,216,217,218,219,220, + 221,222,223,224,224,225,226,227,228,229,230,231,232,233,234,235, + 236,236,237,238,239,240,241,242,243,244,245,246,247,248,249,250, + 251,252,253,254,255 +}; + +static const unsigned char silk_TargetRate_WB_21[201 - 10] = { + 0, 0, 0, 8, 29, 41, + 49, 56, 62, 66, 70, 74, 77, 80, 83, 86, 88, 91, 93, 95, 97, 99, + 101,103,105,107,108,110,112,113,115,116,118,119,121,122,123,125, + 126,127,129,130,131,132,134,135,136,137,138,140,141,142,143,144, + 145,146,147,148,149,150,151,152,153,154,156,157,158,159,159,160, + 161,162,163,164,165,166,167,168,169,170,171,171,172,173,174,175, + 176,177,177,178,179,180,181,181,182,183,184,185,185,186,187,188, + 189,189,190,191,192,192,193,194,195,195,196,197,198,198,199,200, + 200,201,202,203,203,204,205,206,206,207,208,209,209,210,211,211, + 212,213,214,214,215,216,216,217,218,219,219,220,221,221,222,223, + 224,224,225,226,226,227,228,229,229,230,231,232,232,233,234,234, + 235,236,237,237,238,239,240,240,241,242,243,243,244,245,246,246, + 247,248,249,249,250,251,252,253,255 +}; + +/* Control SNR of redidual quantizer */ +opus_int silk_control_SNR( + silk_encoder_state *psEncC, /* I/O Pointer to Silk encoder state */ + opus_int32 TargetRate_bps /* I Target max bitrate (bps) */ +) +{ + int id; + int bound; + const unsigned char *snr_table; + + psEncC->TargetRate_bps = TargetRate_bps; + if( psEncC->nb_subfr == 2 ) { + TargetRate_bps -= 2000 + psEncC->fs_kHz/16; + } + if( psEncC->fs_kHz == 8 ) { + bound = sizeof(silk_TargetRate_NB_21); + snr_table = silk_TargetRate_NB_21; + } else if( psEncC->fs_kHz == 12 ) { + bound = sizeof(silk_TargetRate_MB_21); + snr_table = silk_TargetRate_MB_21; + } else { + bound = sizeof(silk_TargetRate_WB_21); + snr_table = silk_TargetRate_WB_21; + } + id = (TargetRate_bps+200)/400; + id = silk_min(id - 10, bound-1); + if( id <= 0 ) { + psEncC->SNR_dB_Q7 = 0; + } else { + psEncC->SNR_dB_Q7 = snr_table[id]*21; + } + return SILK_NO_ERROR; +} diff --git a/lib/opus/silk/control_audio_bandwidth.c b/lib/opus/silk/control_audio_bandwidth.c new file mode 100644 index 00000000..f6d22d83 --- /dev/null +++ b/lib/opus/silk/control_audio_bandwidth.c @@ -0,0 +1,132 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "tuning_parameters.h" + +/* Control internal sampling rate */ +opus_int silk_control_audio_bandwidth( + silk_encoder_state *psEncC, /* I/O Pointer to Silk encoder state */ + silk_EncControlStruct *encControl /* I Control structure */ +) +{ + opus_int fs_kHz; + opus_int orig_kHz; + opus_int32 fs_Hz; + + orig_kHz = psEncC->fs_kHz; + /* Handle a bandwidth-switching reset where we need to be aware what the last sampling rate was. */ + if( orig_kHz == 0 ) { + orig_kHz = psEncC->sLP.saved_fs_kHz; + } + fs_kHz = orig_kHz; + fs_Hz = silk_SMULBB( fs_kHz, 1000 ); + if( fs_Hz == 0 ) { + /* Encoder has just been initialized */ + fs_Hz = silk_min( psEncC->desiredInternal_fs_Hz, psEncC->API_fs_Hz ); + fs_kHz = silk_DIV32_16( fs_Hz, 1000 ); + } else if( fs_Hz > psEncC->API_fs_Hz || fs_Hz > psEncC->maxInternal_fs_Hz || fs_Hz < psEncC->minInternal_fs_Hz ) { + /* Make sure internal rate is not higher than external rate or maximum allowed, or lower than minimum allowed */ + fs_Hz = psEncC->API_fs_Hz; + fs_Hz = silk_min( fs_Hz, psEncC->maxInternal_fs_Hz ); + fs_Hz = silk_max( fs_Hz, psEncC->minInternal_fs_Hz ); + fs_kHz = silk_DIV32_16( fs_Hz, 1000 ); + } else { + /* State machine for the internal sampling rate switching */ + if( psEncC->sLP.transition_frame_no >= TRANSITION_FRAMES ) { + /* Stop transition phase */ + psEncC->sLP.mode = 0; + } + if( psEncC->allow_bandwidth_switch || encControl->opusCanSwitch ) { + /* Check if we should switch down */ + if( silk_SMULBB( orig_kHz, 1000 ) > psEncC->desiredInternal_fs_Hz ) + { + /* Switch down */ + if( psEncC->sLP.mode == 0 ) { + /* New transition */ + psEncC->sLP.transition_frame_no = TRANSITION_FRAMES; + + /* Reset transition filter state */ + silk_memset( psEncC->sLP.In_LP_State, 0, sizeof( psEncC->sLP.In_LP_State ) ); + } + if( encControl->opusCanSwitch ) { + /* Stop transition phase */ + psEncC->sLP.mode = 0; + + /* Switch to a lower sample frequency */ + fs_kHz = orig_kHz == 16 ? 12 : 8; + } else { + if( psEncC->sLP.transition_frame_no <= 0 ) { + encControl->switchReady = 1; + /* Make room for redundancy */ + encControl->maxBits -= encControl->maxBits * 5 / ( encControl->payloadSize_ms + 5 ); + } else { + /* Direction: down (at double speed) */ + psEncC->sLP.mode = -2; + } + } + } + else + /* Check if we should switch up */ + if( silk_SMULBB( orig_kHz, 1000 ) < psEncC->desiredInternal_fs_Hz ) + { + /* Switch up */ + if( encControl->opusCanSwitch ) { + /* Switch to a higher sample frequency */ + fs_kHz = orig_kHz == 8 ? 12 : 16; + + /* New transition */ + psEncC->sLP.transition_frame_no = 0; + + /* Reset transition filter state */ + silk_memset( psEncC->sLP.In_LP_State, 0, sizeof( psEncC->sLP.In_LP_State ) ); + + /* Direction: up */ + psEncC->sLP.mode = 1; + } else { + if( psEncC->sLP.mode == 0 ) { + encControl->switchReady = 1; + /* Make room for redundancy */ + encControl->maxBits -= encControl->maxBits * 5 / ( encControl->payloadSize_ms + 5 ); + } else { + /* Direction: up */ + psEncC->sLP.mode = 1; + } + } + } else { + if (psEncC->sLP.mode<0) + psEncC->sLP.mode = 1; + } + } + } + + return fs_kHz; +} diff --git a/lib/opus/silk/control_codec.c b/lib/opus/silk/control_codec.c new file mode 100644 index 00000000..784ffe66 --- /dev/null +++ b/lib/opus/silk/control_codec.c @@ -0,0 +1,423 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#ifdef FIXED_POINT +#include "main_FIX.h" +#define silk_encoder_state_Fxx silk_encoder_state_FIX +#else +#include "main_FLP.h" +#define silk_encoder_state_Fxx silk_encoder_state_FLP +#endif +#include "stack_alloc.h" +#include "tuning_parameters.h" +#include "pitch_est_defines.h" + +static opus_int silk_setup_resamplers( + silk_encoder_state_Fxx *psEnc, /* I/O */ + opus_int fs_kHz /* I */ +); + +static opus_int silk_setup_fs( + silk_encoder_state_Fxx *psEnc, /* I/O */ + opus_int fs_kHz, /* I */ + opus_int PacketSize_ms /* I */ +); + +static opus_int silk_setup_complexity( + silk_encoder_state *psEncC, /* I/O */ + opus_int Complexity /* I */ +); + +static OPUS_INLINE opus_int silk_setup_LBRR( + silk_encoder_state *psEncC, /* I/O */ + const silk_EncControlStruct *encControl /* I */ +); + + +/* Control encoder */ +opus_int silk_control_encoder( + silk_encoder_state_Fxx *psEnc, /* I/O Pointer to Silk encoder state */ + silk_EncControlStruct *encControl, /* I Control structure */ + const opus_int allow_bw_switch, /* I Flag to allow switching audio bandwidth */ + const opus_int channelNb, /* I Channel number */ + const opus_int force_fs_kHz +) +{ + opus_int fs_kHz, ret = 0; + + psEnc->sCmn.useDTX = encControl->useDTX; + psEnc->sCmn.useCBR = encControl->useCBR; + psEnc->sCmn.API_fs_Hz = encControl->API_sampleRate; + psEnc->sCmn.maxInternal_fs_Hz = encControl->maxInternalSampleRate; + psEnc->sCmn.minInternal_fs_Hz = encControl->minInternalSampleRate; + psEnc->sCmn.desiredInternal_fs_Hz = encControl->desiredInternalSampleRate; + psEnc->sCmn.useInBandFEC = encControl->useInBandFEC; + psEnc->sCmn.nChannelsAPI = encControl->nChannelsAPI; + psEnc->sCmn.nChannelsInternal = encControl->nChannelsInternal; + psEnc->sCmn.allow_bandwidth_switch = allow_bw_switch; + psEnc->sCmn.channelNb = channelNb; + + if( psEnc->sCmn.controlled_since_last_payload != 0 && psEnc->sCmn.prefillFlag == 0 ) { + if( psEnc->sCmn.API_fs_Hz != psEnc->sCmn.prev_API_fs_Hz && psEnc->sCmn.fs_kHz > 0 ) { + /* Change in API sampling rate in the middle of encoding a packet */ + ret += silk_setup_resamplers( psEnc, psEnc->sCmn.fs_kHz ); + } + return ret; + } + + /* Beyond this point we know that there are no previously coded frames in the payload buffer */ + + /********************************************/ + /* Determine internal sampling rate */ + /********************************************/ + fs_kHz = silk_control_audio_bandwidth( &psEnc->sCmn, encControl ); + if( force_fs_kHz ) { + fs_kHz = force_fs_kHz; + } + /********************************************/ + /* Prepare resampler and buffered data */ + /********************************************/ + ret += silk_setup_resamplers( psEnc, fs_kHz ); + + /********************************************/ + /* Set internal sampling frequency */ + /********************************************/ + ret += silk_setup_fs( psEnc, fs_kHz, encControl->payloadSize_ms ); + + /********************************************/ + /* Set encoding complexity */ + /********************************************/ + ret += silk_setup_complexity( &psEnc->sCmn, encControl->complexity ); + + /********************************************/ + /* Set packet loss rate measured by farend */ + /********************************************/ + psEnc->sCmn.PacketLoss_perc = encControl->packetLossPercentage; + + /********************************************/ + /* Set LBRR usage */ + /********************************************/ + ret += silk_setup_LBRR( &psEnc->sCmn, encControl ); + + psEnc->sCmn.controlled_since_last_payload = 1; + + return ret; +} + +static opus_int silk_setup_resamplers( + silk_encoder_state_Fxx *psEnc, /* I/O */ + opus_int fs_kHz /* I */ +) +{ + opus_int ret = SILK_NO_ERROR; + SAVE_STACK; + + if( psEnc->sCmn.fs_kHz != fs_kHz || psEnc->sCmn.prev_API_fs_Hz != psEnc->sCmn.API_fs_Hz ) + { + if( psEnc->sCmn.fs_kHz == 0 ) { + /* Initialize the resampler for enc_API.c preparing resampling from API_fs_Hz to fs_kHz */ + ret += silk_resampler_init( &psEnc->sCmn.resampler_state, psEnc->sCmn.API_fs_Hz, fs_kHz * 1000, 1 ); + } else { + VARDECL( opus_int16, x_buf_API_fs_Hz ); + VARDECL( silk_resampler_state_struct, temp_resampler_state ); +#ifdef FIXED_POINT + opus_int16 *x_bufFIX = psEnc->x_buf; +#else + VARDECL( opus_int16, x_bufFIX ); + opus_int32 new_buf_samples; +#endif + opus_int32 api_buf_samples; + opus_int32 old_buf_samples; + opus_int32 buf_length_ms; + + buf_length_ms = silk_LSHIFT( psEnc->sCmn.nb_subfr * 5, 1 ) + LA_SHAPE_MS; + old_buf_samples = buf_length_ms * psEnc->sCmn.fs_kHz; + +#ifndef FIXED_POINT + new_buf_samples = buf_length_ms * fs_kHz; + ALLOC( x_bufFIX, silk_max( old_buf_samples, new_buf_samples ), + opus_int16 ); + silk_float2short_array( x_bufFIX, psEnc->x_buf, old_buf_samples ); +#endif + + /* Initialize resampler for temporary resampling of x_buf data to API_fs_Hz */ + ALLOC( temp_resampler_state, 1, silk_resampler_state_struct ); + ret += silk_resampler_init( temp_resampler_state, silk_SMULBB( psEnc->sCmn.fs_kHz, 1000 ), psEnc->sCmn.API_fs_Hz, 0 ); + + /* Calculate number of samples to temporarily upsample */ + api_buf_samples = buf_length_ms * silk_DIV32_16( psEnc->sCmn.API_fs_Hz, 1000 ); + + /* Temporary resampling of x_buf data to API_fs_Hz */ + ALLOC( x_buf_API_fs_Hz, api_buf_samples, opus_int16 ); + ret += silk_resampler( temp_resampler_state, x_buf_API_fs_Hz, x_bufFIX, old_buf_samples ); + + /* Initialize the resampler for enc_API.c preparing resampling from API_fs_Hz to fs_kHz */ + ret += silk_resampler_init( &psEnc->sCmn.resampler_state, psEnc->sCmn.API_fs_Hz, silk_SMULBB( fs_kHz, 1000 ), 1 ); + + /* Correct resampler state by resampling buffered data from API_fs_Hz to fs_kHz */ + ret += silk_resampler( &psEnc->sCmn.resampler_state, x_bufFIX, x_buf_API_fs_Hz, api_buf_samples ); + +#ifndef FIXED_POINT + silk_short2float_array( psEnc->x_buf, x_bufFIX, new_buf_samples); +#endif + } + } + + psEnc->sCmn.prev_API_fs_Hz = psEnc->sCmn.API_fs_Hz; + + RESTORE_STACK; + return ret; +} + +static opus_int silk_setup_fs( + silk_encoder_state_Fxx *psEnc, /* I/O */ + opus_int fs_kHz, /* I */ + opus_int PacketSize_ms /* I */ +) +{ + opus_int ret = SILK_NO_ERROR; + + /* Set packet size */ + if( PacketSize_ms != psEnc->sCmn.PacketSize_ms ) { + if( ( PacketSize_ms != 10 ) && + ( PacketSize_ms != 20 ) && + ( PacketSize_ms != 40 ) && + ( PacketSize_ms != 60 ) ) { + ret = SILK_ENC_PACKET_SIZE_NOT_SUPPORTED; + } + if( PacketSize_ms <= 10 ) { + psEnc->sCmn.nFramesPerPacket = 1; + psEnc->sCmn.nb_subfr = PacketSize_ms == 10 ? 2 : 1; + psEnc->sCmn.frame_length = silk_SMULBB( PacketSize_ms, fs_kHz ); + psEnc->sCmn.pitch_LPC_win_length = silk_SMULBB( FIND_PITCH_LPC_WIN_MS_2_SF, fs_kHz ); + if( psEnc->sCmn.fs_kHz == 8 ) { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_10_ms_NB_iCDF; + } else { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_10_ms_iCDF; + } + } else { + psEnc->sCmn.nFramesPerPacket = silk_DIV32_16( PacketSize_ms, MAX_FRAME_LENGTH_MS ); + psEnc->sCmn.nb_subfr = MAX_NB_SUBFR; + psEnc->sCmn.frame_length = silk_SMULBB( 20, fs_kHz ); + psEnc->sCmn.pitch_LPC_win_length = silk_SMULBB( FIND_PITCH_LPC_WIN_MS, fs_kHz ); + if( psEnc->sCmn.fs_kHz == 8 ) { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_NB_iCDF; + } else { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_iCDF; + } + } + psEnc->sCmn.PacketSize_ms = PacketSize_ms; + psEnc->sCmn.TargetRate_bps = 0; /* trigger new SNR computation */ + } + + /* Set internal sampling frequency */ + celt_assert( fs_kHz == 8 || fs_kHz == 12 || fs_kHz == 16 ); + celt_assert( psEnc->sCmn.nb_subfr == 2 || psEnc->sCmn.nb_subfr == 4 ); + if( psEnc->sCmn.fs_kHz != fs_kHz ) { + /* reset part of the state */ + silk_memset( &psEnc->sShape, 0, sizeof( psEnc->sShape ) ); + silk_memset( &psEnc->sCmn.sNSQ, 0, sizeof( psEnc->sCmn.sNSQ ) ); + silk_memset( psEnc->sCmn.prev_NLSFq_Q15, 0, sizeof( psEnc->sCmn.prev_NLSFq_Q15 ) ); + silk_memset( &psEnc->sCmn.sLP.In_LP_State, 0, sizeof( psEnc->sCmn.sLP.In_LP_State ) ); + psEnc->sCmn.inputBufIx = 0; + psEnc->sCmn.nFramesEncoded = 0; + psEnc->sCmn.TargetRate_bps = 0; /* trigger new SNR computation */ + + /* Initialize non-zero parameters */ + psEnc->sCmn.prevLag = 100; + psEnc->sCmn.first_frame_after_reset = 1; + psEnc->sShape.LastGainIndex = 10; + psEnc->sCmn.sNSQ.lagPrev = 100; + psEnc->sCmn.sNSQ.prev_gain_Q16 = 65536; + psEnc->sCmn.prevSignalType = TYPE_NO_VOICE_ACTIVITY; + + psEnc->sCmn.fs_kHz = fs_kHz; + if( psEnc->sCmn.fs_kHz == 8 ) { + if( psEnc->sCmn.nb_subfr == MAX_NB_SUBFR ) { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_NB_iCDF; + } else { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_10_ms_NB_iCDF; + } + } else { + if( psEnc->sCmn.nb_subfr == MAX_NB_SUBFR ) { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_iCDF; + } else { + psEnc->sCmn.pitch_contour_iCDF = silk_pitch_contour_10_ms_iCDF; + } + } + if( psEnc->sCmn.fs_kHz == 8 || psEnc->sCmn.fs_kHz == 12 ) { + psEnc->sCmn.predictLPCOrder = MIN_LPC_ORDER; + psEnc->sCmn.psNLSF_CB = &silk_NLSF_CB_NB_MB; + } else { + psEnc->sCmn.predictLPCOrder = MAX_LPC_ORDER; + psEnc->sCmn.psNLSF_CB = &silk_NLSF_CB_WB; + } + psEnc->sCmn.subfr_length = SUB_FRAME_LENGTH_MS * fs_kHz; + psEnc->sCmn.frame_length = silk_SMULBB( psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr ); + psEnc->sCmn.ltp_mem_length = silk_SMULBB( LTP_MEM_LENGTH_MS, fs_kHz ); + psEnc->sCmn.la_pitch = silk_SMULBB( LA_PITCH_MS, fs_kHz ); + psEnc->sCmn.max_pitch_lag = silk_SMULBB( 18, fs_kHz ); + if( psEnc->sCmn.nb_subfr == MAX_NB_SUBFR ) { + psEnc->sCmn.pitch_LPC_win_length = silk_SMULBB( FIND_PITCH_LPC_WIN_MS, fs_kHz ); + } else { + psEnc->sCmn.pitch_LPC_win_length = silk_SMULBB( FIND_PITCH_LPC_WIN_MS_2_SF, fs_kHz ); + } + if( psEnc->sCmn.fs_kHz == 16 ) { + psEnc->sCmn.pitch_lag_low_bits_iCDF = silk_uniform8_iCDF; + } else if( psEnc->sCmn.fs_kHz == 12 ) { + psEnc->sCmn.pitch_lag_low_bits_iCDF = silk_uniform6_iCDF; + } else { + psEnc->sCmn.pitch_lag_low_bits_iCDF = silk_uniform4_iCDF; + } + } + + /* Check that settings are valid */ + celt_assert( ( psEnc->sCmn.subfr_length * psEnc->sCmn.nb_subfr ) == psEnc->sCmn.frame_length ); + + return ret; +} + +static opus_int silk_setup_complexity( + silk_encoder_state *psEncC, /* I/O */ + opus_int Complexity /* I */ +) +{ + opus_int ret = 0; + + /* Set encoding complexity */ + celt_assert( Complexity >= 0 && Complexity <= 10 ); + if( Complexity < 1 ) { + psEncC->pitchEstimationComplexity = SILK_PE_MIN_COMPLEX; + psEncC->pitchEstimationThreshold_Q16 = SILK_FIX_CONST( 0.8, 16 ); + psEncC->pitchEstimationLPCOrder = 6; + psEncC->shapingLPCOrder = 12; + psEncC->la_shape = 3 * psEncC->fs_kHz; + psEncC->nStatesDelayedDecision = 1; + psEncC->useInterpolatedNLSFs = 0; + psEncC->NLSF_MSVQ_Survivors = 2; + psEncC->warping_Q16 = 0; + } else if( Complexity < 2 ) { + psEncC->pitchEstimationComplexity = SILK_PE_MID_COMPLEX; + psEncC->pitchEstimationThreshold_Q16 = SILK_FIX_CONST( 0.76, 16 ); + psEncC->pitchEstimationLPCOrder = 8; + psEncC->shapingLPCOrder = 14; + psEncC->la_shape = 5 * psEncC->fs_kHz; + psEncC->nStatesDelayedDecision = 1; + psEncC->useInterpolatedNLSFs = 0; + psEncC->NLSF_MSVQ_Survivors = 3; + psEncC->warping_Q16 = 0; + } else if( Complexity < 3 ) { + psEncC->pitchEstimationComplexity = SILK_PE_MIN_COMPLEX; + psEncC->pitchEstimationThreshold_Q16 = SILK_FIX_CONST( 0.8, 16 ); + psEncC->pitchEstimationLPCOrder = 6; + psEncC->shapingLPCOrder = 12; + psEncC->la_shape = 3 * psEncC->fs_kHz; + psEncC->nStatesDelayedDecision = 2; + psEncC->useInterpolatedNLSFs = 0; + psEncC->NLSF_MSVQ_Survivors = 2; + psEncC->warping_Q16 = 0; + } else if( Complexity < 4 ) { + psEncC->pitchEstimationComplexity = SILK_PE_MID_COMPLEX; + psEncC->pitchEstimationThreshold_Q16 = SILK_FIX_CONST( 0.76, 16 ); + psEncC->pitchEstimationLPCOrder = 8; + psEncC->shapingLPCOrder = 14; + psEncC->la_shape = 5 * psEncC->fs_kHz; + psEncC->nStatesDelayedDecision = 2; + psEncC->useInterpolatedNLSFs = 0; + psEncC->NLSF_MSVQ_Survivors = 4; + psEncC->warping_Q16 = 0; + } else if( Complexity < 6 ) { + psEncC->pitchEstimationComplexity = SILK_PE_MID_COMPLEX; + psEncC->pitchEstimationThreshold_Q16 = SILK_FIX_CONST( 0.74, 16 ); + psEncC->pitchEstimationLPCOrder = 10; + psEncC->shapingLPCOrder = 16; + psEncC->la_shape = 5 * psEncC->fs_kHz; + psEncC->nStatesDelayedDecision = 2; + psEncC->useInterpolatedNLSFs = 1; + psEncC->NLSF_MSVQ_Survivors = 6; + psEncC->warping_Q16 = psEncC->fs_kHz * SILK_FIX_CONST( WARPING_MULTIPLIER, 16 ); + } else if( Complexity < 8 ) { + psEncC->pitchEstimationComplexity = SILK_PE_MID_COMPLEX; + psEncC->pitchEstimationThreshold_Q16 = SILK_FIX_CONST( 0.72, 16 ); + psEncC->pitchEstimationLPCOrder = 12; + psEncC->shapingLPCOrder = 20; + psEncC->la_shape = 5 * psEncC->fs_kHz; + psEncC->nStatesDelayedDecision = 3; + psEncC->useInterpolatedNLSFs = 1; + psEncC->NLSF_MSVQ_Survivors = 8; + psEncC->warping_Q16 = psEncC->fs_kHz * SILK_FIX_CONST( WARPING_MULTIPLIER, 16 ); + } else { + psEncC->pitchEstimationComplexity = SILK_PE_MAX_COMPLEX; + psEncC->pitchEstimationThreshold_Q16 = SILK_FIX_CONST( 0.7, 16 ); + psEncC->pitchEstimationLPCOrder = 16; + psEncC->shapingLPCOrder = 24; + psEncC->la_shape = 5 * psEncC->fs_kHz; + psEncC->nStatesDelayedDecision = MAX_DEL_DEC_STATES; + psEncC->useInterpolatedNLSFs = 1; + psEncC->NLSF_MSVQ_Survivors = 16; + psEncC->warping_Q16 = psEncC->fs_kHz * SILK_FIX_CONST( WARPING_MULTIPLIER, 16 ); + } + + /* Do not allow higher pitch estimation LPC order than predict LPC order */ + psEncC->pitchEstimationLPCOrder = silk_min_int( psEncC->pitchEstimationLPCOrder, psEncC->predictLPCOrder ); + psEncC->shapeWinLength = SUB_FRAME_LENGTH_MS * psEncC->fs_kHz + 2 * psEncC->la_shape; + psEncC->Complexity = Complexity; + + celt_assert( psEncC->pitchEstimationLPCOrder <= MAX_FIND_PITCH_LPC_ORDER ); + celt_assert( psEncC->shapingLPCOrder <= MAX_SHAPE_LPC_ORDER ); + celt_assert( psEncC->nStatesDelayedDecision <= MAX_DEL_DEC_STATES ); + celt_assert( psEncC->warping_Q16 <= 32767 ); + celt_assert( psEncC->la_shape <= LA_SHAPE_MAX ); + celt_assert( psEncC->shapeWinLength <= SHAPE_LPC_WIN_MAX ); + + return ret; +} + +static OPUS_INLINE opus_int silk_setup_LBRR( + silk_encoder_state *psEncC, /* I/O */ + const silk_EncControlStruct *encControl /* I */ +) +{ + opus_int LBRR_in_previous_packet, ret = SILK_NO_ERROR; + + LBRR_in_previous_packet = psEncC->LBRR_enabled; + psEncC->LBRR_enabled = encControl->LBRR_coded; + if( psEncC->LBRR_enabled ) { + /* Set gain increase for coding LBRR excitation */ + if( LBRR_in_previous_packet == 0 ) { + /* Previous packet did not have LBRR, and was therefore coded at a higher bitrate */ + psEncC->LBRR_GainIncreases = 7; + } else { + psEncC->LBRR_GainIncreases = silk_max_int( 7 - silk_SMULWB( (opus_int32)psEncC->PacketLoss_perc, SILK_FIX_CONST( 0.2, 16 ) ), 3 ); + } + } + + return ret; +} diff --git a/lib/opus/silk/debug.c b/lib/opus/silk/debug.c new file mode 100644 index 00000000..46a24a47 --- /dev/null +++ b/lib/opus/silk/debug.c @@ -0,0 +1,174 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +typedef int prevent_empty_translation_unit_warning; + +#include "debug.h" + +#if SILK_DEBUG || SILK_TIC_TOC +#include "SigProc_FIX.h" +#endif + +#if SILK_TIC_TOC + +#if (defined(_WIN32) || defined(_WINCE)) +#include /* timer */ +#else /* Linux or Mac*/ +#include +#endif + +#ifdef _WIN32 +unsigned long silk_GetHighResolutionTime(void) /* O time in usec*/ +{ + /* Returns a time counter in microsec */ + /* the resolution is platform dependent */ + /* but is typically 1.62 us resolution */ + LARGE_INTEGER lpPerformanceCount; + LARGE_INTEGER lpFrequency; + QueryPerformanceCounter(&lpPerformanceCount); + QueryPerformanceFrequency(&lpFrequency); + return (unsigned long)((1000000*(lpPerformanceCount.QuadPart)) / lpFrequency.QuadPart); +} +#else /* Linux or Mac*/ +unsigned long GetHighResolutionTime(void) /* O time in usec*/ +{ + struct timeval tv; + gettimeofday(&tv, 0); + return((tv.tv_sec*1000000)+(tv.tv_usec)); +} +#endif + +int silk_Timer_nTimers = 0; +int silk_Timer_depth_ctr = 0; +char silk_Timer_tags[silk_NUM_TIMERS_MAX][silk_NUM_TIMERS_MAX_TAG_LEN]; +#ifdef _WIN32 +LARGE_INTEGER silk_Timer_start[silk_NUM_TIMERS_MAX]; +#else +unsigned long silk_Timer_start[silk_NUM_TIMERS_MAX]; +#endif +unsigned int silk_Timer_cnt[silk_NUM_TIMERS_MAX]; +opus_int64 silk_Timer_min[silk_NUM_TIMERS_MAX]; +opus_int64 silk_Timer_sum[silk_NUM_TIMERS_MAX]; +opus_int64 silk_Timer_max[silk_NUM_TIMERS_MAX]; +opus_int64 silk_Timer_depth[silk_NUM_TIMERS_MAX]; + +#ifdef _WIN32 +void silk_TimerSave(char *file_name) +{ + if( silk_Timer_nTimers > 0 ) + { + int k; + FILE *fp; + LARGE_INTEGER lpFrequency; + LARGE_INTEGER lpPerformanceCount1, lpPerformanceCount2; + int del = 0x7FFFFFFF; + double avg, sum_avg; + /* estimate overhead of calling performance counters */ + for( k = 0; k < 1000; k++ ) { + QueryPerformanceCounter(&lpPerformanceCount1); + QueryPerformanceCounter(&lpPerformanceCount2); + lpPerformanceCount2.QuadPart -= lpPerformanceCount1.QuadPart; + if( (int)lpPerformanceCount2.LowPart < del ) + del = lpPerformanceCount2.LowPart; + } + QueryPerformanceFrequency(&lpFrequency); + /* print results to file */ + sum_avg = 0.0f; + for( k = 0; k < silk_Timer_nTimers; k++ ) { + if (silk_Timer_depth[k] == 0) { + sum_avg += (1e6 * silk_Timer_sum[k] / silk_Timer_cnt[k] - del) / lpFrequency.QuadPart * silk_Timer_cnt[k]; + } + } + fp = fopen(file_name, "w"); + fprintf(fp, " min avg %% max count\n"); + for( k = 0; k < silk_Timer_nTimers; k++ ) { + if (silk_Timer_depth[k] == 0) { + fprintf(fp, "%-28s", silk_Timer_tags[k]); + } else if (silk_Timer_depth[k] == 1) { + fprintf(fp, " %-27s", silk_Timer_tags[k]); + } else if (silk_Timer_depth[k] == 2) { + fprintf(fp, " %-26s", silk_Timer_tags[k]); + } else if (silk_Timer_depth[k] == 3) { + fprintf(fp, " %-25s", silk_Timer_tags[k]); + } else { + fprintf(fp, " %-24s", silk_Timer_tags[k]); + } + avg = (1e6 * silk_Timer_sum[k] / silk_Timer_cnt[k] - del) / lpFrequency.QuadPart; + fprintf(fp, "%8.2f", (1e6 * (silk_max_64(silk_Timer_min[k] - del, 0))) / lpFrequency.QuadPart); + fprintf(fp, "%12.2f %6.2f", avg, 100.0 * avg / sum_avg * silk_Timer_cnt[k]); + fprintf(fp, "%12.2f", (1e6 * (silk_max_64(silk_Timer_max[k] - del, 0))) / lpFrequency.QuadPart); + fprintf(fp, "%10d\n", silk_Timer_cnt[k]); + } + fprintf(fp, " microseconds\n"); + fclose(fp); + } +} +#else +void silk_TimerSave(char *file_name) +{ + if( silk_Timer_nTimers > 0 ) + { + int k; + FILE *fp; + /* print results to file */ + fp = fopen(file_name, "w"); + fprintf(fp, " min avg max count\n"); + for( k = 0; k < silk_Timer_nTimers; k++ ) + { + if (silk_Timer_depth[k] == 0) { + fprintf(fp, "%-28s", silk_Timer_tags[k]); + } else if (silk_Timer_depth[k] == 1) { + fprintf(fp, " %-27s", silk_Timer_tags[k]); + } else if (silk_Timer_depth[k] == 2) { + fprintf(fp, " %-26s", silk_Timer_tags[k]); + } else if (silk_Timer_depth[k] == 3) { + fprintf(fp, " %-25s", silk_Timer_tags[k]); + } else { + fprintf(fp, " %-24s", silk_Timer_tags[k]); + } + fprintf(fp, "%d ", silk_Timer_min[k]); + fprintf(fp, "%f ", (double)silk_Timer_sum[k] / (double)silk_Timer_cnt[k]); + fprintf(fp, "%d ", silk_Timer_max[k]); + fprintf(fp, "%10d\n", silk_Timer_cnt[k]); + } + fprintf(fp, " microseconds\n"); + fclose(fp); + } +} +#endif + +#endif /* SILK_TIC_TOC */ + +#if SILK_DEBUG +FILE *silk_debug_store_fp[ silk_NUM_STORES_MAX ]; +int silk_debug_store_count = 0; +#endif /* SILK_DEBUG */ + diff --git a/lib/opus/silk/debug.h b/lib/opus/silk/debug.h new file mode 100644 index 00000000..36163e47 --- /dev/null +++ b/lib/opus/silk/debug.h @@ -0,0 +1,267 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_DEBUG_H +#define SILK_DEBUG_H + +/* Set to 1 to enable DEBUG_STORE_DATA() macros for dumping + * intermediate signals from the codec. + */ +#define SILK_DEBUG 0 + +/* Flag for using timers */ +#define SILK_TIC_TOC 0 + +#if SILK_DEBUG || SILK_TIC_TOC +#include "typedef.h" +#include /* strcpy, strcmp */ +#include /* file writing */ +#endif + +#ifdef __cplusplus +extern "C" +{ +#endif + +#if SILK_TIC_TOC + +unsigned long GetHighResolutionTime(void); /* O time in usec*/ + +#if (defined(_WIN32) || defined(_WINCE)) +#include /* timer */ +#else /* Linux or Mac*/ +#include +#endif + +/*********************************/ +/* timer functions for profiling */ +/*********************************/ +/* example: */ +/* */ +/* TIC(LPC) */ +/* do_LPC(in_vec, order, acoef); // do LPC analysis */ +/* TOC(LPC) */ +/* */ +/* and call the following just before exiting (from main) */ +/* */ +/* silk_TimerSave("silk_TimingData.txt"); */ +/* */ +/* results are now in silk_TimingData.txt */ + +void silk_TimerSave(char *file_name); + +/* max number of timers (in different locations) */ +#define silk_NUM_TIMERS_MAX 50 +/* max length of name tags in TIC(..), TOC(..) */ +#define silk_NUM_TIMERS_MAX_TAG_LEN 30 + +extern int silk_Timer_nTimers; +extern int silk_Timer_depth_ctr; +extern char silk_Timer_tags[silk_NUM_TIMERS_MAX][silk_NUM_TIMERS_MAX_TAG_LEN]; +#ifdef _WIN32 +extern LARGE_INTEGER silk_Timer_start[silk_NUM_TIMERS_MAX]; +#else +extern unsigned long silk_Timer_start[silk_NUM_TIMERS_MAX]; +#endif +extern unsigned int silk_Timer_cnt[silk_NUM_TIMERS_MAX]; +extern opus_int64 silk_Timer_sum[silk_NUM_TIMERS_MAX]; +extern opus_int64 silk_Timer_max[silk_NUM_TIMERS_MAX]; +extern opus_int64 silk_Timer_min[silk_NUM_TIMERS_MAX]; +extern opus_int64 silk_Timer_depth[silk_NUM_TIMERS_MAX]; + +/* WARNING: TIC()/TOC can measure only up to 0.1 seconds at a time */ +#ifdef _WIN32 +#define TIC(TAG_NAME) { \ + static int init = 0; \ + static int ID = -1; \ + if( init == 0 ) \ + { \ + int k; \ + init = 1; \ + for( k = 0; k < silk_Timer_nTimers; k++ ) { \ + if( strcmp(silk_Timer_tags[k], #TAG_NAME) == 0 ) { \ + ID = k; \ + break; \ + } \ + } \ + if (ID == -1) { \ + ID = silk_Timer_nTimers; \ + silk_Timer_nTimers++; \ + silk_Timer_depth[ID] = silk_Timer_depth_ctr; \ + strcpy(silk_Timer_tags[ID], #TAG_NAME); \ + silk_Timer_cnt[ID] = 0; \ + silk_Timer_sum[ID] = 0; \ + silk_Timer_min[ID] = 0xFFFFFFFF; \ + silk_Timer_max[ID] = 0; \ + } \ + } \ + silk_Timer_depth_ctr++; \ + QueryPerformanceCounter(&silk_Timer_start[ID]); \ +} +#else +#define TIC(TAG_NAME) { \ + static int init = 0; \ + static int ID = -1; \ + if( init == 0 ) \ + { \ + int k; \ + init = 1; \ + for( k = 0; k < silk_Timer_nTimers; k++ ) { \ + if( strcmp(silk_Timer_tags[k], #TAG_NAME) == 0 ) { \ + ID = k; \ + break; \ + } \ + } \ + if (ID == -1) { \ + ID = silk_Timer_nTimers; \ + silk_Timer_nTimers++; \ + silk_Timer_depth[ID] = silk_Timer_depth_ctr; \ + strcpy(silk_Timer_tags[ID], #TAG_NAME); \ + silk_Timer_cnt[ID] = 0; \ + silk_Timer_sum[ID] = 0; \ + silk_Timer_min[ID] = 0xFFFFFFFF; \ + silk_Timer_max[ID] = 0; \ + } \ + } \ + silk_Timer_depth_ctr++; \ + silk_Timer_start[ID] = GetHighResolutionTime(); \ +} +#endif + +#ifdef _WIN32 +#define TOC(TAG_NAME) { \ + LARGE_INTEGER lpPerformanceCount; \ + static int init = 0; \ + static int ID = 0; \ + if( init == 0 ) \ + { \ + int k; \ + init = 1; \ + for( k = 0; k < silk_Timer_nTimers; k++ ) { \ + if( strcmp(silk_Timer_tags[k], #TAG_NAME) == 0 ) { \ + ID = k; \ + break; \ + } \ + } \ + } \ + QueryPerformanceCounter(&lpPerformanceCount); \ + lpPerformanceCount.QuadPart -= silk_Timer_start[ID].QuadPart; \ + if((lpPerformanceCount.QuadPart < 100000000) && \ + (lpPerformanceCount.QuadPart >= 0)) { \ + silk_Timer_cnt[ID]++; \ + silk_Timer_sum[ID] += lpPerformanceCount.QuadPart; \ + if( lpPerformanceCount.QuadPart > silk_Timer_max[ID] ) \ + silk_Timer_max[ID] = lpPerformanceCount.QuadPart; \ + if( lpPerformanceCount.QuadPart < silk_Timer_min[ID] ) \ + silk_Timer_min[ID] = lpPerformanceCount.QuadPart; \ + } \ + silk_Timer_depth_ctr--; \ +} +#else +#define TOC(TAG_NAME) { \ + unsigned long endTime; \ + static int init = 0; \ + static int ID = 0; \ + if( init == 0 ) \ + { \ + int k; \ + init = 1; \ + for( k = 0; k < silk_Timer_nTimers; k++ ) { \ + if( strcmp(silk_Timer_tags[k], #TAG_NAME) == 0 ) { \ + ID = k; \ + break; \ + } \ + } \ + } \ + endTime = GetHighResolutionTime(); \ + endTime -= silk_Timer_start[ID]; \ + if((endTime < 100000000) && \ + (endTime >= 0)) { \ + silk_Timer_cnt[ID]++; \ + silk_Timer_sum[ID] += endTime; \ + if( endTime > silk_Timer_max[ID] ) \ + silk_Timer_max[ID] = endTime; \ + if( endTime < silk_Timer_min[ID] ) \ + silk_Timer_min[ID] = endTime; \ + } \ + silk_Timer_depth_ctr--; \ +} +#endif + +#else /* SILK_TIC_TOC */ + +/* define macros as empty strings */ +#define TIC(TAG_NAME) +#define TOC(TAG_NAME) +#define silk_TimerSave(FILE_NAME) + +#endif /* SILK_TIC_TOC */ + + +#if SILK_DEBUG +/************************************/ +/* write data to file for debugging */ +/************************************/ +/* Example: DEBUG_STORE_DATA(testfile.pcm, &RIN[0], 160*sizeof(opus_int16)); */ + +#define silk_NUM_STORES_MAX 100 +extern FILE *silk_debug_store_fp[ silk_NUM_STORES_MAX ]; +extern int silk_debug_store_count; + +/* Faster way of storing the data */ +#define DEBUG_STORE_DATA( FILE_NAME, DATA_PTR, N_BYTES ) { \ + static opus_int init = 0, cnt = 0; \ + static FILE **fp; \ + if (init == 0) { \ + init = 1; \ + cnt = silk_debug_store_count++; \ + silk_debug_store_fp[ cnt ] = fopen(#FILE_NAME, "wb"); \ + } \ + fwrite((DATA_PTR), (N_BYTES), 1, silk_debug_store_fp[ cnt ]); \ +} + +/* Call this at the end of main() */ +#define SILK_DEBUG_STORE_CLOSE_FILES { \ + opus_int i; \ + for( i = 0; i < silk_debug_store_count; i++ ) { \ + fclose( silk_debug_store_fp[ i ] ); \ + } \ +} + +#else /* SILK_DEBUG */ + +/* define macros as empty strings */ +#define DEBUG_STORE_DATA(FILE_NAME, DATA_PTR, N_BYTES) +#define SILK_DEBUG_STORE_CLOSE_FILES + +#endif /* SILK_DEBUG */ + +#ifdef __cplusplus +} +#endif + +#endif /* SILK_DEBUG_H */ diff --git a/lib/opus/silk/dec_API.c b/lib/opus/silk/dec_API.c new file mode 100644 index 00000000..7d5ca7fb --- /dev/null +++ b/lib/opus/silk/dec_API.c @@ -0,0 +1,419 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#include "API.h" +#include "main.h" +#include "stack_alloc.h" +#include "os_support.h" + +/************************/ +/* Decoder Super Struct */ +/************************/ +typedef struct { + silk_decoder_state channel_state[ DECODER_NUM_CHANNELS ]; + stereo_dec_state sStereo; + opus_int nChannelsAPI; + opus_int nChannelsInternal; + opus_int prev_decode_only_middle; +} silk_decoder; + +/*********************/ +/* Decoder functions */ +/*********************/ + +opus_int silk_Get_Decoder_Size( /* O Returns error code */ + opus_int *decSizeBytes /* O Number of bytes in SILK decoder state */ +) +{ + opus_int ret = SILK_NO_ERROR; + + *decSizeBytes = sizeof( silk_decoder ); + + return ret; +} + +/* Reset decoder state */ +opus_int silk_InitDecoder( /* O Returns error code */ + void *decState /* I/O State */ +) +{ + opus_int n, ret = SILK_NO_ERROR; + silk_decoder_state *channel_state = ((silk_decoder *)decState)->channel_state; + + for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { + ret = silk_init_decoder( &channel_state[ n ] ); + } + silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo)); + /* Not strictly needed, but it's cleaner that way */ + ((silk_decoder *)decState)->prev_decode_only_middle = 0; + + return ret; +} + +/* Decode a frame */ +opus_int silk_Decode( /* O Returns error code */ + void* decState, /* I/O State */ + silk_DecControlStruct* decControl, /* I/O Control Structure */ + opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int newPacketFlag, /* I Indicates first decoder call for this packet */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 *samplesOut, /* O Decoded output speech vector */ + opus_int32 *nSamplesOut, /* O Number of samples decoded */ + int arch /* I Run-time architecture */ +) +{ + opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR; + opus_int32 nSamplesOutDec, LBRR_symbol; + opus_int16 *samplesOut1_tmp[ 2 ]; + VARDECL( opus_int16, samplesOut1_tmp_storage1 ); + VARDECL( opus_int16, samplesOut1_tmp_storage2 ); + VARDECL( opus_int16, samplesOut2_tmp ); + opus_int32 MS_pred_Q13[ 2 ] = { 0 }; + opus_int16 *resample_out_ptr; + silk_decoder *psDec = ( silk_decoder * )decState; + silk_decoder_state *channel_state = psDec->channel_state; + opus_int has_side; + opus_int stereo_to_mono; + int delay_stack_alloc; + SAVE_STACK; + + celt_assert( decControl->nChannelsInternal == 1 || decControl->nChannelsInternal == 2 ); + + /**********************************/ + /* Test if first frame in payload */ + /**********************************/ + if( newPacketFlag ) { + for( n = 0; n < decControl->nChannelsInternal; n++ ) { + channel_state[ n ].nFramesDecoded = 0; /* Used to count frames in packet */ + } + } + + /* If Mono -> Stereo transition in bitstream: init state of second channel */ + if( decControl->nChannelsInternal > psDec->nChannelsInternal ) { + ret += silk_init_decoder( &channel_state[ 1 ] ); + } + + stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 && + ( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz ); + + if( channel_state[ 0 ].nFramesDecoded == 0 ) { + for( n = 0; n < decControl->nChannelsInternal; n++ ) { + opus_int fs_kHz_dec; + if( decControl->payloadSize_ms == 0 ) { + /* Assuming packet loss, use 10 ms */ + channel_state[ n ].nFramesPerPacket = 1; + channel_state[ n ].nb_subfr = 2; + } else if( decControl->payloadSize_ms == 10 ) { + channel_state[ n ].nFramesPerPacket = 1; + channel_state[ n ].nb_subfr = 2; + } else if( decControl->payloadSize_ms == 20 ) { + channel_state[ n ].nFramesPerPacket = 1; + channel_state[ n ].nb_subfr = 4; + } else if( decControl->payloadSize_ms == 40 ) { + channel_state[ n ].nFramesPerPacket = 2; + channel_state[ n ].nb_subfr = 4; + } else if( decControl->payloadSize_ms == 60 ) { + channel_state[ n ].nFramesPerPacket = 3; + channel_state[ n ].nb_subfr = 4; + } else { + celt_assert( 0 ); + RESTORE_STACK; + return SILK_DEC_INVALID_FRAME_SIZE; + } + fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1; + if( fs_kHz_dec != 8 && fs_kHz_dec != 12 && fs_kHz_dec != 16 ) { + celt_assert( 0 ); + RESTORE_STACK; + return SILK_DEC_INVALID_SAMPLING_FREQUENCY; + } + ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate ); + } + } + + if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 || psDec->nChannelsInternal == 1 ) ) { + silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) ); + silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) ); + silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) ); + } + psDec->nChannelsAPI = decControl->nChannelsAPI; + psDec->nChannelsInternal = decControl->nChannelsInternal; + + if( decControl->API_sampleRate > (opus_int32)MAX_API_FS_KHZ * 1000 || decControl->API_sampleRate < 8000 ) { + ret = SILK_DEC_INVALID_SAMPLING_FREQUENCY; + RESTORE_STACK; + return( ret ); + } + + if( lostFlag != FLAG_PACKET_LOST && channel_state[ 0 ].nFramesDecoded == 0 ) { + /* First decoder call for this payload */ + /* Decode VAD flags and LBRR flag */ + for( n = 0; n < decControl->nChannelsInternal; n++ ) { + for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { + channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1); + } + channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1); + } + /* Decode LBRR flags */ + for( n = 0; n < decControl->nChannelsInternal; n++ ) { + silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) ); + if( channel_state[ n ].LBRR_flag ) { + if( channel_state[ n ].nFramesPerPacket == 1 ) { + channel_state[ n ].LBRR_flags[ 0 ] = 1; + } else { + LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1; + for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { + channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1; + } + } + } + } + + if( lostFlag == FLAG_DECODE_NORMAL ) { + /* Regular decoding: skip all LBRR data */ + for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) { + for( n = 0; n < decControl->nChannelsInternal; n++ ) { + if( channel_state[ n ].LBRR_flags[ i ] ) { + opus_int16 pulses[ MAX_FRAME_LENGTH ]; + opus_int condCoding; + + if( decControl->nChannelsInternal == 2 && n == 0 ) { + silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); + if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) { + silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); + } + } + /* Use conditional coding if previous frame available */ + if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) { + condCoding = CODE_CONDITIONALLY; + } else { + condCoding = CODE_INDEPENDENTLY; + } + silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding ); + silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType, + channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length ); + } + } + } + } + } + + /* Get MS predictor index */ + if( decControl->nChannelsInternal == 2 ) { + if( lostFlag == FLAG_DECODE_NORMAL || + ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) ) + { + silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); + /* For LBRR data, decode mid-only flag only if side-channel's LBRR flag is false */ + if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) || + ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) ) + { + silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); + } else { + decode_only_middle = 0; + } + } else { + for( n = 0; n < 2; n++ ) { + MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ]; + } + } + } + + /* Reset side channel decoder prediction memory for first frame with side coding */ + if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) { + silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) ); + silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) ); + psDec->channel_state[ 1 ].lagPrev = 100; + psDec->channel_state[ 1 ].LastGainIndex = 10; + psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY; + psDec->channel_state[ 1 ].first_frame_after_reset = 1; + } + + /* Check if the temp buffer fits into the output PCM buffer. If it fits, + we can delay allocating the temp buffer until after the SILK peak stack + usage. We need to use a < and not a <= because of the two extra samples. */ + delay_stack_alloc = decControl->internalSampleRate*decControl->nChannelsInternal + < decControl->API_sampleRate*decControl->nChannelsAPI; + ALLOC( samplesOut1_tmp_storage1, delay_stack_alloc ? ALLOC_NONE + : decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ), + opus_int16 ); + if ( delay_stack_alloc ) + { + samplesOut1_tmp[ 0 ] = samplesOut; + samplesOut1_tmp[ 1 ] = samplesOut + channel_state[ 0 ].frame_length + 2; + } else { + samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1; + samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2; + } + + if( lostFlag == FLAG_DECODE_NORMAL ) { + has_side = !decode_only_middle; + } else { + has_side = !psDec->prev_decode_only_middle + || (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 ); + } + /* Call decoder for one frame */ + for( n = 0; n < decControl->nChannelsInternal; n++ ) { + if( n == 0 || has_side ) { + opus_int FrameIndex; + opus_int condCoding; + + FrameIndex = channel_state[ 0 ].nFramesDecoded - n; + /* Use independent coding if no previous frame available */ + if( FrameIndex <= 0 ) { + condCoding = CODE_INDEPENDENTLY; + } else if( lostFlag == FLAG_DECODE_LBRR ) { + condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY; + } else if( n > 0 && psDec->prev_decode_only_middle ) { + /* If we skipped a side frame in this packet, we don't + need LTP scaling; the LTP state is well-defined. */ + condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; + } else { + condCoding = CODE_CONDITIONALLY; + } + ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding, arch); + } else { + silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) ); + } + channel_state[ n ].nFramesDecoded++; + } + + if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) { + /* Convert Mid/Side to Left/Right */ + silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec ); + } else { + /* Buffering */ + silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) ); + silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) ); + } + + /* Number of output samples */ + *nSamplesOut = silk_DIV32( nSamplesOutDec * decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) ); + + /* Set up pointers to temp buffers */ + ALLOC( samplesOut2_tmp, + decControl->nChannelsAPI == 2 ? *nSamplesOut : ALLOC_NONE, opus_int16 ); + if( decControl->nChannelsAPI == 2 ) { + resample_out_ptr = samplesOut2_tmp; + } else { + resample_out_ptr = samplesOut; + } + + ALLOC( samplesOut1_tmp_storage2, delay_stack_alloc + ? decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ) + : ALLOC_NONE, + opus_int16 ); + if ( delay_stack_alloc ) { + OPUS_COPY(samplesOut1_tmp_storage2, samplesOut, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2)); + samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage2; + samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage2 + channel_state[ 0 ].frame_length + 2; + } + for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) { + + /* Resample decoded signal to API_sampleRate */ + ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); + + /* Interleave if stereo output and stereo stream */ + if( decControl->nChannelsAPI == 2 ) { + for( i = 0; i < *nSamplesOut; i++ ) { + samplesOut[ n + 2 * i ] = resample_out_ptr[ i ]; + } + } + } + + /* Create two channel output from mono stream */ + if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) { + if ( stereo_to_mono ){ + /* Resample right channel for newly collapsed stereo just in case + we weren't doing collapsing when switching to mono */ + ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec ); + + for( i = 0; i < *nSamplesOut; i++ ) { + samplesOut[ 1 + 2 * i ] = resample_out_ptr[ i ]; + } + } else { + for( i = 0; i < *nSamplesOut; i++ ) { + samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ]; + } + } + } + + /* Export pitch lag, measured at 48 kHz sampling rate */ + if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) { + int mult_tab[ 3 ] = { 6, 4, 3 }; + decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ]; + } else { + decControl->prevPitchLag = 0; + } + + if( lostFlag == FLAG_PACKET_LOST ) { + /* On packet loss, remove the gain clamping to prevent having the energy "bounce back" + if we lose packets when the energy is going down */ + for ( i = 0; i < psDec->nChannelsInternal; i++ ) + psDec->channel_state[ i ].LastGainIndex = 10; + } else { + psDec->prev_decode_only_middle = decode_only_middle; + } + RESTORE_STACK; + return ret; +} + +#if 0 +/* Getting table of contents for a packet */ +opus_int silk_get_TOC( + const opus_uint8 *payload, /* I Payload data */ + const opus_int nBytesIn, /* I Number of input bytes */ + const opus_int nFramesPerPayload, /* I Number of SILK frames per payload */ + silk_TOC_struct *Silk_TOC /* O Type of content */ +) +{ + opus_int i, flags, ret = SILK_NO_ERROR; + + if( nBytesIn < 1 ) { + return -1; + } + if( nFramesPerPayload < 0 || nFramesPerPayload > 3 ) { + return -1; + } + + silk_memset( Silk_TOC, 0, sizeof( *Silk_TOC ) ); + + /* For stereo, extract the flags for the mid channel */ + flags = silk_RSHIFT( payload[ 0 ], 7 - nFramesPerPayload ) & ( silk_LSHIFT( 1, nFramesPerPayload + 1 ) - 1 ); + + Silk_TOC->inbandFECFlag = flags & 1; + for( i = nFramesPerPayload - 1; i >= 0 ; i-- ) { + flags = silk_RSHIFT( flags, 1 ); + Silk_TOC->VADFlags[ i ] = flags & 1; + Silk_TOC->VADFlag |= flags & 1; + } + + return ret; +} +#endif diff --git a/lib/opus/silk/decode_core.c b/lib/opus/silk/decode_core.c new file mode 100644 index 00000000..1c352a65 --- /dev/null +++ b/lib/opus/silk/decode_core.c @@ -0,0 +1,237 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" + +/**********************************************************/ +/* Core decoder. Performs inverse NSQ operation LTP + LPC */ +/**********************************************************/ +void silk_decode_core( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I Decoder control */ + opus_int16 xq[], /* O Decoded speech */ + const opus_int16 pulses[ MAX_FRAME_LENGTH ], /* I Pulse signal */ + int arch /* I Run-time architecture */ +) +{ + opus_int i, k, lag = 0, start_idx, sLTP_buf_idx, NLSF_interpolation_flag, signalType; + opus_int16 *A_Q12, *B_Q14, *pxq, A_Q12_tmp[ MAX_LPC_ORDER ]; + VARDECL( opus_int16, sLTP ); + VARDECL( opus_int32, sLTP_Q15 ); + opus_int32 LTP_pred_Q13, LPC_pred_Q10, Gain_Q10, inv_gain_Q31, gain_adj_Q16, rand_seed, offset_Q10; + opus_int32 *pred_lag_ptr, *pexc_Q14, *pres_Q14; + VARDECL( opus_int32, res_Q14 ); + VARDECL( opus_int32, sLPC_Q14 ); + SAVE_STACK; + + silk_assert( psDec->prev_gain_Q16 != 0 ); + + ALLOC( sLTP, psDec->ltp_mem_length, opus_int16 ); + ALLOC( sLTP_Q15, psDec->ltp_mem_length + psDec->frame_length, opus_int32 ); + ALLOC( res_Q14, psDec->subfr_length, opus_int32 ); + ALLOC( sLPC_Q14, psDec->subfr_length + MAX_LPC_ORDER, opus_int32 ); + + offset_Q10 = silk_Quantization_Offsets_Q10[ psDec->indices.signalType >> 1 ][ psDec->indices.quantOffsetType ]; + + if( psDec->indices.NLSFInterpCoef_Q2 < 1 << 2 ) { + NLSF_interpolation_flag = 1; + } else { + NLSF_interpolation_flag = 0; + } + + /* Decode excitation */ + rand_seed = psDec->indices.Seed; + for( i = 0; i < psDec->frame_length; i++ ) { + rand_seed = silk_RAND( rand_seed ); + psDec->exc_Q14[ i ] = silk_LSHIFT( (opus_int32)pulses[ i ], 14 ); + if( psDec->exc_Q14[ i ] > 0 ) { + psDec->exc_Q14[ i ] -= QUANT_LEVEL_ADJUST_Q10 << 4; + } else + if( psDec->exc_Q14[ i ] < 0 ) { + psDec->exc_Q14[ i ] += QUANT_LEVEL_ADJUST_Q10 << 4; + } + psDec->exc_Q14[ i ] += offset_Q10 << 4; + if( rand_seed < 0 ) { + psDec->exc_Q14[ i ] = -psDec->exc_Q14[ i ]; + } + + rand_seed = silk_ADD32_ovflw( rand_seed, pulses[ i ] ); + } + + /* Copy LPC state */ + silk_memcpy( sLPC_Q14, psDec->sLPC_Q14_buf, MAX_LPC_ORDER * sizeof( opus_int32 ) ); + + pexc_Q14 = psDec->exc_Q14; + pxq = xq; + sLTP_buf_idx = psDec->ltp_mem_length; + /* Loop over subframes */ + for( k = 0; k < psDec->nb_subfr; k++ ) { + pres_Q14 = res_Q14; + A_Q12 = psDecCtrl->PredCoef_Q12[ k >> 1 ]; + + /* Preload LPC coeficients to array on stack. Gives small performance gain */ + silk_memcpy( A_Q12_tmp, A_Q12, psDec->LPC_order * sizeof( opus_int16 ) ); + B_Q14 = &psDecCtrl->LTPCoef_Q14[ k * LTP_ORDER ]; + signalType = psDec->indices.signalType; + + Gain_Q10 = silk_RSHIFT( psDecCtrl->Gains_Q16[ k ], 6 ); + inv_gain_Q31 = silk_INVERSE32_varQ( psDecCtrl->Gains_Q16[ k ], 47 ); + + /* Calculate gain adjustment factor */ + if( psDecCtrl->Gains_Q16[ k ] != psDec->prev_gain_Q16 ) { + gain_adj_Q16 = silk_DIV32_varQ( psDec->prev_gain_Q16, psDecCtrl->Gains_Q16[ k ], 16 ); + + /* Scale short term state */ + for( i = 0; i < MAX_LPC_ORDER; i++ ) { + sLPC_Q14[ i ] = silk_SMULWW( gain_adj_Q16, sLPC_Q14[ i ] ); + } + } else { + gain_adj_Q16 = (opus_int32)1 << 16; + } + + /* Save inv_gain */ + silk_assert( inv_gain_Q31 != 0 ); + psDec->prev_gain_Q16 = psDecCtrl->Gains_Q16[ k ]; + + /* Avoid abrupt transition from voiced PLC to unvoiced normal decoding */ + if( psDec->lossCnt && psDec->prevSignalType == TYPE_VOICED && + psDec->indices.signalType != TYPE_VOICED && k < MAX_NB_SUBFR/2 ) { + + silk_memset( B_Q14, 0, LTP_ORDER * sizeof( opus_int16 ) ); + B_Q14[ LTP_ORDER/2 ] = SILK_FIX_CONST( 0.25, 14 ); + + signalType = TYPE_VOICED; + psDecCtrl->pitchL[ k ] = psDec->lagPrev; + } + + if( signalType == TYPE_VOICED ) { + /* Voiced */ + lag = psDecCtrl->pitchL[ k ]; + + /* Re-whitening */ + if( k == 0 || ( k == 2 && NLSF_interpolation_flag ) ) { + /* Rewhiten with new A coefs */ + start_idx = psDec->ltp_mem_length - lag - psDec->LPC_order - LTP_ORDER / 2; + celt_assert( start_idx > 0 ); + + if( k == 2 ) { + silk_memcpy( &psDec->outBuf[ psDec->ltp_mem_length ], xq, 2 * psDec->subfr_length * sizeof( opus_int16 ) ); + } + + silk_LPC_analysis_filter( &sLTP[ start_idx ], &psDec->outBuf[ start_idx + k * psDec->subfr_length ], + A_Q12, psDec->ltp_mem_length - start_idx, psDec->LPC_order, arch ); + + /* After rewhitening the LTP state is unscaled */ + if( k == 0 ) { + /* Do LTP downscaling to reduce inter-packet dependency */ + inv_gain_Q31 = silk_LSHIFT( silk_SMULWB( inv_gain_Q31, psDecCtrl->LTP_scale_Q14 ), 2 ); + } + for( i = 0; i < lag + LTP_ORDER/2; i++ ) { + sLTP_Q15[ sLTP_buf_idx - i - 1 ] = silk_SMULWB( inv_gain_Q31, sLTP[ psDec->ltp_mem_length - i - 1 ] ); + } + } else { + /* Update LTP state when Gain changes */ + if( gain_adj_Q16 != (opus_int32)1 << 16 ) { + for( i = 0; i < lag + LTP_ORDER/2; i++ ) { + sLTP_Q15[ sLTP_buf_idx - i - 1 ] = silk_SMULWW( gain_adj_Q16, sLTP_Q15[ sLTP_buf_idx - i - 1 ] ); + } + } + } + } + + /* Long-term prediction */ + if( signalType == TYPE_VOICED ) { + /* Set up pointer */ + pred_lag_ptr = &sLTP_Q15[ sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + for( i = 0; i < psDec->subfr_length; i++ ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LTP_pred_Q13 = 2; + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ 0 ], B_Q14[ 0 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -1 ], B_Q14[ 1 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -2 ], B_Q14[ 2 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -3 ], B_Q14[ 3 ] ); + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -4 ], B_Q14[ 4 ] ); + pred_lag_ptr++; + + /* Generate LPC excitation */ + pres_Q14[ i ] = silk_ADD_LSHIFT32( pexc_Q14[ i ], LTP_pred_Q13, 1 ); + + /* Update states */ + sLTP_Q15[ sLTP_buf_idx ] = silk_LSHIFT( pres_Q14[ i ], 1 ); + sLTP_buf_idx++; + } + } else { + pres_Q14 = pexc_Q14; + } + + for( i = 0; i < psDec->subfr_length; i++ ) { + /* Short-term prediction */ + celt_assert( psDec->LPC_order == 10 || psDec->LPC_order == 16 ); + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LPC_pred_Q10 = silk_RSHIFT( psDec->LPC_order, 1 ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 1 ], A_Q12_tmp[ 0 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 2 ], A_Q12_tmp[ 1 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 3 ], A_Q12_tmp[ 2 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 4 ], A_Q12_tmp[ 3 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 5 ], A_Q12_tmp[ 4 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 6 ], A_Q12_tmp[ 5 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 7 ], A_Q12_tmp[ 6 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 8 ], A_Q12_tmp[ 7 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 9 ], A_Q12_tmp[ 8 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 10 ], A_Q12_tmp[ 9 ] ); + if( psDec->LPC_order == 16 ) { + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 11 ], A_Q12_tmp[ 10 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 12 ], A_Q12_tmp[ 11 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 13 ], A_Q12_tmp[ 12 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 14 ], A_Q12_tmp[ 13 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 15 ], A_Q12_tmp[ 14 ] ); + LPC_pred_Q10 = silk_SMLAWB( LPC_pred_Q10, sLPC_Q14[ MAX_LPC_ORDER + i - 16 ], A_Q12_tmp[ 15 ] ); + } + + /* Add prediction to LPC excitation */ + sLPC_Q14[ MAX_LPC_ORDER + i ] = silk_ADD_SAT32( pres_Q14[ i ], silk_LSHIFT_SAT32( LPC_pred_Q10, 4 ) ); + + /* Scale with gain */ + pxq[ i ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( sLPC_Q14[ MAX_LPC_ORDER + i ], Gain_Q10 ), 8 ) ); + } + + /* Update LPC filter state */ + silk_memcpy( sLPC_Q14, &sLPC_Q14[ psDec->subfr_length ], MAX_LPC_ORDER * sizeof( opus_int32 ) ); + pexc_Q14 += psDec->subfr_length; + pxq += psDec->subfr_length; + } + + /* Save LPC state */ + silk_memcpy( psDec->sLPC_Q14_buf, sLPC_Q14, MAX_LPC_ORDER * sizeof( opus_int32 ) ); + RESTORE_STACK; +} diff --git a/lib/opus/silk/decode_frame.c b/lib/opus/silk/decode_frame.c new file mode 100644 index 00000000..4f36f854 --- /dev/null +++ b/lib/opus/silk/decode_frame.c @@ -0,0 +1,129 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" +#include "PLC.h" + +/****************/ +/* Decode frame */ +/****************/ +opus_int silk_decode_frame( + silk_decoder_state *psDec, /* I/O Pointer to Silk decoder state */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 pOut[], /* O Pointer to output speech frame */ + opus_int32 *pN, /* O Pointer to size of output frame */ + opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int condCoding, /* I The type of conditional coding to use */ + int arch /* I Run-time architecture */ +) +{ + VARDECL( silk_decoder_control, psDecCtrl ); + opus_int L, mv_len, ret = 0; + SAVE_STACK; + + L = psDec->frame_length; + ALLOC( psDecCtrl, 1, silk_decoder_control ); + psDecCtrl->LTP_scale_Q14 = 0; + + /* Safety checks */ + celt_assert( L > 0 && L <= MAX_FRAME_LENGTH ); + + if( lostFlag == FLAG_DECODE_NORMAL || + ( lostFlag == FLAG_DECODE_LBRR && psDec->LBRR_flags[ psDec->nFramesDecoded ] == 1 ) ) + { + VARDECL( opus_int16, pulses ); + ALLOC( pulses, (L + SHELL_CODEC_FRAME_LENGTH - 1) & + ~(SHELL_CODEC_FRAME_LENGTH - 1), opus_int16 ); + /*********************************************/ + /* Decode quantization indices of side info */ + /*********************************************/ + silk_decode_indices( psDec, psRangeDec, psDec->nFramesDecoded, lostFlag, condCoding ); + + /*********************************************/ + /* Decode quantization indices of excitation */ + /*********************************************/ + silk_decode_pulses( psRangeDec, pulses, psDec->indices.signalType, + psDec->indices.quantOffsetType, psDec->frame_length ); + + /********************************************/ + /* Decode parameters and pulse signal */ + /********************************************/ + silk_decode_parameters( psDec, psDecCtrl, condCoding ); + + /********************************************************/ + /* Run inverse NSQ */ + /********************************************************/ + silk_decode_core( psDec, psDecCtrl, pOut, pulses, arch ); + + /********************************************************/ + /* Update PLC state */ + /********************************************************/ + silk_PLC( psDec, psDecCtrl, pOut, 0, arch ); + + psDec->lossCnt = 0; + psDec->prevSignalType = psDec->indices.signalType; + celt_assert( psDec->prevSignalType >= 0 && psDec->prevSignalType <= 2 ); + + /* A frame has been decoded without errors */ + psDec->first_frame_after_reset = 0; + } else { + /* Handle packet loss by extrapolation */ + silk_PLC( psDec, psDecCtrl, pOut, 1, arch ); + } + + /*************************/ + /* Update output buffer. */ + /*************************/ + celt_assert( psDec->ltp_mem_length >= psDec->frame_length ); + mv_len = psDec->ltp_mem_length - psDec->frame_length; + silk_memmove( psDec->outBuf, &psDec->outBuf[ psDec->frame_length ], mv_len * sizeof(opus_int16) ); + silk_memcpy( &psDec->outBuf[ mv_len ], pOut, psDec->frame_length * sizeof( opus_int16 ) ); + + /************************************************/ + /* Comfort noise generation / estimation */ + /************************************************/ + silk_CNG( psDec, psDecCtrl, pOut, L ); + + /****************************************************************/ + /* Ensure smooth connection of extrapolated and good frames */ + /****************************************************************/ + silk_PLC_glue_frames( psDec, pOut, L ); + + /* Update some decoder state variables */ + psDec->lagPrev = psDecCtrl->pitchL[ psDec->nb_subfr - 1 ]; + + /* Set output frame length */ + *pN = L; + + RESTORE_STACK; + return ret; +} diff --git a/lib/opus/silk/decode_indices.c b/lib/opus/silk/decode_indices.c new file mode 100644 index 00000000..0bb4a997 --- /dev/null +++ b/lib/opus/silk/decode_indices.c @@ -0,0 +1,151 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Decode side-information parameters from payload */ +void silk_decode_indices( + silk_decoder_state *psDec, /* I/O State */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int FrameIndex, /* I Frame number */ + opus_int decode_LBRR, /* I Flag indicating LBRR data is being decoded */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + opus_int i, k, Ix; + opus_int decode_absolute_lagIndex, delta_lagIndex; + opus_int16 ec_ix[ MAX_LPC_ORDER ]; + opus_uint8 pred_Q8[ MAX_LPC_ORDER ]; + + /*******************************************/ + /* Decode signal type and quantizer offset */ + /*******************************************/ + if( decode_LBRR || psDec->VAD_flags[ FrameIndex ] ) { + Ix = ec_dec_icdf( psRangeDec, silk_type_offset_VAD_iCDF, 8 ) + 2; + } else { + Ix = ec_dec_icdf( psRangeDec, silk_type_offset_no_VAD_iCDF, 8 ); + } + psDec->indices.signalType = (opus_int8)silk_RSHIFT( Ix, 1 ); + psDec->indices.quantOffsetType = (opus_int8)( Ix & 1 ); + + /****************/ + /* Decode gains */ + /****************/ + /* First subframe */ + if( condCoding == CODE_CONDITIONALLY ) { + /* Conditional coding */ + psDec->indices.GainsIndices[ 0 ] = (opus_int8)ec_dec_icdf( psRangeDec, silk_delta_gain_iCDF, 8 ); + } else { + /* Independent coding, in two stages: MSB bits followed by 3 LSBs */ + psDec->indices.GainsIndices[ 0 ] = (opus_int8)silk_LSHIFT( ec_dec_icdf( psRangeDec, silk_gain_iCDF[ psDec->indices.signalType ], 8 ), 3 ); + psDec->indices.GainsIndices[ 0 ] += (opus_int8)ec_dec_icdf( psRangeDec, silk_uniform8_iCDF, 8 ); + } + + /* Remaining subframes */ + for( i = 1; i < psDec->nb_subfr; i++ ) { + psDec->indices.GainsIndices[ i ] = (opus_int8)ec_dec_icdf( psRangeDec, silk_delta_gain_iCDF, 8 ); + } + + /**********************/ + /* Decode LSF Indices */ + /**********************/ + psDec->indices.NLSFIndices[ 0 ] = (opus_int8)ec_dec_icdf( psRangeDec, &psDec->psNLSF_CB->CB1_iCDF[ ( psDec->indices.signalType >> 1 ) * psDec->psNLSF_CB->nVectors ], 8 ); + silk_NLSF_unpack( ec_ix, pred_Q8, psDec->psNLSF_CB, psDec->indices.NLSFIndices[ 0 ] ); + celt_assert( psDec->psNLSF_CB->order == psDec->LPC_order ); + for( i = 0; i < psDec->psNLSF_CB->order; i++ ) { + Ix = ec_dec_icdf( psRangeDec, &psDec->psNLSF_CB->ec_iCDF[ ec_ix[ i ] ], 8 ); + if( Ix == 0 ) { + Ix -= ec_dec_icdf( psRangeDec, silk_NLSF_EXT_iCDF, 8 ); + } else if( Ix == 2 * NLSF_QUANT_MAX_AMPLITUDE ) { + Ix += ec_dec_icdf( psRangeDec, silk_NLSF_EXT_iCDF, 8 ); + } + psDec->indices.NLSFIndices[ i+1 ] = (opus_int8)( Ix - NLSF_QUANT_MAX_AMPLITUDE ); + } + + /* Decode LSF interpolation factor */ + if( psDec->nb_subfr == MAX_NB_SUBFR ) { + psDec->indices.NLSFInterpCoef_Q2 = (opus_int8)ec_dec_icdf( psRangeDec, silk_NLSF_interpolation_factor_iCDF, 8 ); + } else { + psDec->indices.NLSFInterpCoef_Q2 = 4; + } + + if( psDec->indices.signalType == TYPE_VOICED ) + { + /*********************/ + /* Decode pitch lags */ + /*********************/ + /* Get lag index */ + decode_absolute_lagIndex = 1; + if( condCoding == CODE_CONDITIONALLY && psDec->ec_prevSignalType == TYPE_VOICED ) { + /* Decode Delta index */ + delta_lagIndex = (opus_int16)ec_dec_icdf( psRangeDec, silk_pitch_delta_iCDF, 8 ); + if( delta_lagIndex > 0 ) { + delta_lagIndex = delta_lagIndex - 9; + psDec->indices.lagIndex = (opus_int16)( psDec->ec_prevLagIndex + delta_lagIndex ); + decode_absolute_lagIndex = 0; + } + } + if( decode_absolute_lagIndex ) { + /* Absolute decoding */ + psDec->indices.lagIndex = (opus_int16)ec_dec_icdf( psRangeDec, silk_pitch_lag_iCDF, 8 ) * silk_RSHIFT( psDec->fs_kHz, 1 ); + psDec->indices.lagIndex += (opus_int16)ec_dec_icdf( psRangeDec, psDec->pitch_lag_low_bits_iCDF, 8 ); + } + psDec->ec_prevLagIndex = psDec->indices.lagIndex; + + /* Get countour index */ + psDec->indices.contourIndex = (opus_int8)ec_dec_icdf( psRangeDec, psDec->pitch_contour_iCDF, 8 ); + + /********************/ + /* Decode LTP gains */ + /********************/ + /* Decode PERIndex value */ + psDec->indices.PERIndex = (opus_int8)ec_dec_icdf( psRangeDec, silk_LTP_per_index_iCDF, 8 ); + + for( k = 0; k < psDec->nb_subfr; k++ ) { + psDec->indices.LTPIndex[ k ] = (opus_int8)ec_dec_icdf( psRangeDec, silk_LTP_gain_iCDF_ptrs[ psDec->indices.PERIndex ], 8 ); + } + + /**********************/ + /* Decode LTP scaling */ + /**********************/ + if( condCoding == CODE_INDEPENDENTLY ) { + psDec->indices.LTP_scaleIndex = (opus_int8)ec_dec_icdf( psRangeDec, silk_LTPscale_iCDF, 8 ); + } else { + psDec->indices.LTP_scaleIndex = 0; + } + } + psDec->ec_prevSignalType = psDec->indices.signalType; + + /***************/ + /* Decode seed */ + /***************/ + psDec->indices.Seed = (opus_int8)ec_dec_icdf( psRangeDec, silk_uniform4_iCDF, 8 ); +} diff --git a/lib/opus/silk/decode_parameters.c b/lib/opus/silk/decode_parameters.c new file mode 100644 index 00000000..a56a4098 --- /dev/null +++ b/lib/opus/silk/decode_parameters.c @@ -0,0 +1,115 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Decode parameters from payload */ +void silk_decode_parameters( + silk_decoder_state *psDec, /* I/O State */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + opus_int i, k, Ix; + opus_int16 pNLSF_Q15[ MAX_LPC_ORDER ], pNLSF0_Q15[ MAX_LPC_ORDER ]; + const opus_int8 *cbk_ptr_Q7; + + /* Dequant Gains */ + silk_gains_dequant( psDecCtrl->Gains_Q16, psDec->indices.GainsIndices, + &psDec->LastGainIndex, condCoding == CODE_CONDITIONALLY, psDec->nb_subfr ); + + /****************/ + /* Decode NLSFs */ + /****************/ + silk_NLSF_decode( pNLSF_Q15, psDec->indices.NLSFIndices, psDec->psNLSF_CB ); + + /* Convert NLSF parameters to AR prediction filter coefficients */ + silk_NLSF2A( psDecCtrl->PredCoef_Q12[ 1 ], pNLSF_Q15, psDec->LPC_order, psDec->arch ); + + /* If just reset, e.g., because internal Fs changed, do not allow interpolation */ + /* improves the case of packet loss in the first frame after a switch */ + if( psDec->first_frame_after_reset == 1 ) { + psDec->indices.NLSFInterpCoef_Q2 = 4; + } + + if( psDec->indices.NLSFInterpCoef_Q2 < 4 ) { + /* Calculation of the interpolated NLSF0 vector from the interpolation factor, */ + /* the previous NLSF1, and the current NLSF1 */ + for( i = 0; i < psDec->LPC_order; i++ ) { + pNLSF0_Q15[ i ] = psDec->prevNLSF_Q15[ i ] + silk_RSHIFT( silk_MUL( psDec->indices.NLSFInterpCoef_Q2, + pNLSF_Q15[ i ] - psDec->prevNLSF_Q15[ i ] ), 2 ); + } + + /* Convert NLSF parameters to AR prediction filter coefficients */ + silk_NLSF2A( psDecCtrl->PredCoef_Q12[ 0 ], pNLSF0_Q15, psDec->LPC_order, psDec->arch ); + } else { + /* Copy LPC coefficients for first half from second half */ + silk_memcpy( psDecCtrl->PredCoef_Q12[ 0 ], psDecCtrl->PredCoef_Q12[ 1 ], psDec->LPC_order * sizeof( opus_int16 ) ); + } + + silk_memcpy( psDec->prevNLSF_Q15, pNLSF_Q15, psDec->LPC_order * sizeof( opus_int16 ) ); + + /* After a packet loss do BWE of LPC coefs */ + if( psDec->lossCnt ) { + silk_bwexpander( psDecCtrl->PredCoef_Q12[ 0 ], psDec->LPC_order, BWE_AFTER_LOSS_Q16 ); + silk_bwexpander( psDecCtrl->PredCoef_Q12[ 1 ], psDec->LPC_order, BWE_AFTER_LOSS_Q16 ); + } + + if( psDec->indices.signalType == TYPE_VOICED ) { + /*********************/ + /* Decode pitch lags */ + /*********************/ + + /* Decode pitch values */ + silk_decode_pitch( psDec->indices.lagIndex, psDec->indices.contourIndex, psDecCtrl->pitchL, psDec->fs_kHz, psDec->nb_subfr ); + + /* Decode Codebook Index */ + cbk_ptr_Q7 = silk_LTP_vq_ptrs_Q7[ psDec->indices.PERIndex ]; /* set pointer to start of codebook */ + + for( k = 0; k < psDec->nb_subfr; k++ ) { + Ix = psDec->indices.LTPIndex[ k ]; + for( i = 0; i < LTP_ORDER; i++ ) { + psDecCtrl->LTPCoef_Q14[ k * LTP_ORDER + i ] = silk_LSHIFT( cbk_ptr_Q7[ Ix * LTP_ORDER + i ], 7 ); + } + } + + /**********************/ + /* Decode LTP scaling */ + /**********************/ + Ix = psDec->indices.LTP_scaleIndex; + psDecCtrl->LTP_scale_Q14 = silk_LTPScales_table_Q14[ Ix ]; + } else { + silk_memset( psDecCtrl->pitchL, 0, psDec->nb_subfr * sizeof( opus_int ) ); + silk_memset( psDecCtrl->LTPCoef_Q14, 0, LTP_ORDER * psDec->nb_subfr * sizeof( opus_int16 ) ); + psDec->indices.PERIndex = 0; + psDecCtrl->LTP_scale_Q14 = 0; + } +} diff --git a/lib/opus/silk/decode_pitch.c b/lib/opus/silk/decode_pitch.c new file mode 100644 index 00000000..fd1b6bf5 --- /dev/null +++ b/lib/opus/silk/decode_pitch.c @@ -0,0 +1,77 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/*********************************************************** +* Pitch analyser function +********************************************************** */ +#include "SigProc_FIX.h" +#include "pitch_est_defines.h" + +void silk_decode_pitch( + opus_int16 lagIndex, /* I */ + opus_int8 contourIndex, /* O */ + opus_int pitch_lags[], /* O 4 pitch values */ + const opus_int Fs_kHz, /* I sampling frequency (kHz) */ + const opus_int nb_subfr /* I number of sub frames */ +) +{ + opus_int lag, k, min_lag, max_lag, cbk_size; + const opus_int8 *Lag_CB_ptr; + + if( Fs_kHz == 8 ) { + if( nb_subfr == PE_MAX_NB_SUBFR ) { + Lag_CB_ptr = &silk_CB_lags_stage2[ 0 ][ 0 ]; + cbk_size = PE_NB_CBKS_STAGE2_EXT; + } else { + celt_assert( nb_subfr == PE_MAX_NB_SUBFR >> 1 ); + Lag_CB_ptr = &silk_CB_lags_stage2_10_ms[ 0 ][ 0 ]; + cbk_size = PE_NB_CBKS_STAGE2_10MS; + } + } else { + if( nb_subfr == PE_MAX_NB_SUBFR ) { + Lag_CB_ptr = &silk_CB_lags_stage3[ 0 ][ 0 ]; + cbk_size = PE_NB_CBKS_STAGE3_MAX; + } else { + celt_assert( nb_subfr == PE_MAX_NB_SUBFR >> 1 ); + Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; + cbk_size = PE_NB_CBKS_STAGE3_10MS; + } + } + + min_lag = silk_SMULBB( PE_MIN_LAG_MS, Fs_kHz ); + max_lag = silk_SMULBB( PE_MAX_LAG_MS, Fs_kHz ); + lag = min_lag + lagIndex; + + for( k = 0; k < nb_subfr; k++ ) { + pitch_lags[ k ] = lag + matrix_ptr( Lag_CB_ptr, k, contourIndex, cbk_size ); + pitch_lags[ k ] = silk_LIMIT( pitch_lags[ k ], min_lag, max_lag ); + } +} diff --git a/lib/opus/silk/decode_pulses.c b/lib/opus/silk/decode_pulses.c new file mode 100644 index 00000000..a56d2d30 --- /dev/null +++ b/lib/opus/silk/decode_pulses.c @@ -0,0 +1,115 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/*********************************************/ +/* Decode quantization indices of excitation */ +/*********************************************/ +void silk_decode_pulses( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 pulses[], /* O Excitation signal */ + const opus_int signalType, /* I Sigtype */ + const opus_int quantOffsetType, /* I quantOffsetType */ + const opus_int frame_length /* I Frame length */ +) +{ + opus_int i, j, k, iter, abs_q, nLS, RateLevelIndex; + opus_int sum_pulses[ MAX_NB_SHELL_BLOCKS ], nLshifts[ MAX_NB_SHELL_BLOCKS ]; + opus_int16 *pulses_ptr; + const opus_uint8 *cdf_ptr; + + /*********************/ + /* Decode rate level */ + /*********************/ + RateLevelIndex = ec_dec_icdf( psRangeDec, silk_rate_levels_iCDF[ signalType >> 1 ], 8 ); + + /* Calculate number of shell blocks */ + silk_assert( 1 << LOG2_SHELL_CODEC_FRAME_LENGTH == SHELL_CODEC_FRAME_LENGTH ); + iter = silk_RSHIFT( frame_length, LOG2_SHELL_CODEC_FRAME_LENGTH ); + if( iter * SHELL_CODEC_FRAME_LENGTH < frame_length ) { + celt_assert( frame_length == 12 * 10 ); /* Make sure only happens for 10 ms @ 12 kHz */ + iter++; + } + + /***************************************************/ + /* Sum-Weighted-Pulses Decoding */ + /***************************************************/ + cdf_ptr = silk_pulses_per_block_iCDF[ RateLevelIndex ]; + for( i = 0; i < iter; i++ ) { + nLshifts[ i ] = 0; + sum_pulses[ i ] = ec_dec_icdf( psRangeDec, cdf_ptr, 8 ); + + /* LSB indication */ + while( sum_pulses[ i ] == SILK_MAX_PULSES + 1 ) { + nLshifts[ i ]++; + /* When we've already got 10 LSBs, we shift the table to not allow (SILK_MAX_PULSES + 1) */ + sum_pulses[ i ] = ec_dec_icdf( psRangeDec, + silk_pulses_per_block_iCDF[ N_RATE_LEVELS - 1] + ( nLshifts[ i ] == 10 ), 8 ); + } + } + + /***************************************************/ + /* Shell decoding */ + /***************************************************/ + for( i = 0; i < iter; i++ ) { + if( sum_pulses[ i ] > 0 ) { + silk_shell_decoder( &pulses[ silk_SMULBB( i, SHELL_CODEC_FRAME_LENGTH ) ], psRangeDec, sum_pulses[ i ] ); + } else { + silk_memset( &pulses[ silk_SMULBB( i, SHELL_CODEC_FRAME_LENGTH ) ], 0, SHELL_CODEC_FRAME_LENGTH * sizeof( pulses[0] ) ); + } + } + + /***************************************************/ + /* LSB Decoding */ + /***************************************************/ + for( i = 0; i < iter; i++ ) { + if( nLshifts[ i ] > 0 ) { + nLS = nLshifts[ i ]; + pulses_ptr = &pulses[ silk_SMULBB( i, SHELL_CODEC_FRAME_LENGTH ) ]; + for( k = 0; k < SHELL_CODEC_FRAME_LENGTH; k++ ) { + abs_q = pulses_ptr[ k ]; + for( j = 0; j < nLS; j++ ) { + abs_q = silk_LSHIFT( abs_q, 1 ); + abs_q += ec_dec_icdf( psRangeDec, silk_lsb_iCDF, 8 ); + } + pulses_ptr[ k ] = abs_q; + } + /* Mark the number of pulses non-zero for sign decoding. */ + sum_pulses[ i ] |= nLS << 5; + } + } + + /****************************************/ + /* Decode and add signs to pulse signal */ + /****************************************/ + silk_decode_signs( psRangeDec, pulses, frame_length, signalType, quantOffsetType, sum_pulses ); +} diff --git a/lib/opus/silk/decoder_set_fs.c b/lib/opus/silk/decoder_set_fs.c new file mode 100644 index 00000000..d9a13d0f --- /dev/null +++ b/lib/opus/silk/decoder_set_fs.c @@ -0,0 +1,108 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Set decoder sampling rate */ +opus_int silk_decoder_set_fs( + silk_decoder_state *psDec, /* I/O Decoder state pointer */ + opus_int fs_kHz, /* I Sampling frequency (kHz) */ + opus_int32 fs_API_Hz /* I API Sampling frequency (Hz) */ +) +{ + opus_int frame_length, ret = 0; + + celt_assert( fs_kHz == 8 || fs_kHz == 12 || fs_kHz == 16 ); + celt_assert( psDec->nb_subfr == MAX_NB_SUBFR || psDec->nb_subfr == MAX_NB_SUBFR/2 ); + + /* New (sub)frame length */ + psDec->subfr_length = silk_SMULBB( SUB_FRAME_LENGTH_MS, fs_kHz ); + frame_length = silk_SMULBB( psDec->nb_subfr, psDec->subfr_length ); + + /* Initialize resampler when switching internal or external sampling frequency */ + if( psDec->fs_kHz != fs_kHz || psDec->fs_API_hz != fs_API_Hz ) { + /* Initialize the resampler for dec_API.c preparing resampling from fs_kHz to API_fs_Hz */ + ret += silk_resampler_init( &psDec->resampler_state, silk_SMULBB( fs_kHz, 1000 ), fs_API_Hz, 0 ); + + psDec->fs_API_hz = fs_API_Hz; + } + + if( psDec->fs_kHz != fs_kHz || frame_length != psDec->frame_length ) { + if( fs_kHz == 8 ) { + if( psDec->nb_subfr == MAX_NB_SUBFR ) { + psDec->pitch_contour_iCDF = silk_pitch_contour_NB_iCDF; + } else { + psDec->pitch_contour_iCDF = silk_pitch_contour_10_ms_NB_iCDF; + } + } else { + if( psDec->nb_subfr == MAX_NB_SUBFR ) { + psDec->pitch_contour_iCDF = silk_pitch_contour_iCDF; + } else { + psDec->pitch_contour_iCDF = silk_pitch_contour_10_ms_iCDF; + } + } + if( psDec->fs_kHz != fs_kHz ) { + psDec->ltp_mem_length = silk_SMULBB( LTP_MEM_LENGTH_MS, fs_kHz ); + if( fs_kHz == 8 || fs_kHz == 12 ) { + psDec->LPC_order = MIN_LPC_ORDER; + psDec->psNLSF_CB = &silk_NLSF_CB_NB_MB; + } else { + psDec->LPC_order = MAX_LPC_ORDER; + psDec->psNLSF_CB = &silk_NLSF_CB_WB; + } + if( fs_kHz == 16 ) { + psDec->pitch_lag_low_bits_iCDF = silk_uniform8_iCDF; + } else if( fs_kHz == 12 ) { + psDec->pitch_lag_low_bits_iCDF = silk_uniform6_iCDF; + } else if( fs_kHz == 8 ) { + psDec->pitch_lag_low_bits_iCDF = silk_uniform4_iCDF; + } else { + /* unsupported sampling rate */ + celt_assert( 0 ); + } + psDec->first_frame_after_reset = 1; + psDec->lagPrev = 100; + psDec->LastGainIndex = 10; + psDec->prevSignalType = TYPE_NO_VOICE_ACTIVITY; + silk_memset( psDec->outBuf, 0, sizeof(psDec->outBuf)); + silk_memset( psDec->sLPC_Q14_buf, 0, sizeof(psDec->sLPC_Q14_buf) ); + } + + psDec->fs_kHz = fs_kHz; + psDec->frame_length = frame_length; + } + + /* Check that settings are valid */ + celt_assert( psDec->frame_length > 0 && psDec->frame_length <= MAX_FRAME_LENGTH ); + + return ret; +} + diff --git a/lib/opus/silk/define.h b/lib/opus/silk/define.h new file mode 100644 index 00000000..491c86f3 --- /dev/null +++ b/lib/opus/silk/define.h @@ -0,0 +1,235 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_DEFINE_H +#define SILK_DEFINE_H + +#include "errors.h" +#include "typedef.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +/* Max number of encoder channels (1/2) */ +#define ENCODER_NUM_CHANNELS 2 +/* Number of decoder channels (1/2) */ +#define DECODER_NUM_CHANNELS 2 + +#define MAX_FRAMES_PER_PACKET 3 + +/* Limits on bitrate */ +#define MIN_TARGET_RATE_BPS 5000 +#define MAX_TARGET_RATE_BPS 80000 + +/* LBRR thresholds */ +#define LBRR_NB_MIN_RATE_BPS 12000 +#define LBRR_MB_MIN_RATE_BPS 14000 +#define LBRR_WB_MIN_RATE_BPS 16000 + +/* DTX settings */ +#define NB_SPEECH_FRAMES_BEFORE_DTX 10 /* eq 200 ms */ +#define MAX_CONSECUTIVE_DTX 20 /* eq 400 ms */ +#define DTX_ACTIVITY_THRESHOLD 0.1f + +/* VAD decision */ +#define VAD_NO_DECISION -1 +#define VAD_NO_ACTIVITY 0 +#define VAD_ACTIVITY 1 + +/* Maximum sampling frequency */ +#define MAX_FS_KHZ 16 +#define MAX_API_FS_KHZ 48 + +/* Signal types */ +#define TYPE_NO_VOICE_ACTIVITY 0 +#define TYPE_UNVOICED 1 +#define TYPE_VOICED 2 + +/* Conditional coding types */ +#define CODE_INDEPENDENTLY 0 +#define CODE_INDEPENDENTLY_NO_LTP_SCALING 1 +#define CODE_CONDITIONALLY 2 + +/* Settings for stereo processing */ +#define STEREO_QUANT_TAB_SIZE 16 +#define STEREO_QUANT_SUB_STEPS 5 +#define STEREO_INTERP_LEN_MS 8 /* must be even */ +#define STEREO_RATIO_SMOOTH_COEF 0.01 /* smoothing coef for signal norms and stereo width */ + +/* Range of pitch lag estimates */ +#define PITCH_EST_MIN_LAG_MS 2 /* 2 ms -> 500 Hz */ +#define PITCH_EST_MAX_LAG_MS 18 /* 18 ms -> 56 Hz */ + +/* Maximum number of subframes */ +#define MAX_NB_SUBFR 4 + +/* Number of samples per frame */ +#define LTP_MEM_LENGTH_MS 20 +#define SUB_FRAME_LENGTH_MS 5 +#define MAX_SUB_FRAME_LENGTH ( SUB_FRAME_LENGTH_MS * MAX_FS_KHZ ) +#define MAX_FRAME_LENGTH_MS ( SUB_FRAME_LENGTH_MS * MAX_NB_SUBFR ) +#define MAX_FRAME_LENGTH ( MAX_FRAME_LENGTH_MS * MAX_FS_KHZ ) + +/* Milliseconds of lookahead for pitch analysis */ +#define LA_PITCH_MS 2 +#define LA_PITCH_MAX ( LA_PITCH_MS * MAX_FS_KHZ ) + +/* Order of LPC used in find pitch */ +#define MAX_FIND_PITCH_LPC_ORDER 16 + +/* Length of LPC window used in find pitch */ +#define FIND_PITCH_LPC_WIN_MS ( 20 + (LA_PITCH_MS << 1) ) +#define FIND_PITCH_LPC_WIN_MS_2_SF ( 10 + (LA_PITCH_MS << 1) ) +#define FIND_PITCH_LPC_WIN_MAX ( FIND_PITCH_LPC_WIN_MS * MAX_FS_KHZ ) + +/* Milliseconds of lookahead for noise shape analysis */ +#define LA_SHAPE_MS 5 +#define LA_SHAPE_MAX ( LA_SHAPE_MS * MAX_FS_KHZ ) + +/* Maximum length of LPC window used in noise shape analysis */ +#define SHAPE_LPC_WIN_MAX ( 15 * MAX_FS_KHZ ) + +/* dB level of lowest gain quantization level */ +#define MIN_QGAIN_DB 2 +/* dB level of highest gain quantization level */ +#define MAX_QGAIN_DB 88 +/* Number of gain quantization levels */ +#define N_LEVELS_QGAIN 64 +/* Max increase in gain quantization index */ +#define MAX_DELTA_GAIN_QUANT 36 +/* Max decrease in gain quantization index */ +#define MIN_DELTA_GAIN_QUANT -4 + +/* Quantization offsets (multiples of 4) */ +#define OFFSET_VL_Q10 32 +#define OFFSET_VH_Q10 100 +#define OFFSET_UVL_Q10 100 +#define OFFSET_UVH_Q10 240 + +#define QUANT_LEVEL_ADJUST_Q10 80 + +/* Maximum numbers of iterations used to stabilize an LPC vector */ +#define MAX_LPC_STABILIZE_ITERATIONS 16 +#define MAX_PREDICTION_POWER_GAIN 1e4f +#define MAX_PREDICTION_POWER_GAIN_AFTER_RESET 1e2f + +#define MAX_LPC_ORDER 16 +#define MIN_LPC_ORDER 10 + +/* Find Pred Coef defines */ +#define LTP_ORDER 5 + +/* LTP quantization settings */ +#define NB_LTP_CBKS 3 + +/* Flag to use harmonic noise shaping */ +#define USE_HARM_SHAPING 1 + +/* Max LPC order of noise shaping filters */ +#define MAX_SHAPE_LPC_ORDER 24 + +#define HARM_SHAPE_FIR_TAPS 3 + +/* Maximum number of delayed decision states */ +#define MAX_DEL_DEC_STATES 4 + +#define LTP_BUF_LENGTH 512 +#define LTP_MASK ( LTP_BUF_LENGTH - 1 ) + +#define DECISION_DELAY 40 + +/* Number of subframes for excitation entropy coding */ +#define SHELL_CODEC_FRAME_LENGTH 16 +#define LOG2_SHELL_CODEC_FRAME_LENGTH 4 +#define MAX_NB_SHELL_BLOCKS ( MAX_FRAME_LENGTH / SHELL_CODEC_FRAME_LENGTH ) + +/* Number of rate levels, for entropy coding of excitation */ +#define N_RATE_LEVELS 10 + +/* Maximum sum of pulses per shell coding frame */ +#define SILK_MAX_PULSES 16 + +#define MAX_MATRIX_SIZE MAX_LPC_ORDER /* Max of LPC Order and LTP order */ + +# define NSQ_LPC_BUF_LENGTH MAX_LPC_ORDER + +/***************************/ +/* Voice activity detector */ +/***************************/ +#define VAD_N_BANDS 4 + +#define VAD_INTERNAL_SUBFRAMES_LOG2 2 +#define VAD_INTERNAL_SUBFRAMES ( 1 << VAD_INTERNAL_SUBFRAMES_LOG2 ) + +#define VAD_NOISE_LEVEL_SMOOTH_COEF_Q16 1024 /* Must be < 4096 */ +#define VAD_NOISE_LEVELS_BIAS 50 + +/* Sigmoid settings */ +#define VAD_NEGATIVE_OFFSET_Q5 128 /* sigmoid is 0 at -128 */ +#define VAD_SNR_FACTOR_Q16 45000 + +/* smoothing for SNR measurement */ +#define VAD_SNR_SMOOTH_COEF_Q18 4096 + +/* Size of the piecewise linear cosine approximation table for the LSFs */ +#define LSF_COS_TAB_SZ_FIX 128 + +/******************/ +/* NLSF quantizer */ +/******************/ +#define NLSF_W_Q 2 +#define NLSF_VQ_MAX_VECTORS 32 +#define NLSF_QUANT_MAX_AMPLITUDE 4 +#define NLSF_QUANT_MAX_AMPLITUDE_EXT 10 +#define NLSF_QUANT_LEVEL_ADJ 0.1 +#define NLSF_QUANT_DEL_DEC_STATES_LOG2 2 +#define NLSF_QUANT_DEL_DEC_STATES ( 1 << NLSF_QUANT_DEL_DEC_STATES_LOG2 ) + +/* Transition filtering for mode switching */ +#define TRANSITION_TIME_MS 5120 /* 5120 = 64 * FRAME_LENGTH_MS * ( TRANSITION_INT_NUM - 1 ) = 64*(20*4)*/ +#define TRANSITION_NB 3 /* Hardcoded in tables */ +#define TRANSITION_NA 2 /* Hardcoded in tables */ +#define TRANSITION_INT_NUM 5 /* Hardcoded in tables */ +#define TRANSITION_FRAMES ( TRANSITION_TIME_MS / MAX_FRAME_LENGTH_MS ) +#define TRANSITION_INT_STEPS ( TRANSITION_FRAMES / ( TRANSITION_INT_NUM - 1 ) ) + +/* BWE factors to apply after packet loss */ +#define BWE_AFTER_LOSS_Q16 63570 + +/* Defines for CN generation */ +#define CNG_BUF_MASK_MAX 255 /* 2^floor(log2(MAX_FRAME_LENGTH))-1 */ +#define CNG_GAIN_SMTH_Q16 4634 /* 0.25^(1/4) */ +#define CNG_GAIN_SMTH_THRESHOLD_Q16 46396 /* -3 dB */ +#define CNG_NLSF_SMTH_Q16 16348 /* 0.25 */ + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/enc_API.c b/lib/opus/silk/enc_API.c new file mode 100644 index 00000000..548e0736 --- /dev/null +++ b/lib/opus/silk/enc_API.c @@ -0,0 +1,587 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#include "define.h" +#include "API.h" +#include "control.h" +#include "typedef.h" +#include "stack_alloc.h" +#include "structs.h" +#include "tuning_parameters.h" +#ifdef FIXED_POINT +#include "main_FIX.h" +#else +#include "main_FLP.h" +#endif + +/***************************************/ +/* Read control structure from encoder */ +/***************************************/ +static opus_int silk_QueryEncoder( /* O Returns error code */ + const void *encState, /* I State */ + silk_EncControlStruct *encStatus /* O Encoder Status */ +); + +/****************************************/ +/* Encoder functions */ +/****************************************/ + +opus_int silk_Get_Encoder_Size( /* O Returns error code */ + opus_int *encSizeBytes /* O Number of bytes in SILK encoder state */ +) +{ + opus_int ret = SILK_NO_ERROR; + + *encSizeBytes = sizeof( silk_encoder ); + + return ret; +} + +/*************************/ +/* Init or Reset encoder */ +/*************************/ +opus_int silk_InitEncoder( /* O Returns error code */ + void *encState, /* I/O State */ + int arch, /* I Run-time architecture */ + silk_EncControlStruct *encStatus /* O Encoder Status */ +) +{ + silk_encoder *psEnc; + opus_int n, ret = SILK_NO_ERROR; + + psEnc = (silk_encoder *)encState; + + /* Reset encoder */ + silk_memset( psEnc, 0, sizeof( silk_encoder ) ); + for( n = 0; n < ENCODER_NUM_CHANNELS; n++ ) { + if( ret += silk_init_encoder( &psEnc->state_Fxx[ n ], arch ) ) { + celt_assert( 0 ); + } + } + + psEnc->nChannelsAPI = 1; + psEnc->nChannelsInternal = 1; + + /* Read control structure */ + if( ret += silk_QueryEncoder( encState, encStatus ) ) { + celt_assert( 0 ); + } + + return ret; +} + +/***************************************/ +/* Read control structure from encoder */ +/***************************************/ +static opus_int silk_QueryEncoder( /* O Returns error code */ + const void *encState, /* I State */ + silk_EncControlStruct *encStatus /* O Encoder Status */ +) +{ + opus_int ret = SILK_NO_ERROR; + silk_encoder_state_Fxx *state_Fxx; + silk_encoder *psEnc = (silk_encoder *)encState; + + state_Fxx = psEnc->state_Fxx; + + encStatus->nChannelsAPI = psEnc->nChannelsAPI; + encStatus->nChannelsInternal = psEnc->nChannelsInternal; + encStatus->API_sampleRate = state_Fxx[ 0 ].sCmn.API_fs_Hz; + encStatus->maxInternalSampleRate = state_Fxx[ 0 ].sCmn.maxInternal_fs_Hz; + encStatus->minInternalSampleRate = state_Fxx[ 0 ].sCmn.minInternal_fs_Hz; + encStatus->desiredInternalSampleRate = state_Fxx[ 0 ].sCmn.desiredInternal_fs_Hz; + encStatus->payloadSize_ms = state_Fxx[ 0 ].sCmn.PacketSize_ms; + encStatus->bitRate = state_Fxx[ 0 ].sCmn.TargetRate_bps; + encStatus->packetLossPercentage = state_Fxx[ 0 ].sCmn.PacketLoss_perc; + encStatus->complexity = state_Fxx[ 0 ].sCmn.Complexity; + encStatus->useInBandFEC = state_Fxx[ 0 ].sCmn.useInBandFEC; + encStatus->useDTX = state_Fxx[ 0 ].sCmn.useDTX; + encStatus->useCBR = state_Fxx[ 0 ].sCmn.useCBR; + encStatus->internalSampleRate = silk_SMULBB( state_Fxx[ 0 ].sCmn.fs_kHz, 1000 ); + encStatus->allowBandwidthSwitch = state_Fxx[ 0 ].sCmn.allow_bandwidth_switch; + encStatus->inWBmodeWithoutVariableLP = state_Fxx[ 0 ].sCmn.fs_kHz == 16 && state_Fxx[ 0 ].sCmn.sLP.mode == 0; + + return ret; +} + + +/**************************/ +/* Encode frame with Silk */ +/**************************/ +/* Note: if prefillFlag is set, the input must contain 10 ms of audio, irrespective of what */ +/* encControl->payloadSize_ms is set to */ +opus_int silk_Encode( /* O Returns error code */ + void *encState, /* I/O State */ + silk_EncControlStruct *encControl, /* I Control status */ + const opus_int16 *samplesIn, /* I Speech sample input vector */ + opus_int nSamplesIn, /* I Number of samples in input vector */ + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int32 *nBytesOut, /* I/O Number of bytes in payload (input: Max bytes) */ + const opus_int prefillFlag, /* I Flag to indicate prefilling buffers no coding */ + opus_int activity /* I Decision of Opus voice activity detector */ +) +{ + opus_int n, i, nBits, flags, tmp_payloadSize_ms = 0, tmp_complexity = 0, ret = 0; + opus_int nSamplesToBuffer, nSamplesToBufferMax, nBlocksOf10ms; + opus_int nSamplesFromInput = 0, nSamplesFromInputMax; + opus_int speech_act_thr_for_switch_Q8; + opus_int32 TargetRate_bps, MStargetRates_bps[ 2 ], channelRate_bps, LBRR_symbol, sum; + silk_encoder *psEnc = ( silk_encoder * )encState; + VARDECL( opus_int16, buf ); + opus_int transition, curr_block, tot_blocks; + SAVE_STACK; + + if (encControl->reducedDependency) + { + psEnc->state_Fxx[0].sCmn.first_frame_after_reset = 1; + psEnc->state_Fxx[1].sCmn.first_frame_after_reset = 1; + } + psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded = psEnc->state_Fxx[ 1 ].sCmn.nFramesEncoded = 0; + + /* Check values in encoder control structure */ + if( ( ret = check_control_input( encControl ) ) != 0 ) { + celt_assert( 0 ); + RESTORE_STACK; + return ret; + } + + encControl->switchReady = 0; + + if( encControl->nChannelsInternal > psEnc->nChannelsInternal ) { + /* Mono -> Stereo transition: init state of second channel and stereo state */ + ret += silk_init_encoder( &psEnc->state_Fxx[ 1 ], psEnc->state_Fxx[ 0 ].sCmn.arch ); + silk_memset( psEnc->sStereo.pred_prev_Q13, 0, sizeof( psEnc->sStereo.pred_prev_Q13 ) ); + silk_memset( psEnc->sStereo.sSide, 0, sizeof( psEnc->sStereo.sSide ) ); + psEnc->sStereo.mid_side_amp_Q0[ 0 ] = 0; + psEnc->sStereo.mid_side_amp_Q0[ 1 ] = 1; + psEnc->sStereo.mid_side_amp_Q0[ 2 ] = 0; + psEnc->sStereo.mid_side_amp_Q0[ 3 ] = 1; + psEnc->sStereo.width_prev_Q14 = 0; + psEnc->sStereo.smth_width_Q14 = SILK_FIX_CONST( 1, 14 ); + if( psEnc->nChannelsAPI == 2 ) { + silk_memcpy( &psEnc->state_Fxx[ 1 ].sCmn.resampler_state, &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, sizeof( silk_resampler_state_struct ) ); + silk_memcpy( &psEnc->state_Fxx[ 1 ].sCmn.In_HP_State, &psEnc->state_Fxx[ 0 ].sCmn.In_HP_State, sizeof( psEnc->state_Fxx[ 1 ].sCmn.In_HP_State ) ); + } + } + + transition = (encControl->payloadSize_ms != psEnc->state_Fxx[ 0 ].sCmn.PacketSize_ms) || (psEnc->nChannelsInternal != encControl->nChannelsInternal); + + psEnc->nChannelsAPI = encControl->nChannelsAPI; + psEnc->nChannelsInternal = encControl->nChannelsInternal; + + nBlocksOf10ms = silk_DIV32( 100 * nSamplesIn, encControl->API_sampleRate ); + tot_blocks = ( nBlocksOf10ms > 1 ) ? nBlocksOf10ms >> 1 : 1; + curr_block = 0; + if( prefillFlag ) { + silk_LP_state save_LP; + /* Only accept input length of 10 ms */ + if( nBlocksOf10ms != 1 ) { + celt_assert( 0 ); + RESTORE_STACK; + return SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; + } + if ( prefillFlag == 2 ) { + save_LP = psEnc->state_Fxx[ 0 ].sCmn.sLP; + /* Save the sampling rate so the bandwidth switching code can keep handling transitions. */ + save_LP.saved_fs_kHz = psEnc->state_Fxx[ 0 ].sCmn.fs_kHz; + } + /* Reset Encoder */ + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + ret = silk_init_encoder( &psEnc->state_Fxx[ n ], psEnc->state_Fxx[ n ].sCmn.arch ); + /* Restore the variable LP state. */ + if ( prefillFlag == 2 ) { + psEnc->state_Fxx[ n ].sCmn.sLP = save_LP; + } + celt_assert( !ret ); + } + tmp_payloadSize_ms = encControl->payloadSize_ms; + encControl->payloadSize_ms = 10; + tmp_complexity = encControl->complexity; + encControl->complexity = 0; + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + psEnc->state_Fxx[ n ].sCmn.controlled_since_last_payload = 0; + psEnc->state_Fxx[ n ].sCmn.prefillFlag = 1; + } + } else { + /* Only accept input lengths that are a multiple of 10 ms */ + if( nBlocksOf10ms * encControl->API_sampleRate != 100 * nSamplesIn || nSamplesIn < 0 ) { + celt_assert( 0 ); + RESTORE_STACK; + return SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; + } + /* Make sure no more than one packet can be produced */ + if( 1000 * (opus_int32)nSamplesIn > encControl->payloadSize_ms * encControl->API_sampleRate ) { + celt_assert( 0 ); + RESTORE_STACK; + return SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES; + } + } + + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + /* Force the side channel to the same rate as the mid */ + opus_int force_fs_kHz = (n==1) ? psEnc->state_Fxx[0].sCmn.fs_kHz : 0; + if( ( ret = silk_control_encoder( &psEnc->state_Fxx[ n ], encControl, psEnc->allowBandwidthSwitch, n, force_fs_kHz ) ) != 0 ) { + silk_assert( 0 ); + RESTORE_STACK; + return ret; + } + if( psEnc->state_Fxx[n].sCmn.first_frame_after_reset || transition ) { + for( i = 0; i < psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket; i++ ) { + psEnc->state_Fxx[ n ].sCmn.LBRR_flags[ i ] = 0; + } + } + psEnc->state_Fxx[ n ].sCmn.inDTX = psEnc->state_Fxx[ n ].sCmn.useDTX; + } + celt_assert( encControl->nChannelsInternal == 1 || psEnc->state_Fxx[ 0 ].sCmn.fs_kHz == psEnc->state_Fxx[ 1 ].sCmn.fs_kHz ); + + /* Input buffering/resampling and encoding */ + nSamplesToBufferMax = + 10 * nBlocksOf10ms * psEnc->state_Fxx[ 0 ].sCmn.fs_kHz; + nSamplesFromInputMax = + silk_DIV32_16( nSamplesToBufferMax * + psEnc->state_Fxx[ 0 ].sCmn.API_fs_Hz, + psEnc->state_Fxx[ 0 ].sCmn.fs_kHz * 1000 ); + ALLOC( buf, nSamplesFromInputMax, opus_int16 ); + while( 1 ) { + int curr_nBitsUsedLBRR = 0; + nSamplesToBuffer = psEnc->state_Fxx[ 0 ].sCmn.frame_length - psEnc->state_Fxx[ 0 ].sCmn.inputBufIx; + nSamplesToBuffer = silk_min( nSamplesToBuffer, nSamplesToBufferMax ); + nSamplesFromInput = silk_DIV32_16( nSamplesToBuffer * psEnc->state_Fxx[ 0 ].sCmn.API_fs_Hz, psEnc->state_Fxx[ 0 ].sCmn.fs_kHz * 1000 ); + /* Resample and write to buffer */ + if( encControl->nChannelsAPI == 2 && encControl->nChannelsInternal == 2 ) { + opus_int id = psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded; + for( n = 0; n < nSamplesFromInput; n++ ) { + buf[ n ] = samplesIn[ 2 * n ]; + } + /* Making sure to start both resamplers from the same state when switching from mono to stereo */ + if( psEnc->nPrevChannelsInternal == 1 && id==0 ) { + silk_memcpy( &psEnc->state_Fxx[ 1 ].sCmn.resampler_state, &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, sizeof(psEnc->state_Fxx[ 1 ].sCmn.resampler_state)); + } + + ret += silk_resampler( &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, + &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); + psEnc->state_Fxx[ 0 ].sCmn.inputBufIx += nSamplesToBuffer; + + nSamplesToBuffer = psEnc->state_Fxx[ 1 ].sCmn.frame_length - psEnc->state_Fxx[ 1 ].sCmn.inputBufIx; + nSamplesToBuffer = silk_min( nSamplesToBuffer, 10 * nBlocksOf10ms * psEnc->state_Fxx[ 1 ].sCmn.fs_kHz ); + for( n = 0; n < nSamplesFromInput; n++ ) { + buf[ n ] = samplesIn[ 2 * n + 1 ]; + } + ret += silk_resampler( &psEnc->state_Fxx[ 1 ].sCmn.resampler_state, + &psEnc->state_Fxx[ 1 ].sCmn.inputBuf[ psEnc->state_Fxx[ 1 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); + + psEnc->state_Fxx[ 1 ].sCmn.inputBufIx += nSamplesToBuffer; + } else if( encControl->nChannelsAPI == 2 && encControl->nChannelsInternal == 1 ) { + /* Combine left and right channels before resampling */ + for( n = 0; n < nSamplesFromInput; n++ ) { + sum = samplesIn[ 2 * n ] + samplesIn[ 2 * n + 1 ]; + buf[ n ] = (opus_int16)silk_RSHIFT_ROUND( sum, 1 ); + } + ret += silk_resampler( &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, + &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); + /* On the first mono frame, average the results for the two resampler states */ + if( psEnc->nPrevChannelsInternal == 2 && psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded == 0 ) { + ret += silk_resampler( &psEnc->state_Fxx[ 1 ].sCmn.resampler_state, + &psEnc->state_Fxx[ 1 ].sCmn.inputBuf[ psEnc->state_Fxx[ 1 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); + for( n = 0; n < psEnc->state_Fxx[ 0 ].sCmn.frame_length; n++ ) { + psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx+n+2 ] = + silk_RSHIFT(psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx+n+2 ] + + psEnc->state_Fxx[ 1 ].sCmn.inputBuf[ psEnc->state_Fxx[ 1 ].sCmn.inputBufIx+n+2 ], 1); + } + } + psEnc->state_Fxx[ 0 ].sCmn.inputBufIx += nSamplesToBuffer; + } else { + celt_assert( encControl->nChannelsAPI == 1 && encControl->nChannelsInternal == 1 ); + silk_memcpy(buf, samplesIn, nSamplesFromInput*sizeof(opus_int16)); + ret += silk_resampler( &psEnc->state_Fxx[ 0 ].sCmn.resampler_state, + &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.inputBufIx + 2 ], buf, nSamplesFromInput ); + psEnc->state_Fxx[ 0 ].sCmn.inputBufIx += nSamplesToBuffer; + } + + samplesIn += nSamplesFromInput * encControl->nChannelsAPI; + nSamplesIn -= nSamplesFromInput; + + /* Default */ + psEnc->allowBandwidthSwitch = 0; + + /* Silk encoder */ + if( psEnc->state_Fxx[ 0 ].sCmn.inputBufIx >= psEnc->state_Fxx[ 0 ].sCmn.frame_length ) { + /* Enough data in input buffer, so encode */ + celt_assert( psEnc->state_Fxx[ 0 ].sCmn.inputBufIx == psEnc->state_Fxx[ 0 ].sCmn.frame_length ); + celt_assert( encControl->nChannelsInternal == 1 || psEnc->state_Fxx[ 1 ].sCmn.inputBufIx == psEnc->state_Fxx[ 1 ].sCmn.frame_length ); + + /* Deal with LBRR data */ + if( psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded == 0 && !prefillFlag ) { + /* Create space at start of payload for VAD and FEC flags */ + opus_uint8 iCDF[ 2 ] = { 0, 0 }; + iCDF[ 0 ] = 256 - silk_RSHIFT( 256, ( psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket + 1 ) * encControl->nChannelsInternal ); + ec_enc_icdf( psRangeEnc, 0, iCDF, 8 ); + curr_nBitsUsedLBRR = ec_tell( psRangeEnc ); + + /* Encode any LBRR data from previous packet */ + /* Encode LBRR flags */ + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + LBRR_symbol = 0; + for( i = 0; i < psEnc->state_Fxx[ n ].sCmn.nFramesPerPacket; i++ ) { + LBRR_symbol |= silk_LSHIFT( psEnc->state_Fxx[ n ].sCmn.LBRR_flags[ i ], i ); + } + psEnc->state_Fxx[ n ].sCmn.LBRR_flag = LBRR_symbol > 0 ? 1 : 0; + if( LBRR_symbol && psEnc->state_Fxx[ n ].sCmn.nFramesPerPacket > 1 ) { + ec_enc_icdf( psRangeEnc, LBRR_symbol - 1, silk_LBRR_flags_iCDF_ptr[ psEnc->state_Fxx[ n ].sCmn.nFramesPerPacket - 2 ], 8 ); + } + } + + /* Code LBRR indices and excitation signals */ + for( i = 0; i < psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket; i++ ) { + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + if( psEnc->state_Fxx[ n ].sCmn.LBRR_flags[ i ] ) { + opus_int condCoding; + + if( encControl->nChannelsInternal == 2 && n == 0 ) { + silk_stereo_encode_pred( psRangeEnc, psEnc->sStereo.predIx[ i ] ); + /* For LBRR data there's no need to code the mid-only flag if the side-channel LBRR flag is set */ + if( psEnc->state_Fxx[ 1 ].sCmn.LBRR_flags[ i ] == 0 ) { + silk_stereo_encode_mid_only( psRangeEnc, psEnc->sStereo.mid_only_flags[ i ] ); + } + } + /* Use conditional coding if previous frame available */ + if( i > 0 && psEnc->state_Fxx[ n ].sCmn.LBRR_flags[ i - 1 ] ) { + condCoding = CODE_CONDITIONALLY; + } else { + condCoding = CODE_INDEPENDENTLY; + } + silk_encode_indices( &psEnc->state_Fxx[ n ].sCmn, psRangeEnc, i, 1, condCoding ); + silk_encode_pulses( psRangeEnc, psEnc->state_Fxx[ n ].sCmn.indices_LBRR[i].signalType, psEnc->state_Fxx[ n ].sCmn.indices_LBRR[i].quantOffsetType, + psEnc->state_Fxx[ n ].sCmn.pulses_LBRR[ i ], psEnc->state_Fxx[ n ].sCmn.frame_length ); + } + } + } + + /* Reset LBRR flags */ + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + silk_memset( psEnc->state_Fxx[ n ].sCmn.LBRR_flags, 0, sizeof( psEnc->state_Fxx[ n ].sCmn.LBRR_flags ) ); + } + curr_nBitsUsedLBRR = ec_tell( psRangeEnc ) - curr_nBitsUsedLBRR; + } + + silk_HP_variable_cutoff( psEnc->state_Fxx ); + + /* Total target bits for packet */ + nBits = silk_DIV32_16( silk_MUL( encControl->bitRate, encControl->payloadSize_ms ), 1000 ); + /* Subtract bits used for LBRR */ + if( !prefillFlag ) { + /* psEnc->nBitsUsedLBRR is an exponential moving average of the LBRR usage, + except that for the first LBRR frame it does no averaging and for the first + frame after after LBRR, it goes back to zero immediately. */ + if ( curr_nBitsUsedLBRR < 10 ) { + psEnc->nBitsUsedLBRR = 0; + } else if ( psEnc->nBitsUsedLBRR < 10) { + psEnc->nBitsUsedLBRR = curr_nBitsUsedLBRR; + } else { + psEnc->nBitsUsedLBRR = ( psEnc->nBitsUsedLBRR + curr_nBitsUsedLBRR ) / 2; + } + nBits -= psEnc->nBitsUsedLBRR; + } + /* Divide by number of uncoded frames left in packet */ + nBits = silk_DIV32_16( nBits, psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket ); + /* Convert to bits/second */ + if( encControl->payloadSize_ms == 10 ) { + TargetRate_bps = silk_SMULBB( nBits, 100 ); + } else { + TargetRate_bps = silk_SMULBB( nBits, 50 ); + } + /* Subtract fraction of bits in excess of target in previous frames and packets */ + TargetRate_bps -= silk_DIV32_16( silk_MUL( psEnc->nBitsExceeded, 1000 ), BITRESERVOIR_DECAY_TIME_MS ); + if( !prefillFlag && psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded > 0 ) { + /* Compare actual vs target bits so far in this packet */ + opus_int32 bitsBalance = ec_tell( psRangeEnc ) - psEnc->nBitsUsedLBRR - nBits * psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded; + TargetRate_bps -= silk_DIV32_16( silk_MUL( bitsBalance, 1000 ), BITRESERVOIR_DECAY_TIME_MS ); + } + /* Never exceed input bitrate */ + TargetRate_bps = silk_LIMIT( TargetRate_bps, encControl->bitRate, 5000 ); + + /* Convert Left/Right to Mid/Side */ + if( encControl->nChannelsInternal == 2 ) { + silk_stereo_LR_to_MS( &psEnc->sStereo, &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ 2 ], &psEnc->state_Fxx[ 1 ].sCmn.inputBuf[ 2 ], + psEnc->sStereo.predIx[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ], &psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ], + MStargetRates_bps, TargetRate_bps, psEnc->state_Fxx[ 0 ].sCmn.speech_activity_Q8, encControl->toMono, + psEnc->state_Fxx[ 0 ].sCmn.fs_kHz, psEnc->state_Fxx[ 0 ].sCmn.frame_length ); + if( psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ] == 0 ) { + /* Reset side channel encoder memory for first frame with side coding */ + if( psEnc->prev_decode_only_middle == 1 ) { + silk_memset( &psEnc->state_Fxx[ 1 ].sShape, 0, sizeof( psEnc->state_Fxx[ 1 ].sShape ) ); + silk_memset( &psEnc->state_Fxx[ 1 ].sCmn.sNSQ, 0, sizeof( psEnc->state_Fxx[ 1 ].sCmn.sNSQ ) ); + silk_memset( psEnc->state_Fxx[ 1 ].sCmn.prev_NLSFq_Q15, 0, sizeof( psEnc->state_Fxx[ 1 ].sCmn.prev_NLSFq_Q15 ) ); + silk_memset( &psEnc->state_Fxx[ 1 ].sCmn.sLP.In_LP_State, 0, sizeof( psEnc->state_Fxx[ 1 ].sCmn.sLP.In_LP_State ) ); + psEnc->state_Fxx[ 1 ].sCmn.prevLag = 100; + psEnc->state_Fxx[ 1 ].sCmn.sNSQ.lagPrev = 100; + psEnc->state_Fxx[ 1 ].sShape.LastGainIndex = 10; + psEnc->state_Fxx[ 1 ].sCmn.prevSignalType = TYPE_NO_VOICE_ACTIVITY; + psEnc->state_Fxx[ 1 ].sCmn.sNSQ.prev_gain_Q16 = 65536; + psEnc->state_Fxx[ 1 ].sCmn.first_frame_after_reset = 1; + } + silk_encode_do_VAD_Fxx( &psEnc->state_Fxx[ 1 ], activity ); + } else { + psEnc->state_Fxx[ 1 ].sCmn.VAD_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ] = 0; + } + if( !prefillFlag ) { + silk_stereo_encode_pred( psRangeEnc, psEnc->sStereo.predIx[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ] ); + if( psEnc->state_Fxx[ 1 ].sCmn.VAD_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ] == 0 ) { + silk_stereo_encode_mid_only( psRangeEnc, psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ] ); + } + } + } else { + /* Buffering */ + silk_memcpy( psEnc->state_Fxx[ 0 ].sCmn.inputBuf, psEnc->sStereo.sMid, 2 * sizeof( opus_int16 ) ); + silk_memcpy( psEnc->sStereo.sMid, &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[ psEnc->state_Fxx[ 0 ].sCmn.frame_length ], 2 * sizeof( opus_int16 ) ); + } + silk_encode_do_VAD_Fxx( &psEnc->state_Fxx[ 0 ], activity ); + + /* Encode */ + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + opus_int maxBits, useCBR; + + /* Handling rate constraints */ + maxBits = encControl->maxBits; + if( tot_blocks == 2 && curr_block == 0 ) { + maxBits = maxBits * 3 / 5; + } else if( tot_blocks == 3 ) { + if( curr_block == 0 ) { + maxBits = maxBits * 2 / 5; + } else if( curr_block == 1 ) { + maxBits = maxBits * 3 / 4; + } + } + useCBR = encControl->useCBR && curr_block == tot_blocks - 1; + + if( encControl->nChannelsInternal == 1 ) { + channelRate_bps = TargetRate_bps; + } else { + channelRate_bps = MStargetRates_bps[ n ]; + if( n == 0 && MStargetRates_bps[ 1 ] > 0 ) { + useCBR = 0; + /* Give mid up to 1/2 of the max bits for that frame */ + maxBits -= encControl->maxBits / ( tot_blocks * 2 ); + } + } + + if( channelRate_bps > 0 ) { + opus_int condCoding; + + silk_control_SNR( &psEnc->state_Fxx[ n ].sCmn, channelRate_bps ); + + /* Use independent coding if no previous frame available */ + if( psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded - n <= 0 ) { + condCoding = CODE_INDEPENDENTLY; + } else if( n > 0 && psEnc->prev_decode_only_middle ) { + /* If we skipped a side frame in this packet, we don't + need LTP scaling; the LTP state is well-defined. */ + condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; + } else { + condCoding = CODE_CONDITIONALLY; + } + if( ( ret = silk_encode_frame_Fxx( &psEnc->state_Fxx[ n ], nBytesOut, psRangeEnc, condCoding, maxBits, useCBR ) ) != 0 ) { + silk_assert( 0 ); + } + } + psEnc->state_Fxx[ n ].sCmn.controlled_since_last_payload = 0; + psEnc->state_Fxx[ n ].sCmn.inputBufIx = 0; + psEnc->state_Fxx[ n ].sCmn.nFramesEncoded++; + } + psEnc->prev_decode_only_middle = psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded - 1 ]; + + /* Insert VAD and FEC flags at beginning of bitstream */ + if( *nBytesOut > 0 && psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded == psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket) { + flags = 0; + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + for( i = 0; i < psEnc->state_Fxx[ n ].sCmn.nFramesPerPacket; i++ ) { + flags = silk_LSHIFT( flags, 1 ); + flags |= psEnc->state_Fxx[ n ].sCmn.VAD_flags[ i ]; + } + flags = silk_LSHIFT( flags, 1 ); + flags |= psEnc->state_Fxx[ n ].sCmn.LBRR_flag; + } + if( !prefillFlag ) { + ec_enc_patch_initial_bits( psRangeEnc, flags, ( psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket + 1 ) * encControl->nChannelsInternal ); + } + + /* Return zero bytes if all channels DTXed */ + if( psEnc->state_Fxx[ 0 ].sCmn.inDTX && ( encControl->nChannelsInternal == 1 || psEnc->state_Fxx[ 1 ].sCmn.inDTX ) ) { + *nBytesOut = 0; + } + + psEnc->nBitsExceeded += *nBytesOut * 8; + psEnc->nBitsExceeded -= silk_DIV32_16( silk_MUL( encControl->bitRate, encControl->payloadSize_ms ), 1000 ); + psEnc->nBitsExceeded = silk_LIMIT( psEnc->nBitsExceeded, 0, 10000 ); + + /* Update flag indicating if bandwidth switching is allowed */ + speech_act_thr_for_switch_Q8 = silk_SMLAWB( SILK_FIX_CONST( SPEECH_ACTIVITY_DTX_THRES, 8 ), + SILK_FIX_CONST( ( 1 - SPEECH_ACTIVITY_DTX_THRES ) / MAX_BANDWIDTH_SWITCH_DELAY_MS, 16 + 8 ), psEnc->timeSinceSwitchAllowed_ms ); + if( psEnc->state_Fxx[ 0 ].sCmn.speech_activity_Q8 < speech_act_thr_for_switch_Q8 ) { + psEnc->allowBandwidthSwitch = 1; + psEnc->timeSinceSwitchAllowed_ms = 0; + } else { + psEnc->allowBandwidthSwitch = 0; + psEnc->timeSinceSwitchAllowed_ms += encControl->payloadSize_ms; + } + } + + if( nSamplesIn == 0 ) { + break; + } + } else { + break; + } + curr_block++; + } + + psEnc->nPrevChannelsInternal = encControl->nChannelsInternal; + + encControl->allowBandwidthSwitch = psEnc->allowBandwidthSwitch; + encControl->inWBmodeWithoutVariableLP = psEnc->state_Fxx[ 0 ].sCmn.fs_kHz == 16 && psEnc->state_Fxx[ 0 ].sCmn.sLP.mode == 0; + encControl->internalSampleRate = silk_SMULBB( psEnc->state_Fxx[ 0 ].sCmn.fs_kHz, 1000 ); + encControl->stereoWidth_Q14 = encControl->toMono ? 0 : psEnc->sStereo.smth_width_Q14; + if( prefillFlag ) { + encControl->payloadSize_ms = tmp_payloadSize_ms; + encControl->complexity = tmp_complexity; + for( n = 0; n < encControl->nChannelsInternal; n++ ) { + psEnc->state_Fxx[ n ].sCmn.controlled_since_last_payload = 0; + psEnc->state_Fxx[ n ].sCmn.prefillFlag = 0; + } + } + + encControl->signalType = psEnc->state_Fxx[0].sCmn.indices.signalType; + encControl->offset = silk_Quantization_Offsets_Q10 + [ psEnc->state_Fxx[0].sCmn.indices.signalType >> 1 ] + [ psEnc->state_Fxx[0].sCmn.indices.quantOffsetType ]; + RESTORE_STACK; + return ret; +} + diff --git a/lib/opus/silk/encode_indices.c b/lib/opus/silk/encode_indices.c new file mode 100644 index 00000000..4bcbc334 --- /dev/null +++ b/lib/opus/silk/encode_indices.c @@ -0,0 +1,181 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Encode side-information parameters to payload */ +void silk_encode_indices( + silk_encoder_state *psEncC, /* I/O Encoder state */ + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int FrameIndex, /* I Frame number */ + opus_int encode_LBRR, /* I Flag indicating LBRR data is being encoded */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + opus_int i, k, typeOffset; + opus_int encode_absolute_lagIndex, delta_lagIndex; + opus_int16 ec_ix[ MAX_LPC_ORDER ]; + opus_uint8 pred_Q8[ MAX_LPC_ORDER ]; + const SideInfoIndices *psIndices; + + if( encode_LBRR ) { + psIndices = &psEncC->indices_LBRR[ FrameIndex ]; + } else { + psIndices = &psEncC->indices; + } + + /*******************************************/ + /* Encode signal type and quantizer offset */ + /*******************************************/ + typeOffset = 2 * psIndices->signalType + psIndices->quantOffsetType; + celt_assert( typeOffset >= 0 && typeOffset < 6 ); + celt_assert( encode_LBRR == 0 || typeOffset >= 2 ); + if( encode_LBRR || typeOffset >= 2 ) { + ec_enc_icdf( psRangeEnc, typeOffset - 2, silk_type_offset_VAD_iCDF, 8 ); + } else { + ec_enc_icdf( psRangeEnc, typeOffset, silk_type_offset_no_VAD_iCDF, 8 ); + } + + /****************/ + /* Encode gains */ + /****************/ + /* first subframe */ + if( condCoding == CODE_CONDITIONALLY ) { + /* conditional coding */ + silk_assert( psIndices->GainsIndices[ 0 ] >= 0 && psIndices->GainsIndices[ 0 ] < MAX_DELTA_GAIN_QUANT - MIN_DELTA_GAIN_QUANT + 1 ); + ec_enc_icdf( psRangeEnc, psIndices->GainsIndices[ 0 ], silk_delta_gain_iCDF, 8 ); + } else { + /* independent coding, in two stages: MSB bits followed by 3 LSBs */ + silk_assert( psIndices->GainsIndices[ 0 ] >= 0 && psIndices->GainsIndices[ 0 ] < N_LEVELS_QGAIN ); + ec_enc_icdf( psRangeEnc, silk_RSHIFT( psIndices->GainsIndices[ 0 ], 3 ), silk_gain_iCDF[ psIndices->signalType ], 8 ); + ec_enc_icdf( psRangeEnc, psIndices->GainsIndices[ 0 ] & 7, silk_uniform8_iCDF, 8 ); + } + + /* remaining subframes */ + for( i = 1; i < psEncC->nb_subfr; i++ ) { + silk_assert( psIndices->GainsIndices[ i ] >= 0 && psIndices->GainsIndices[ i ] < MAX_DELTA_GAIN_QUANT - MIN_DELTA_GAIN_QUANT + 1 ); + ec_enc_icdf( psRangeEnc, psIndices->GainsIndices[ i ], silk_delta_gain_iCDF, 8 ); + } + + /****************/ + /* Encode NLSFs */ + /****************/ + ec_enc_icdf( psRangeEnc, psIndices->NLSFIndices[ 0 ], &psEncC->psNLSF_CB->CB1_iCDF[ ( psIndices->signalType >> 1 ) * psEncC->psNLSF_CB->nVectors ], 8 ); + silk_NLSF_unpack( ec_ix, pred_Q8, psEncC->psNLSF_CB, psIndices->NLSFIndices[ 0 ] ); + celt_assert( psEncC->psNLSF_CB->order == psEncC->predictLPCOrder ); + for( i = 0; i < psEncC->psNLSF_CB->order; i++ ) { + if( psIndices->NLSFIndices[ i+1 ] >= NLSF_QUANT_MAX_AMPLITUDE ) { + ec_enc_icdf( psRangeEnc, 2 * NLSF_QUANT_MAX_AMPLITUDE, &psEncC->psNLSF_CB->ec_iCDF[ ec_ix[ i ] ], 8 ); + ec_enc_icdf( psRangeEnc, psIndices->NLSFIndices[ i+1 ] - NLSF_QUANT_MAX_AMPLITUDE, silk_NLSF_EXT_iCDF, 8 ); + } else if( psIndices->NLSFIndices[ i+1 ] <= -NLSF_QUANT_MAX_AMPLITUDE ) { + ec_enc_icdf( psRangeEnc, 0, &psEncC->psNLSF_CB->ec_iCDF[ ec_ix[ i ] ], 8 ); + ec_enc_icdf( psRangeEnc, -psIndices->NLSFIndices[ i+1 ] - NLSF_QUANT_MAX_AMPLITUDE, silk_NLSF_EXT_iCDF, 8 ); + } else { + ec_enc_icdf( psRangeEnc, psIndices->NLSFIndices[ i+1 ] + NLSF_QUANT_MAX_AMPLITUDE, &psEncC->psNLSF_CB->ec_iCDF[ ec_ix[ i ] ], 8 ); + } + } + + /* Encode NLSF interpolation factor */ + if( psEncC->nb_subfr == MAX_NB_SUBFR ) { + silk_assert( psIndices->NLSFInterpCoef_Q2 >= 0 && psIndices->NLSFInterpCoef_Q2 < 5 ); + ec_enc_icdf( psRangeEnc, psIndices->NLSFInterpCoef_Q2, silk_NLSF_interpolation_factor_iCDF, 8 ); + } + + if( psIndices->signalType == TYPE_VOICED ) + { + /*********************/ + /* Encode pitch lags */ + /*********************/ + /* lag index */ + encode_absolute_lagIndex = 1; + if( condCoding == CODE_CONDITIONALLY && psEncC->ec_prevSignalType == TYPE_VOICED ) { + /* Delta Encoding */ + delta_lagIndex = psIndices->lagIndex - psEncC->ec_prevLagIndex; + if( delta_lagIndex < -8 || delta_lagIndex > 11 ) { + delta_lagIndex = 0; + } else { + delta_lagIndex = delta_lagIndex + 9; + encode_absolute_lagIndex = 0; /* Only use delta */ + } + silk_assert( delta_lagIndex >= 0 && delta_lagIndex < 21 ); + ec_enc_icdf( psRangeEnc, delta_lagIndex, silk_pitch_delta_iCDF, 8 ); + } + if( encode_absolute_lagIndex ) { + /* Absolute encoding */ + opus_int32 pitch_high_bits, pitch_low_bits; + pitch_high_bits = silk_DIV32_16( psIndices->lagIndex, silk_RSHIFT( psEncC->fs_kHz, 1 ) ); + pitch_low_bits = psIndices->lagIndex - silk_SMULBB( pitch_high_bits, silk_RSHIFT( psEncC->fs_kHz, 1 ) ); + silk_assert( pitch_low_bits < psEncC->fs_kHz / 2 ); + silk_assert( pitch_high_bits < 32 ); + ec_enc_icdf( psRangeEnc, pitch_high_bits, silk_pitch_lag_iCDF, 8 ); + ec_enc_icdf( psRangeEnc, pitch_low_bits, psEncC->pitch_lag_low_bits_iCDF, 8 ); + } + psEncC->ec_prevLagIndex = psIndices->lagIndex; + + /* Countour index */ + silk_assert( psIndices->contourIndex >= 0 ); + silk_assert( ( psIndices->contourIndex < 34 && psEncC->fs_kHz > 8 && psEncC->nb_subfr == 4 ) || + ( psIndices->contourIndex < 11 && psEncC->fs_kHz == 8 && psEncC->nb_subfr == 4 ) || + ( psIndices->contourIndex < 12 && psEncC->fs_kHz > 8 && psEncC->nb_subfr == 2 ) || + ( psIndices->contourIndex < 3 && psEncC->fs_kHz == 8 && psEncC->nb_subfr == 2 ) ); + ec_enc_icdf( psRangeEnc, psIndices->contourIndex, psEncC->pitch_contour_iCDF, 8 ); + + /********************/ + /* Encode LTP gains */ + /********************/ + /* PERIndex value */ + silk_assert( psIndices->PERIndex >= 0 && psIndices->PERIndex < 3 ); + ec_enc_icdf( psRangeEnc, psIndices->PERIndex, silk_LTP_per_index_iCDF, 8 ); + + /* Codebook Indices */ + for( k = 0; k < psEncC->nb_subfr; k++ ) { + silk_assert( psIndices->LTPIndex[ k ] >= 0 && psIndices->LTPIndex[ k ] < ( 8 << psIndices->PERIndex ) ); + ec_enc_icdf( psRangeEnc, psIndices->LTPIndex[ k ], silk_LTP_gain_iCDF_ptrs[ psIndices->PERIndex ], 8 ); + } + + /**********************/ + /* Encode LTP scaling */ + /**********************/ + if( condCoding == CODE_INDEPENDENTLY ) { + silk_assert( psIndices->LTP_scaleIndex >= 0 && psIndices->LTP_scaleIndex < 3 ); + ec_enc_icdf( psRangeEnc, psIndices->LTP_scaleIndex, silk_LTPscale_iCDF, 8 ); + } + silk_assert( !condCoding || psIndices->LTP_scaleIndex == 0 ); + } + + psEncC->ec_prevSignalType = psIndices->signalType; + + /***************/ + /* Encode seed */ + /***************/ + silk_assert( psIndices->Seed >= 0 && psIndices->Seed < 4 ); + ec_enc_icdf( psRangeEnc, psIndices->Seed, silk_uniform4_iCDF, 8 ); +} diff --git a/lib/opus/silk/encode_pulses.c b/lib/opus/silk/encode_pulses.c new file mode 100644 index 00000000..8a199913 --- /dev/null +++ b/lib/opus/silk/encode_pulses.c @@ -0,0 +1,206 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" + +/*********************************************/ +/* Encode quantization indices of excitation */ +/*********************************************/ + +static OPUS_INLINE opus_int combine_and_check( /* return ok */ + opus_int *pulses_comb, /* O */ + const opus_int *pulses_in, /* I */ + opus_int max_pulses, /* I max value for sum of pulses */ + opus_int len /* I number of output values */ +) +{ + opus_int k, sum; + + for( k = 0; k < len; k++ ) { + sum = pulses_in[ 2 * k ] + pulses_in[ 2 * k + 1 ]; + if( sum > max_pulses ) { + return 1; + } + pulses_comb[ k ] = sum; + } + + return 0; +} + +/* Encode quantization indices of excitation */ +void silk_encode_pulses( + ec_enc *psRangeEnc, /* I/O compressor data structure */ + const opus_int signalType, /* I Signal type */ + const opus_int quantOffsetType, /* I quantOffsetType */ + opus_int8 pulses[], /* I quantization indices */ + const opus_int frame_length /* I Frame length */ +) +{ + opus_int i, k, j, iter, bit, nLS, scale_down, RateLevelIndex = 0; + opus_int32 abs_q, minSumBits_Q5, sumBits_Q5; + VARDECL( opus_int, abs_pulses ); + VARDECL( opus_int, sum_pulses ); + VARDECL( opus_int, nRshifts ); + opus_int pulses_comb[ 8 ]; + opus_int *abs_pulses_ptr; + const opus_int8 *pulses_ptr; + const opus_uint8 *cdf_ptr; + const opus_uint8 *nBits_ptr; + SAVE_STACK; + + silk_memset( pulses_comb, 0, 8 * sizeof( opus_int ) ); /* Fixing Valgrind reported problem*/ + + /****************************/ + /* Prepare for shell coding */ + /****************************/ + /* Calculate number of shell blocks */ + silk_assert( 1 << LOG2_SHELL_CODEC_FRAME_LENGTH == SHELL_CODEC_FRAME_LENGTH ); + iter = silk_RSHIFT( frame_length, LOG2_SHELL_CODEC_FRAME_LENGTH ); + if( iter * SHELL_CODEC_FRAME_LENGTH < frame_length ) { + celt_assert( frame_length == 12 * 10 ); /* Make sure only happens for 10 ms @ 12 kHz */ + iter++; + silk_memset( &pulses[ frame_length ], 0, SHELL_CODEC_FRAME_LENGTH * sizeof(opus_int8)); + } + + /* Take the absolute value of the pulses */ + ALLOC( abs_pulses, iter * SHELL_CODEC_FRAME_LENGTH, opus_int ); + silk_assert( !( SHELL_CODEC_FRAME_LENGTH & 3 ) ); + for( i = 0; i < iter * SHELL_CODEC_FRAME_LENGTH; i+=4 ) { + abs_pulses[i+0] = ( opus_int )silk_abs( pulses[ i + 0 ] ); + abs_pulses[i+1] = ( opus_int )silk_abs( pulses[ i + 1 ] ); + abs_pulses[i+2] = ( opus_int )silk_abs( pulses[ i + 2 ] ); + abs_pulses[i+3] = ( opus_int )silk_abs( pulses[ i + 3 ] ); + } + + /* Calc sum pulses per shell code frame */ + ALLOC( sum_pulses, iter, opus_int ); + ALLOC( nRshifts, iter, opus_int ); + abs_pulses_ptr = abs_pulses; + for( i = 0; i < iter; i++ ) { + nRshifts[ i ] = 0; + + while( 1 ) { + /* 1+1 -> 2 */ + scale_down = combine_and_check( pulses_comb, abs_pulses_ptr, silk_max_pulses_table[ 0 ], 8 ); + /* 2+2 -> 4 */ + scale_down += combine_and_check( pulses_comb, pulses_comb, silk_max_pulses_table[ 1 ], 4 ); + /* 4+4 -> 8 */ + scale_down += combine_and_check( pulses_comb, pulses_comb, silk_max_pulses_table[ 2 ], 2 ); + /* 8+8 -> 16 */ + scale_down += combine_and_check( &sum_pulses[ i ], pulses_comb, silk_max_pulses_table[ 3 ], 1 ); + + if( scale_down ) { + /* We need to downscale the quantization signal */ + nRshifts[ i ]++; + for( k = 0; k < SHELL_CODEC_FRAME_LENGTH; k++ ) { + abs_pulses_ptr[ k ] = silk_RSHIFT( abs_pulses_ptr[ k ], 1 ); + } + } else { + /* Jump out of while(1) loop and go to next shell coding frame */ + break; + } + } + abs_pulses_ptr += SHELL_CODEC_FRAME_LENGTH; + } + + /**************/ + /* Rate level */ + /**************/ + /* find rate level that leads to fewest bits for coding of pulses per block info */ + minSumBits_Q5 = silk_int32_MAX; + for( k = 0; k < N_RATE_LEVELS - 1; k++ ) { + nBits_ptr = silk_pulses_per_block_BITS_Q5[ k ]; + sumBits_Q5 = silk_rate_levels_BITS_Q5[ signalType >> 1 ][ k ]; + for( i = 0; i < iter; i++ ) { + if( nRshifts[ i ] > 0 ) { + sumBits_Q5 += nBits_ptr[ SILK_MAX_PULSES + 1 ]; + } else { + sumBits_Q5 += nBits_ptr[ sum_pulses[ i ] ]; + } + } + if( sumBits_Q5 < minSumBits_Q5 ) { + minSumBits_Q5 = sumBits_Q5; + RateLevelIndex = k; + } + } + ec_enc_icdf( psRangeEnc, RateLevelIndex, silk_rate_levels_iCDF[ signalType >> 1 ], 8 ); + + /***************************************************/ + /* Sum-Weighted-Pulses Encoding */ + /***************************************************/ + cdf_ptr = silk_pulses_per_block_iCDF[ RateLevelIndex ]; + for( i = 0; i < iter; i++ ) { + if( nRshifts[ i ] == 0 ) { + ec_enc_icdf( psRangeEnc, sum_pulses[ i ], cdf_ptr, 8 ); + } else { + ec_enc_icdf( psRangeEnc, SILK_MAX_PULSES + 1, cdf_ptr, 8 ); + for( k = 0; k < nRshifts[ i ] - 1; k++ ) { + ec_enc_icdf( psRangeEnc, SILK_MAX_PULSES + 1, silk_pulses_per_block_iCDF[ N_RATE_LEVELS - 1 ], 8 ); + } + ec_enc_icdf( psRangeEnc, sum_pulses[ i ], silk_pulses_per_block_iCDF[ N_RATE_LEVELS - 1 ], 8 ); + } + } + + /******************/ + /* Shell Encoding */ + /******************/ + for( i = 0; i < iter; i++ ) { + if( sum_pulses[ i ] > 0 ) { + silk_shell_encoder( psRangeEnc, &abs_pulses[ i * SHELL_CODEC_FRAME_LENGTH ] ); + } + } + + /****************/ + /* LSB Encoding */ + /****************/ + for( i = 0; i < iter; i++ ) { + if( nRshifts[ i ] > 0 ) { + pulses_ptr = &pulses[ i * SHELL_CODEC_FRAME_LENGTH ]; + nLS = nRshifts[ i ] - 1; + for( k = 0; k < SHELL_CODEC_FRAME_LENGTH; k++ ) { + abs_q = (opus_int8)silk_abs( pulses_ptr[ k ] ); + for( j = nLS; j > 0; j-- ) { + bit = silk_RSHIFT( abs_q, j ) & 1; + ec_enc_icdf( psRangeEnc, bit, silk_lsb_iCDF, 8 ); + } + bit = abs_q & 1; + ec_enc_icdf( psRangeEnc, bit, silk_lsb_iCDF, 8 ); + } + } + } + + /****************/ + /* Encode signs */ + /****************/ + silk_encode_signs( psRangeEnc, pulses, frame_length, signalType, quantOffsetType, sum_pulses ); + RESTORE_STACK; +} diff --git a/lib/opus/silk/errors.h b/lib/opus/silk/errors.h new file mode 100644 index 00000000..45070800 --- /dev/null +++ b/lib/opus/silk/errors.h @@ -0,0 +1,98 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_ERRORS_H +#define SILK_ERRORS_H + +#ifdef __cplusplus +extern "C" +{ +#endif + +/******************/ +/* Error messages */ +/******************/ +#define SILK_NO_ERROR 0 + +/**************************/ +/* Encoder error messages */ +/**************************/ + +/* Input length is not a multiple of 10 ms, or length is longer than the packet length */ +#define SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES -101 + +/* Sampling frequency not 8000, 12000 or 16000 Hertz */ +#define SILK_ENC_FS_NOT_SUPPORTED -102 + +/* Packet size not 10, 20, 40, or 60 ms */ +#define SILK_ENC_PACKET_SIZE_NOT_SUPPORTED -103 + +/* Allocated payload buffer too short */ +#define SILK_ENC_PAYLOAD_BUF_TOO_SHORT -104 + +/* Loss rate not between 0 and 100 percent */ +#define SILK_ENC_INVALID_LOSS_RATE -105 + +/* Complexity setting not valid, use 0...10 */ +#define SILK_ENC_INVALID_COMPLEXITY_SETTING -106 + +/* Inband FEC setting not valid, use 0 or 1 */ +#define SILK_ENC_INVALID_INBAND_FEC_SETTING -107 + +/* DTX setting not valid, use 0 or 1 */ +#define SILK_ENC_INVALID_DTX_SETTING -108 + +/* CBR setting not valid, use 0 or 1 */ +#define SILK_ENC_INVALID_CBR_SETTING -109 + +/* Internal encoder error */ +#define SILK_ENC_INTERNAL_ERROR -110 + +/* Internal encoder error */ +#define SILK_ENC_INVALID_NUMBER_OF_CHANNELS_ERROR -111 + +/**************************/ +/* Decoder error messages */ +/**************************/ + +/* Output sampling frequency lower than internal decoded sampling frequency */ +#define SILK_DEC_INVALID_SAMPLING_FREQUENCY -200 + +/* Payload size exceeded the maximum allowed 1024 bytes */ +#define SILK_DEC_PAYLOAD_TOO_LARGE -201 + +/* Payload has bit errors */ +#define SILK_DEC_PAYLOAD_ERROR -202 + +/* Payload has bit errors */ +#define SILK_DEC_INVALID_FRAME_SIZE -203 + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/fixed/LTP_analysis_filter_FIX.c b/lib/opus/silk/fixed/LTP_analysis_filter_FIX.c new file mode 100644 index 00000000..5574e706 --- /dev/null +++ b/lib/opus/silk/fixed/LTP_analysis_filter_FIX.c @@ -0,0 +1,90 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" + +void silk_LTP_analysis_filter_FIX( + opus_int16 *LTP_res, /* O LTP residual signal of length MAX_NB_SUBFR * ( pre_length + subfr_length ) */ + const opus_int16 *x, /* I Pointer to input signal with at least max( pitchL ) preceding samples */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ],/* I LTP_ORDER LTP coefficients for each MAX_NB_SUBFR subframe */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag, one for each subframe */ + const opus_int32 invGains_Q16[ MAX_NB_SUBFR ], /* I Inverse quantization gains, one for each subframe */ + const opus_int subfr_length, /* I Length of each subframe */ + const opus_int nb_subfr, /* I Number of subframes */ + const opus_int pre_length /* I Length of the preceding samples starting at &x[0] for each subframe */ +) +{ + const opus_int16 *x_ptr, *x_lag_ptr; + opus_int16 Btmp_Q14[ LTP_ORDER ]; + opus_int16 *LTP_res_ptr; + opus_int k, i; + opus_int32 LTP_est; + + x_ptr = x; + LTP_res_ptr = LTP_res; + for( k = 0; k < nb_subfr; k++ ) { + + x_lag_ptr = x_ptr - pitchL[ k ]; + + Btmp_Q14[ 0 ] = LTPCoef_Q14[ k * LTP_ORDER ]; + Btmp_Q14[ 1 ] = LTPCoef_Q14[ k * LTP_ORDER + 1 ]; + Btmp_Q14[ 2 ] = LTPCoef_Q14[ k * LTP_ORDER + 2 ]; + Btmp_Q14[ 3 ] = LTPCoef_Q14[ k * LTP_ORDER + 3 ]; + Btmp_Q14[ 4 ] = LTPCoef_Q14[ k * LTP_ORDER + 4 ]; + + /* LTP analysis FIR filter */ + for( i = 0; i < subfr_length + pre_length; i++ ) { + LTP_res_ptr[ i ] = x_ptr[ i ]; + + /* Long-term prediction */ + LTP_est = silk_SMULBB( x_lag_ptr[ LTP_ORDER / 2 ], Btmp_Q14[ 0 ] ); + LTP_est = silk_SMLABB_ovflw( LTP_est, x_lag_ptr[ 1 ], Btmp_Q14[ 1 ] ); + LTP_est = silk_SMLABB_ovflw( LTP_est, x_lag_ptr[ 0 ], Btmp_Q14[ 2 ] ); + LTP_est = silk_SMLABB_ovflw( LTP_est, x_lag_ptr[ -1 ], Btmp_Q14[ 3 ] ); + LTP_est = silk_SMLABB_ovflw( LTP_est, x_lag_ptr[ -2 ], Btmp_Q14[ 4 ] ); + + LTP_est = silk_RSHIFT_ROUND( LTP_est, 14 ); /* round and -> Q0*/ + + /* Subtract long-term prediction */ + LTP_res_ptr[ i ] = (opus_int16)silk_SAT16( (opus_int32)x_ptr[ i ] - LTP_est ); + + /* Scale residual */ + LTP_res_ptr[ i ] = silk_SMULWB( invGains_Q16[ k ], LTP_res_ptr[ i ] ); + + x_lag_ptr++; + } + + /* Update pointers */ + LTP_res_ptr += subfr_length + pre_length; + x_ptr += subfr_length; + } +} + diff --git a/lib/opus/silk/fixed/LTP_scale_ctrl_FIX.c b/lib/opus/silk/fixed/LTP_scale_ctrl_FIX.c new file mode 100644 index 00000000..db1016e0 --- /dev/null +++ b/lib/opus/silk/fixed/LTP_scale_ctrl_FIX.c @@ -0,0 +1,58 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" + +/* Calculation of LTP state scaling */ +void silk_LTP_scale_ctrl_FIX( + silk_encoder_state_FIX *psEnc, /* I/O encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + opus_int round_loss; + + if( condCoding == CODE_INDEPENDENTLY ) { + /* Only scale if first frame in packet */ + round_loss = psEnc->sCmn.PacketLoss_perc * psEnc->sCmn.nFramesPerPacket; + if ( psEnc->sCmn.LBRR_flag ) { + /* LBRR reduces the effective loss. In practice, it does not square the loss because + losses aren't independent, but that still seems to work best. We also never go below 2%. */ + round_loss = 2 + silk_SMULBB( round_loss, round_loss ) / 100; + } + psEnc->sCmn.indices.LTP_scaleIndex = silk_SMULBB( psEncCtrl->LTPredCodGain_Q7, round_loss ) > silk_log2lin( 128*7 + 2900-psEnc->sCmn.SNR_dB_Q7 ); + psEnc->sCmn.indices.LTP_scaleIndex += silk_SMULBB( psEncCtrl->LTPredCodGain_Q7, round_loss ) > silk_log2lin( 128*7 + 3900-psEnc->sCmn.SNR_dB_Q7 ); + } else { + /* Default is minimum scaling */ + psEnc->sCmn.indices.LTP_scaleIndex = 0; + } + psEncCtrl->LTP_scale_Q14 = silk_LTPScales_table_Q14[ psEnc->sCmn.indices.LTP_scaleIndex ]; +} diff --git a/lib/opus/silk/fixed/apply_sine_window_FIX.c b/lib/opus/silk/fixed/apply_sine_window_FIX.c new file mode 100644 index 00000000..03e088a6 --- /dev/null +++ b/lib/opus/silk/fixed/apply_sine_window_FIX.c @@ -0,0 +1,101 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Apply sine window to signal vector. */ +/* Window types: */ +/* 1 -> sine window from 0 to pi/2 */ +/* 2 -> sine window from pi/2 to pi */ +/* Every other sample is linearly interpolated, for speed. */ +/* Window length must be between 16 and 120 (incl) and a multiple of 4. */ + +/* Matlab code for table: + for k=16:9*4:16+2*9*4, fprintf(' %7.d,', -round(65536*pi ./ (k:4:k+8*4))); fprintf('\n'); end +*/ +static const opus_int16 freq_table_Q16[ 27 ] = { + 12111, 9804, 8235, 7100, 6239, 5565, 5022, 4575, 4202, + 3885, 3612, 3375, 3167, 2984, 2820, 2674, 2542, 2422, + 2313, 2214, 2123, 2038, 1961, 1889, 1822, 1760, 1702, +}; + +void silk_apply_sine_window( + opus_int16 px_win[], /* O Pointer to windowed signal */ + const opus_int16 px[], /* I Pointer to input signal */ + const opus_int win_type, /* I Selects a window type */ + const opus_int length /* I Window length, multiple of 4 */ +) +{ + opus_int k, f_Q16, c_Q16; + opus_int32 S0_Q16, S1_Q16; + + celt_assert( win_type == 1 || win_type == 2 ); + + /* Length must be in a range from 16 to 120 and a multiple of 4 */ + celt_assert( length >= 16 && length <= 120 ); + celt_assert( ( length & 3 ) == 0 ); + + /* Frequency */ + k = ( length >> 2 ) - 4; + celt_assert( k >= 0 && k <= 26 ); + f_Q16 = (opus_int)freq_table_Q16[ k ]; + + /* Factor used for cosine approximation */ + c_Q16 = silk_SMULWB( (opus_int32)f_Q16, -f_Q16 ); + silk_assert( c_Q16 >= -32768 ); + + /* initialize state */ + if( win_type == 1 ) { + /* start from 0 */ + S0_Q16 = 0; + /* approximation of sin(f) */ + S1_Q16 = f_Q16 + silk_RSHIFT( length, 3 ); + } else { + /* start from 1 */ + S0_Q16 = ( (opus_int32)1 << 16 ); + /* approximation of cos(f) */ + S1_Q16 = ( (opus_int32)1 << 16 ) + silk_RSHIFT( c_Q16, 1 ) + silk_RSHIFT( length, 4 ); + } + + /* Uses the recursive equation: sin(n*f) = 2 * cos(f) * sin((n-1)*f) - sin((n-2)*f) */ + /* 4 samples at a time */ + for( k = 0; k < length; k += 4 ) { + px_win[ k ] = (opus_int16)silk_SMULWB( silk_RSHIFT( S0_Q16 + S1_Q16, 1 ), px[ k ] ); + px_win[ k + 1 ] = (opus_int16)silk_SMULWB( S1_Q16, px[ k + 1] ); + S0_Q16 = silk_SMULWB( S1_Q16, c_Q16 ) + silk_LSHIFT( S1_Q16, 1 ) - S0_Q16 + 1; + S0_Q16 = silk_min( S0_Q16, ( (opus_int32)1 << 16 ) ); + + px_win[ k + 2 ] = (opus_int16)silk_SMULWB( silk_RSHIFT( S0_Q16 + S1_Q16, 1 ), px[ k + 2] ); + px_win[ k + 3 ] = (opus_int16)silk_SMULWB( S0_Q16, px[ k + 3 ] ); + S1_Q16 = silk_SMULWB( S0_Q16, c_Q16 ) + silk_LSHIFT( S0_Q16, 1 ) - S1_Q16; + S1_Q16 = silk_min( S1_Q16, ( (opus_int32)1 << 16 ) ); + } +} diff --git a/lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_arm.h b/lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_arm.h new file mode 100644 index 00000000..1992e432 --- /dev/null +++ b/lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_arm.h @@ -0,0 +1,68 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_WARPED_AUTOCORRELATION_FIX_ARM_H +# define SILK_WARPED_AUTOCORRELATION_FIX_ARM_H + +# include "celt/arm/armcpu.h" + +# if defined(FIXED_POINT) + +# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) +void silk_warped_autocorrelation_FIX_neon( + opus_int32 *corr, /* O Result [order + 1] */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *input, /* I Input data to correlate */ + const opus_int warping_Q16, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +); + +# if !defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_PRESUME_NEON) +# define OVERRIDE_silk_warped_autocorrelation_FIX (1) +# define silk_warped_autocorrelation_FIX(corr, scale, input, warping_Q16, length, order, arch) \ + ((void)(arch), PRESUME_NEON(silk_warped_autocorrelation_FIX)(corr, scale, input, warping_Q16, length, order)) +# endif +# endif + +# if !defined(OVERRIDE_silk_warped_autocorrelation_FIX) +/*Is run-time CPU detection enabled on this platform?*/ +# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)) +extern void (*const SILK_WARPED_AUTOCORRELATION_FIX_IMPL[OPUS_ARCHMASK+1])(opus_int32*, opus_int*, const opus_int16*, const opus_int, const opus_int, const opus_int); +# define OVERRIDE_silk_warped_autocorrelation_FIX (1) +# define silk_warped_autocorrelation_FIX(corr, scale, input, warping_Q16, length, order, arch) \ + ((*SILK_WARPED_AUTOCORRELATION_FIX_IMPL[(arch)&OPUS_ARCHMASK])(corr, scale, input, warping_Q16, length, order)) +# elif defined(OPUS_ARM_PRESUME_NEON_INTR) +# define OVERRIDE_silk_warped_autocorrelation_FIX (1) +# define silk_warped_autocorrelation_FIX(corr, scale, input, warping_Q16, length, order, arch) \ + ((void)(arch), silk_warped_autocorrelation_FIX_neon(corr, scale, input, warping_Q16, length, order)) +# endif +# endif + +# endif /* end FIXED_POINT */ + +#endif /* end SILK_WARPED_AUTOCORRELATION_FIX_ARM_H */ diff --git a/lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_neon_intr.c b/lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_neon_intr.c new file mode 100644 index 00000000..6f3be025 --- /dev/null +++ b/lib/opus/silk/fixed/arm/warped_autocorrelation_FIX_neon_intr.c @@ -0,0 +1,265 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc., Jean-Marc Valin +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#ifdef OPUS_CHECK_ASM +# include +#endif +#include "stack_alloc.h" +#include "main_FIX.h" + +static OPUS_INLINE void calc_corr( const opus_int32 *const input_QS, opus_int64 *const corr_QC, const opus_int offset, const int32x4_t state_QS_s32x4 ) +{ + int64x2_t corr_QC_s64x2[ 2 ], t_s64x2[ 2 ]; + const int32x4_t input_QS_s32x4 = vld1q_s32( input_QS + offset ); + corr_QC_s64x2[ 0 ] = vld1q_s64( corr_QC + offset + 0 ); + corr_QC_s64x2[ 1 ] = vld1q_s64( corr_QC + offset + 2 ); + t_s64x2[ 0 ] = vmull_s32( vget_low_s32( state_QS_s32x4 ), vget_low_s32( input_QS_s32x4 ) ); + t_s64x2[ 1 ] = vmull_s32( vget_high_s32( state_QS_s32x4 ), vget_high_s32( input_QS_s32x4 ) ); + corr_QC_s64x2[ 0 ] = vsraq_n_s64( corr_QC_s64x2[ 0 ], t_s64x2[ 0 ], 2 * QS - QC ); + corr_QC_s64x2[ 1 ] = vsraq_n_s64( corr_QC_s64x2[ 1 ], t_s64x2[ 1 ], 2 * QS - QC ); + vst1q_s64( corr_QC + offset + 0, corr_QC_s64x2[ 0 ] ); + vst1q_s64( corr_QC + offset + 2, corr_QC_s64x2[ 1 ] ); +} + +static OPUS_INLINE int32x4_t calc_state( const int32x4_t state_QS0_s32x4, const int32x4_t state_QS0_1_s32x4, const int32x4_t state_QS1_1_s32x4, const int32x4_t warping_Q16_s32x4 ) +{ + int32x4_t t_s32x4 = vsubq_s32( state_QS0_s32x4, state_QS0_1_s32x4 ); + t_s32x4 = vqdmulhq_s32( t_s32x4, warping_Q16_s32x4 ); + return vaddq_s32( state_QS1_1_s32x4, t_s32x4 ); +} + +void silk_warped_autocorrelation_FIX_neon( + opus_int32 *corr, /* O Result [order + 1] */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *input, /* I Input data to correlate */ + const opus_int warping_Q16, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +) +{ + if( ( MAX_SHAPE_LPC_ORDER > 24 ) || ( order < 6 ) ) { + silk_warped_autocorrelation_FIX_c( corr, scale, input, warping_Q16, length, order ); + } else { + opus_int n, i, lsh; + opus_int64 corr_QC[ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; /* In reverse order */ + opus_int64 corr_QC_orderT; + int64x2_t lsh_s64x2; + const opus_int orderT = ( order + 3 ) & ~3; + opus_int64 *corr_QCT; + opus_int32 *input_QS; + VARDECL( opus_int32, input_QST ); + VARDECL( opus_int32, state ); + SAVE_STACK; + + /* Order must be even */ + silk_assert( ( order & 1 ) == 0 ); + silk_assert( 2 * QS - QC >= 0 ); + + /* The additional +4 is to ensure a later vld1q_s32 call does not overflow. */ + /* Strictly, only +3 is needed but +4 simplifies initialization using the 4x32 neon load. */ + ALLOC( input_QST, length + 2 * MAX_SHAPE_LPC_ORDER + 4, opus_int32 ); + + input_QS = input_QST; + /* input_QS has zero paddings in the beginning and end. */ + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + + /* Loop over samples */ + for( n = 0; n < length - 7; n += 8, input_QS += 8 ) { + const int16x8_t t0_s16x4 = vld1q_s16( input + n ); + vst1q_s32( input_QS + 0, vshll_n_s16( vget_low_s16( t0_s16x4 ), QS ) ); + vst1q_s32( input_QS + 4, vshll_n_s16( vget_high_s16( t0_s16x4 ), QS ) ); + } + for( ; n < length; n++, input_QS++ ) { + input_QS[ 0 ] = silk_LSHIFT32( (opus_int32)input[ n ], QS ); + } + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS += 4; + vst1q_s32( input_QS, vdupq_n_s32( 0 ) ); + input_QS = input_QST + MAX_SHAPE_LPC_ORDER - orderT; + + /* The following loop runs ( length + order ) times, with ( order ) extra epilogues. */ + /* The zero paddings in input_QS guarantee corr_QC's correctness even with the extra epilogues. */ + /* The values of state_QS will be polluted by the extra epilogues, however they are temporary values. */ + + /* Keep the C code here to help understand the intrinsics optimization. */ + /* + { + opus_int32 state_QS[ 2 ][ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; + opus_int32 *state_QST[ 3 ]; + state_QST[ 0 ] = state_QS[ 0 ]; + state_QST[ 1 ] = state_QS[ 1 ]; + for( n = 0; n < length + order; n++, input_QS++ ) { + state_QST[ 0 ][ orderT ] = input_QS[ orderT ]; + for( i = 0; i < orderT; i++ ) { + corr_QC[ i ] += silk_RSHIFT64( silk_SMULL( state_QST[ 0 ][ i ], input_QS[ i ] ), 2 * QS - QC ); + state_QST[ 1 ][ i ] = silk_SMLAWB( state_QST[ 1 ][ i + 1 ], state_QST[ 0 ][ i ] - state_QST[ 0 ][ i + 1 ], warping_Q16 ); + } + state_QST[ 2 ] = state_QST[ 0 ]; + state_QST[ 0 ] = state_QST[ 1 ]; + state_QST[ 1 ] = state_QST[ 2 ]; + } + } + */ + + { + const int32x4_t warping_Q16_s32x4 = vdupq_n_s32( warping_Q16 << 15 ); + const opus_int32 *in = input_QS + orderT; + opus_int o = orderT; + int32x4_t state_QS_s32x4[ 3 ][ 2 ]; + + /* The additional +4 is to ensure a later vld1q_s32 call does not overflow. */ + ALLOC( state, length + order + 4, opus_int32 ); + state_QS_s32x4[ 2 ][ 1 ] = vdupq_n_s32( 0 ); + + /* Calculate 8 taps of all inputs in each loop. */ + do { + state_QS_s32x4[ 0 ][ 0 ] = state_QS_s32x4[ 0 ][ 1 ] = + state_QS_s32x4[ 1 ][ 0 ] = state_QS_s32x4[ 1 ][ 1 ] = vdupq_n_s32( 0 ); + n = 0; + do { + calc_corr( input_QS + n, corr_QC, o - 8, state_QS_s32x4[ 0 ][ 0 ] ); + calc_corr( input_QS + n, corr_QC, o - 4, state_QS_s32x4[ 0 ][ 1 ] ); + state_QS_s32x4[ 2 ][ 1 ] = vld1q_s32( in + n ); + vst1q_lane_s32( state + n, state_QS_s32x4[ 0 ][ 0 ], 0 ); + state_QS_s32x4[ 2 ][ 0 ] = vextq_s32( state_QS_s32x4[ 0 ][ 0 ], state_QS_s32x4[ 0 ][ 1 ], 1 ); + state_QS_s32x4[ 2 ][ 1 ] = vextq_s32( state_QS_s32x4[ 0 ][ 1 ], state_QS_s32x4[ 2 ][ 1 ], 1 ); + state_QS_s32x4[ 0 ][ 0 ] = calc_state( state_QS_s32x4[ 0 ][ 0 ], state_QS_s32x4[ 2 ][ 0 ], state_QS_s32x4[ 1 ][ 0 ], warping_Q16_s32x4 ); + state_QS_s32x4[ 0 ][ 1 ] = calc_state( state_QS_s32x4[ 0 ][ 1 ], state_QS_s32x4[ 2 ][ 1 ], state_QS_s32x4[ 1 ][ 1 ], warping_Q16_s32x4 ); + state_QS_s32x4[ 1 ][ 0 ] = state_QS_s32x4[ 2 ][ 0 ]; + state_QS_s32x4[ 1 ][ 1 ] = state_QS_s32x4[ 2 ][ 1 ]; + } while( ++n < ( length + order ) ); + in = state; + o -= 8; + } while( o > 4 ); + + if( o ) { + /* Calculate the last 4 taps of all inputs. */ + opus_int32 *stateT = state; + silk_assert( o == 4 ); + state_QS_s32x4[ 0 ][ 0 ] = state_QS_s32x4[ 1 ][ 0 ] = vdupq_n_s32( 0 ); + n = length + order; + do { + calc_corr( input_QS, corr_QC, 0, state_QS_s32x4[ 0 ][ 0 ] ); + state_QS_s32x4[ 2 ][ 0 ] = vld1q_s32( stateT ); + vst1q_lane_s32( stateT, state_QS_s32x4[ 0 ][ 0 ], 0 ); + state_QS_s32x4[ 2 ][ 0 ] = vextq_s32( state_QS_s32x4[ 0 ][ 0 ], state_QS_s32x4[ 2 ][ 0 ], 1 ); + state_QS_s32x4[ 0 ][ 0 ] = calc_state( state_QS_s32x4[ 0 ][ 0 ], state_QS_s32x4[ 2 ][ 0 ], state_QS_s32x4[ 1 ][ 0 ], warping_Q16_s32x4 ); + state_QS_s32x4[ 1 ][ 0 ] = state_QS_s32x4[ 2 ][ 0 ]; + input_QS++; + stateT++; + } while( --n ); + } + } + + { + const opus_int16 *inputT = input; + int32x4_t t_s32x4; + int64x1_t t_s64x1; + int64x2_t t_s64x2 = vdupq_n_s64( 0 ); + for( n = 0; n <= length - 8; n += 8 ) { + int16x8_t input_s16x8 = vld1q_s16( inputT ); + t_s32x4 = vmull_s16( vget_low_s16( input_s16x8 ), vget_low_s16( input_s16x8 ) ); + t_s32x4 = vmlal_s16( t_s32x4, vget_high_s16( input_s16x8 ), vget_high_s16( input_s16x8 ) ); + t_s64x2 = vaddw_s32( t_s64x2, vget_low_s32( t_s32x4 ) ); + t_s64x2 = vaddw_s32( t_s64x2, vget_high_s32( t_s32x4 ) ); + inputT += 8; + } + t_s64x1 = vadd_s64( vget_low_s64( t_s64x2 ), vget_high_s64( t_s64x2 ) ); + corr_QC_orderT = vget_lane_s64( t_s64x1, 0 ); + for( ; n < length; n++ ) { + corr_QC_orderT += silk_SMULL( input[ n ], input[ n ] ); + } + corr_QC_orderT = silk_LSHIFT64( corr_QC_orderT, QC ); + corr_QC[ orderT ] = corr_QC_orderT; + } + + corr_QCT = corr_QC + orderT - order; + lsh = silk_CLZ64( corr_QC_orderT ) - 35; + lsh = silk_LIMIT( lsh, -12 - QC, 30 - QC ); + *scale = -( QC + lsh ); + silk_assert( *scale >= -30 && *scale <= 12 ); + lsh_s64x2 = vdupq_n_s64( lsh ); + for( i = 0; i <= order - 3; i += 4 ) { + int32x4_t corr_s32x4; + int64x2_t corr_QC0_s64x2, corr_QC1_s64x2; + corr_QC0_s64x2 = vld1q_s64( corr_QCT + i ); + corr_QC1_s64x2 = vld1q_s64( corr_QCT + i + 2 ); + corr_QC0_s64x2 = vshlq_s64( corr_QC0_s64x2, lsh_s64x2 ); + corr_QC1_s64x2 = vshlq_s64( corr_QC1_s64x2, lsh_s64x2 ); + corr_s32x4 = vcombine_s32( vmovn_s64( corr_QC1_s64x2 ), vmovn_s64( corr_QC0_s64x2 ) ); + corr_s32x4 = vrev64q_s32( corr_s32x4 ); + vst1q_s32( corr + order - i - 3, corr_s32x4 ); + } + if( lsh >= 0 ) { + for( ; i < order + 1; i++ ) { + corr[ order - i ] = (opus_int32)silk_CHECK_FIT32( silk_LSHIFT64( corr_QCT[ i ], lsh ) ); + } + } else { + for( ; i < order + 1; i++ ) { + corr[ order - i ] = (opus_int32)silk_CHECK_FIT32( silk_RSHIFT64( corr_QCT[ i ], -lsh ) ); + } + } + silk_assert( corr_QCT[ order ] >= 0 ); /* If breaking, decrease QC*/ + RESTORE_STACK; + } + +#ifdef OPUS_CHECK_ASM + { + opus_int32 corr_c[ MAX_SHAPE_LPC_ORDER + 1 ]; + opus_int scale_c; + silk_warped_autocorrelation_FIX_c( corr_c, &scale_c, input, warping_Q16, length, order ); + silk_assert( !memcmp( corr_c, corr, sizeof( corr_c[ 0 ] ) * ( order + 1 ) ) ); + silk_assert( scale_c == *scale ); + } +#endif +} diff --git a/lib/opus/silk/fixed/autocorr_FIX.c b/lib/opus/silk/fixed/autocorr_FIX.c new file mode 100644 index 00000000..de95c986 --- /dev/null +++ b/lib/opus/silk/fixed/autocorr_FIX.c @@ -0,0 +1,48 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "celt_lpc.h" + +/* Compute autocorrelation */ +void silk_autocorr( + opus_int32 *results, /* O Result (length correlationCount) */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *inputData, /* I Input data to correlate */ + const opus_int inputDataSize, /* I Length of input */ + const opus_int correlationCount, /* I Number of correlation taps to compute */ + int arch /* I Run-time architecture */ +) +{ + opus_int corrCount; + corrCount = silk_min_int( inputDataSize, correlationCount ); + *scale = _celt_autocorr(inputData, results, NULL, 0, corrCount-1, inputDataSize, arch); +} diff --git a/lib/opus/silk/fixed/burg_modified_FIX.c b/lib/opus/silk/fixed/burg_modified_FIX.c new file mode 100644 index 00000000..185a12b1 --- /dev/null +++ b/lib/opus/silk/fixed/burg_modified_FIX.c @@ -0,0 +1,280 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "define.h" +#include "tuning_parameters.h" +#include "pitch.h" + +#define MAX_FRAME_SIZE 384 /* subfr_length * nb_subfr = ( 0.005 * 16000 + 16 ) * 4 = 384 */ + +#define QA 25 +#define N_BITS_HEAD_ROOM 3 +#define MIN_RSHIFTS -16 +#define MAX_RSHIFTS (32 - QA) + +/* Compute reflection coefficients from input signal */ +void silk_burg_modified_c( + opus_int32 *res_nrg, /* O Residual energy */ + opus_int *res_nrg_Q, /* O Residual energy Q value */ + opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ + const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ + const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ + const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I Number of subframes stacked in x */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */ +) +{ + opus_int k, n, s, lz, rshifts, reached_max_gain; + opus_int32 C0, num, nrg, rc_Q31, invGain_Q30, Atmp_QA, Atmp1, tmp1, tmp2, x1, x2; + const opus_int16 *x_ptr; + opus_int32 C_first_row[ SILK_MAX_ORDER_LPC ]; + opus_int32 C_last_row[ SILK_MAX_ORDER_LPC ]; + opus_int32 Af_QA[ SILK_MAX_ORDER_LPC ]; + opus_int32 CAf[ SILK_MAX_ORDER_LPC + 1 ]; + opus_int32 CAb[ SILK_MAX_ORDER_LPC + 1 ]; + opus_int32 xcorr[ SILK_MAX_ORDER_LPC ]; + opus_int64 C0_64; + + celt_assert( subfr_length * nb_subfr <= MAX_FRAME_SIZE ); + + /* Compute autocorrelations, added over subframes */ + C0_64 = silk_inner_prod16( x, x, subfr_length*nb_subfr, arch ); + lz = silk_CLZ64(C0_64); + rshifts = 32 + 1 + N_BITS_HEAD_ROOM - lz; + if (rshifts > MAX_RSHIFTS) rshifts = MAX_RSHIFTS; + if (rshifts < MIN_RSHIFTS) rshifts = MIN_RSHIFTS; + + if (rshifts > 0) { + C0 = (opus_int32)silk_RSHIFT64(C0_64, rshifts ); + } else { + C0 = silk_LSHIFT32((opus_int32)C0_64, -rshifts ); + } + + CAb[ 0 ] = CAf[ 0 ] = C0 + silk_SMMUL( SILK_FIX_CONST( FIND_LPC_COND_FAC, 32 ), C0 ) + 1; /* Q(-rshifts) */ + silk_memset( C_first_row, 0, SILK_MAX_ORDER_LPC * sizeof( opus_int32 ) ); + if( rshifts > 0 ) { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + for( n = 1; n < D + 1; n++ ) { + C_first_row[ n - 1 ] += (opus_int32)silk_RSHIFT64( + silk_inner_prod16( x_ptr, x_ptr + n, subfr_length - n, arch ), rshifts ); + } + } + } else { + for( s = 0; s < nb_subfr; s++ ) { + int i; + opus_int32 d; + x_ptr = x + s * subfr_length; + celt_pitch_xcorr(x_ptr, x_ptr + 1, xcorr, subfr_length - D, D, arch ); + for( n = 1; n < D + 1; n++ ) { + for ( i = n + subfr_length - D, d = 0; i < subfr_length; i++ ) + d = MAC16_16( d, x_ptr[ i ], x_ptr[ i - n ] ); + xcorr[ n - 1 ] += d; + } + for( n = 1; n < D + 1; n++ ) { + C_first_row[ n - 1 ] += silk_LSHIFT32( xcorr[ n - 1 ], -rshifts ); + } + } + } + silk_memcpy( C_last_row, C_first_row, SILK_MAX_ORDER_LPC * sizeof( opus_int32 ) ); + + /* Initialize */ + CAb[ 0 ] = CAf[ 0 ] = C0 + silk_SMMUL( SILK_FIX_CONST( FIND_LPC_COND_FAC, 32 ), C0 ) + 1; /* Q(-rshifts) */ + + invGain_Q30 = (opus_int32)1 << 30; + reached_max_gain = 0; + for( n = 0; n < D; n++ ) { + /* Update first row of correlation matrix (without first element) */ + /* Update last row of correlation matrix (without last element, stored in reversed order) */ + /* Update C * Af */ + /* Update C * flipud(Af) (stored in reversed order) */ + if( rshifts > -2 ) { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + x1 = -silk_LSHIFT32( (opus_int32)x_ptr[ n ], 16 - rshifts ); /* Q(16-rshifts) */ + x2 = -silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], 16 - rshifts ); /* Q(16-rshifts) */ + tmp1 = silk_LSHIFT32( (opus_int32)x_ptr[ n ], QA - 16 ); /* Q(QA-16) */ + tmp2 = silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], QA - 16 ); /* Q(QA-16) */ + for( k = 0; k < n; k++ ) { + C_first_row[ k ] = silk_SMLAWB( C_first_row[ k ], x1, x_ptr[ n - k - 1 ] ); /* Q( -rshifts ) */ + C_last_row[ k ] = silk_SMLAWB( C_last_row[ k ], x2, x_ptr[ subfr_length - n + k ] ); /* Q( -rshifts ) */ + Atmp_QA = Af_QA[ k ]; + tmp1 = silk_SMLAWB( tmp1, Atmp_QA, x_ptr[ n - k - 1 ] ); /* Q(QA-16) */ + tmp2 = silk_SMLAWB( tmp2, Atmp_QA, x_ptr[ subfr_length - n + k ] ); /* Q(QA-16) */ + } + tmp1 = silk_LSHIFT32( -tmp1, 32 - QA - rshifts ); /* Q(16-rshifts) */ + tmp2 = silk_LSHIFT32( -tmp2, 32 - QA - rshifts ); /* Q(16-rshifts) */ + for( k = 0; k <= n; k++ ) { + CAf[ k ] = silk_SMLAWB( CAf[ k ], tmp1, x_ptr[ n - k ] ); /* Q( -rshift ) */ + CAb[ k ] = silk_SMLAWB( CAb[ k ], tmp2, x_ptr[ subfr_length - n + k - 1 ] ); /* Q( -rshift ) */ + } + } + } else { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + x1 = -silk_LSHIFT32( (opus_int32)x_ptr[ n ], -rshifts ); /* Q( -rshifts ) */ + x2 = -silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], -rshifts ); /* Q( -rshifts ) */ + tmp1 = silk_LSHIFT32( (opus_int32)x_ptr[ n ], 17 ); /* Q17 */ + tmp2 = silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], 17 ); /* Q17 */ + for( k = 0; k < n; k++ ) { + C_first_row[ k ] = silk_MLA( C_first_row[ k ], x1, x_ptr[ n - k - 1 ] ); /* Q( -rshifts ) */ + C_last_row[ k ] = silk_MLA( C_last_row[ k ], x2, x_ptr[ subfr_length - n + k ] ); /* Q( -rshifts ) */ + Atmp1 = silk_RSHIFT_ROUND( Af_QA[ k ], QA - 17 ); /* Q17 */ + /* We sometimes get overflows in the multiplications (even beyond +/- 2^32), + but they cancel each other and the real result seems to always fit in a 32-bit + signed integer. This was determined experimentally, not theoretically (unfortunately). */ + tmp1 = silk_MLA_ovflw( tmp1, x_ptr[ n - k - 1 ], Atmp1 ); /* Q17 */ + tmp2 = silk_MLA_ovflw( tmp2, x_ptr[ subfr_length - n + k ], Atmp1 ); /* Q17 */ + } + tmp1 = -tmp1; /* Q17 */ + tmp2 = -tmp2; /* Q17 */ + for( k = 0; k <= n; k++ ) { + CAf[ k ] = silk_SMLAWW( CAf[ k ], tmp1, + silk_LSHIFT32( (opus_int32)x_ptr[ n - k ], -rshifts - 1 ) ); /* Q( -rshift ) */ + CAb[ k ] = silk_SMLAWW( CAb[ k ], tmp2, + silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n + k - 1 ], -rshifts - 1 ) ); /* Q( -rshift ) */ + } + } + } + + /* Calculate nominator and denominator for the next order reflection (parcor) coefficient */ + tmp1 = C_first_row[ n ]; /* Q( -rshifts ) */ + tmp2 = C_last_row[ n ]; /* Q( -rshifts ) */ + num = 0; /* Q( -rshifts ) */ + nrg = silk_ADD32( CAb[ 0 ], CAf[ 0 ] ); /* Q( 1-rshifts ) */ + for( k = 0; k < n; k++ ) { + Atmp_QA = Af_QA[ k ]; + lz = silk_CLZ32( silk_abs( Atmp_QA ) ) - 1; + lz = silk_min( 32 - QA, lz ); + Atmp1 = silk_LSHIFT32( Atmp_QA, lz ); /* Q( QA + lz ) */ + + tmp1 = silk_ADD_LSHIFT32( tmp1, silk_SMMUL( C_last_row[ n - k - 1 ], Atmp1 ), 32 - QA - lz ); /* Q( -rshifts ) */ + tmp2 = silk_ADD_LSHIFT32( tmp2, silk_SMMUL( C_first_row[ n - k - 1 ], Atmp1 ), 32 - QA - lz ); /* Q( -rshifts ) */ + num = silk_ADD_LSHIFT32( num, silk_SMMUL( CAb[ n - k ], Atmp1 ), 32 - QA - lz ); /* Q( -rshifts ) */ + nrg = silk_ADD_LSHIFT32( nrg, silk_SMMUL( silk_ADD32( CAb[ k + 1 ], CAf[ k + 1 ] ), + Atmp1 ), 32 - QA - lz ); /* Q( 1-rshifts ) */ + } + CAf[ n + 1 ] = tmp1; /* Q( -rshifts ) */ + CAb[ n + 1 ] = tmp2; /* Q( -rshifts ) */ + num = silk_ADD32( num, tmp2 ); /* Q( -rshifts ) */ + num = silk_LSHIFT32( -num, 1 ); /* Q( 1-rshifts ) */ + + /* Calculate the next order reflection (parcor) coefficient */ + if( silk_abs( num ) < nrg ) { + rc_Q31 = silk_DIV32_varQ( num, nrg, 31 ); + } else { + rc_Q31 = ( num > 0 ) ? silk_int32_MAX : silk_int32_MIN; + } + + /* Update inverse prediction gain */ + tmp1 = ( (opus_int32)1 << 30 ) - silk_SMMUL( rc_Q31, rc_Q31 ); + tmp1 = silk_LSHIFT( silk_SMMUL( invGain_Q30, tmp1 ), 2 ); + if( tmp1 <= minInvGain_Q30 ) { + /* Max prediction gain exceeded; set reflection coefficient such that max prediction gain is exactly hit */ + tmp2 = ( (opus_int32)1 << 30 ) - silk_DIV32_varQ( minInvGain_Q30, invGain_Q30, 30 ); /* Q30 */ + rc_Q31 = silk_SQRT_APPROX( tmp2 ); /* Q15 */ + if( rc_Q31 > 0 ) { + /* Newton-Raphson iteration */ + rc_Q31 = silk_RSHIFT32( rc_Q31 + silk_DIV32( tmp2, rc_Q31 ), 1 ); /* Q15 */ + rc_Q31 = silk_LSHIFT32( rc_Q31, 16 ); /* Q31 */ + if( num < 0 ) { + /* Ensure adjusted reflection coefficients has the original sign */ + rc_Q31 = -rc_Q31; + } + } + invGain_Q30 = minInvGain_Q30; + reached_max_gain = 1; + } else { + invGain_Q30 = tmp1; + } + + /* Update the AR coefficients */ + for( k = 0; k < (n + 1) >> 1; k++ ) { + tmp1 = Af_QA[ k ]; /* QA */ + tmp2 = Af_QA[ n - k - 1 ]; /* QA */ + Af_QA[ k ] = silk_ADD_LSHIFT32( tmp1, silk_SMMUL( tmp2, rc_Q31 ), 1 ); /* QA */ + Af_QA[ n - k - 1 ] = silk_ADD_LSHIFT32( tmp2, silk_SMMUL( tmp1, rc_Q31 ), 1 ); /* QA */ + } + Af_QA[ n ] = silk_RSHIFT32( rc_Q31, 31 - QA ); /* QA */ + + if( reached_max_gain ) { + /* Reached max prediction gain; set remaining coefficients to zero and exit loop */ + for( k = n + 1; k < D; k++ ) { + Af_QA[ k ] = 0; + } + break; + } + + /* Update C * Af and C * Ab */ + for( k = 0; k <= n + 1; k++ ) { + tmp1 = CAf[ k ]; /* Q( -rshifts ) */ + tmp2 = CAb[ n - k + 1 ]; /* Q( -rshifts ) */ + CAf[ k ] = silk_ADD_LSHIFT32( tmp1, silk_SMMUL( tmp2, rc_Q31 ), 1 ); /* Q( -rshifts ) */ + CAb[ n - k + 1 ] = silk_ADD_LSHIFT32( tmp2, silk_SMMUL( tmp1, rc_Q31 ), 1 ); /* Q( -rshifts ) */ + } + } + + if( reached_max_gain ) { + for( k = 0; k < D; k++ ) { + /* Scale coefficients */ + A_Q16[ k ] = -silk_RSHIFT_ROUND( Af_QA[ k ], QA - 16 ); + } + /* Subtract energy of preceding samples from C0 */ + if( rshifts > 0 ) { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + C0 -= (opus_int32)silk_RSHIFT64( silk_inner_prod16( x_ptr, x_ptr, D, arch ), rshifts ); + } + } else { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + C0 -= silk_LSHIFT32( silk_inner_prod_aligned( x_ptr, x_ptr, D, arch), -rshifts); + } + } + /* Approximate residual energy */ + *res_nrg = silk_LSHIFT( silk_SMMUL( invGain_Q30, C0 ), 2 ); + *res_nrg_Q = -rshifts; + } else { + /* Return residual energy */ + nrg = CAf[ 0 ]; /* Q( -rshifts ) */ + tmp1 = (opus_int32)1 << 16; /* Q16 */ + for( k = 0; k < D; k++ ) { + Atmp1 = silk_RSHIFT_ROUND( Af_QA[ k ], QA - 16 ); /* Q16 */ + nrg = silk_SMLAWW( nrg, CAf[ k + 1 ], Atmp1 ); /* Q( -rshifts ) */ + tmp1 = silk_SMLAWW( tmp1, Atmp1, Atmp1 ); /* Q16 */ + A_Q16[ k ] = -Atmp1; + } + *res_nrg = silk_SMLAWW( nrg, silk_SMMUL( SILK_FIX_CONST( FIND_LPC_COND_FAC, 32 ), C0 ), -tmp1 );/* Q( -rshifts ) */ + *res_nrg_Q = -rshifts; + } +} diff --git a/lib/opus/silk/fixed/corrMatrix_FIX.c b/lib/opus/silk/fixed/corrMatrix_FIX.c new file mode 100644 index 00000000..1b4a29c2 --- /dev/null +++ b/lib/opus/silk/fixed/corrMatrix_FIX.c @@ -0,0 +1,150 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/********************************************************************** + * Correlation Matrix Computations for LS estimate. + **********************************************************************/ + +#include "main_FIX.h" + +/* Calculates correlation vector X'*t */ +void silk_corrVector_FIX( + const opus_int16 *x, /* I x vector [L + order - 1] used to form data matrix X */ + const opus_int16 *t, /* I Target vector [L] */ + const opus_int L, /* I Length of vectors */ + const opus_int order, /* I Max lag for correlation */ + opus_int32 *Xt, /* O Pointer to X'*t correlation vector [order] */ + const opus_int rshifts, /* I Right shifts of correlations */ + int arch /* I Run-time architecture */ +) +{ + opus_int lag, i; + const opus_int16 *ptr1, *ptr2; + opus_int32 inner_prod; + + ptr1 = &x[ order - 1 ]; /* Points to first sample of column 0 of X: X[:,0] */ + ptr2 = t; + /* Calculate X'*t */ + if( rshifts > 0 ) { + /* Right shifting used */ + for( lag = 0; lag < order; lag++ ) { + inner_prod = 0; + for( i = 0; i < L; i++ ) { + inner_prod = silk_ADD_RSHIFT32( inner_prod, silk_SMULBB( ptr1[ i ], ptr2[i] ), rshifts ); + } + Xt[ lag ] = inner_prod; /* X[:,lag]'*t */ + ptr1--; /* Go to next column of X */ + } + } else { + silk_assert( rshifts == 0 ); + for( lag = 0; lag < order; lag++ ) { + Xt[ lag ] = silk_inner_prod_aligned( ptr1, ptr2, L, arch ); /* X[:,lag]'*t */ + ptr1--; /* Go to next column of X */ + } + } +} + +/* Calculates correlation matrix X'*X */ +void silk_corrMatrix_FIX( + const opus_int16 *x, /* I x vector [L + order - 1] used to form data matrix X */ + const opus_int L, /* I Length of vectors */ + const opus_int order, /* I Max lag for correlation */ + opus_int32 *XX, /* O Pointer to X'*X correlation matrix [ order x order ] */ + opus_int32 *nrg, /* O Energy of x vector */ + opus_int *rshifts, /* O Right shifts of correlations and energy */ + int arch /* I Run-time architecture */ +) +{ + opus_int i, j, lag; + opus_int32 energy; + const opus_int16 *ptr1, *ptr2; + + /* Calculate energy to find shift used to fit in 32 bits */ + silk_sum_sqr_shift( nrg, rshifts, x, L + order - 1 ); + energy = *nrg; + + /* Calculate energy of first column (0) of X: X[:,0]'*X[:,0] */ + /* Remove contribution of first order - 1 samples */ + for( i = 0; i < order - 1; i++ ) { + energy -= silk_RSHIFT32( silk_SMULBB( x[ i ], x[ i ] ), *rshifts ); + } + + /* Calculate energy of remaining columns of X: X[:,j]'*X[:,j] */ + /* Fill out the diagonal of the correlation matrix */ + matrix_ptr( XX, 0, 0, order ) = energy; + silk_assert( energy >= 0 ); + ptr1 = &x[ order - 1 ]; /* First sample of column 0 of X */ + for( j = 1; j < order; j++ ) { + energy = silk_SUB32( energy, silk_RSHIFT32( silk_SMULBB( ptr1[ L - j ], ptr1[ L - j ] ), *rshifts ) ); + energy = silk_ADD32( energy, silk_RSHIFT32( silk_SMULBB( ptr1[ -j ], ptr1[ -j ] ), *rshifts ) ); + matrix_ptr( XX, j, j, order ) = energy; + silk_assert( energy >= 0 ); + } + + ptr2 = &x[ order - 2 ]; /* First sample of column 1 of X */ + /* Calculate the remaining elements of the correlation matrix */ + if( *rshifts > 0 ) { + /* Right shifting used */ + for( lag = 1; lag < order; lag++ ) { + /* Inner product of column 0 and column lag: X[:,0]'*X[:,lag] */ + energy = 0; + for( i = 0; i < L; i++ ) { + energy += silk_RSHIFT32( silk_SMULBB( ptr1[ i ], ptr2[i] ), *rshifts ); + } + /* Calculate remaining off diagonal: X[:,j]'*X[:,j + lag] */ + matrix_ptr( XX, lag, 0, order ) = energy; + matrix_ptr( XX, 0, lag, order ) = energy; + for( j = 1; j < ( order - lag ); j++ ) { + energy = silk_SUB32( energy, silk_RSHIFT32( silk_SMULBB( ptr1[ L - j ], ptr2[ L - j ] ), *rshifts ) ); + energy = silk_ADD32( energy, silk_RSHIFT32( silk_SMULBB( ptr1[ -j ], ptr2[ -j ] ), *rshifts ) ); + matrix_ptr( XX, lag + j, j, order ) = energy; + matrix_ptr( XX, j, lag + j, order ) = energy; + } + ptr2--; /* Update pointer to first sample of next column (lag) in X */ + } + } else { + for( lag = 1; lag < order; lag++ ) { + /* Inner product of column 0 and column lag: X[:,0]'*X[:,lag] */ + energy = silk_inner_prod_aligned( ptr1, ptr2, L, arch ); + matrix_ptr( XX, lag, 0, order ) = energy; + matrix_ptr( XX, 0, lag, order ) = energy; + /* Calculate remaining off diagonal: X[:,j]'*X[:,j + lag] */ + for( j = 1; j < ( order - lag ); j++ ) { + energy = silk_SUB32( energy, silk_SMULBB( ptr1[ L - j ], ptr2[ L - j ] ) ); + energy = silk_SMLABB( energy, ptr1[ -j ], ptr2[ -j ] ); + matrix_ptr( XX, lag + j, j, order ) = energy; + matrix_ptr( XX, j, lag + j, order ) = energy; + } + ptr2--;/* Update pointer to first sample of next column (lag) in X */ + } + } +} + diff --git a/lib/opus/silk/fixed/encode_frame_FIX.c b/lib/opus/silk/fixed/encode_frame_FIX.c new file mode 100644 index 00000000..a02bf87d --- /dev/null +++ b/lib/opus/silk/fixed/encode_frame_FIX.c @@ -0,0 +1,448 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "main_FIX.h" +#include "stack_alloc.h" +#include "tuning_parameters.h" + +/* Low Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode with lower bitrate */ +static OPUS_INLINE void silk_LBRR_encode_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Pointer to Silk FIX encoder control struct */ + const opus_int16 x16[], /* I Input signal */ + opus_int condCoding /* I The type of conditional coding used so far for this frame */ +); + +void silk_encode_do_VAD_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ + opus_int activity /* I Decision of Opus voice activity detector */ +) +{ + const opus_int activity_threshold = SILK_FIX_CONST( SPEECH_ACTIVITY_DTX_THRES, 8 ); + + /****************************/ + /* Voice Activity Detection */ + /****************************/ + silk_VAD_GetSA_Q8( &psEnc->sCmn, psEnc->sCmn.inputBuf + 1, psEnc->sCmn.arch ); + /* If Opus VAD is inactive and Silk VAD is active: lower Silk VAD to just under the threshold */ + if( activity == VAD_NO_ACTIVITY && psEnc->sCmn.speech_activity_Q8 >= activity_threshold ) { + psEnc->sCmn.speech_activity_Q8 = activity_threshold - 1; + } + + /**************************************************/ + /* Convert speech activity into VAD and DTX flags */ + /**************************************************/ + if( psEnc->sCmn.speech_activity_Q8 < activity_threshold ) { + psEnc->sCmn.indices.signalType = TYPE_NO_VOICE_ACTIVITY; + psEnc->sCmn.noSpeechCounter++; + if( psEnc->sCmn.noSpeechCounter <= NB_SPEECH_FRAMES_BEFORE_DTX ) { + psEnc->sCmn.inDTX = 0; + } else if( psEnc->sCmn.noSpeechCounter > MAX_CONSECUTIVE_DTX + NB_SPEECH_FRAMES_BEFORE_DTX ) { + psEnc->sCmn.noSpeechCounter = NB_SPEECH_FRAMES_BEFORE_DTX; + psEnc->sCmn.inDTX = 0; + } + psEnc->sCmn.VAD_flags[ psEnc->sCmn.nFramesEncoded ] = 0; + } else { + psEnc->sCmn.noSpeechCounter = 0; + psEnc->sCmn.inDTX = 0; + psEnc->sCmn.indices.signalType = TYPE_UNVOICED; + psEnc->sCmn.VAD_flags[ psEnc->sCmn.nFramesEncoded ] = 1; + } +} + +/****************/ +/* Encode frame */ +/****************/ +opus_int silk_encode_frame_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ + opus_int32 *pnBytesOut, /* O Pointer to number of payload bytes; */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding, /* I The type of conditional coding to use */ + opus_int maxBits, /* I If > 0: maximum number of output bits */ + opus_int useCBR /* I Flag to force constant-bitrate operation */ +) +{ + silk_encoder_control_FIX sEncCtrl; + opus_int i, iter, maxIter, found_upper, found_lower, ret = 0; + opus_int16 *x_frame; + ec_enc sRangeEnc_copy, sRangeEnc_copy2; + silk_nsq_state sNSQ_copy, sNSQ_copy2; + opus_int32 seed_copy, nBits, nBits_lower, nBits_upper, gainMult_lower, gainMult_upper; + opus_int32 gainsID, gainsID_lower, gainsID_upper; + opus_int16 gainMult_Q8; + opus_int16 ec_prevLagIndex_copy; + opus_int ec_prevSignalType_copy; + opus_int8 LastGainIndex_copy2; + opus_int gain_lock[ MAX_NB_SUBFR ] = {0}; + opus_int16 best_gain_mult[ MAX_NB_SUBFR ]; + opus_int best_sum[ MAX_NB_SUBFR ]; + SAVE_STACK; + + /* This is totally unnecessary but many compilers (including gcc) are too dumb to realise it */ + LastGainIndex_copy2 = nBits_lower = nBits_upper = gainMult_lower = gainMult_upper = 0; + + psEnc->sCmn.indices.Seed = psEnc->sCmn.frameCounter++ & 3; + + /**************************************************************/ + /* Set up Input Pointers, and insert frame in input buffer */ + /*************************************************************/ + /* start of frame to encode */ + x_frame = psEnc->x_buf + psEnc->sCmn.ltp_mem_length; + + /***************************************/ + /* Ensure smooth bandwidth transitions */ + /***************************************/ + silk_LP_variable_cutoff( &psEnc->sCmn.sLP, psEnc->sCmn.inputBuf + 1, psEnc->sCmn.frame_length ); + + /*******************************************/ + /* Copy new frame to front of input buffer */ + /*******************************************/ + silk_memcpy( x_frame + LA_SHAPE_MS * psEnc->sCmn.fs_kHz, psEnc->sCmn.inputBuf + 1, psEnc->sCmn.frame_length * sizeof( opus_int16 ) ); + + if( !psEnc->sCmn.prefillFlag ) { + VARDECL( opus_int16, res_pitch ); + VARDECL( opus_uint8, ec_buf_copy ); + opus_int16 *res_pitch_frame; + + ALLOC( res_pitch, + psEnc->sCmn.la_pitch + psEnc->sCmn.frame_length + + psEnc->sCmn.ltp_mem_length, opus_int16 ); + /* start of pitch LPC residual frame */ + res_pitch_frame = res_pitch + psEnc->sCmn.ltp_mem_length; + + /*****************************************/ + /* Find pitch lags, initial LPC analysis */ + /*****************************************/ + silk_find_pitch_lags_FIX( psEnc, &sEncCtrl, res_pitch, x_frame - psEnc->sCmn.ltp_mem_length, psEnc->sCmn.arch ); + + /************************/ + /* Noise shape analysis */ + /************************/ + silk_noise_shape_analysis_FIX( psEnc, &sEncCtrl, res_pitch_frame, x_frame, psEnc->sCmn.arch ); + + /***************************************************/ + /* Find linear prediction coefficients (LPC + LTP) */ + /***************************************************/ + silk_find_pred_coefs_FIX( psEnc, &sEncCtrl, res_pitch_frame, x_frame, condCoding ); + + /****************************************/ + /* Process gains */ + /****************************************/ + silk_process_gains_FIX( psEnc, &sEncCtrl, condCoding ); + + /****************************************/ + /* Low Bitrate Redundant Encoding */ + /****************************************/ + silk_LBRR_encode_FIX( psEnc, &sEncCtrl, x_frame, condCoding ); + + /* Loop over quantizer and entropy coding to control bitrate */ + maxIter = 6; + gainMult_Q8 = SILK_FIX_CONST( 1, 8 ); + found_lower = 0; + found_upper = 0; + gainsID = silk_gains_ID( psEnc->sCmn.indices.GainsIndices, psEnc->sCmn.nb_subfr ); + gainsID_lower = -1; + gainsID_upper = -1; + /* Copy part of the input state */ + silk_memcpy( &sRangeEnc_copy, psRangeEnc, sizeof( ec_enc ) ); + silk_memcpy( &sNSQ_copy, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) ); + seed_copy = psEnc->sCmn.indices.Seed; + ec_prevLagIndex_copy = psEnc->sCmn.ec_prevLagIndex; + ec_prevSignalType_copy = psEnc->sCmn.ec_prevSignalType; + ALLOC( ec_buf_copy, 1275, opus_uint8 ); + for( iter = 0; ; iter++ ) { + if( gainsID == gainsID_lower ) { + nBits = nBits_lower; + } else if( gainsID == gainsID_upper ) { + nBits = nBits_upper; + } else { + /* Restore part of the input state */ + if( iter > 0 ) { + silk_memcpy( psRangeEnc, &sRangeEnc_copy, sizeof( ec_enc ) ); + silk_memcpy( &psEnc->sCmn.sNSQ, &sNSQ_copy, sizeof( silk_nsq_state ) ); + psEnc->sCmn.indices.Seed = seed_copy; + psEnc->sCmn.ec_prevLagIndex = ec_prevLagIndex_copy; + psEnc->sCmn.ec_prevSignalType = ec_prevSignalType_copy; + } + + /*****************************************/ + /* Noise shaping quantization */ + /*****************************************/ + if( psEnc->sCmn.nStatesDelayedDecision > 1 || psEnc->sCmn.warping_Q16 > 0 ) { + silk_NSQ_del_dec( &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, x_frame, psEnc->sCmn.pulses, + sEncCtrl.PredCoef_Q12[ 0 ], sEncCtrl.LTPCoef_Q14, sEncCtrl.AR_Q13, sEncCtrl.HarmShapeGain_Q14, + sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14, + psEnc->sCmn.arch ); + } else { + silk_NSQ( &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, x_frame, psEnc->sCmn.pulses, + sEncCtrl.PredCoef_Q12[ 0 ], sEncCtrl.LTPCoef_Q14, sEncCtrl.AR_Q13, sEncCtrl.HarmShapeGain_Q14, + sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14, + psEnc->sCmn.arch); + } + + if ( iter == maxIter && !found_lower ) { + silk_memcpy( &sRangeEnc_copy2, psRangeEnc, sizeof( ec_enc ) ); + } + + /****************************************/ + /* Encode Parameters */ + /****************************************/ + silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding ); + + /****************************************/ + /* Encode Excitation Signal */ + /****************************************/ + silk_encode_pulses( psRangeEnc, psEnc->sCmn.indices.signalType, psEnc->sCmn.indices.quantOffsetType, + psEnc->sCmn.pulses, psEnc->sCmn.frame_length ); + + nBits = ec_tell( psRangeEnc ); + + /* If we still bust after the last iteration, do some damage control. */ + if ( iter == maxIter && !found_lower && nBits > maxBits ) { + silk_memcpy( psRangeEnc, &sRangeEnc_copy2, sizeof( ec_enc ) ); + + /* Keep gains the same as the last frame. */ + psEnc->sShape.LastGainIndex = sEncCtrl.lastGainIndexPrev; + for ( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + psEnc->sCmn.indices.GainsIndices[ i ] = 4; + } + if (condCoding != CODE_CONDITIONALLY) { + psEnc->sCmn.indices.GainsIndices[ 0 ] = sEncCtrl.lastGainIndexPrev; + } + psEnc->sCmn.ec_prevLagIndex = ec_prevLagIndex_copy; + psEnc->sCmn.ec_prevSignalType = ec_prevSignalType_copy; + /* Clear all pulses. */ + for ( i = 0; i < psEnc->sCmn.frame_length; i++ ) { + psEnc->sCmn.pulses[ i ] = 0; + } + + silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding ); + + silk_encode_pulses( psRangeEnc, psEnc->sCmn.indices.signalType, psEnc->sCmn.indices.quantOffsetType, + psEnc->sCmn.pulses, psEnc->sCmn.frame_length ); + + nBits = ec_tell( psRangeEnc ); + } + + if( useCBR == 0 && iter == 0 && nBits <= maxBits ) { + break; + } + } + + if( iter == maxIter ) { + if( found_lower && ( gainsID == gainsID_lower || nBits > maxBits ) ) { + /* Restore output state from earlier iteration that did meet the bitrate budget */ + silk_memcpy( psRangeEnc, &sRangeEnc_copy2, sizeof( ec_enc ) ); + celt_assert( sRangeEnc_copy2.offs <= 1275 ); + silk_memcpy( psRangeEnc->buf, ec_buf_copy, sRangeEnc_copy2.offs ); + silk_memcpy( &psEnc->sCmn.sNSQ, &sNSQ_copy2, sizeof( silk_nsq_state ) ); + psEnc->sShape.LastGainIndex = LastGainIndex_copy2; + } + break; + } + + if( nBits > maxBits ) { + if( found_lower == 0 && iter >= 2 ) { + /* Adjust the quantizer's rate/distortion tradeoff and discard previous "upper" results */ + sEncCtrl.Lambda_Q10 = silk_ADD_RSHIFT32( sEncCtrl.Lambda_Q10, sEncCtrl.Lambda_Q10, 1 ); + found_upper = 0; + gainsID_upper = -1; + } else { + found_upper = 1; + nBits_upper = nBits; + gainMult_upper = gainMult_Q8; + gainsID_upper = gainsID; + } + } else if( nBits < maxBits - 5 ) { + found_lower = 1; + nBits_lower = nBits; + gainMult_lower = gainMult_Q8; + if( gainsID != gainsID_lower ) { + gainsID_lower = gainsID; + /* Copy part of the output state */ + silk_memcpy( &sRangeEnc_copy2, psRangeEnc, sizeof( ec_enc ) ); + celt_assert( psRangeEnc->offs <= 1275 ); + silk_memcpy( ec_buf_copy, psRangeEnc->buf, psRangeEnc->offs ); + silk_memcpy( &sNSQ_copy2, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) ); + LastGainIndex_copy2 = psEnc->sShape.LastGainIndex; + } + } else { + /* Within 5 bits of budget: close enough */ + break; + } + + if ( !found_lower && nBits > maxBits ) { + int j; + for ( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + int sum=0; + for ( j = i*psEnc->sCmn.subfr_length; j < (i+1)*psEnc->sCmn.subfr_length; j++ ) { + sum += abs( psEnc->sCmn.pulses[j] ); + } + if ( iter == 0 || (sum < best_sum[i] && !gain_lock[i]) ) { + best_sum[i] = sum; + best_gain_mult[i] = gainMult_Q8; + } else { + gain_lock[i] = 1; + } + } + } + if( ( found_lower & found_upper ) == 0 ) { + /* Adjust gain according to high-rate rate/distortion curve */ + if( nBits > maxBits ) { + if (gainMult_Q8 < 16384) { + gainMult_Q8 *= 2; + } else { + gainMult_Q8 = 32767; + } + } else { + opus_int32 gain_factor_Q16; + gain_factor_Q16 = silk_log2lin( silk_LSHIFT( nBits - maxBits, 7 ) / psEnc->sCmn.frame_length + SILK_FIX_CONST( 16, 7 ) ); + gainMult_Q8 = silk_SMULWB( gain_factor_Q16, gainMult_Q8 ); + } + + } else { + /* Adjust gain by interpolating */ + gainMult_Q8 = gainMult_lower + silk_DIV32_16( silk_MUL( gainMult_upper - gainMult_lower, maxBits - nBits_lower ), nBits_upper - nBits_lower ); + /* New gain multplier must be between 25% and 75% of old range (note that gainMult_upper < gainMult_lower) */ + if( gainMult_Q8 > silk_ADD_RSHIFT32( gainMult_lower, gainMult_upper - gainMult_lower, 2 ) ) { + gainMult_Q8 = silk_ADD_RSHIFT32( gainMult_lower, gainMult_upper - gainMult_lower, 2 ); + } else + if( gainMult_Q8 < silk_SUB_RSHIFT32( gainMult_upper, gainMult_upper - gainMult_lower, 2 ) ) { + gainMult_Q8 = silk_SUB_RSHIFT32( gainMult_upper, gainMult_upper - gainMult_lower, 2 ); + } + } + + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + opus_int16 tmp; + if ( gain_lock[i] ) { + tmp = best_gain_mult[i]; + } else { + tmp = gainMult_Q8; + } + sEncCtrl.Gains_Q16[ i ] = silk_LSHIFT_SAT32( silk_SMULWB( sEncCtrl.GainsUnq_Q16[ i ], tmp ), 8 ); + } + + /* Quantize gains */ + psEnc->sShape.LastGainIndex = sEncCtrl.lastGainIndexPrev; + silk_gains_quant( psEnc->sCmn.indices.GainsIndices, sEncCtrl.Gains_Q16, + &psEnc->sShape.LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); + + /* Unique identifier of gains vector */ + gainsID = silk_gains_ID( psEnc->sCmn.indices.GainsIndices, psEnc->sCmn.nb_subfr ); + } + } + + /* Update input buffer */ + silk_memmove( psEnc->x_buf, &psEnc->x_buf[ psEnc->sCmn.frame_length ], + ( psEnc->sCmn.ltp_mem_length + LA_SHAPE_MS * psEnc->sCmn.fs_kHz ) * sizeof( opus_int16 ) ); + + /* Exit without entropy coding */ + if( psEnc->sCmn.prefillFlag ) { + /* No payload */ + *pnBytesOut = 0; + RESTORE_STACK; + return ret; + } + + /* Parameters needed for next frame */ + psEnc->sCmn.prevLag = sEncCtrl.pitchL[ psEnc->sCmn.nb_subfr - 1 ]; + psEnc->sCmn.prevSignalType = psEnc->sCmn.indices.signalType; + + /****************************************/ + /* Finalize payload */ + /****************************************/ + psEnc->sCmn.first_frame_after_reset = 0; + /* Payload size */ + *pnBytesOut = silk_RSHIFT( ec_tell( psRangeEnc ) + 7, 3 ); + + RESTORE_STACK; + return ret; +} + +/* Low-Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode excitation at lower bitrate */ +static OPUS_INLINE void silk_LBRR_encode_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Pointer to Silk FIX encoder control struct */ + const opus_int16 x16[], /* I Input signal */ + opus_int condCoding /* I The type of conditional coding used so far for this frame */ +) +{ + opus_int32 TempGains_Q16[ MAX_NB_SUBFR ]; + SideInfoIndices *psIndices_LBRR = &psEnc->sCmn.indices_LBRR[ psEnc->sCmn.nFramesEncoded ]; + silk_nsq_state sNSQ_LBRR; + + /*******************************************/ + /* Control use of inband LBRR */ + /*******************************************/ + if( psEnc->sCmn.LBRR_enabled && psEnc->sCmn.speech_activity_Q8 > SILK_FIX_CONST( LBRR_SPEECH_ACTIVITY_THRES, 8 ) ) { + psEnc->sCmn.LBRR_flags[ psEnc->sCmn.nFramesEncoded ] = 1; + + /* Copy noise shaping quantizer state and quantization indices from regular encoding */ + silk_memcpy( &sNSQ_LBRR, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) ); + silk_memcpy( psIndices_LBRR, &psEnc->sCmn.indices, sizeof( SideInfoIndices ) ); + + /* Save original gains */ + silk_memcpy( TempGains_Q16, psEncCtrl->Gains_Q16, psEnc->sCmn.nb_subfr * sizeof( opus_int32 ) ); + + if( psEnc->sCmn.nFramesEncoded == 0 || psEnc->sCmn.LBRR_flags[ psEnc->sCmn.nFramesEncoded - 1 ] == 0 ) { + /* First frame in packet or previous frame not LBRR coded */ + psEnc->sCmn.LBRRprevLastGainIndex = psEnc->sShape.LastGainIndex; + + /* Increase Gains to get target LBRR rate */ + psIndices_LBRR->GainsIndices[ 0 ] = psIndices_LBRR->GainsIndices[ 0 ] + psEnc->sCmn.LBRR_GainIncreases; + psIndices_LBRR->GainsIndices[ 0 ] = silk_min_int( psIndices_LBRR->GainsIndices[ 0 ], N_LEVELS_QGAIN - 1 ); + } + + /* Decode to get gains in sync with decoder */ + /* Overwrite unquantized gains with quantized gains */ + silk_gains_dequant( psEncCtrl->Gains_Q16, psIndices_LBRR->GainsIndices, + &psEnc->sCmn.LBRRprevLastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); + + /*****************************************/ + /* Noise shaping quantization */ + /*****************************************/ + if( psEnc->sCmn.nStatesDelayedDecision > 1 || psEnc->sCmn.warping_Q16 > 0 ) { + silk_NSQ_del_dec( &psEnc->sCmn, &sNSQ_LBRR, psIndices_LBRR, x16, + psEnc->sCmn.pulses_LBRR[ psEnc->sCmn.nFramesEncoded ], psEncCtrl->PredCoef_Q12[ 0 ], psEncCtrl->LTPCoef_Q14, + psEncCtrl->AR_Q13, psEncCtrl->HarmShapeGain_Q14, psEncCtrl->Tilt_Q14, psEncCtrl->LF_shp_Q14, + psEncCtrl->Gains_Q16, psEncCtrl->pitchL, psEncCtrl->Lambda_Q10, psEncCtrl->LTP_scale_Q14, psEnc->sCmn.arch ); + } else { + silk_NSQ( &psEnc->sCmn, &sNSQ_LBRR, psIndices_LBRR, x16, + psEnc->sCmn.pulses_LBRR[ psEnc->sCmn.nFramesEncoded ], psEncCtrl->PredCoef_Q12[ 0 ], psEncCtrl->LTPCoef_Q14, + psEncCtrl->AR_Q13, psEncCtrl->HarmShapeGain_Q14, psEncCtrl->Tilt_Q14, psEncCtrl->LF_shp_Q14, + psEncCtrl->Gains_Q16, psEncCtrl->pitchL, psEncCtrl->Lambda_Q10, psEncCtrl->LTP_scale_Q14, psEnc->sCmn.arch ); + } + + /* Restore original gains */ + silk_memcpy( psEncCtrl->Gains_Q16, TempGains_Q16, psEnc->sCmn.nb_subfr * sizeof( opus_int32 ) ); + } +} diff --git a/lib/opus/silk/fixed/find_LPC_FIX.c b/lib/opus/silk/fixed/find_LPC_FIX.c new file mode 100644 index 00000000..c762a0f2 --- /dev/null +++ b/lib/opus/silk/fixed/find_LPC_FIX.c @@ -0,0 +1,151 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "stack_alloc.h" +#include "tuning_parameters.h" + +/* Finds LPC vector from correlations, and converts to NLSF */ +void silk_find_LPC_FIX( + silk_encoder_state *psEncC, /* I/O Encoder state */ + opus_int16 NLSF_Q15[], /* O NLSFs */ + const opus_int16 x[], /* I Input signal */ + const opus_int32 minInvGain_Q30 /* I Inverse of max prediction gain */ +) +{ + opus_int k, subfr_length; + opus_int32 a_Q16[ MAX_LPC_ORDER ]; + opus_int isInterpLower, shift; + opus_int32 res_nrg0, res_nrg1; + opus_int rshift0, rshift1; + + /* Used only for LSF interpolation */ + opus_int32 a_tmp_Q16[ MAX_LPC_ORDER ], res_nrg_interp, res_nrg, res_tmp_nrg; + opus_int res_nrg_interp_Q, res_nrg_Q, res_tmp_nrg_Q; + opus_int16 a_tmp_Q12[ MAX_LPC_ORDER ]; + opus_int16 NLSF0_Q15[ MAX_LPC_ORDER ]; + SAVE_STACK; + + subfr_length = psEncC->subfr_length + psEncC->predictLPCOrder; + + /* Default: no interpolation */ + psEncC->indices.NLSFInterpCoef_Q2 = 4; + + /* Burg AR analysis for the full frame */ + silk_burg_modified( &res_nrg, &res_nrg_Q, a_Q16, x, minInvGain_Q30, subfr_length, psEncC->nb_subfr, psEncC->predictLPCOrder, psEncC->arch ); + + if( psEncC->useInterpolatedNLSFs && !psEncC->first_frame_after_reset && psEncC->nb_subfr == MAX_NB_SUBFR ) { + VARDECL( opus_int16, LPC_res ); + + /* Optimal solution for last 10 ms */ + silk_burg_modified( &res_tmp_nrg, &res_tmp_nrg_Q, a_tmp_Q16, x + 2 * subfr_length, minInvGain_Q30, subfr_length, 2, psEncC->predictLPCOrder, psEncC->arch ); + + /* subtract residual energy here, as that's easier than adding it to the */ + /* residual energy of the first 10 ms in each iteration of the search below */ + shift = res_tmp_nrg_Q - res_nrg_Q; + if( shift >= 0 ) { + if( shift < 32 ) { + res_nrg = res_nrg - silk_RSHIFT( res_tmp_nrg, shift ); + } + } else { + silk_assert( shift > -32 ); + res_nrg = silk_RSHIFT( res_nrg, -shift ) - res_tmp_nrg; + res_nrg_Q = res_tmp_nrg_Q; + } + + /* Convert to NLSFs */ + silk_A2NLSF( NLSF_Q15, a_tmp_Q16, psEncC->predictLPCOrder ); + + ALLOC( LPC_res, 2 * subfr_length, opus_int16 ); + + /* Search over interpolation indices to find the one with lowest residual energy */ + for( k = 3; k >= 0; k-- ) { + /* Interpolate NLSFs for first half */ + silk_interpolate( NLSF0_Q15, psEncC->prev_NLSFq_Q15, NLSF_Q15, k, psEncC->predictLPCOrder ); + + /* Convert to LPC for residual energy evaluation */ + silk_NLSF2A( a_tmp_Q12, NLSF0_Q15, psEncC->predictLPCOrder, psEncC->arch ); + + /* Calculate residual energy with NLSF interpolation */ + silk_LPC_analysis_filter( LPC_res, x, a_tmp_Q12, 2 * subfr_length, psEncC->predictLPCOrder, psEncC->arch ); + + silk_sum_sqr_shift( &res_nrg0, &rshift0, LPC_res + psEncC->predictLPCOrder, subfr_length - psEncC->predictLPCOrder ); + silk_sum_sqr_shift( &res_nrg1, &rshift1, LPC_res + psEncC->predictLPCOrder + subfr_length, subfr_length - psEncC->predictLPCOrder ); + + /* Add subframe energies from first half frame */ + shift = rshift0 - rshift1; + if( shift >= 0 ) { + res_nrg1 = silk_RSHIFT( res_nrg1, shift ); + res_nrg_interp_Q = -rshift0; + } else { + res_nrg0 = silk_RSHIFT( res_nrg0, -shift ); + res_nrg_interp_Q = -rshift1; + } + res_nrg_interp = silk_ADD32( res_nrg0, res_nrg1 ); + + /* Compare with first half energy without NLSF interpolation, or best interpolated value so far */ + shift = res_nrg_interp_Q - res_nrg_Q; + if( shift >= 0 ) { + if( silk_RSHIFT( res_nrg_interp, shift ) < res_nrg ) { + isInterpLower = silk_TRUE; + } else { + isInterpLower = silk_FALSE; + } + } else { + if( -shift < 32 ) { + if( res_nrg_interp < silk_RSHIFT( res_nrg, -shift ) ) { + isInterpLower = silk_TRUE; + } else { + isInterpLower = silk_FALSE; + } + } else { + isInterpLower = silk_FALSE; + } + } + + /* Determine whether current interpolated NLSFs are best so far */ + if( isInterpLower == silk_TRUE ) { + /* Interpolation has lower residual energy */ + res_nrg = res_nrg_interp; + res_nrg_Q = res_nrg_interp_Q; + psEncC->indices.NLSFInterpCoef_Q2 = (opus_int8)k; + } + } + } + + if( psEncC->indices.NLSFInterpCoef_Q2 == 4 ) { + /* NLSF interpolation is currently inactive, calculate NLSFs from full frame AR coefficients */ + silk_A2NLSF( NLSF_Q15, a_Q16, psEncC->predictLPCOrder ); + } + + celt_assert( psEncC->indices.NLSFInterpCoef_Q2 == 4 || ( psEncC->useInterpolatedNLSFs && !psEncC->first_frame_after_reset && psEncC->nb_subfr == MAX_NB_SUBFR ) ); + RESTORE_STACK; +} diff --git a/lib/opus/silk/fixed/find_LTP_FIX.c b/lib/opus/silk/fixed/find_LTP_FIX.c new file mode 100644 index 00000000..62d4afb2 --- /dev/null +++ b/lib/opus/silk/fixed/find_LTP_FIX.c @@ -0,0 +1,99 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "tuning_parameters.h" + +void silk_find_LTP_FIX( + opus_int32 XXLTP_Q17[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* O Correlation matrix */ + opus_int32 xXLTP_Q17[ MAX_NB_SUBFR * LTP_ORDER ], /* O Correlation vector */ + const opus_int16 r_ptr[], /* I Residual signal after LPC */ + const opus_int lag[ MAX_NB_SUBFR ], /* I LTP lags */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr, /* I Number of subframes */ + int arch /* I Run-time architecture */ +) +{ + opus_int i, k, extra_shifts; + opus_int xx_shifts, xX_shifts, XX_shifts; + const opus_int16 *lag_ptr; + opus_int32 *XXLTP_Q17_ptr, *xXLTP_Q17_ptr; + opus_int32 xx, nrg, temp; + + xXLTP_Q17_ptr = xXLTP_Q17; + XXLTP_Q17_ptr = XXLTP_Q17; + for( k = 0; k < nb_subfr; k++ ) { + lag_ptr = r_ptr - ( lag[ k ] + LTP_ORDER / 2 ); + + silk_sum_sqr_shift( &xx, &xx_shifts, r_ptr, subfr_length + LTP_ORDER ); /* xx in Q( -xx_shifts ) */ + silk_corrMatrix_FIX( lag_ptr, subfr_length, LTP_ORDER, XXLTP_Q17_ptr, &nrg, &XX_shifts, arch ); /* XXLTP_Q17_ptr and nrg in Q( -XX_shifts ) */ + extra_shifts = xx_shifts - XX_shifts; + if( extra_shifts > 0 ) { + /* Shift XX */ + xX_shifts = xx_shifts; + for( i = 0; i < LTP_ORDER * LTP_ORDER; i++ ) { + XXLTP_Q17_ptr[ i ] = silk_RSHIFT32( XXLTP_Q17_ptr[ i ], extra_shifts ); /* Q( -xX_shifts ) */ + } + nrg = silk_RSHIFT32( nrg, extra_shifts ); /* Q( -xX_shifts ) */ + } else if( extra_shifts < 0 ) { + /* Shift xx */ + xX_shifts = XX_shifts; + xx = silk_RSHIFT32( xx, -extra_shifts ); /* Q( -xX_shifts ) */ + } else { + xX_shifts = xx_shifts; + } + silk_corrVector_FIX( lag_ptr, r_ptr, subfr_length, LTP_ORDER, xXLTP_Q17_ptr, xX_shifts, arch ); /* xXLTP_Q17_ptr in Q( -xX_shifts ) */ + + /* At this point all correlations are in Q(-xX_shifts) */ + temp = silk_SMLAWB( 1, nrg, SILK_FIX_CONST( LTP_CORR_INV_MAX, 16 ) ); + temp = silk_max( temp, xx ); +TIC(div) +#if 0 + for( i = 0; i < LTP_ORDER * LTP_ORDER; i++ ) { + XXLTP_Q17_ptr[ i ] = silk_DIV32_varQ( XXLTP_Q17_ptr[ i ], temp, 17 ); + } + for( i = 0; i < LTP_ORDER; i++ ) { + xXLTP_Q17_ptr[ i ] = silk_DIV32_varQ( xXLTP_Q17_ptr[ i ], temp, 17 ); + } +#else + for( i = 0; i < LTP_ORDER * LTP_ORDER; i++ ) { + XXLTP_Q17_ptr[ i ] = (opus_int32)( silk_LSHIFT64( (opus_int64)XXLTP_Q17_ptr[ i ], 17 ) / temp ); + } + for( i = 0; i < LTP_ORDER; i++ ) { + xXLTP_Q17_ptr[ i ] = (opus_int32)( silk_LSHIFT64( (opus_int64)xXLTP_Q17_ptr[ i ], 17 ) / temp ); + } +#endif +TOC(div) + r_ptr += subfr_length; + XXLTP_Q17_ptr += LTP_ORDER * LTP_ORDER; + xXLTP_Q17_ptr += LTP_ORDER; + } +} diff --git a/lib/opus/silk/fixed/find_pitch_lags_FIX.c b/lib/opus/silk/fixed/find_pitch_lags_FIX.c new file mode 100644 index 00000000..6c3379f2 --- /dev/null +++ b/lib/opus/silk/fixed/find_pitch_lags_FIX.c @@ -0,0 +1,143 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "stack_alloc.h" +#include "tuning_parameters.h" + +/* Find pitch lags */ +void silk_find_pitch_lags_FIX( + silk_encoder_state_FIX *psEnc, /* I/O encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ + opus_int16 res[], /* O residual */ + const opus_int16 x[], /* I Speech signal */ + int arch /* I Run-time architecture */ +) +{ + opus_int buf_len, i, scale; + opus_int32 thrhld_Q13, res_nrg; + const opus_int16 *x_ptr; + VARDECL( opus_int16, Wsig ); + opus_int16 *Wsig_ptr; + opus_int32 auto_corr[ MAX_FIND_PITCH_LPC_ORDER + 1 ]; + opus_int16 rc_Q15[ MAX_FIND_PITCH_LPC_ORDER ]; + opus_int32 A_Q24[ MAX_FIND_PITCH_LPC_ORDER ]; + opus_int16 A_Q12[ MAX_FIND_PITCH_LPC_ORDER ]; + SAVE_STACK; + + /******************************************/ + /* Set up buffer lengths etc based on Fs */ + /******************************************/ + buf_len = psEnc->sCmn.la_pitch + psEnc->sCmn.frame_length + psEnc->sCmn.ltp_mem_length; + + /* Safety check */ + celt_assert( buf_len >= psEnc->sCmn.pitch_LPC_win_length ); + + /*************************************/ + /* Estimate LPC AR coefficients */ + /*************************************/ + + /* Calculate windowed signal */ + + ALLOC( Wsig, psEnc->sCmn.pitch_LPC_win_length, opus_int16 ); + + /* First LA_LTP samples */ + x_ptr = x + buf_len - psEnc->sCmn.pitch_LPC_win_length; + Wsig_ptr = Wsig; + silk_apply_sine_window( Wsig_ptr, x_ptr, 1, psEnc->sCmn.la_pitch ); + + /* Middle un - windowed samples */ + Wsig_ptr += psEnc->sCmn.la_pitch; + x_ptr += psEnc->sCmn.la_pitch; + silk_memcpy( Wsig_ptr, x_ptr, ( psEnc->sCmn.pitch_LPC_win_length - silk_LSHIFT( psEnc->sCmn.la_pitch, 1 ) ) * sizeof( opus_int16 ) ); + + /* Last LA_LTP samples */ + Wsig_ptr += psEnc->sCmn.pitch_LPC_win_length - silk_LSHIFT( psEnc->sCmn.la_pitch, 1 ); + x_ptr += psEnc->sCmn.pitch_LPC_win_length - silk_LSHIFT( psEnc->sCmn.la_pitch, 1 ); + silk_apply_sine_window( Wsig_ptr, x_ptr, 2, psEnc->sCmn.la_pitch ); + + /* Calculate autocorrelation sequence */ + silk_autocorr( auto_corr, &scale, Wsig, psEnc->sCmn.pitch_LPC_win_length, psEnc->sCmn.pitchEstimationLPCOrder + 1, arch ); + + /* Add white noise, as fraction of energy */ + auto_corr[ 0 ] = silk_SMLAWB( auto_corr[ 0 ], auto_corr[ 0 ], SILK_FIX_CONST( FIND_PITCH_WHITE_NOISE_FRACTION, 16 ) ) + 1; + + /* Calculate the reflection coefficients using schur */ + res_nrg = silk_schur( rc_Q15, auto_corr, psEnc->sCmn.pitchEstimationLPCOrder ); + + /* Prediction gain */ + psEncCtrl->predGain_Q16 = silk_DIV32_varQ( auto_corr[ 0 ], silk_max_int( res_nrg, 1 ), 16 ); + + /* Convert reflection coefficients to prediction coefficients */ + silk_k2a( A_Q24, rc_Q15, psEnc->sCmn.pitchEstimationLPCOrder ); + + /* Convert From 32 bit Q24 to 16 bit Q12 coefs */ + for( i = 0; i < psEnc->sCmn.pitchEstimationLPCOrder; i++ ) { + A_Q12[ i ] = (opus_int16)silk_SAT16( silk_RSHIFT( A_Q24[ i ], 12 ) ); + } + + /* Do BWE */ + silk_bwexpander( A_Q12, psEnc->sCmn.pitchEstimationLPCOrder, SILK_FIX_CONST( FIND_PITCH_BANDWIDTH_EXPANSION, 16 ) ); + + /*****************************************/ + /* LPC analysis filtering */ + /*****************************************/ + silk_LPC_analysis_filter( res, x, A_Q12, buf_len, psEnc->sCmn.pitchEstimationLPCOrder, psEnc->sCmn.arch ); + + if( psEnc->sCmn.indices.signalType != TYPE_NO_VOICE_ACTIVITY && psEnc->sCmn.first_frame_after_reset == 0 ) { + /* Threshold for pitch estimator */ + thrhld_Q13 = SILK_FIX_CONST( 0.6, 13 ); + thrhld_Q13 = silk_SMLABB( thrhld_Q13, SILK_FIX_CONST( -0.004, 13 ), psEnc->sCmn.pitchEstimationLPCOrder ); + thrhld_Q13 = silk_SMLAWB( thrhld_Q13, SILK_FIX_CONST( -0.1, 21 ), psEnc->sCmn.speech_activity_Q8 ); + thrhld_Q13 = silk_SMLABB( thrhld_Q13, SILK_FIX_CONST( -0.15, 13 ), silk_RSHIFT( psEnc->sCmn.prevSignalType, 1 ) ); + thrhld_Q13 = silk_SMLAWB( thrhld_Q13, SILK_FIX_CONST( -0.1, 14 ), psEnc->sCmn.input_tilt_Q15 ); + thrhld_Q13 = silk_SAT16( thrhld_Q13 ); + + /*****************************************/ + /* Call pitch estimator */ + /*****************************************/ + if( silk_pitch_analysis_core( res, psEncCtrl->pitchL, &psEnc->sCmn.indices.lagIndex, &psEnc->sCmn.indices.contourIndex, + &psEnc->LTPCorr_Q15, psEnc->sCmn.prevLag, psEnc->sCmn.pitchEstimationThreshold_Q16, + (opus_int)thrhld_Q13, psEnc->sCmn.fs_kHz, psEnc->sCmn.pitchEstimationComplexity, psEnc->sCmn.nb_subfr, + psEnc->sCmn.arch) == 0 ) + { + psEnc->sCmn.indices.signalType = TYPE_VOICED; + } else { + psEnc->sCmn.indices.signalType = TYPE_UNVOICED; + } + } else { + silk_memset( psEncCtrl->pitchL, 0, sizeof( psEncCtrl->pitchL ) ); + psEnc->sCmn.indices.lagIndex = 0; + psEnc->sCmn.indices.contourIndex = 0; + psEnc->LTPCorr_Q15 = 0; + } + RESTORE_STACK; +} diff --git a/lib/opus/silk/fixed/find_pred_coefs_FIX.c b/lib/opus/silk/fixed/find_pred_coefs_FIX.c new file mode 100644 index 00000000..ad363fb7 --- /dev/null +++ b/lib/opus/silk/fixed/find_pred_coefs_FIX.c @@ -0,0 +1,146 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "stack_alloc.h" + +void silk_find_pred_coefs_FIX( + silk_encoder_state_FIX *psEnc, /* I/O encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ + const opus_int16 res_pitch[], /* I Residual from pitch analysis */ + const opus_int16 x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + opus_int i; + opus_int32 invGains_Q16[ MAX_NB_SUBFR ], local_gains[ MAX_NB_SUBFR ]; + /* Set to NLSF_Q15 to zero so we don't copy junk to the state. */ + opus_int16 NLSF_Q15[ MAX_LPC_ORDER ]={0}; + const opus_int16 *x_ptr; + opus_int16 *x_pre_ptr; + VARDECL( opus_int16, LPC_in_pre ); + opus_int32 min_gain_Q16, minInvGain_Q30; + SAVE_STACK; + + /* weighting for weighted least squares */ + min_gain_Q16 = silk_int32_MAX >> 6; + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + min_gain_Q16 = silk_min( min_gain_Q16, psEncCtrl->Gains_Q16[ i ] ); + } + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + /* Divide to Q16 */ + silk_assert( psEncCtrl->Gains_Q16[ i ] > 0 ); + /* Invert and normalize gains, and ensure that maximum invGains_Q16 is within range of a 16 bit int */ + invGains_Q16[ i ] = silk_DIV32_varQ( min_gain_Q16, psEncCtrl->Gains_Q16[ i ], 16 - 2 ); + + /* Limit inverse */ + invGains_Q16[ i ] = silk_max( invGains_Q16[ i ], 100 ); + + /* Square the inverted gains */ + silk_assert( invGains_Q16[ i ] == silk_SAT16( invGains_Q16[ i ] ) ); + + /* Invert the inverted and normalized gains */ + local_gains[ i ] = silk_DIV32( ( (opus_int32)1 << 16 ), invGains_Q16[ i ] ); + } + + ALLOC( LPC_in_pre, + psEnc->sCmn.nb_subfr * psEnc->sCmn.predictLPCOrder + + psEnc->sCmn.frame_length, opus_int16 ); + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + VARDECL( opus_int32, xXLTP_Q17 ); + VARDECL( opus_int32, XXLTP_Q17 ); + + /**********/ + /* VOICED */ + /**********/ + celt_assert( psEnc->sCmn.ltp_mem_length - psEnc->sCmn.predictLPCOrder >= psEncCtrl->pitchL[ 0 ] + LTP_ORDER / 2 ); + + ALLOC( xXLTP_Q17, psEnc->sCmn.nb_subfr * LTP_ORDER, opus_int32 ); + ALLOC( XXLTP_Q17, psEnc->sCmn.nb_subfr * LTP_ORDER * LTP_ORDER, opus_int32 ); + + /* LTP analysis */ + silk_find_LTP_FIX( XXLTP_Q17, xXLTP_Q17, res_pitch, + psEncCtrl->pitchL, psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.arch ); + + /* Quantize LTP gain parameters */ + silk_quant_LTP_gains( psEncCtrl->LTPCoef_Q14, psEnc->sCmn.indices.LTPIndex, &psEnc->sCmn.indices.PERIndex, + &psEnc->sCmn.sum_log_gain_Q7, &psEncCtrl->LTPredCodGain_Q7, XXLTP_Q17, xXLTP_Q17, psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.arch ); + + /* Control LTP scaling */ + silk_LTP_scale_ctrl_FIX( psEnc, psEncCtrl, condCoding ); + + /* Create LTP residual */ + silk_LTP_analysis_filter_FIX( LPC_in_pre, x - psEnc->sCmn.predictLPCOrder, psEncCtrl->LTPCoef_Q14, + psEncCtrl->pitchL, invGains_Q16, psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.predictLPCOrder ); + + } else { + /************/ + /* UNVOICED */ + /************/ + /* Create signal with prepended subframes, scaled by inverse gains */ + x_ptr = x - psEnc->sCmn.predictLPCOrder; + x_pre_ptr = LPC_in_pre; + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + silk_scale_copy_vector16( x_pre_ptr, x_ptr, invGains_Q16[ i ], + psEnc->sCmn.subfr_length + psEnc->sCmn.predictLPCOrder ); + x_pre_ptr += psEnc->sCmn.subfr_length + psEnc->sCmn.predictLPCOrder; + x_ptr += psEnc->sCmn.subfr_length; + } + + silk_memset( psEncCtrl->LTPCoef_Q14, 0, psEnc->sCmn.nb_subfr * LTP_ORDER * sizeof( opus_int16 ) ); + psEncCtrl->LTPredCodGain_Q7 = 0; + psEnc->sCmn.sum_log_gain_Q7 = 0; + } + + /* Limit on total predictive coding gain */ + if( psEnc->sCmn.first_frame_after_reset ) { + minInvGain_Q30 = SILK_FIX_CONST( 1.0f / MAX_PREDICTION_POWER_GAIN_AFTER_RESET, 30 ); + } else { + minInvGain_Q30 = silk_log2lin( silk_SMLAWB( 16 << 7, (opus_int32)psEncCtrl->LTPredCodGain_Q7, SILK_FIX_CONST( 1.0 / 3, 16 ) ) ); /* Q16 */ + minInvGain_Q30 = silk_DIV32_varQ( minInvGain_Q30, + silk_SMULWW( SILK_FIX_CONST( MAX_PREDICTION_POWER_GAIN, 0 ), + silk_SMLAWB( SILK_FIX_CONST( 0.25, 18 ), SILK_FIX_CONST( 0.75, 18 ), psEncCtrl->coding_quality_Q14 ) ), 14 ); + } + + /* LPC_in_pre contains the LTP-filtered input for voiced, and the unfiltered input for unvoiced */ + silk_find_LPC_FIX( &psEnc->sCmn, NLSF_Q15, LPC_in_pre, minInvGain_Q30 ); + + /* Quantize LSFs */ + silk_process_NLSFs( &psEnc->sCmn, psEncCtrl->PredCoef_Q12, NLSF_Q15, psEnc->sCmn.prev_NLSFq_Q15 ); + + /* Calculate residual energy using quantized LPC coefficients */ + silk_residual_energy_FIX( psEncCtrl->ResNrg, psEncCtrl->ResNrgQ, LPC_in_pre, psEncCtrl->PredCoef_Q12, local_gains, + psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.predictLPCOrder, psEnc->sCmn.arch ); + + /* Copy to prediction struct for use in next frame for interpolation */ + silk_memcpy( psEnc->sCmn.prev_NLSFq_Q15, NLSF_Q15, sizeof( psEnc->sCmn.prev_NLSFq_Q15 ) ); + RESTORE_STACK; +} diff --git a/lib/opus/silk/fixed/k2a_FIX.c b/lib/opus/silk/fixed/k2a_FIX.c new file mode 100644 index 00000000..549f6ead --- /dev/null +++ b/lib/opus/silk/fixed/k2a_FIX.c @@ -0,0 +1,54 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Step up function, converts reflection coefficients to prediction coefficients */ +void silk_k2a( + opus_int32 *A_Q24, /* O Prediction coefficients [order] Q24 */ + const opus_int16 *rc_Q15, /* I Reflection coefficients [order] Q15 */ + const opus_int32 order /* I Prediction order */ +) +{ + opus_int k, n; + opus_int32 rc, tmp1, tmp2; + + for( k = 0; k < order; k++ ) { + rc = rc_Q15[ k ]; + for( n = 0; n < (k + 1) >> 1; n++ ) { + tmp1 = A_Q24[ n ]; + tmp2 = A_Q24[ k - n - 1 ]; + A_Q24[ n ] = silk_SMLAWB( tmp1, silk_LSHIFT( tmp2, 1 ), rc ); + A_Q24[ k - n - 1 ] = silk_SMLAWB( tmp2, silk_LSHIFT( tmp1, 1 ), rc ); + } + A_Q24[ k ] = -silk_LSHIFT( (opus_int32)rc_Q15[ k ], 9 ); + } +} diff --git a/lib/opus/silk/fixed/k2a_Q16_FIX.c b/lib/opus/silk/fixed/k2a_Q16_FIX.c new file mode 100644 index 00000000..1595aa62 --- /dev/null +++ b/lib/opus/silk/fixed/k2a_Q16_FIX.c @@ -0,0 +1,54 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Step up function, converts reflection coefficients to prediction coefficients */ +void silk_k2a_Q16( + opus_int32 *A_Q24, /* O Prediction coefficients [order] Q24 */ + const opus_int32 *rc_Q16, /* I Reflection coefficients [order] Q16 */ + const opus_int32 order /* I Prediction order */ +) +{ + opus_int k, n; + opus_int32 rc, tmp1, tmp2; + + for( k = 0; k < order; k++ ) { + rc = rc_Q16[ k ]; + for( n = 0; n < (k + 1) >> 1; n++ ) { + tmp1 = A_Q24[ n ]; + tmp2 = A_Q24[ k - n - 1 ]; + A_Q24[ n ] = silk_SMLAWW( tmp1, tmp2, rc ); + A_Q24[ k - n - 1 ] = silk_SMLAWW( tmp2, tmp1, rc ); + } + A_Q24[ k ] = -silk_LSHIFT( rc, 8 ); + } +} diff --git a/lib/opus/silk/fixed/main_FIX.h b/lib/opus/silk/fixed/main_FIX.h new file mode 100644 index 00000000..6d2112e5 --- /dev/null +++ b/lib/opus/silk/fixed/main_FIX.h @@ -0,0 +1,244 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MAIN_FIX_H +#define SILK_MAIN_FIX_H + +#include "SigProc_FIX.h" +#include "structs_FIX.h" +#include "control.h" +#include "main.h" +#include "PLC.h" +#include "debug.h" +#include "entenc.h" + +#if ((defined(OPUS_ARM_ASM) && defined(FIXED_POINT)) \ + || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#include "fixed/arm/warped_autocorrelation_FIX_arm.h" +#endif + +#ifndef FORCE_CPP_BUILD +#ifdef __cplusplus +extern "C" +{ +#endif +#endif + +#define silk_encoder_state_Fxx silk_encoder_state_FIX +#define silk_encode_do_VAD_Fxx silk_encode_do_VAD_FIX +#define silk_encode_frame_Fxx silk_encode_frame_FIX + +#define QC 10 +#define QS 13 + +/*********************/ +/* Encoder Functions */ +/*********************/ + +/* High-pass filter with cutoff frequency adaptation based on pitch lag statistics */ +void silk_HP_variable_cutoff( + silk_encoder_state_Fxx state_Fxx[] /* I/O Encoder states */ +); + +/* Encoder main function */ +void silk_encode_do_VAD_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ + opus_int activity /* I Decision of Opus voice activity detector */ +); + +/* Encoder main function */ +opus_int silk_encode_frame_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ + opus_int32 *pnBytesOut, /* O Pointer to number of payload bytes; */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding, /* I The type of conditional coding to use */ + opus_int maxBits, /* I If > 0: maximum number of output bits */ + opus_int useCBR /* I Flag to force constant-bitrate operation */ +); + +/* Initializes the Silk encoder state */ +opus_int silk_init_encoder( + silk_encoder_state_Fxx *psEnc, /* I/O Pointer to Silk FIX encoder state */ + int arch /* I Run-time architecture */ +); + +/* Control the Silk encoder */ +opus_int silk_control_encoder( + silk_encoder_state_Fxx *psEnc, /* I/O Pointer to Silk encoder state */ + silk_EncControlStruct *encControl, /* I Control structure */ + const opus_int allow_bw_switch, /* I Flag to allow switching audio bandwidth */ + const opus_int channelNb, /* I Channel number */ + const opus_int force_fs_kHz +); + +/**************************/ +/* Noise shaping analysis */ +/**************************/ +/* Compute noise shaping coefficients and initial gain values */ +void silk_noise_shape_analysis_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Encoder state FIX */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control FIX */ + const opus_int16 *pitch_res, /* I LPC residual from pitch analysis */ + const opus_int16 *x, /* I Input signal [ frame_length + la_shape ] */ + int arch /* I Run-time architecture */ +); + +/* Autocorrelations for a warped frequency axis */ +void silk_warped_autocorrelation_FIX_c( + opus_int32 *corr, /* O Result [order + 1] */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *input, /* I Input data to correlate */ + const opus_int warping_Q16, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +); + +#if !defined(OVERRIDE_silk_warped_autocorrelation_FIX) +#define silk_warped_autocorrelation_FIX(corr, scale, input, warping_Q16, length, order, arch) \ + ((void)(arch), silk_warped_autocorrelation_FIX_c(corr, scale, input, warping_Q16, length, order)) +#endif + +/* Calculation of LTP state scaling */ +void silk_LTP_scale_ctrl_FIX( + silk_encoder_state_FIX *psEnc, /* I/O encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/**********************************************/ +/* Prediction Analysis */ +/**********************************************/ +/* Find pitch lags */ +void silk_find_pitch_lags_FIX( + silk_encoder_state_FIX *psEnc, /* I/O encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ + opus_int16 res[], /* O residual */ + const opus_int16 x[], /* I Speech signal */ + int arch /* I Run-time architecture */ +); + +/* Find LPC and LTP coefficients */ +void silk_find_pred_coefs_FIX( + silk_encoder_state_FIX *psEnc, /* I/O encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ + const opus_int16 res_pitch[], /* I Residual from pitch analysis */ + const opus_int16 x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/* LPC analysis */ +void silk_find_LPC_FIX( + silk_encoder_state *psEncC, /* I/O Encoder state */ + opus_int16 NLSF_Q15[], /* O NLSFs */ + const opus_int16 x[], /* I Input signal */ + const opus_int32 minInvGain_Q30 /* I Inverse of max prediction gain */ +); + +/* LTP analysis */ +void silk_find_LTP_FIX( + opus_int32 XXLTP_Q17[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* O Correlation matrix */ + opus_int32 xXLTP_Q17[ MAX_NB_SUBFR * LTP_ORDER ], /* O Correlation vector */ + const opus_int16 r_lpc[], /* I Residual signal after LPC */ + const opus_int lag[ MAX_NB_SUBFR ], /* I LTP lags */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr, /* I Number of subframes */ + int arch /* I Run-time architecture */ +); + +void silk_LTP_analysis_filter_FIX( + opus_int16 *LTP_res, /* O LTP residual signal of length MAX_NB_SUBFR * ( pre_length + subfr_length ) */ + const opus_int16 *x, /* I Pointer to input signal with at least max( pitchL ) preceding samples */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ],/* I LTP_ORDER LTP coefficients for each MAX_NB_SUBFR subframe */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag, one for each subframe */ + const opus_int32 invGains_Q16[ MAX_NB_SUBFR ], /* I Inverse quantization gains, one for each subframe */ + const opus_int subfr_length, /* I Length of each subframe */ + const opus_int nb_subfr, /* I Number of subframes */ + const opus_int pre_length /* I Length of the preceding samples starting at &x[0] for each subframe */ +); + +/* Calculates residual energies of input subframes where all subframes have LPC_order */ +/* of preceding samples */ +void silk_residual_energy_FIX( + opus_int32 nrgs[ MAX_NB_SUBFR ], /* O Residual energy per subframe */ + opus_int nrgsQ[ MAX_NB_SUBFR ], /* O Q value per subframe */ + const opus_int16 x[], /* I Input signal */ + opus_int16 a_Q12[ 2 ][ MAX_LPC_ORDER ], /* I AR coefs for each frame half */ + const opus_int32 gains[ MAX_NB_SUBFR ], /* I Quantization gains */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr, /* I Number of subframes */ + const opus_int LPC_order, /* I LPC order */ + int arch /* I Run-time architecture */ +); + +/* Residual energy: nrg = wxx - 2 * wXx * c + c' * wXX * c */ +opus_int32 silk_residual_energy16_covar_FIX( + const opus_int16 *c, /* I Prediction vector */ + const opus_int32 *wXX, /* I Correlation matrix */ + const opus_int32 *wXx, /* I Correlation vector */ + opus_int32 wxx, /* I Signal energy */ + opus_int D, /* I Dimension */ + opus_int cQ /* I Q value for c vector 0 - 15 */ +); + +/* Processing of gains */ +void silk_process_gains_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/******************/ +/* Linear Algebra */ +/******************/ +/* Calculates correlation matrix X'*X */ +void silk_corrMatrix_FIX( + const opus_int16 *x, /* I x vector [L + order - 1] used to form data matrix X */ + const opus_int L, /* I Length of vectors */ + const opus_int order, /* I Max lag for correlation */ + opus_int32 *XX, /* O Pointer to X'*X correlation matrix [ order x order ] */ + opus_int32 *nrg, /* O Energy of x vector */ + opus_int *rshifts, /* O Right shifts of correlations */ + int arch /* I Run-time architecture */ +); + +/* Calculates correlation vector X'*t */ +void silk_corrVector_FIX( + const opus_int16 *x, /* I x vector [L + order - 1] used to form data matrix X */ + const opus_int16 *t, /* I Target vector [L] */ + const opus_int L, /* I Length of vectors */ + const opus_int order, /* I Max lag for correlation */ + opus_int32 *Xt, /* O Pointer to X'*t correlation vector [order] */ + const opus_int rshifts, /* I Right shifts of correlations */ + int arch /* I Run-time architecture */ +); + +#ifndef FORCE_CPP_BUILD +#ifdef __cplusplus +} +#endif /* __cplusplus */ +#endif /* FORCE_CPP_BUILD */ +#endif /* SILK_MAIN_FIX_H */ diff --git a/lib/opus/silk/fixed/mips/noise_shape_analysis_FIX_mipsr1.h b/lib/opus/silk/fixed/mips/noise_shape_analysis_FIX_mipsr1.h new file mode 100644 index 00000000..3999b5bd --- /dev/null +++ b/lib/opus/silk/fixed/mips/noise_shape_analysis_FIX_mipsr1.h @@ -0,0 +1,336 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + + +/**************************************************************/ +/* Compute noise shaping coefficients and initial gain values */ +/**************************************************************/ +#define OVERRIDE_silk_noise_shape_analysis_FIX + +void silk_noise_shape_analysis_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Encoder state FIX */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control FIX */ + const opus_int16 *pitch_res, /* I LPC residual from pitch analysis */ + const opus_int16 *x, /* I Input signal [ frame_length + la_shape ] */ + int arch /* I Run-time architecture */ +) +{ + silk_shape_state_FIX *psShapeSt = &psEnc->sShape; + opus_int k, i, nSamples, Qnrg, b_Q14, warping_Q16, scale = 0; + opus_int32 SNR_adj_dB_Q7, HarmBoost_Q16, HarmShapeGain_Q16, Tilt_Q16, tmp32; + opus_int32 nrg, pre_nrg_Q30, log_energy_Q7, log_energy_prev_Q7, energy_variation_Q7; + opus_int32 delta_Q16, BWExp1_Q16, BWExp2_Q16, gain_mult_Q16, gain_add_Q16, strength_Q16, b_Q8; + opus_int32 auto_corr[ MAX_SHAPE_LPC_ORDER + 1 ]; + opus_int32 refl_coef_Q16[ MAX_SHAPE_LPC_ORDER ]; + opus_int32 AR1_Q24[ MAX_SHAPE_LPC_ORDER ]; + opus_int32 AR2_Q24[ MAX_SHAPE_LPC_ORDER ]; + VARDECL( opus_int16, x_windowed ); + const opus_int16 *x_ptr, *pitch_res_ptr; + SAVE_STACK; + + /* Point to start of first LPC analysis block */ + x_ptr = x - psEnc->sCmn.la_shape; + + /****************/ + /* GAIN CONTROL */ + /****************/ + SNR_adj_dB_Q7 = psEnc->sCmn.SNR_dB_Q7; + + /* Input quality is the average of the quality in the lowest two VAD bands */ + psEncCtrl->input_quality_Q14 = ( opus_int )silk_RSHIFT( (opus_int32)psEnc->sCmn.input_quality_bands_Q15[ 0 ] + + psEnc->sCmn.input_quality_bands_Q15[ 1 ], 2 ); + + /* Coding quality level, between 0.0_Q0 and 1.0_Q0, but in Q14 */ + psEncCtrl->coding_quality_Q14 = silk_RSHIFT( silk_sigm_Q15( silk_RSHIFT_ROUND( SNR_adj_dB_Q7 - + SILK_FIX_CONST( 20.0, 7 ), 4 ) ), 1 ); + + /* Reduce coding SNR during low speech activity */ + if( psEnc->sCmn.useCBR == 0 ) { + b_Q8 = SILK_FIX_CONST( 1.0, 8 ) - psEnc->sCmn.speech_activity_Q8; + b_Q8 = silk_SMULWB( silk_LSHIFT( b_Q8, 8 ), b_Q8 ); + SNR_adj_dB_Q7 = silk_SMLAWB( SNR_adj_dB_Q7, + silk_SMULBB( SILK_FIX_CONST( -BG_SNR_DECR_dB, 7 ) >> ( 4 + 1 ), b_Q8 ), /* Q11*/ + silk_SMULWB( SILK_FIX_CONST( 1.0, 14 ) + psEncCtrl->input_quality_Q14, psEncCtrl->coding_quality_Q14 ) ); /* Q12*/ + } + + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Reduce gains for periodic signals */ + SNR_adj_dB_Q7 = silk_SMLAWB( SNR_adj_dB_Q7, SILK_FIX_CONST( HARM_SNR_INCR_dB, 8 ), psEnc->LTPCorr_Q15 ); + } else { + /* For unvoiced signals and low-quality input, adjust the quality slower than SNR_dB setting */ + SNR_adj_dB_Q7 = silk_SMLAWB( SNR_adj_dB_Q7, + silk_SMLAWB( SILK_FIX_CONST( 6.0, 9 ), -SILK_FIX_CONST( 0.4, 18 ), psEnc->sCmn.SNR_dB_Q7 ), + SILK_FIX_CONST( 1.0, 14 ) - psEncCtrl->input_quality_Q14 ); + } + + /*************************/ + /* SPARSENESS PROCESSING */ + /*************************/ + /* Set quantizer offset */ + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Initially set to 0; may be overruled in process_gains(..) */ + psEnc->sCmn.indices.quantOffsetType = 0; + psEncCtrl->sparseness_Q8 = 0; + } else { + /* Sparseness measure, based on relative fluctuations of energy per 2 milliseconds */ + nSamples = silk_LSHIFT( psEnc->sCmn.fs_kHz, 1 ); + energy_variation_Q7 = 0; + log_energy_prev_Q7 = 0; + pitch_res_ptr = pitch_res; + for( k = 0; k < silk_SMULBB( SUB_FRAME_LENGTH_MS, psEnc->sCmn.nb_subfr ) / 2; k++ ) { + silk_sum_sqr_shift( &nrg, &scale, pitch_res_ptr, nSamples ); + nrg += silk_RSHIFT( nSamples, scale ); /* Q(-scale)*/ + + log_energy_Q7 = silk_lin2log( nrg ); + if( k > 0 ) { + energy_variation_Q7 += silk_abs( log_energy_Q7 - log_energy_prev_Q7 ); + } + log_energy_prev_Q7 = log_energy_Q7; + pitch_res_ptr += nSamples; + } + + psEncCtrl->sparseness_Q8 = silk_RSHIFT( silk_sigm_Q15( silk_SMULWB( energy_variation_Q7 - + SILK_FIX_CONST( 5.0, 7 ), SILK_FIX_CONST( 0.1, 16 ) ) ), 7 ); + + /* Set quantization offset depending on sparseness measure */ + if( psEncCtrl->sparseness_Q8 > SILK_FIX_CONST( SPARSENESS_THRESHOLD_QNT_OFFSET, 8 ) ) { + psEnc->sCmn.indices.quantOffsetType = 0; + } else { + psEnc->sCmn.indices.quantOffsetType = 1; + } + + /* Increase coding SNR for sparse signals */ + SNR_adj_dB_Q7 = silk_SMLAWB( SNR_adj_dB_Q7, SILK_FIX_CONST( SPARSE_SNR_INCR_dB, 15 ), psEncCtrl->sparseness_Q8 - SILK_FIX_CONST( 0.5, 8 ) ); + } + + /*******************************/ + /* Control bandwidth expansion */ + /*******************************/ + /* More BWE for signals with high prediction gain */ + strength_Q16 = silk_SMULWB( psEncCtrl->predGain_Q16, SILK_FIX_CONST( FIND_PITCH_WHITE_NOISE_FRACTION, 16 ) ); + BWExp1_Q16 = BWExp2_Q16 = silk_DIV32_varQ( SILK_FIX_CONST( BANDWIDTH_EXPANSION, 16 ), + silk_SMLAWW( SILK_FIX_CONST( 1.0, 16 ), strength_Q16, strength_Q16 ), 16 ); + delta_Q16 = silk_SMULWB( SILK_FIX_CONST( 1.0, 16 ) - silk_SMULBB( 3, psEncCtrl->coding_quality_Q14 ), + SILK_FIX_CONST( LOW_RATE_BANDWIDTH_EXPANSION_DELTA, 16 ) ); + BWExp1_Q16 = silk_SUB32( BWExp1_Q16, delta_Q16 ); + BWExp2_Q16 = silk_ADD32( BWExp2_Q16, delta_Q16 ); + /* BWExp1 will be applied after BWExp2, so make it relative */ + BWExp1_Q16 = silk_DIV32_16( silk_LSHIFT( BWExp1_Q16, 14 ), silk_RSHIFT( BWExp2_Q16, 2 ) ); + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Slightly more warping in analysis will move quantization noise up in frequency, where it's better masked */ + warping_Q16 = silk_SMLAWB( psEnc->sCmn.warping_Q16, (opus_int32)psEncCtrl->coding_quality_Q14, SILK_FIX_CONST( 0.01, 18 ) ); + } else { + warping_Q16 = 0; + } + + /********************************************/ + /* Compute noise shaping AR coefs and gains */ + /********************************************/ + ALLOC( x_windowed, psEnc->sCmn.shapeWinLength, opus_int16 ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + /* Apply window: sine slope followed by flat part followed by cosine slope */ + opus_int shift, slope_part, flat_part; + flat_part = psEnc->sCmn.fs_kHz * 3; + slope_part = silk_RSHIFT( psEnc->sCmn.shapeWinLength - flat_part, 1 ); + + silk_apply_sine_window( x_windowed, x_ptr, 1, slope_part ); + shift = slope_part; + silk_memcpy( x_windowed + shift, x_ptr + shift, flat_part * sizeof(opus_int16) ); + shift += flat_part; + silk_apply_sine_window( x_windowed + shift, x_ptr + shift, 2, slope_part ); + + /* Update pointer: next LPC analysis block */ + x_ptr += psEnc->sCmn.subfr_length; + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Calculate warped auto correlation */ + silk_warped_autocorrelation_FIX( auto_corr, &scale, x_windowed, warping_Q16, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder, arch ); + } else { + /* Calculate regular auto correlation */ + silk_autocorr( auto_corr, &scale, x_windowed, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder + 1, arch ); + } + + /* Add white noise, as a fraction of energy */ + auto_corr[0] = silk_ADD32( auto_corr[0], silk_max_32( silk_SMULWB( silk_RSHIFT( auto_corr[ 0 ], 4 ), + SILK_FIX_CONST( SHAPE_WHITE_NOISE_FRACTION, 20 ) ), 1 ) ); + + /* Calculate the reflection coefficients using schur */ + nrg = silk_schur64( refl_coef_Q16, auto_corr, psEnc->sCmn.shapingLPCOrder ); + silk_assert( nrg >= 0 ); + + /* Convert reflection coefficients to prediction coefficients */ + silk_k2a_Q16( AR2_Q24, refl_coef_Q16, psEnc->sCmn.shapingLPCOrder ); + + Qnrg = -scale; /* range: -12...30*/ + silk_assert( Qnrg >= -12 ); + silk_assert( Qnrg <= 30 ); + + /* Make sure that Qnrg is an even number */ + if( Qnrg & 1 ) { + Qnrg -= 1; + nrg >>= 1; + } + + tmp32 = silk_SQRT_APPROX( nrg ); + Qnrg >>= 1; /* range: -6...15*/ + + psEncCtrl->Gains_Q16[ k ] = (silk_LSHIFT32( silk_LIMIT( (tmp32), silk_RSHIFT32( silk_int32_MIN, (16 - Qnrg) ), \ + silk_RSHIFT32( silk_int32_MAX, (16 - Qnrg) ) ), (16 - Qnrg) )); + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Adjust gain for warping */ + gain_mult_Q16 = warped_gain( AR2_Q24, warping_Q16, psEnc->sCmn.shapingLPCOrder ); + silk_assert( psEncCtrl->Gains_Q16[ k ] >= 0 ); + if ( silk_SMULWW( silk_RSHIFT_ROUND( psEncCtrl->Gains_Q16[ k ], 1 ), gain_mult_Q16 ) >= ( silk_int32_MAX >> 1 ) ) { + psEncCtrl->Gains_Q16[ k ] = silk_int32_MAX; + } else { + psEncCtrl->Gains_Q16[ k ] = silk_SMULWW( psEncCtrl->Gains_Q16[ k ], gain_mult_Q16 ); + } + } + + /* Bandwidth expansion for synthesis filter shaping */ + silk_bwexpander_32( AR2_Q24, psEnc->sCmn.shapingLPCOrder, BWExp2_Q16 ); + + /* Compute noise shaping filter coefficients */ + silk_memcpy( AR1_Q24, AR2_Q24, psEnc->sCmn.shapingLPCOrder * sizeof( opus_int32 ) ); + + /* Bandwidth expansion for analysis filter shaping */ + silk_assert( BWExp1_Q16 <= SILK_FIX_CONST( 1.0, 16 ) ); + silk_bwexpander_32( AR1_Q24, psEnc->sCmn.shapingLPCOrder, BWExp1_Q16 ); + + /* Ratio of prediction gains, in energy domain */ + pre_nrg_Q30 = silk_LPC_inverse_pred_gain_Q24( AR2_Q24, psEnc->sCmn.shapingLPCOrder, arch ); + nrg = silk_LPC_inverse_pred_gain_Q24( AR1_Q24, psEnc->sCmn.shapingLPCOrder, arch ); + + /*psEncCtrl->GainsPre[ k ] = 1.0f - 0.7f * ( 1.0f - pre_nrg / nrg ) = 0.3f + 0.7f * pre_nrg / nrg;*/ + pre_nrg_Q30 = silk_LSHIFT32( silk_SMULWB( pre_nrg_Q30, SILK_FIX_CONST( 0.7, 15 ) ), 1 ); + psEncCtrl->GainsPre_Q14[ k ] = ( opus_int ) SILK_FIX_CONST( 0.3, 14 ) + silk_DIV32_varQ( pre_nrg_Q30, nrg, 14 ); + + /* Convert to monic warped prediction coefficients and limit absolute values */ + limit_warped_coefs( AR2_Q24, AR1_Q24, warping_Q16, SILK_FIX_CONST( 3.999, 24 ), psEnc->sCmn.shapingLPCOrder ); + + /* Convert from Q24 to Q13 and store in int16 */ + for( i = 0; i < psEnc->sCmn.shapingLPCOrder; i++ ) { + psEncCtrl->AR1_Q13[ k * MAX_SHAPE_LPC_ORDER + i ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( AR1_Q24[ i ], 11 ) ); + psEncCtrl->AR2_Q13[ k * MAX_SHAPE_LPC_ORDER + i ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( AR2_Q24[ i ], 11 ) ); + } + } + + /*****************/ + /* Gain tweaking */ + /*****************/ + /* Increase gains during low speech activity and put lower limit on gains */ + gain_mult_Q16 = silk_log2lin( -silk_SMLAWB( -SILK_FIX_CONST( 16.0, 7 ), SNR_adj_dB_Q7, SILK_FIX_CONST( 0.16, 16 ) ) ); + gain_add_Q16 = silk_log2lin( silk_SMLAWB( SILK_FIX_CONST( 16.0, 7 ), SILK_FIX_CONST( MIN_QGAIN_DB, 7 ), SILK_FIX_CONST( 0.16, 16 ) ) ); + silk_assert( gain_mult_Q16 > 0 ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->Gains_Q16[ k ] = silk_SMULWW( psEncCtrl->Gains_Q16[ k ], gain_mult_Q16 ); + silk_assert( psEncCtrl->Gains_Q16[ k ] >= 0 ); + psEncCtrl->Gains_Q16[ k ] = silk_ADD_POS_SAT32( psEncCtrl->Gains_Q16[ k ], gain_add_Q16 ); + } + + gain_mult_Q16 = SILK_FIX_CONST( 1.0, 16 ) + silk_RSHIFT_ROUND( silk_MLA( SILK_FIX_CONST( INPUT_TILT, 26 ), + psEncCtrl->coding_quality_Q14, SILK_FIX_CONST( HIGH_RATE_INPUT_TILT, 12 ) ), 10 ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->GainsPre_Q14[ k ] = silk_SMULWB( gain_mult_Q16, psEncCtrl->GainsPre_Q14[ k ] ); + } + + /************************************************/ + /* Control low-frequency shaping and noise tilt */ + /************************************************/ + /* Less low frequency shaping for noisy inputs */ + strength_Q16 = silk_MUL( SILK_FIX_CONST( LOW_FREQ_SHAPING, 4 ), silk_SMLAWB( SILK_FIX_CONST( 1.0, 12 ), + SILK_FIX_CONST( LOW_QUALITY_LOW_FREQ_SHAPING_DECR, 13 ), psEnc->sCmn.input_quality_bands_Q15[ 0 ] - SILK_FIX_CONST( 1.0, 15 ) ) ); + strength_Q16 = silk_RSHIFT( silk_MUL( strength_Q16, psEnc->sCmn.speech_activity_Q8 ), 8 ); + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Reduce low frequencies quantization noise for periodic signals, depending on pitch lag */ + /*f = 400; freqz([1, -0.98 + 2e-4 * f], [1, -0.97 + 7e-4 * f], 2^12, Fs); axis([0, 1000, -10, 1])*/ + opus_int fs_kHz_inv = silk_DIV32_16( SILK_FIX_CONST( 0.2, 14 ), psEnc->sCmn.fs_kHz ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + b_Q14 = fs_kHz_inv + silk_DIV32_16( SILK_FIX_CONST( 3.0, 14 ), psEncCtrl->pitchL[ k ] ); + /* Pack two coefficients in one int32 */ + psEncCtrl->LF_shp_Q14[ k ] = silk_LSHIFT( SILK_FIX_CONST( 1.0, 14 ) - b_Q14 - silk_SMULWB( strength_Q16, b_Q14 ), 16 ); + psEncCtrl->LF_shp_Q14[ k ] |= (opus_uint16)( b_Q14 - SILK_FIX_CONST( 1.0, 14 ) ); + } + silk_assert( SILK_FIX_CONST( HARM_HP_NOISE_COEF, 24 ) < SILK_FIX_CONST( 0.5, 24 ) ); /* Guarantees that second argument to SMULWB() is within range of an opus_int16*/ + Tilt_Q16 = - SILK_FIX_CONST( HP_NOISE_COEF, 16 ) - + silk_SMULWB( SILK_FIX_CONST( 1.0, 16 ) - SILK_FIX_CONST( HP_NOISE_COEF, 16 ), + silk_SMULWB( SILK_FIX_CONST( HARM_HP_NOISE_COEF, 24 ), psEnc->sCmn.speech_activity_Q8 ) ); + } else { + b_Q14 = silk_DIV32_16( 21299, psEnc->sCmn.fs_kHz ); /* 1.3_Q0 = 21299_Q14*/ + /* Pack two coefficients in one int32 */ + psEncCtrl->LF_shp_Q14[ 0 ] = silk_LSHIFT( SILK_FIX_CONST( 1.0, 14 ) - b_Q14 - + silk_SMULWB( strength_Q16, silk_SMULWB( SILK_FIX_CONST( 0.6, 16 ), b_Q14 ) ), 16 ); + psEncCtrl->LF_shp_Q14[ 0 ] |= (opus_uint16)( b_Q14 - SILK_FIX_CONST( 1.0, 14 ) ); + for( k = 1; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->LF_shp_Q14[ k ] = psEncCtrl->LF_shp_Q14[ 0 ]; + } + Tilt_Q16 = -SILK_FIX_CONST( HP_NOISE_COEF, 16 ); + } + + /****************************/ + /* HARMONIC SHAPING CONTROL */ + /****************************/ + /* Control boosting of harmonic frequencies */ + HarmBoost_Q16 = silk_SMULWB( silk_SMULWB( SILK_FIX_CONST( 1.0, 17 ) - silk_LSHIFT( psEncCtrl->coding_quality_Q14, 3 ), + psEnc->LTPCorr_Q15 ), SILK_FIX_CONST( LOW_RATE_HARMONIC_BOOST, 16 ) ); + + /* More harmonic boost for noisy input signals */ + HarmBoost_Q16 = silk_SMLAWB( HarmBoost_Q16, + SILK_FIX_CONST( 1.0, 16 ) - silk_LSHIFT( psEncCtrl->input_quality_Q14, 2 ), SILK_FIX_CONST( LOW_INPUT_QUALITY_HARMONIC_BOOST, 16 ) ); + + if( USE_HARM_SHAPING && psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* More harmonic noise shaping for high bitrates or noisy input */ + HarmShapeGain_Q16 = silk_SMLAWB( SILK_FIX_CONST( HARMONIC_SHAPING, 16 ), + SILK_FIX_CONST( 1.0, 16 ) - silk_SMULWB( SILK_FIX_CONST( 1.0, 18 ) - silk_LSHIFT( psEncCtrl->coding_quality_Q14, 4 ), + psEncCtrl->input_quality_Q14 ), SILK_FIX_CONST( HIGH_RATE_OR_LOW_QUALITY_HARMONIC_SHAPING, 16 ) ); + + /* Less harmonic noise shaping for less periodic signals */ + HarmShapeGain_Q16 = silk_SMULWB( silk_LSHIFT( HarmShapeGain_Q16, 1 ), + silk_SQRT_APPROX( silk_LSHIFT( psEnc->LTPCorr_Q15, 15 ) ) ); + } else { + HarmShapeGain_Q16 = 0; + } + + /*************************/ + /* Smooth over subframes */ + /*************************/ + for( k = 0; k < MAX_NB_SUBFR; k++ ) { + psShapeSt->HarmBoost_smth_Q16 = + silk_SMLAWB( psShapeSt->HarmBoost_smth_Q16, HarmBoost_Q16 - psShapeSt->HarmBoost_smth_Q16, SILK_FIX_CONST( SUBFR_SMTH_COEF, 16 ) ); + psShapeSt->HarmShapeGain_smth_Q16 = + silk_SMLAWB( psShapeSt->HarmShapeGain_smth_Q16, HarmShapeGain_Q16 - psShapeSt->HarmShapeGain_smth_Q16, SILK_FIX_CONST( SUBFR_SMTH_COEF, 16 ) ); + psShapeSt->Tilt_smth_Q16 = + silk_SMLAWB( psShapeSt->Tilt_smth_Q16, Tilt_Q16 - psShapeSt->Tilt_smth_Q16, SILK_FIX_CONST( SUBFR_SMTH_COEF, 16 ) ); + + psEncCtrl->HarmBoost_Q14[ k ] = ( opus_int )silk_RSHIFT_ROUND( psShapeSt->HarmBoost_smth_Q16, 2 ); + psEncCtrl->HarmShapeGain_Q14[ k ] = ( opus_int )silk_RSHIFT_ROUND( psShapeSt->HarmShapeGain_smth_Q16, 2 ); + psEncCtrl->Tilt_Q14[ k ] = ( opus_int )silk_RSHIFT_ROUND( psShapeSt->Tilt_smth_Q16, 2 ); + } + RESTORE_STACK; +} diff --git a/lib/opus/silk/fixed/mips/prefilter_FIX_mipsr1.h b/lib/opus/silk/fixed/mips/prefilter_FIX_mipsr1.h new file mode 100644 index 00000000..21b25688 --- /dev/null +++ b/lib/opus/silk/fixed/mips/prefilter_FIX_mipsr1.h @@ -0,0 +1,184 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ +#ifndef __PREFILTER_FIX_MIPSR1_H__ +#define __PREFILTER_FIX_MIPSR1_H__ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "stack_alloc.h" +#include "tuning_parameters.h" + +#define OVERRIDE_silk_warped_LPC_analysis_filter_FIX +void silk_warped_LPC_analysis_filter_FIX( + opus_int32 state[], /* I/O State [order + 1] */ + opus_int32 res_Q2[], /* O Residual signal [length] */ + const opus_int16 coef_Q13[], /* I Coefficients [order] */ + const opus_int16 input[], /* I Input signal [length] */ + const opus_int16 lambda_Q16, /* I Warping factor */ + const opus_int length, /* I Length of input signal */ + const opus_int order, /* I Filter order (even) */ + int arch +) +{ + opus_int n, i; + opus_int32 acc_Q11, acc_Q22, tmp1, tmp2, tmp3, tmp4; + opus_int32 state_cur, state_next; + + (void)arch; + + /* Order must be even */ + /* Length must be even */ + + silk_assert( ( order & 1 ) == 0 ); + silk_assert( ( length & 1 ) == 0 ); + + for( n = 0; n < length; n+=2 ) { + /* Output of lowpass section */ + tmp2 = silk_SMLAWB( state[ 0 ], state[ 1 ], lambda_Q16 ); + state_cur = silk_LSHIFT( input[ n ], 14 ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( state[ 1 ], state[ 2 ] - tmp2, lambda_Q16 ); + state_next = tmp2; + acc_Q11 = silk_RSHIFT( order, 1 ); + acc_Q11 = silk_SMLAWB( acc_Q11, tmp2, coef_Q13[ 0 ] ); + + + /* Output of lowpass section */ + tmp4 = silk_SMLAWB( state_cur, state_next, lambda_Q16 ); + state[ 0 ] = silk_LSHIFT( input[ n+1 ], 14 ); + /* Output of allpass section */ + tmp3 = silk_SMLAWB( state_next, tmp1 - tmp4, lambda_Q16 ); + state[ 1 ] = tmp4; + acc_Q22 = silk_RSHIFT( order, 1 ); + acc_Q22 = silk_SMLAWB( acc_Q22, tmp4, coef_Q13[ 0 ] ); + + /* Loop over allpass sections */ + for( i = 2; i < order; i += 2 ) { + /* Output of allpass section */ + tmp2 = silk_SMLAWB( state[ i ], state[ i + 1 ] - tmp1, lambda_Q16 ); + state_cur = tmp1; + acc_Q11 = silk_SMLAWB( acc_Q11, tmp1, coef_Q13[ i - 1 ] ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( state[ i + 1 ], state[ i + 2 ] - tmp2, lambda_Q16 ); + state_next = tmp2; + acc_Q11 = silk_SMLAWB( acc_Q11, tmp2, coef_Q13[ i ] ); + + + /* Output of allpass section */ + tmp4 = silk_SMLAWB( state_cur, state_next - tmp3, lambda_Q16 ); + state[ i ] = tmp3; + acc_Q22 = silk_SMLAWB( acc_Q22, tmp3, coef_Q13[ i - 1 ] ); + /* Output of allpass section */ + tmp3 = silk_SMLAWB( state_next, tmp1 - tmp4, lambda_Q16 ); + state[ i + 1 ] = tmp4; + acc_Q22 = silk_SMLAWB( acc_Q22, tmp4, coef_Q13[ i ] ); + } + acc_Q11 = silk_SMLAWB( acc_Q11, tmp1, coef_Q13[ order - 1 ] ); + res_Q2[ n ] = silk_LSHIFT( (opus_int32)input[ n ], 2 ) - silk_RSHIFT_ROUND( acc_Q11, 9 ); + + state[ order ] = tmp3; + acc_Q22 = silk_SMLAWB( acc_Q22, tmp3, coef_Q13[ order - 1 ] ); + res_Q2[ n+1 ] = silk_LSHIFT( (opus_int32)input[ n+1 ], 2 ) - silk_RSHIFT_ROUND( acc_Q22, 9 ); + } +} + + + +/* Prefilter for finding Quantizer input signal */ +#define OVERRIDE_silk_prefilt_FIX +static inline void silk_prefilt_FIX( + silk_prefilter_state_FIX *P, /* I/O state */ + opus_int32 st_res_Q12[], /* I short term residual signal */ + opus_int32 xw_Q3[], /* O prefiltered signal */ + opus_int32 HarmShapeFIRPacked_Q12, /* I Harmonic shaping coeficients */ + opus_int Tilt_Q14, /* I Tilt shaping coeficient */ + opus_int32 LF_shp_Q14, /* I Low-frequancy shaping coeficients */ + opus_int lag, /* I Lag for harmonic shaping */ + opus_int length /* I Length of signals */ +) +{ + opus_int i, idx, LTP_shp_buf_idx; + opus_int32 n_LTP_Q12, n_Tilt_Q10, n_LF_Q10; + opus_int32 sLF_MA_shp_Q12, sLF_AR_shp_Q12; + opus_int16 *LTP_shp_buf; + + /* To speed up use temp variables instead of using the struct */ + LTP_shp_buf = P->sLTP_shp; + LTP_shp_buf_idx = P->sLTP_shp_buf_idx; + sLF_AR_shp_Q12 = P->sLF_AR_shp_Q12; + sLF_MA_shp_Q12 = P->sLF_MA_shp_Q12; + + if( lag > 0 ) { + for( i = 0; i < length; i++ ) { + /* unrolled loop */ + silk_assert( HARM_SHAPE_FIR_TAPS == 3 ); + idx = lag + LTP_shp_buf_idx; + n_LTP_Q12 = silk_SMULBB( LTP_shp_buf[ ( idx - HARM_SHAPE_FIR_TAPS / 2 - 1) & LTP_MASK ], HarmShapeFIRPacked_Q12 ); + n_LTP_Q12 = silk_SMLABT( n_LTP_Q12, LTP_shp_buf[ ( idx - HARM_SHAPE_FIR_TAPS / 2 ) & LTP_MASK ], HarmShapeFIRPacked_Q12 ); + n_LTP_Q12 = silk_SMLABB( n_LTP_Q12, LTP_shp_buf[ ( idx - HARM_SHAPE_FIR_TAPS / 2 + 1) & LTP_MASK ], HarmShapeFIRPacked_Q12 ); + + n_Tilt_Q10 = silk_SMULWB( sLF_AR_shp_Q12, Tilt_Q14 ); + n_LF_Q10 = silk_SMLAWB( silk_SMULWT( sLF_AR_shp_Q12, LF_shp_Q14 ), sLF_MA_shp_Q12, LF_shp_Q14 ); + + sLF_AR_shp_Q12 = silk_SUB32( st_res_Q12[ i ], silk_LSHIFT( n_Tilt_Q10, 2 ) ); + sLF_MA_shp_Q12 = silk_SUB32( sLF_AR_shp_Q12, silk_LSHIFT( n_LF_Q10, 2 ) ); + + LTP_shp_buf_idx = ( LTP_shp_buf_idx - 1 ) & LTP_MASK; + LTP_shp_buf[ LTP_shp_buf_idx ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( sLF_MA_shp_Q12, 12 ) ); + + xw_Q3[i] = silk_RSHIFT_ROUND( silk_SUB32( sLF_MA_shp_Q12, n_LTP_Q12 ), 9 ); + } + } + else + { + for( i = 0; i < length; i++ ) { + + n_LTP_Q12 = 0; + + n_Tilt_Q10 = silk_SMULWB( sLF_AR_shp_Q12, Tilt_Q14 ); + n_LF_Q10 = silk_SMLAWB( silk_SMULWT( sLF_AR_shp_Q12, LF_shp_Q14 ), sLF_MA_shp_Q12, LF_shp_Q14 ); + + sLF_AR_shp_Q12 = silk_SUB32( st_res_Q12[ i ], silk_LSHIFT( n_Tilt_Q10, 2 ) ); + sLF_MA_shp_Q12 = silk_SUB32( sLF_AR_shp_Q12, silk_LSHIFT( n_LF_Q10, 2 ) ); + + LTP_shp_buf_idx = ( LTP_shp_buf_idx - 1 ) & LTP_MASK; + LTP_shp_buf[ LTP_shp_buf_idx ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( sLF_MA_shp_Q12, 12 ) ); + + xw_Q3[i] = silk_RSHIFT_ROUND( sLF_MA_shp_Q12, 9 ); + } + } + + /* Copy temp variable back to state */ + P->sLF_AR_shp_Q12 = sLF_AR_shp_Q12; + P->sLF_MA_shp_Q12 = sLF_MA_shp_Q12; + P->sLTP_shp_buf_idx = LTP_shp_buf_idx; +} + +#endif /* __PREFILTER_FIX_MIPSR1_H__ */ diff --git a/lib/opus/silk/fixed/mips/warped_autocorrelation_FIX_mipsr1.h b/lib/opus/silk/fixed/mips/warped_autocorrelation_FIX_mipsr1.h new file mode 100644 index 00000000..66eb2ed2 --- /dev/null +++ b/lib/opus/silk/fixed/mips/warped_autocorrelation_FIX_mipsr1.h @@ -0,0 +1,165 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef __WARPED_AUTOCORRELATION_FIX_MIPSR1_H__ +#define __WARPED_AUTOCORRELATION_FIX_MIPSR1_H__ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" + +#undef QC +#define QC 10 + +#undef QS +#define QS 14 + +/* Autocorrelations for a warped frequency axis */ +#define OVERRIDE_silk_warped_autocorrelation_FIX_c +void silk_warped_autocorrelation_FIX_c( + opus_int32 *corr, /* O Result [order + 1] */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *input, /* I Input data to correlate */ + const opus_int warping_Q16, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +) +{ + opus_int n, i, lsh; + opus_int32 tmp1_QS=0, tmp2_QS=0, tmp3_QS=0, tmp4_QS=0, tmp5_QS=0, tmp6_QS=0, tmp7_QS=0, tmp8_QS=0, start_1=0, start_2=0, start_3=0; + opus_int32 state_QS[ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; + opus_int64 corr_QC[ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; + opus_int64 temp64; + + opus_int32 val; + val = 2 * QS - QC; + + /* Order must be even */ + silk_assert( ( order & 1 ) == 0 ); + silk_assert( 2 * QS - QC >= 0 ); + + /* Loop over samples */ + for( n = 0; n < length; n=n+4 ) { + + tmp1_QS = silk_LSHIFT32( (opus_int32)input[ n ], QS ); + start_1 = tmp1_QS; + tmp3_QS = silk_LSHIFT32( (opus_int32)input[ n+1], QS ); + start_2 = tmp3_QS; + tmp5_QS = silk_LSHIFT32( (opus_int32)input[ n+2], QS ); + start_3 = tmp5_QS; + tmp7_QS = silk_LSHIFT32( (opus_int32)input[ n+3], QS ); + + /* Loop over allpass sections */ + for( i = 0; i < order; i += 2 ) { + /* Output of allpass section */ + tmp2_QS = silk_SMLAWB( state_QS[ i ], state_QS[ i + 1 ] - tmp1_QS, warping_Q16 ); + corr_QC[ i ] = __builtin_mips_madd( corr_QC[ i ], tmp1_QS, start_1); + + tmp4_QS = silk_SMLAWB( tmp1_QS, tmp2_QS - tmp3_QS, warping_Q16 ); + corr_QC[ i ] = __builtin_mips_madd( corr_QC[ i ], tmp3_QS, start_2); + + tmp6_QS = silk_SMLAWB( tmp3_QS, tmp4_QS - tmp5_QS, warping_Q16 ); + corr_QC[ i ] = __builtin_mips_madd( corr_QC[ i ], tmp5_QS, start_3); + + tmp8_QS = silk_SMLAWB( tmp5_QS, tmp6_QS - tmp7_QS, warping_Q16 ); + state_QS[ i ] = tmp7_QS; + corr_QC[ i ] = __builtin_mips_madd( corr_QC[ i ], tmp7_QS, state_QS[0]); + + /* Output of allpass section */ + tmp1_QS = silk_SMLAWB( state_QS[ i + 1 ], state_QS[ i + 2 ] - tmp2_QS, warping_Q16 ); + corr_QC[ i+1 ] = __builtin_mips_madd( corr_QC[ i+1 ], tmp2_QS, start_1); + + tmp3_QS = silk_SMLAWB( tmp2_QS, tmp1_QS - tmp4_QS, warping_Q16 ); + corr_QC[ i+1 ] = __builtin_mips_madd( corr_QC[ i+1 ], tmp4_QS, start_2); + + tmp5_QS = silk_SMLAWB( tmp4_QS, tmp3_QS - tmp6_QS, warping_Q16 ); + corr_QC[ i+1 ] = __builtin_mips_madd( corr_QC[ i+1 ], tmp6_QS, start_3); + + tmp7_QS = silk_SMLAWB( tmp6_QS, tmp5_QS - tmp8_QS, warping_Q16 ); + state_QS[ i + 1 ] = tmp8_QS; + corr_QC[ i+1 ] = __builtin_mips_madd( corr_QC[ i+1 ], tmp8_QS, state_QS[ 0 ]); + + } + state_QS[ order ] = tmp7_QS; + + corr_QC[ order ] = __builtin_mips_madd( corr_QC[ order ], tmp1_QS, start_1); + corr_QC[ order ] = __builtin_mips_madd( corr_QC[ order ], tmp3_QS, start_2); + corr_QC[ order ] = __builtin_mips_madd( corr_QC[ order ], tmp5_QS, start_3); + corr_QC[ order ] = __builtin_mips_madd( corr_QC[ order ], tmp7_QS, state_QS[ 0 ]); + } + + for(;n< length; n++ ) { + + tmp1_QS = silk_LSHIFT32( (opus_int32)input[ n ], QS ); + + /* Loop over allpass sections */ + for( i = 0; i < order; i += 2 ) { + + /* Output of allpass section */ + tmp2_QS = silk_SMLAWB( state_QS[ i ], state_QS[ i + 1 ] - tmp1_QS, warping_Q16 ); + state_QS[ i ] = tmp1_QS; + corr_QC[ i ] = __builtin_mips_madd( corr_QC[ i ], tmp1_QS, state_QS[ 0 ]); + + /* Output of allpass section */ + tmp1_QS = silk_SMLAWB( state_QS[ i + 1 ], state_QS[ i + 2 ] - tmp2_QS, warping_Q16 ); + state_QS[ i + 1 ] = tmp2_QS; + corr_QC[ i+1 ] = __builtin_mips_madd( corr_QC[ i+1 ], tmp2_QS, state_QS[ 0 ]); + } + state_QS[ order ] = tmp1_QS; + corr_QC[ order ] = __builtin_mips_madd( corr_QC[ order ], tmp1_QS, state_QS[ 0 ]); + } + + temp64 = corr_QC[ 0 ]; + temp64 = __builtin_mips_shilo(temp64, val); + + lsh = silk_CLZ64( temp64 ) - 35; + lsh = silk_LIMIT( lsh, -12 - QC, 30 - QC ); + *scale = -( QC + lsh ); + silk_assert( *scale >= -30 && *scale <= 12 ); + if( lsh >= 0 ) { + for( i = 0; i < order + 1; i++ ) { + temp64 = corr_QC[ i ]; + //temp64 = __builtin_mips_shilo(temp64, val); + temp64 = (val >= 0) ? (temp64 >> val) : (temp64 << -val); + corr[ i ] = (opus_int32)silk_CHECK_FIT32( __builtin_mips_shilo( temp64, -lsh ) ); + } + } else { + for( i = 0; i < order + 1; i++ ) { + temp64 = corr_QC[ i ]; + //temp64 = __builtin_mips_shilo(temp64, val); + temp64 = (val >= 0) ? (temp64 >> val) : (temp64 << -val); + corr[ i ] = (opus_int32)silk_CHECK_FIT32( __builtin_mips_shilo( temp64, -lsh ) ); + } + } + + corr_QC[ 0 ] = __builtin_mips_shilo(corr_QC[ 0 ], val); + + silk_assert( corr_QC[ 0 ] >= 0 ); /* If breaking, decrease QC*/ +} +#endif /* __WARPED_AUTOCORRELATION_FIX_MIPSR1_H__ */ diff --git a/lib/opus/silk/fixed/noise_shape_analysis_FIX.c b/lib/opus/silk/fixed/noise_shape_analysis_FIX.c new file mode 100644 index 00000000..85fea0bf --- /dev/null +++ b/lib/opus/silk/fixed/noise_shape_analysis_FIX.c @@ -0,0 +1,407 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "stack_alloc.h" +#include "tuning_parameters.h" + +/* Compute gain to make warped filter coefficients have a zero mean log frequency response on a */ +/* non-warped frequency scale. (So that it can be implemented with a minimum-phase monic filter.) */ +/* Note: A monic filter is one with the first coefficient equal to 1.0. In Silk we omit the first */ +/* coefficient in an array of coefficients, for monic filters. */ +static OPUS_INLINE opus_int32 warped_gain( /* gain in Q16*/ + const opus_int32 *coefs_Q24, + opus_int lambda_Q16, + opus_int order +) { + opus_int i; + opus_int32 gain_Q24; + + lambda_Q16 = -lambda_Q16; + gain_Q24 = coefs_Q24[ order - 1 ]; + for( i = order - 2; i >= 0; i-- ) { + gain_Q24 = silk_SMLAWB( coefs_Q24[ i ], gain_Q24, lambda_Q16 ); + } + gain_Q24 = silk_SMLAWB( SILK_FIX_CONST( 1.0, 24 ), gain_Q24, -lambda_Q16 ); + return silk_INVERSE32_varQ( gain_Q24, 40 ); +} + +/* Convert warped filter coefficients to monic pseudo-warped coefficients and limit maximum */ +/* amplitude of monic warped coefficients by using bandwidth expansion on the true coefficients */ +static OPUS_INLINE void limit_warped_coefs( + opus_int32 *coefs_Q24, + opus_int lambda_Q16, + opus_int32 limit_Q24, + opus_int order +) { + opus_int i, iter, ind = 0; + opus_int32 tmp, maxabs_Q24, chirp_Q16, gain_Q16; + opus_int32 nom_Q16, den_Q24; + opus_int32 limit_Q20, maxabs_Q20; + + /* Convert to monic coefficients */ + lambda_Q16 = -lambda_Q16; + for( i = order - 1; i > 0; i-- ) { + coefs_Q24[ i - 1 ] = silk_SMLAWB( coefs_Q24[ i - 1 ], coefs_Q24[ i ], lambda_Q16 ); + } + lambda_Q16 = -lambda_Q16; + nom_Q16 = silk_SMLAWB( SILK_FIX_CONST( 1.0, 16 ), -(opus_int32)lambda_Q16, lambda_Q16 ); + den_Q24 = silk_SMLAWB( SILK_FIX_CONST( 1.0, 24 ), coefs_Q24[ 0 ], lambda_Q16 ); + gain_Q16 = silk_DIV32_varQ( nom_Q16, den_Q24, 24 ); + for( i = 0; i < order; i++ ) { + coefs_Q24[ i ] = silk_SMULWW( gain_Q16, coefs_Q24[ i ] ); + } + limit_Q20 = silk_RSHIFT(limit_Q24, 4); + for( iter = 0; iter < 10; iter++ ) { + /* Find maximum absolute value */ + maxabs_Q24 = -1; + for( i = 0; i < order; i++ ) { + tmp = silk_abs_int32( coefs_Q24[ i ] ); + if( tmp > maxabs_Q24 ) { + maxabs_Q24 = tmp; + ind = i; + } + } + /* Use Q20 to avoid any overflow when multiplying by (ind + 1) later. */ + maxabs_Q20 = silk_RSHIFT(maxabs_Q24, 4); + if( maxabs_Q20 <= limit_Q20 ) { + /* Coefficients are within range - done */ + return; + } + + /* Convert back to true warped coefficients */ + for( i = 1; i < order; i++ ) { + coefs_Q24[ i - 1 ] = silk_SMLAWB( coefs_Q24[ i - 1 ], coefs_Q24[ i ], lambda_Q16 ); + } + gain_Q16 = silk_INVERSE32_varQ( gain_Q16, 32 ); + for( i = 0; i < order; i++ ) { + coefs_Q24[ i ] = silk_SMULWW( gain_Q16, coefs_Q24[ i ] ); + } + + /* Apply bandwidth expansion */ + chirp_Q16 = SILK_FIX_CONST( 0.99, 16 ) - silk_DIV32_varQ( + silk_SMULWB( maxabs_Q20 - limit_Q20, silk_SMLABB( SILK_FIX_CONST( 0.8, 10 ), SILK_FIX_CONST( 0.1, 10 ), iter ) ), + silk_MUL( maxabs_Q20, ind + 1 ), 22 ); + silk_bwexpander_32( coefs_Q24, order, chirp_Q16 ); + + /* Convert to monic warped coefficients */ + lambda_Q16 = -lambda_Q16; + for( i = order - 1; i > 0; i-- ) { + coefs_Q24[ i - 1 ] = silk_SMLAWB( coefs_Q24[ i - 1 ], coefs_Q24[ i ], lambda_Q16 ); + } + lambda_Q16 = -lambda_Q16; + nom_Q16 = silk_SMLAWB( SILK_FIX_CONST( 1.0, 16 ), -(opus_int32)lambda_Q16, lambda_Q16 ); + den_Q24 = silk_SMLAWB( SILK_FIX_CONST( 1.0, 24 ), coefs_Q24[ 0 ], lambda_Q16 ); + gain_Q16 = silk_DIV32_varQ( nom_Q16, den_Q24, 24 ); + for( i = 0; i < order; i++ ) { + coefs_Q24[ i ] = silk_SMULWW( gain_Q16, coefs_Q24[ i ] ); + } + } + silk_assert( 0 ); +} + +/* Disable MIPS version until it's updated. */ +#if 0 && defined(MIPSr1_ASM) +#include "mips/noise_shape_analysis_FIX_mipsr1.h" +#endif + +/**************************************************************/ +/* Compute noise shaping coefficients and initial gain values */ +/**************************************************************/ +#ifndef OVERRIDE_silk_noise_shape_analysis_FIX +void silk_noise_shape_analysis_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Encoder state FIX */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control FIX */ + const opus_int16 *pitch_res, /* I LPC residual from pitch analysis */ + const opus_int16 *x, /* I Input signal [ frame_length + la_shape ] */ + int arch /* I Run-time architecture */ +) +{ + silk_shape_state_FIX *psShapeSt = &psEnc->sShape; + opus_int k, i, nSamples, nSegs, Qnrg, b_Q14, warping_Q16, scale = 0; + opus_int32 SNR_adj_dB_Q7, HarmShapeGain_Q16, Tilt_Q16, tmp32; + opus_int32 nrg, log_energy_Q7, log_energy_prev_Q7, energy_variation_Q7; + opus_int32 BWExp_Q16, gain_mult_Q16, gain_add_Q16, strength_Q16, b_Q8; + opus_int32 auto_corr[ MAX_SHAPE_LPC_ORDER + 1 ]; + opus_int32 refl_coef_Q16[ MAX_SHAPE_LPC_ORDER ]; + opus_int32 AR_Q24[ MAX_SHAPE_LPC_ORDER ]; + VARDECL( opus_int16, x_windowed ); + const opus_int16 *x_ptr, *pitch_res_ptr; + SAVE_STACK; + + /* Point to start of first LPC analysis block */ + x_ptr = x - psEnc->sCmn.la_shape; + + /****************/ + /* GAIN CONTROL */ + /****************/ + SNR_adj_dB_Q7 = psEnc->sCmn.SNR_dB_Q7; + + /* Input quality is the average of the quality in the lowest two VAD bands */ + psEncCtrl->input_quality_Q14 = ( opus_int )silk_RSHIFT( (opus_int32)psEnc->sCmn.input_quality_bands_Q15[ 0 ] + + psEnc->sCmn.input_quality_bands_Q15[ 1 ], 2 ); + + /* Coding quality level, between 0.0_Q0 and 1.0_Q0, but in Q14 */ + psEncCtrl->coding_quality_Q14 = silk_RSHIFT( silk_sigm_Q15( silk_RSHIFT_ROUND( SNR_adj_dB_Q7 - + SILK_FIX_CONST( 20.0, 7 ), 4 ) ), 1 ); + + /* Reduce coding SNR during low speech activity */ + if( psEnc->sCmn.useCBR == 0 ) { + b_Q8 = SILK_FIX_CONST( 1.0, 8 ) - psEnc->sCmn.speech_activity_Q8; + b_Q8 = silk_SMULWB( silk_LSHIFT( b_Q8, 8 ), b_Q8 ); + SNR_adj_dB_Q7 = silk_SMLAWB( SNR_adj_dB_Q7, + silk_SMULBB( SILK_FIX_CONST( -BG_SNR_DECR_dB, 7 ) >> ( 4 + 1 ), b_Q8 ), /* Q11*/ + silk_SMULWB( SILK_FIX_CONST( 1.0, 14 ) + psEncCtrl->input_quality_Q14, psEncCtrl->coding_quality_Q14 ) ); /* Q12*/ + } + + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Reduce gains for periodic signals */ + SNR_adj_dB_Q7 = silk_SMLAWB( SNR_adj_dB_Q7, SILK_FIX_CONST( HARM_SNR_INCR_dB, 8 ), psEnc->LTPCorr_Q15 ); + } else { + /* For unvoiced signals and low-quality input, adjust the quality slower than SNR_dB setting */ + SNR_adj_dB_Q7 = silk_SMLAWB( SNR_adj_dB_Q7, + silk_SMLAWB( SILK_FIX_CONST( 6.0, 9 ), -SILK_FIX_CONST( 0.4, 18 ), psEnc->sCmn.SNR_dB_Q7 ), + SILK_FIX_CONST( 1.0, 14 ) - psEncCtrl->input_quality_Q14 ); + } + + /*************************/ + /* SPARSENESS PROCESSING */ + /*************************/ + /* Set quantizer offset */ + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Initially set to 0; may be overruled in process_gains(..) */ + psEnc->sCmn.indices.quantOffsetType = 0; + } else { + /* Sparseness measure, based on relative fluctuations of energy per 2 milliseconds */ + nSamples = silk_LSHIFT( psEnc->sCmn.fs_kHz, 1 ); + energy_variation_Q7 = 0; + log_energy_prev_Q7 = 0; + pitch_res_ptr = pitch_res; + nSegs = silk_SMULBB( SUB_FRAME_LENGTH_MS, psEnc->sCmn.nb_subfr ) / 2; + for( k = 0; k < nSegs; k++ ) { + silk_sum_sqr_shift( &nrg, &scale, pitch_res_ptr, nSamples ); + nrg += silk_RSHIFT( nSamples, scale ); /* Q(-scale)*/ + + log_energy_Q7 = silk_lin2log( nrg ); + if( k > 0 ) { + energy_variation_Q7 += silk_abs( log_energy_Q7 - log_energy_prev_Q7 ); + } + log_energy_prev_Q7 = log_energy_Q7; + pitch_res_ptr += nSamples; + } + + /* Set quantization offset depending on sparseness measure */ + if( energy_variation_Q7 > SILK_FIX_CONST( ENERGY_VARIATION_THRESHOLD_QNT_OFFSET, 7 ) * (nSegs-1) ) { + psEnc->sCmn.indices.quantOffsetType = 0; + } else { + psEnc->sCmn.indices.quantOffsetType = 1; + } + } + + /*******************************/ + /* Control bandwidth expansion */ + /*******************************/ + /* More BWE for signals with high prediction gain */ + strength_Q16 = silk_SMULWB( psEncCtrl->predGain_Q16, SILK_FIX_CONST( FIND_PITCH_WHITE_NOISE_FRACTION, 16 ) ); + BWExp_Q16 = silk_DIV32_varQ( SILK_FIX_CONST( BANDWIDTH_EXPANSION, 16 ), + silk_SMLAWW( SILK_FIX_CONST( 1.0, 16 ), strength_Q16, strength_Q16 ), 16 ); + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Slightly more warping in analysis will move quantization noise up in frequency, where it's better masked */ + warping_Q16 = silk_SMLAWB( psEnc->sCmn.warping_Q16, (opus_int32)psEncCtrl->coding_quality_Q14, SILK_FIX_CONST( 0.01, 18 ) ); + } else { + warping_Q16 = 0; + } + + /********************************************/ + /* Compute noise shaping AR coefs and gains */ + /********************************************/ + ALLOC( x_windowed, psEnc->sCmn.shapeWinLength, opus_int16 ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + /* Apply window: sine slope followed by flat part followed by cosine slope */ + opus_int shift, slope_part, flat_part; + flat_part = psEnc->sCmn.fs_kHz * 3; + slope_part = silk_RSHIFT( psEnc->sCmn.shapeWinLength - flat_part, 1 ); + + silk_apply_sine_window( x_windowed, x_ptr, 1, slope_part ); + shift = slope_part; + silk_memcpy( x_windowed + shift, x_ptr + shift, flat_part * sizeof(opus_int16) ); + shift += flat_part; + silk_apply_sine_window( x_windowed + shift, x_ptr + shift, 2, slope_part ); + + /* Update pointer: next LPC analysis block */ + x_ptr += psEnc->sCmn.subfr_length; + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Calculate warped auto correlation */ + silk_warped_autocorrelation_FIX( auto_corr, &scale, x_windowed, warping_Q16, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder, arch ); + } else { + /* Calculate regular auto correlation */ + silk_autocorr( auto_corr, &scale, x_windowed, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder + 1, arch ); + } + + /* Add white noise, as a fraction of energy */ + auto_corr[0] = silk_ADD32( auto_corr[0], silk_max_32( silk_SMULWB( silk_RSHIFT( auto_corr[ 0 ], 4 ), + SILK_FIX_CONST( SHAPE_WHITE_NOISE_FRACTION, 20 ) ), 1 ) ); + + /* Calculate the reflection coefficients using schur */ + nrg = silk_schur64( refl_coef_Q16, auto_corr, psEnc->sCmn.shapingLPCOrder ); + silk_assert( nrg >= 0 ); + + /* Convert reflection coefficients to prediction coefficients */ + silk_k2a_Q16( AR_Q24, refl_coef_Q16, psEnc->sCmn.shapingLPCOrder ); + + Qnrg = -scale; /* range: -12...30*/ + silk_assert( Qnrg >= -12 ); + silk_assert( Qnrg <= 30 ); + + /* Make sure that Qnrg is an even number */ + if( Qnrg & 1 ) { + Qnrg -= 1; + nrg >>= 1; + } + + tmp32 = silk_SQRT_APPROX( nrg ); + Qnrg >>= 1; /* range: -6...15*/ + + psEncCtrl->Gains_Q16[ k ] = silk_LSHIFT_SAT32( tmp32, 16 - Qnrg ); + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Adjust gain for warping */ + gain_mult_Q16 = warped_gain( AR_Q24, warping_Q16, psEnc->sCmn.shapingLPCOrder ); + silk_assert( psEncCtrl->Gains_Q16[ k ] > 0 ); + if( psEncCtrl->Gains_Q16[ k ] < SILK_FIX_CONST( 0.25, 16 ) ) { + psEncCtrl->Gains_Q16[ k ] = silk_SMULWW( psEncCtrl->Gains_Q16[ k ], gain_mult_Q16 ); + } else { + psEncCtrl->Gains_Q16[ k ] = silk_SMULWW( silk_RSHIFT_ROUND( psEncCtrl->Gains_Q16[ k ], 1 ), gain_mult_Q16 ); + if ( psEncCtrl->Gains_Q16[ k ] >= ( silk_int32_MAX >> 1 ) ) { + psEncCtrl->Gains_Q16[ k ] = silk_int32_MAX; + } else { + psEncCtrl->Gains_Q16[ k ] = silk_LSHIFT32( psEncCtrl->Gains_Q16[ k ], 1 ); + } + } + silk_assert( psEncCtrl->Gains_Q16[ k ] > 0 ); + } + + /* Bandwidth expansion */ + silk_bwexpander_32( AR_Q24, psEnc->sCmn.shapingLPCOrder, BWExp_Q16 ); + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Convert to monic warped prediction coefficients and limit absolute values */ + limit_warped_coefs( AR_Q24, warping_Q16, SILK_FIX_CONST( 3.999, 24 ), psEnc->sCmn.shapingLPCOrder ); + + /* Convert from Q24 to Q13 and store in int16 */ + for( i = 0; i < psEnc->sCmn.shapingLPCOrder; i++ ) { + psEncCtrl->AR_Q13[ k * MAX_SHAPE_LPC_ORDER + i ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( AR_Q24[ i ], 11 ) ); + } + } else { + silk_LPC_fit( &psEncCtrl->AR_Q13[ k * MAX_SHAPE_LPC_ORDER ], AR_Q24, 13, 24, psEnc->sCmn.shapingLPCOrder ); + } + } + + /*****************/ + /* Gain tweaking */ + /*****************/ + /* Increase gains during low speech activity and put lower limit on gains */ + gain_mult_Q16 = silk_log2lin( -silk_SMLAWB( -SILK_FIX_CONST( 16.0, 7 ), SNR_adj_dB_Q7, SILK_FIX_CONST( 0.16, 16 ) ) ); + gain_add_Q16 = silk_log2lin( silk_SMLAWB( SILK_FIX_CONST( 16.0, 7 ), SILK_FIX_CONST( MIN_QGAIN_DB, 7 ), SILK_FIX_CONST( 0.16, 16 ) ) ); + silk_assert( gain_mult_Q16 > 0 ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->Gains_Q16[ k ] = silk_SMULWW( psEncCtrl->Gains_Q16[ k ], gain_mult_Q16 ); + silk_assert( psEncCtrl->Gains_Q16[ k ] >= 0 ); + psEncCtrl->Gains_Q16[ k ] = silk_ADD_POS_SAT32( psEncCtrl->Gains_Q16[ k ], gain_add_Q16 ); + } + + + /************************************************/ + /* Control low-frequency shaping and noise tilt */ + /************************************************/ + /* Less low frequency shaping for noisy inputs */ + strength_Q16 = silk_MUL( SILK_FIX_CONST( LOW_FREQ_SHAPING, 4 ), silk_SMLAWB( SILK_FIX_CONST( 1.0, 12 ), + SILK_FIX_CONST( LOW_QUALITY_LOW_FREQ_SHAPING_DECR, 13 ), psEnc->sCmn.input_quality_bands_Q15[ 0 ] - SILK_FIX_CONST( 1.0, 15 ) ) ); + strength_Q16 = silk_RSHIFT( silk_MUL( strength_Q16, psEnc->sCmn.speech_activity_Q8 ), 8 ); + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Reduce low frequencies quantization noise for periodic signals, depending on pitch lag */ + /*f = 400; freqz([1, -0.98 + 2e-4 * f], [1, -0.97 + 7e-4 * f], 2^12, Fs); axis([0, 1000, -10, 1])*/ + opus_int fs_kHz_inv = silk_DIV32_16( SILK_FIX_CONST( 0.2, 14 ), psEnc->sCmn.fs_kHz ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + b_Q14 = fs_kHz_inv + silk_DIV32_16( SILK_FIX_CONST( 3.0, 14 ), psEncCtrl->pitchL[ k ] ); + /* Pack two coefficients in one int32 */ + psEncCtrl->LF_shp_Q14[ k ] = silk_LSHIFT( SILK_FIX_CONST( 1.0, 14 ) - b_Q14 - silk_SMULWB( strength_Q16, b_Q14 ), 16 ); + psEncCtrl->LF_shp_Q14[ k ] |= (opus_uint16)( b_Q14 - SILK_FIX_CONST( 1.0, 14 ) ); + } + silk_assert( SILK_FIX_CONST( HARM_HP_NOISE_COEF, 24 ) < SILK_FIX_CONST( 0.5, 24 ) ); /* Guarantees that second argument to SMULWB() is within range of an opus_int16*/ + Tilt_Q16 = - SILK_FIX_CONST( HP_NOISE_COEF, 16 ) - + silk_SMULWB( SILK_FIX_CONST( 1.0, 16 ) - SILK_FIX_CONST( HP_NOISE_COEF, 16 ), + silk_SMULWB( SILK_FIX_CONST( HARM_HP_NOISE_COEF, 24 ), psEnc->sCmn.speech_activity_Q8 ) ); + } else { + b_Q14 = silk_DIV32_16( 21299, psEnc->sCmn.fs_kHz ); /* 1.3_Q0 = 21299_Q14*/ + /* Pack two coefficients in one int32 */ + psEncCtrl->LF_shp_Q14[ 0 ] = silk_LSHIFT( SILK_FIX_CONST( 1.0, 14 ) - b_Q14 - + silk_SMULWB( strength_Q16, silk_SMULWB( SILK_FIX_CONST( 0.6, 16 ), b_Q14 ) ), 16 ); + psEncCtrl->LF_shp_Q14[ 0 ] |= (opus_uint16)( b_Q14 - SILK_FIX_CONST( 1.0, 14 ) ); + for( k = 1; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->LF_shp_Q14[ k ] = psEncCtrl->LF_shp_Q14[ 0 ]; + } + Tilt_Q16 = -SILK_FIX_CONST( HP_NOISE_COEF, 16 ); + } + + /****************************/ + /* HARMONIC SHAPING CONTROL */ + /****************************/ + if( USE_HARM_SHAPING && psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* More harmonic noise shaping for high bitrates or noisy input */ + HarmShapeGain_Q16 = silk_SMLAWB( SILK_FIX_CONST( HARMONIC_SHAPING, 16 ), + SILK_FIX_CONST( 1.0, 16 ) - silk_SMULWB( SILK_FIX_CONST( 1.0, 18 ) - silk_LSHIFT( psEncCtrl->coding_quality_Q14, 4 ), + psEncCtrl->input_quality_Q14 ), SILK_FIX_CONST( HIGH_RATE_OR_LOW_QUALITY_HARMONIC_SHAPING, 16 ) ); + + /* Less harmonic noise shaping for less periodic signals */ + HarmShapeGain_Q16 = silk_SMULWB( silk_LSHIFT( HarmShapeGain_Q16, 1 ), + silk_SQRT_APPROX( silk_LSHIFT( psEnc->LTPCorr_Q15, 15 ) ) ); + } else { + HarmShapeGain_Q16 = 0; + } + + /*************************/ + /* Smooth over subframes */ + /*************************/ + for( k = 0; k < MAX_NB_SUBFR; k++ ) { + psShapeSt->HarmShapeGain_smth_Q16 = + silk_SMLAWB( psShapeSt->HarmShapeGain_smth_Q16, HarmShapeGain_Q16 - psShapeSt->HarmShapeGain_smth_Q16, SILK_FIX_CONST( SUBFR_SMTH_COEF, 16 ) ); + psShapeSt->Tilt_smth_Q16 = + silk_SMLAWB( psShapeSt->Tilt_smth_Q16, Tilt_Q16 - psShapeSt->Tilt_smth_Q16, SILK_FIX_CONST( SUBFR_SMTH_COEF, 16 ) ); + + psEncCtrl->HarmShapeGain_Q14[ k ] = ( opus_int )silk_RSHIFT_ROUND( psShapeSt->HarmShapeGain_smth_Q16, 2 ); + psEncCtrl->Tilt_Q14[ k ] = ( opus_int )silk_RSHIFT_ROUND( psShapeSt->Tilt_smth_Q16, 2 ); + } + RESTORE_STACK; +} +#endif /* OVERRIDE_silk_noise_shape_analysis_FIX */ diff --git a/lib/opus/silk/fixed/pitch_analysis_core_FIX.c b/lib/opus/silk/fixed/pitch_analysis_core_FIX.c new file mode 100644 index 00000000..14729046 --- /dev/null +++ b/lib/opus/silk/fixed/pitch_analysis_core_FIX.c @@ -0,0 +1,721 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/*********************************************************** +* Pitch analyser function +********************************************************** */ +#include "SigProc_FIX.h" +#include "pitch_est_defines.h" +#include "stack_alloc.h" +#include "debug.h" +#include "pitch.h" + +#define SCRATCH_SIZE 22 +#define SF_LENGTH_4KHZ ( PE_SUBFR_LENGTH_MS * 4 ) +#define SF_LENGTH_8KHZ ( PE_SUBFR_LENGTH_MS * 8 ) +#define MIN_LAG_4KHZ ( PE_MIN_LAG_MS * 4 ) +#define MIN_LAG_8KHZ ( PE_MIN_LAG_MS * 8 ) +#define MAX_LAG_4KHZ ( PE_MAX_LAG_MS * 4 ) +#define MAX_LAG_8KHZ ( PE_MAX_LAG_MS * 8 - 1 ) +#define CSTRIDE_4KHZ ( MAX_LAG_4KHZ + 1 - MIN_LAG_4KHZ ) +#define CSTRIDE_8KHZ ( MAX_LAG_8KHZ + 3 - ( MIN_LAG_8KHZ - 2 ) ) +#define D_COMP_MIN ( MIN_LAG_8KHZ - 3 ) +#define D_COMP_MAX ( MAX_LAG_8KHZ + 4 ) +#define D_COMP_STRIDE ( D_COMP_MAX - D_COMP_MIN ) + +typedef opus_int32 silk_pe_stage3_vals[ PE_NB_STAGE3_LAGS ]; + +/************************************************************/ +/* Internally used functions */ +/************************************************************/ +static void silk_P_Ana_calc_corr_st3( + silk_pe_stage3_vals cross_corr_st3[], /* O 3 DIM correlation array */ + const opus_int16 frame[], /* I vector to correlate */ + opus_int start_lag, /* I lag offset to search around */ + opus_int sf_length, /* I length of a 5 ms subframe */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ +); + +static void silk_P_Ana_calc_energy_st3( + silk_pe_stage3_vals energies_st3[], /* O 3 DIM energy array */ + const opus_int16 frame[], /* I vector to calc energy in */ + opus_int start_lag, /* I lag offset to search around */ + opus_int sf_length, /* I length of one 5 ms subframe */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ +); + +/*************************************************************/ +/* FIXED POINT CORE PITCH ANALYSIS FUNCTION */ +/*************************************************************/ +opus_int silk_pitch_analysis_core( /* O Voicing estimate: 0 voiced, 1 unvoiced */ + const opus_int16 *frame_unscaled, /* I Signal of length PE_FRAME_LENGTH_MS*Fs_kHz */ + opus_int *pitch_out, /* O 4 pitch lag values */ + opus_int16 *lagIndex, /* O Lag Index */ + opus_int8 *contourIndex, /* O Pitch contour Index */ + opus_int *LTPCorr_Q15, /* I/O Normalized correlation; input: value from previous frame */ + opus_int prevLag, /* I Last lag of previous frame; set to zero is unvoiced */ + const opus_int32 search_thres1_Q16, /* I First stage threshold for lag candidates 0 - 1 */ + const opus_int search_thres2_Q13, /* I Final threshold for lag candidates 0 - 1 */ + const opus_int Fs_kHz, /* I Sample frequency (kHz) */ + const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ + const opus_int nb_subfr, /* I number of 5 ms subframes */ + int arch /* I Run-time architecture */ +) +{ + VARDECL( opus_int16, frame_8kHz_buf ); + VARDECL( opus_int16, frame_4kHz ); + VARDECL( opus_int16, frame_scaled ); + opus_int32 filt_state[ 6 ]; + const opus_int16 *frame, *frame_8kHz; + opus_int i, k, d, j; + VARDECL( opus_int16, C ); + VARDECL( opus_int32, xcorr32 ); + const opus_int16 *target_ptr, *basis_ptr; + opus_int32 cross_corr, normalizer, energy, energy_basis, energy_target; + opus_int d_srch[ PE_D_SRCH_LENGTH ], Cmax, length_d_srch, length_d_comp, shift; + VARDECL( opus_int16, d_comp ); + opus_int32 sum, threshold, lag_counter; + opus_int CBimax, CBimax_new, CBimax_old, lag, start_lag, end_lag, lag_new; + opus_int32 CC[ PE_NB_CBKS_STAGE2_EXT ], CCmax, CCmax_b, CCmax_new_b, CCmax_new; + VARDECL( silk_pe_stage3_vals, energies_st3 ); + VARDECL( silk_pe_stage3_vals, cross_corr_st3 ); + opus_int frame_length, frame_length_8kHz, frame_length_4kHz; + opus_int sf_length; + opus_int min_lag; + opus_int max_lag; + opus_int32 contour_bias_Q15, diff; + opus_int nb_cbk_search, cbk_size; + opus_int32 delta_lag_log2_sqr_Q7, lag_log2_Q7, prevLag_log2_Q7, prev_lag_bias_Q13; + const opus_int8 *Lag_CB_ptr; + SAVE_STACK; + + /* Check for valid sampling frequency */ + celt_assert( Fs_kHz == 8 || Fs_kHz == 12 || Fs_kHz == 16 ); + + /* Check for valid complexity setting */ + celt_assert( complexity >= SILK_PE_MIN_COMPLEX ); + celt_assert( complexity <= SILK_PE_MAX_COMPLEX ); + + silk_assert( search_thres1_Q16 >= 0 && search_thres1_Q16 <= (1<<16) ); + silk_assert( search_thres2_Q13 >= 0 && search_thres2_Q13 <= (1<<13) ); + + /* Set up frame lengths max / min lag for the sampling frequency */ + frame_length = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * Fs_kHz; + frame_length_4kHz = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * 4; + frame_length_8kHz = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * 8; + sf_length = PE_SUBFR_LENGTH_MS * Fs_kHz; + min_lag = PE_MIN_LAG_MS * Fs_kHz; + max_lag = PE_MAX_LAG_MS * Fs_kHz - 1; + + /* Downscale input if necessary */ + silk_sum_sqr_shift( &energy, &shift, frame_unscaled, frame_length ); + shift += 3 - silk_CLZ32( energy ); /* at least two bits headroom */ + ALLOC( frame_scaled, frame_length, opus_int16 ); + if( shift > 0 ) { + shift = silk_RSHIFT( shift + 1, 1 ); + for( i = 0; i < frame_length; i++ ) { + frame_scaled[ i ] = silk_RSHIFT( frame_unscaled[ i ], shift ); + } + frame = frame_scaled; + } else { + frame = frame_unscaled; + } + + ALLOC( frame_8kHz_buf, ( Fs_kHz == 8 ) ? 1 : frame_length_8kHz, opus_int16 ); + /* Resample from input sampled at Fs_kHz to 8 kHz */ + if( Fs_kHz == 16 ) { + silk_memset( filt_state, 0, 2 * sizeof( opus_int32 ) ); + silk_resampler_down2( filt_state, frame_8kHz_buf, frame, frame_length ); + frame_8kHz = frame_8kHz_buf; + } else if( Fs_kHz == 12 ) { + silk_memset( filt_state, 0, 6 * sizeof( opus_int32 ) ); + silk_resampler_down2_3( filt_state, frame_8kHz_buf, frame, frame_length ); + frame_8kHz = frame_8kHz_buf; + } else { + celt_assert( Fs_kHz == 8 ); + frame_8kHz = frame; + } + + /* Decimate again to 4 kHz */ + silk_memset( filt_state, 0, 2 * sizeof( opus_int32 ) );/* Set state to zero */ + ALLOC( frame_4kHz, frame_length_4kHz, opus_int16 ); + silk_resampler_down2( filt_state, frame_4kHz, frame_8kHz, frame_length_8kHz ); + + /* Low-pass filter */ + for( i = frame_length_4kHz - 1; i > 0; i-- ) { + frame_4kHz[ i ] = silk_ADD_SAT16( frame_4kHz[ i ], frame_4kHz[ i - 1 ] ); + } + + + /****************************************************************************** + * FIRST STAGE, operating in 4 khz + ******************************************************************************/ + ALLOC( C, nb_subfr * CSTRIDE_8KHZ, opus_int16 ); + ALLOC( xcorr32, MAX_LAG_4KHZ-MIN_LAG_4KHZ+1, opus_int32 ); + silk_memset( C, 0, (nb_subfr >> 1) * CSTRIDE_4KHZ * sizeof( opus_int16 ) ); + target_ptr = &frame_4kHz[ silk_LSHIFT( SF_LENGTH_4KHZ, 2 ) ]; + for( k = 0; k < nb_subfr >> 1; k++ ) { + /* Check that we are within range of the array */ + celt_assert( target_ptr >= frame_4kHz ); + celt_assert( target_ptr + SF_LENGTH_8KHZ <= frame_4kHz + frame_length_4kHz ); + + basis_ptr = target_ptr - MIN_LAG_4KHZ; + + /* Check that we are within range of the array */ + celt_assert( basis_ptr >= frame_4kHz ); + celt_assert( basis_ptr + SF_LENGTH_8KHZ <= frame_4kHz + frame_length_4kHz ); + + celt_pitch_xcorr( target_ptr, target_ptr - MAX_LAG_4KHZ, xcorr32, SF_LENGTH_8KHZ, MAX_LAG_4KHZ - MIN_LAG_4KHZ + 1, arch ); + + /* Calculate first vector products before loop */ + cross_corr = xcorr32[ MAX_LAG_4KHZ - MIN_LAG_4KHZ ]; + normalizer = silk_inner_prod_aligned( target_ptr, target_ptr, SF_LENGTH_8KHZ, arch ); + normalizer = silk_ADD32( normalizer, silk_inner_prod_aligned( basis_ptr, basis_ptr, SF_LENGTH_8KHZ, arch ) ); + normalizer = silk_ADD32( normalizer, silk_SMULBB( SF_LENGTH_8KHZ, 4000 ) ); + + matrix_ptr( C, k, 0, CSTRIDE_4KHZ ) = + (opus_int16)silk_DIV32_varQ( cross_corr, normalizer, 13 + 1 ); /* Q13 */ + + /* From now on normalizer is computed recursively */ + for( d = MIN_LAG_4KHZ + 1; d <= MAX_LAG_4KHZ; d++ ) { + basis_ptr--; + + /* Check that we are within range of the array */ + silk_assert( basis_ptr >= frame_4kHz ); + silk_assert( basis_ptr + SF_LENGTH_8KHZ <= frame_4kHz + frame_length_4kHz ); + + cross_corr = xcorr32[ MAX_LAG_4KHZ - d ]; + + /* Add contribution of new sample and remove contribution from oldest sample */ + normalizer = silk_ADD32( normalizer, + silk_SMULBB( basis_ptr[ 0 ], basis_ptr[ 0 ] ) - + silk_SMULBB( basis_ptr[ SF_LENGTH_8KHZ ], basis_ptr[ SF_LENGTH_8KHZ ] ) ); + + matrix_ptr( C, k, d - MIN_LAG_4KHZ, CSTRIDE_4KHZ) = + (opus_int16)silk_DIV32_varQ( cross_corr, normalizer, 13 + 1 ); /* Q13 */ + } + /* Update target pointer */ + target_ptr += SF_LENGTH_8KHZ; + } + + /* Combine two subframes into single correlation measure and apply short-lag bias */ + if( nb_subfr == PE_MAX_NB_SUBFR ) { + for( i = MAX_LAG_4KHZ; i >= MIN_LAG_4KHZ; i-- ) { + sum = (opus_int32)matrix_ptr( C, 0, i - MIN_LAG_4KHZ, CSTRIDE_4KHZ ) + + (opus_int32)matrix_ptr( C, 1, i - MIN_LAG_4KHZ, CSTRIDE_4KHZ ); /* Q14 */ + sum = silk_SMLAWB( sum, sum, silk_LSHIFT( -i, 4 ) ); /* Q14 */ + C[ i - MIN_LAG_4KHZ ] = (opus_int16)sum; /* Q14 */ + } + } else { + /* Only short-lag bias */ + for( i = MAX_LAG_4KHZ; i >= MIN_LAG_4KHZ; i-- ) { + sum = silk_LSHIFT( (opus_int32)C[ i - MIN_LAG_4KHZ ], 1 ); /* Q14 */ + sum = silk_SMLAWB( sum, sum, silk_LSHIFT( -i, 4 ) ); /* Q14 */ + C[ i - MIN_LAG_4KHZ ] = (opus_int16)sum; /* Q14 */ + } + } + + /* Sort */ + length_d_srch = silk_ADD_LSHIFT32( 4, complexity, 1 ); + celt_assert( 3 * length_d_srch <= PE_D_SRCH_LENGTH ); + silk_insertion_sort_decreasing_int16( C, d_srch, CSTRIDE_4KHZ, + length_d_srch ); + + /* Escape if correlation is very low already here */ + Cmax = (opus_int)C[ 0 ]; /* Q14 */ + if( Cmax < SILK_FIX_CONST( 0.2, 14 ) ) { + silk_memset( pitch_out, 0, nb_subfr * sizeof( opus_int ) ); + *LTPCorr_Q15 = 0; + *lagIndex = 0; + *contourIndex = 0; + RESTORE_STACK; + return 1; + } + + threshold = silk_SMULWB( search_thres1_Q16, Cmax ); + for( i = 0; i < length_d_srch; i++ ) { + /* Convert to 8 kHz indices for the sorted correlation that exceeds the threshold */ + if( C[ i ] > threshold ) { + d_srch[ i ] = silk_LSHIFT( d_srch[ i ] + MIN_LAG_4KHZ, 1 ); + } else { + length_d_srch = i; + break; + } + } + celt_assert( length_d_srch > 0 ); + + ALLOC( d_comp, D_COMP_STRIDE, opus_int16 ); + for( i = D_COMP_MIN; i < D_COMP_MAX; i++ ) { + d_comp[ i - D_COMP_MIN ] = 0; + } + for( i = 0; i < length_d_srch; i++ ) { + d_comp[ d_srch[ i ] - D_COMP_MIN ] = 1; + } + + /* Convolution */ + for( i = D_COMP_MAX - 1; i >= MIN_LAG_8KHZ; i-- ) { + d_comp[ i - D_COMP_MIN ] += + d_comp[ i - 1 - D_COMP_MIN ] + d_comp[ i - 2 - D_COMP_MIN ]; + } + + length_d_srch = 0; + for( i = MIN_LAG_8KHZ; i < MAX_LAG_8KHZ + 1; i++ ) { + if( d_comp[ i + 1 - D_COMP_MIN ] > 0 ) { + d_srch[ length_d_srch ] = i; + length_d_srch++; + } + } + + /* Convolution */ + for( i = D_COMP_MAX - 1; i >= MIN_LAG_8KHZ; i-- ) { + d_comp[ i - D_COMP_MIN ] += d_comp[ i - 1 - D_COMP_MIN ] + + d_comp[ i - 2 - D_COMP_MIN ] + d_comp[ i - 3 - D_COMP_MIN ]; + } + + length_d_comp = 0; + for( i = MIN_LAG_8KHZ; i < D_COMP_MAX; i++ ) { + if( d_comp[ i - D_COMP_MIN ] > 0 ) { + d_comp[ length_d_comp ] = i - 2; + length_d_comp++; + } + } + + /********************************************************************************** + ** SECOND STAGE, operating at 8 kHz, on lag sections with high correlation + *************************************************************************************/ + + /********************************************************************************* + * Find energy of each subframe projected onto its history, for a range of delays + *********************************************************************************/ + silk_memset( C, 0, nb_subfr * CSTRIDE_8KHZ * sizeof( opus_int16 ) ); + + target_ptr = &frame_8kHz[ PE_LTP_MEM_LENGTH_MS * 8 ]; + for( k = 0; k < nb_subfr; k++ ) { + + /* Check that we are within range of the array */ + celt_assert( target_ptr >= frame_8kHz ); + celt_assert( target_ptr + SF_LENGTH_8KHZ <= frame_8kHz + frame_length_8kHz ); + + energy_target = silk_ADD32( silk_inner_prod_aligned( target_ptr, target_ptr, SF_LENGTH_8KHZ, arch ), 1 ); + for( j = 0; j < length_d_comp; j++ ) { + d = d_comp[ j ]; + basis_ptr = target_ptr - d; + + /* Check that we are within range of the array */ + silk_assert( basis_ptr >= frame_8kHz ); + silk_assert( basis_ptr + SF_LENGTH_8KHZ <= frame_8kHz + frame_length_8kHz ); + + cross_corr = silk_inner_prod_aligned( target_ptr, basis_ptr, SF_LENGTH_8KHZ, arch ); + if( cross_corr > 0 ) { + energy_basis = silk_inner_prod_aligned( basis_ptr, basis_ptr, SF_LENGTH_8KHZ, arch ); + matrix_ptr( C, k, d - ( MIN_LAG_8KHZ - 2 ), CSTRIDE_8KHZ ) = + (opus_int16)silk_DIV32_varQ( cross_corr, + silk_ADD32( energy_target, + energy_basis ), + 13 + 1 ); /* Q13 */ + } else { + matrix_ptr( C, k, d - ( MIN_LAG_8KHZ - 2 ), CSTRIDE_8KHZ ) = 0; + } + } + target_ptr += SF_LENGTH_8KHZ; + } + + /* search over lag range and lags codebook */ + /* scale factor for lag codebook, as a function of center lag */ + + CCmax = silk_int32_MIN; + CCmax_b = silk_int32_MIN; + + CBimax = 0; /* To avoid returning undefined lag values */ + lag = -1; /* To check if lag with strong enough correlation has been found */ + + if( prevLag > 0 ) { + if( Fs_kHz == 12 ) { + prevLag = silk_DIV32_16( silk_LSHIFT( prevLag, 1 ), 3 ); + } else if( Fs_kHz == 16 ) { + prevLag = silk_RSHIFT( prevLag, 1 ); + } + prevLag_log2_Q7 = silk_lin2log( (opus_int32)prevLag ); + } else { + prevLag_log2_Q7 = 0; + } + silk_assert( search_thres2_Q13 == silk_SAT16( search_thres2_Q13 ) ); + /* Set up stage 2 codebook based on number of subframes */ + if( nb_subfr == PE_MAX_NB_SUBFR ) { + cbk_size = PE_NB_CBKS_STAGE2_EXT; + Lag_CB_ptr = &silk_CB_lags_stage2[ 0 ][ 0 ]; + if( Fs_kHz == 8 && complexity > SILK_PE_MIN_COMPLEX ) { + /* If input is 8 khz use a larger codebook here because it is last stage */ + nb_cbk_search = PE_NB_CBKS_STAGE2_EXT; + } else { + nb_cbk_search = PE_NB_CBKS_STAGE2; + } + } else { + cbk_size = PE_NB_CBKS_STAGE2_10MS; + Lag_CB_ptr = &silk_CB_lags_stage2_10_ms[ 0 ][ 0 ]; + nb_cbk_search = PE_NB_CBKS_STAGE2_10MS; + } + + for( k = 0; k < length_d_srch; k++ ) { + d = d_srch[ k ]; + for( j = 0; j < nb_cbk_search; j++ ) { + CC[ j ] = 0; + for( i = 0; i < nb_subfr; i++ ) { + opus_int d_subfr; + /* Try all codebooks */ + d_subfr = d + matrix_ptr( Lag_CB_ptr, i, j, cbk_size ); + CC[ j ] = CC[ j ] + + (opus_int32)matrix_ptr( C, i, + d_subfr - ( MIN_LAG_8KHZ - 2 ), + CSTRIDE_8KHZ ); + } + } + /* Find best codebook */ + CCmax_new = silk_int32_MIN; + CBimax_new = 0; + for( i = 0; i < nb_cbk_search; i++ ) { + if( CC[ i ] > CCmax_new ) { + CCmax_new = CC[ i ]; + CBimax_new = i; + } + } + + /* Bias towards shorter lags */ + lag_log2_Q7 = silk_lin2log( d ); /* Q7 */ + silk_assert( lag_log2_Q7 == silk_SAT16( lag_log2_Q7 ) ); + silk_assert( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 13 ) == silk_SAT16( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 13 ) ) ); + CCmax_new_b = CCmax_new - silk_RSHIFT( silk_SMULBB( nb_subfr * SILK_FIX_CONST( PE_SHORTLAG_BIAS, 13 ), lag_log2_Q7 ), 7 ); /* Q13 */ + + /* Bias towards previous lag */ + silk_assert( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 13 ) == silk_SAT16( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 13 ) ) ); + if( prevLag > 0 ) { + delta_lag_log2_sqr_Q7 = lag_log2_Q7 - prevLag_log2_Q7; + silk_assert( delta_lag_log2_sqr_Q7 == silk_SAT16( delta_lag_log2_sqr_Q7 ) ); + delta_lag_log2_sqr_Q7 = silk_RSHIFT( silk_SMULBB( delta_lag_log2_sqr_Q7, delta_lag_log2_sqr_Q7 ), 7 ); + prev_lag_bias_Q13 = silk_RSHIFT( silk_SMULBB( nb_subfr * SILK_FIX_CONST( PE_PREVLAG_BIAS, 13 ), *LTPCorr_Q15 ), 15 ); /* Q13 */ + prev_lag_bias_Q13 = silk_DIV32( silk_MUL( prev_lag_bias_Q13, delta_lag_log2_sqr_Q7 ), delta_lag_log2_sqr_Q7 + SILK_FIX_CONST( 0.5, 7 ) ); + CCmax_new_b -= prev_lag_bias_Q13; /* Q13 */ + } + + if( CCmax_new_b > CCmax_b && /* Find maximum biased correlation */ + CCmax_new > silk_SMULBB( nb_subfr, search_thres2_Q13 ) && /* Correlation needs to be high enough to be voiced */ + silk_CB_lags_stage2[ 0 ][ CBimax_new ] <= MIN_LAG_8KHZ /* Lag must be in range */ + ) { + CCmax_b = CCmax_new_b; + CCmax = CCmax_new; + lag = d; + CBimax = CBimax_new; + } + } + + if( lag == -1 ) { + /* No suitable candidate found */ + silk_memset( pitch_out, 0, nb_subfr * sizeof( opus_int ) ); + *LTPCorr_Q15 = 0; + *lagIndex = 0; + *contourIndex = 0; + RESTORE_STACK; + return 1; + } + + /* Output normalized correlation */ + *LTPCorr_Q15 = (opus_int)silk_LSHIFT( silk_DIV32_16( CCmax, nb_subfr ), 2 ); + silk_assert( *LTPCorr_Q15 >= 0 ); + + if( Fs_kHz > 8 ) { + /* Search in original signal */ + + CBimax_old = CBimax; + /* Compensate for decimation */ + silk_assert( lag == silk_SAT16( lag ) ); + if( Fs_kHz == 12 ) { + lag = silk_RSHIFT( silk_SMULBB( lag, 3 ), 1 ); + } else if( Fs_kHz == 16 ) { + lag = silk_LSHIFT( lag, 1 ); + } else { + lag = silk_SMULBB( lag, 3 ); + } + + lag = silk_LIMIT_int( lag, min_lag, max_lag ); + start_lag = silk_max_int( lag - 2, min_lag ); + end_lag = silk_min_int( lag + 2, max_lag ); + lag_new = lag; /* to avoid undefined lag */ + CBimax = 0; /* to avoid undefined lag */ + + CCmax = silk_int32_MIN; + /* pitch lags according to second stage */ + for( k = 0; k < nb_subfr; k++ ) { + pitch_out[ k ] = lag + 2 * silk_CB_lags_stage2[ k ][ CBimax_old ]; + } + + /* Set up codebook parameters according to complexity setting and frame length */ + if( nb_subfr == PE_MAX_NB_SUBFR ) { + nb_cbk_search = (opus_int)silk_nb_cbk_searchs_stage3[ complexity ]; + cbk_size = PE_NB_CBKS_STAGE3_MAX; + Lag_CB_ptr = &silk_CB_lags_stage3[ 0 ][ 0 ]; + } else { + nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; + cbk_size = PE_NB_CBKS_STAGE3_10MS; + Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; + } + + /* Calculate the correlations and energies needed in stage 3 */ + ALLOC( energies_st3, nb_subfr * nb_cbk_search, silk_pe_stage3_vals ); + ALLOC( cross_corr_st3, nb_subfr * nb_cbk_search, silk_pe_stage3_vals ); + silk_P_Ana_calc_corr_st3( cross_corr_st3, frame, start_lag, sf_length, nb_subfr, complexity, arch ); + silk_P_Ana_calc_energy_st3( energies_st3, frame, start_lag, sf_length, nb_subfr, complexity, arch ); + + lag_counter = 0; + silk_assert( lag == silk_SAT16( lag ) ); + contour_bias_Q15 = silk_DIV32_16( SILK_FIX_CONST( PE_FLATCONTOUR_BIAS, 15 ), lag ); + + target_ptr = &frame[ PE_LTP_MEM_LENGTH_MS * Fs_kHz ]; + energy_target = silk_ADD32( silk_inner_prod_aligned( target_ptr, target_ptr, nb_subfr * sf_length, arch ), 1 ); + for( d = start_lag; d <= end_lag; d++ ) { + for( j = 0; j < nb_cbk_search; j++ ) { + cross_corr = 0; + energy = energy_target; + for( k = 0; k < nb_subfr; k++ ) { + cross_corr = silk_ADD32( cross_corr, + matrix_ptr( cross_corr_st3, k, j, + nb_cbk_search )[ lag_counter ] ); + energy = silk_ADD32( energy, + matrix_ptr( energies_st3, k, j, + nb_cbk_search )[ lag_counter ] ); + silk_assert( energy >= 0 ); + } + if( cross_corr > 0 ) { + CCmax_new = silk_DIV32_varQ( cross_corr, energy, 13 + 1 ); /* Q13 */ + /* Reduce depending on flatness of contour */ + diff = silk_int16_MAX - silk_MUL( contour_bias_Q15, j ); /* Q15 */ + silk_assert( diff == silk_SAT16( diff ) ); + CCmax_new = silk_SMULWB( CCmax_new, diff ); /* Q14 */ + } else { + CCmax_new = 0; + } + + if( CCmax_new > CCmax && ( d + silk_CB_lags_stage3[ 0 ][ j ] ) <= max_lag ) { + CCmax = CCmax_new; + lag_new = d; + CBimax = j; + } + } + lag_counter++; + } + + for( k = 0; k < nb_subfr; k++ ) { + pitch_out[ k ] = lag_new + matrix_ptr( Lag_CB_ptr, k, CBimax, cbk_size ); + pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], min_lag, PE_MAX_LAG_MS * Fs_kHz ); + } + *lagIndex = (opus_int16)( lag_new - min_lag); + *contourIndex = (opus_int8)CBimax; + } else { /* Fs_kHz == 8 */ + /* Save Lags */ + for( k = 0; k < nb_subfr; k++ ) { + pitch_out[ k ] = lag + matrix_ptr( Lag_CB_ptr, k, CBimax, cbk_size ); + pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], MIN_LAG_8KHZ, PE_MAX_LAG_MS * 8 ); + } + *lagIndex = (opus_int16)( lag - MIN_LAG_8KHZ ); + *contourIndex = (opus_int8)CBimax; + } + celt_assert( *lagIndex >= 0 ); + /* return as voiced */ + RESTORE_STACK; + return 0; +} + +/*********************************************************************** + * Calculates the correlations used in stage 3 search. In order to cover + * the whole lag codebook for all the searched offset lags (lag +- 2), + * the following correlations are needed in each sub frame: + * + * sf1: lag range [-8,...,7] total 16 correlations + * sf2: lag range [-4,...,4] total 9 correlations + * sf3: lag range [-3,....4] total 8 correltions + * sf4: lag range [-6,....8] total 15 correlations + * + * In total 48 correlations. The direct implementation computed in worst + * case 4*12*5 = 240 correlations, but more likely around 120. + ***********************************************************************/ +static void silk_P_Ana_calc_corr_st3( + silk_pe_stage3_vals cross_corr_st3[], /* O 3 DIM correlation array */ + const opus_int16 frame[], /* I vector to correlate */ + opus_int start_lag, /* I lag offset to search around */ + opus_int sf_length, /* I length of a 5 ms subframe */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ +) +{ + const opus_int16 *target_ptr; + opus_int i, j, k, lag_counter, lag_low, lag_high; + opus_int nb_cbk_search, delta, idx, cbk_size; + VARDECL( opus_int32, scratch_mem ); + VARDECL( opus_int32, xcorr32 ); + const opus_int8 *Lag_range_ptr, *Lag_CB_ptr; + SAVE_STACK; + + celt_assert( complexity >= SILK_PE_MIN_COMPLEX ); + celt_assert( complexity <= SILK_PE_MAX_COMPLEX ); + + if( nb_subfr == PE_MAX_NB_SUBFR ) { + Lag_range_ptr = &silk_Lag_range_stage3[ complexity ][ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3[ 0 ][ 0 ]; + nb_cbk_search = silk_nb_cbk_searchs_stage3[ complexity ]; + cbk_size = PE_NB_CBKS_STAGE3_MAX; + } else { + celt_assert( nb_subfr == PE_MAX_NB_SUBFR >> 1); + Lag_range_ptr = &silk_Lag_range_stage3_10_ms[ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; + nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; + cbk_size = PE_NB_CBKS_STAGE3_10MS; + } + ALLOC( scratch_mem, SCRATCH_SIZE, opus_int32 ); + ALLOC( xcorr32, SCRATCH_SIZE, opus_int32 ); + + target_ptr = &frame[ silk_LSHIFT( sf_length, 2 ) ]; /* Pointer to middle of frame */ + for( k = 0; k < nb_subfr; k++ ) { + lag_counter = 0; + + /* Calculate the correlations for each subframe */ + lag_low = matrix_ptr( Lag_range_ptr, k, 0, 2 ); + lag_high = matrix_ptr( Lag_range_ptr, k, 1, 2 ); + celt_assert(lag_high-lag_low+1 <= SCRATCH_SIZE); + celt_pitch_xcorr( target_ptr, target_ptr - start_lag - lag_high, xcorr32, sf_length, lag_high - lag_low + 1, arch ); + for( j = lag_low; j <= lag_high; j++ ) { + silk_assert( lag_counter < SCRATCH_SIZE ); + scratch_mem[ lag_counter ] = xcorr32[ lag_high - j ]; + lag_counter++; + } + + delta = matrix_ptr( Lag_range_ptr, k, 0, 2 ); + for( i = 0; i < nb_cbk_search; i++ ) { + /* Fill out the 3 dim array that stores the correlations for */ + /* each code_book vector for each start lag */ + idx = matrix_ptr( Lag_CB_ptr, k, i, cbk_size ) - delta; + for( j = 0; j < PE_NB_STAGE3_LAGS; j++ ) { + silk_assert( idx + j < SCRATCH_SIZE ); + silk_assert( idx + j < lag_counter ); + matrix_ptr( cross_corr_st3, k, i, nb_cbk_search )[ j ] = + scratch_mem[ idx + j ]; + } + } + target_ptr += sf_length; + } + RESTORE_STACK; +} + +/********************************************************************/ +/* Calculate the energies for first two subframes. The energies are */ +/* calculated recursively. */ +/********************************************************************/ +static void silk_P_Ana_calc_energy_st3( + silk_pe_stage3_vals energies_st3[], /* O 3 DIM energy array */ + const opus_int16 frame[], /* I vector to calc energy in */ + opus_int start_lag, /* I lag offset to search around */ + opus_int sf_length, /* I length of one 5 ms subframe */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ +) +{ + const opus_int16 *target_ptr, *basis_ptr; + opus_int32 energy; + opus_int k, i, j, lag_counter; + opus_int nb_cbk_search, delta, idx, cbk_size, lag_diff; + VARDECL( opus_int32, scratch_mem ); + const opus_int8 *Lag_range_ptr, *Lag_CB_ptr; + SAVE_STACK; + + celt_assert( complexity >= SILK_PE_MIN_COMPLEX ); + celt_assert( complexity <= SILK_PE_MAX_COMPLEX ); + + if( nb_subfr == PE_MAX_NB_SUBFR ) { + Lag_range_ptr = &silk_Lag_range_stage3[ complexity ][ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3[ 0 ][ 0 ]; + nb_cbk_search = silk_nb_cbk_searchs_stage3[ complexity ]; + cbk_size = PE_NB_CBKS_STAGE3_MAX; + } else { + celt_assert( nb_subfr == PE_MAX_NB_SUBFR >> 1); + Lag_range_ptr = &silk_Lag_range_stage3_10_ms[ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; + nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; + cbk_size = PE_NB_CBKS_STAGE3_10MS; + } + ALLOC( scratch_mem, SCRATCH_SIZE, opus_int32 ); + + target_ptr = &frame[ silk_LSHIFT( sf_length, 2 ) ]; + for( k = 0; k < nb_subfr; k++ ) { + lag_counter = 0; + + /* Calculate the energy for first lag */ + basis_ptr = target_ptr - ( start_lag + matrix_ptr( Lag_range_ptr, k, 0, 2 ) ); + energy = silk_inner_prod_aligned( basis_ptr, basis_ptr, sf_length, arch ); + silk_assert( energy >= 0 ); + scratch_mem[ lag_counter ] = energy; + lag_counter++; + + lag_diff = ( matrix_ptr( Lag_range_ptr, k, 1, 2 ) - matrix_ptr( Lag_range_ptr, k, 0, 2 ) + 1 ); + for( i = 1; i < lag_diff; i++ ) { + /* remove part outside new window */ + energy -= silk_SMULBB( basis_ptr[ sf_length - i ], basis_ptr[ sf_length - i ] ); + silk_assert( energy >= 0 ); + + /* add part that comes into window */ + energy = silk_ADD_SAT32( energy, silk_SMULBB( basis_ptr[ -i ], basis_ptr[ -i ] ) ); + silk_assert( energy >= 0 ); + silk_assert( lag_counter < SCRATCH_SIZE ); + scratch_mem[ lag_counter ] = energy; + lag_counter++; + } + + delta = matrix_ptr( Lag_range_ptr, k, 0, 2 ); + for( i = 0; i < nb_cbk_search; i++ ) { + /* Fill out the 3 dim array that stores the correlations for */ + /* each code_book vector for each start lag */ + idx = matrix_ptr( Lag_CB_ptr, k, i, cbk_size ) - delta; + for( j = 0; j < PE_NB_STAGE3_LAGS; j++ ) { + silk_assert( idx + j < SCRATCH_SIZE ); + silk_assert( idx + j < lag_counter ); + matrix_ptr( energies_st3, k, i, nb_cbk_search )[ j ] = + scratch_mem[ idx + j ]; + silk_assert( + matrix_ptr( energies_st3, k, i, nb_cbk_search )[ j ] >= 0 ); + } + } + target_ptr += sf_length; + } + RESTORE_STACK; +} diff --git a/lib/opus/silk/fixed/process_gains_FIX.c b/lib/opus/silk/fixed/process_gains_FIX.c new file mode 100644 index 00000000..05aba317 --- /dev/null +++ b/lib/opus/silk/fixed/process_gains_FIX.c @@ -0,0 +1,117 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "tuning_parameters.h" + +/* Processing of gains */ +void silk_process_gains_FIX( + silk_encoder_state_FIX *psEnc, /* I/O Encoder state */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + silk_shape_state_FIX *psShapeSt = &psEnc->sShape; + opus_int k; + opus_int32 s_Q16, InvMaxSqrVal_Q16, gain, gain_squared, ResNrg, ResNrgPart, quant_offset_Q10; + + /* Gain reduction when LTP coding gain is high */ + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /*s = -0.5f * silk_sigmoid( 0.25f * ( psEncCtrl->LTPredCodGain - 12.0f ) ); */ + s_Q16 = -silk_sigm_Q15( silk_RSHIFT_ROUND( psEncCtrl->LTPredCodGain_Q7 - SILK_FIX_CONST( 12.0, 7 ), 4 ) ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->Gains_Q16[ k ] = silk_SMLAWB( psEncCtrl->Gains_Q16[ k ], psEncCtrl->Gains_Q16[ k ], s_Q16 ); + } + } + + /* Limit the quantized signal */ + /* InvMaxSqrVal = pow( 2.0f, 0.33f * ( 21.0f - SNR_dB ) ) / subfr_length; */ + InvMaxSqrVal_Q16 = silk_DIV32_16( silk_log2lin( + silk_SMULWB( SILK_FIX_CONST( 21 + 16 / 0.33, 7 ) - psEnc->sCmn.SNR_dB_Q7, SILK_FIX_CONST( 0.33, 16 ) ) ), psEnc->sCmn.subfr_length ); + + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + /* Soft limit on ratio residual energy and squared gains */ + ResNrg = psEncCtrl->ResNrg[ k ]; + ResNrgPart = silk_SMULWW( ResNrg, InvMaxSqrVal_Q16 ); + if( psEncCtrl->ResNrgQ[ k ] > 0 ) { + ResNrgPart = silk_RSHIFT_ROUND( ResNrgPart, psEncCtrl->ResNrgQ[ k ] ); + } else { + if( ResNrgPart >= silk_RSHIFT( silk_int32_MAX, -psEncCtrl->ResNrgQ[ k ] ) ) { + ResNrgPart = silk_int32_MAX; + } else { + ResNrgPart = silk_LSHIFT( ResNrgPart, -psEncCtrl->ResNrgQ[ k ] ); + } + } + gain = psEncCtrl->Gains_Q16[ k ]; + gain_squared = silk_ADD_SAT32( ResNrgPart, silk_SMMUL( gain, gain ) ); + if( gain_squared < silk_int16_MAX ) { + /* recalculate with higher precision */ + gain_squared = silk_SMLAWW( silk_LSHIFT( ResNrgPart, 16 ), gain, gain ); + silk_assert( gain_squared > 0 ); + gain = silk_SQRT_APPROX( gain_squared ); /* Q8 */ + gain = silk_min( gain, silk_int32_MAX >> 8 ); + psEncCtrl->Gains_Q16[ k ] = silk_LSHIFT_SAT32( gain, 8 ); /* Q16 */ + } else { + gain = silk_SQRT_APPROX( gain_squared ); /* Q0 */ + gain = silk_min( gain, silk_int32_MAX >> 16 ); + psEncCtrl->Gains_Q16[ k ] = silk_LSHIFT_SAT32( gain, 16 ); /* Q16 */ + } + } + + /* Save unquantized gains and gain Index */ + silk_memcpy( psEncCtrl->GainsUnq_Q16, psEncCtrl->Gains_Q16, psEnc->sCmn.nb_subfr * sizeof( opus_int32 ) ); + psEncCtrl->lastGainIndexPrev = psShapeSt->LastGainIndex; + + /* Quantize gains */ + silk_gains_quant( psEnc->sCmn.indices.GainsIndices, psEncCtrl->Gains_Q16, + &psShapeSt->LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); + + /* Set quantizer offset for voiced signals. Larger offset when LTP coding gain is low or tilt is high (ie low-pass) */ + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + if( psEncCtrl->LTPredCodGain_Q7 + silk_RSHIFT( psEnc->sCmn.input_tilt_Q15, 8 ) > SILK_FIX_CONST( 1.0, 7 ) ) { + psEnc->sCmn.indices.quantOffsetType = 0; + } else { + psEnc->sCmn.indices.quantOffsetType = 1; + } + } + + /* Quantizer boundary adjustment */ + quant_offset_Q10 = silk_Quantization_Offsets_Q10[ psEnc->sCmn.indices.signalType >> 1 ][ psEnc->sCmn.indices.quantOffsetType ]; + psEncCtrl->Lambda_Q10 = SILK_FIX_CONST( LAMBDA_OFFSET, 10 ) + + silk_SMULBB( SILK_FIX_CONST( LAMBDA_DELAYED_DECISIONS, 10 ), psEnc->sCmn.nStatesDelayedDecision ) + + silk_SMULWB( SILK_FIX_CONST( LAMBDA_SPEECH_ACT, 18 ), psEnc->sCmn.speech_activity_Q8 ) + + silk_SMULWB( SILK_FIX_CONST( LAMBDA_INPUT_QUALITY, 12 ), psEncCtrl->input_quality_Q14 ) + + silk_SMULWB( SILK_FIX_CONST( LAMBDA_CODING_QUALITY, 12 ), psEncCtrl->coding_quality_Q14 ) + + silk_SMULWB( SILK_FIX_CONST( LAMBDA_QUANT_OFFSET, 16 ), quant_offset_Q10 ); + + silk_assert( psEncCtrl->Lambda_Q10 > 0 ); + silk_assert( psEncCtrl->Lambda_Q10 < SILK_FIX_CONST( 2, 10 ) ); +} diff --git a/lib/opus/silk/fixed/regularize_correlations_FIX.c b/lib/opus/silk/fixed/regularize_correlations_FIX.c new file mode 100644 index 00000000..a2836b05 --- /dev/null +++ b/lib/opus/silk/fixed/regularize_correlations_FIX.c @@ -0,0 +1,47 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" + +/* Add noise to matrix diagonal */ +void silk_regularize_correlations_FIX( + opus_int32 *XX, /* I/O Correlation matrices */ + opus_int32 *xx, /* I/O Correlation values */ + opus_int32 noise, /* I Noise to add */ + opus_int D /* I Dimension of XX */ +) +{ + opus_int i; + for( i = 0; i < D; i++ ) { + matrix_ptr( &XX[ 0 ], i, i, D ) = silk_ADD32( matrix_ptr( &XX[ 0 ], i, i, D ), noise ); + } + xx[ 0 ] += noise; +} diff --git a/lib/opus/silk/fixed/residual_energy16_FIX.c b/lib/opus/silk/fixed/residual_energy16_FIX.c new file mode 100644 index 00000000..7f130f3d --- /dev/null +++ b/lib/opus/silk/fixed/residual_energy16_FIX.c @@ -0,0 +1,103 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" + +/* Residual energy: nrg = wxx - 2 * wXx * c + c' * wXX * c */ +opus_int32 silk_residual_energy16_covar_FIX( + const opus_int16 *c, /* I Prediction vector */ + const opus_int32 *wXX, /* I Correlation matrix */ + const opus_int32 *wXx, /* I Correlation vector */ + opus_int32 wxx, /* I Signal energy */ + opus_int D, /* I Dimension */ + opus_int cQ /* I Q value for c vector 0 - 15 */ +) +{ + opus_int i, j, lshifts, Qxtra; + opus_int32 c_max, w_max, tmp, tmp2, nrg; + opus_int cn[ MAX_MATRIX_SIZE ]; + const opus_int32 *pRow; + + /* Safety checks */ + celt_assert( D >= 0 ); + celt_assert( D <= 16 ); + celt_assert( cQ > 0 ); + celt_assert( cQ < 16 ); + + lshifts = 16 - cQ; + Qxtra = lshifts; + + c_max = 0; + for( i = 0; i < D; i++ ) { + c_max = silk_max_32( c_max, silk_abs( (opus_int32)c[ i ] ) ); + } + Qxtra = silk_min_int( Qxtra, silk_CLZ32( c_max ) - 17 ); + + w_max = silk_max_32( wXX[ 0 ], wXX[ D * D - 1 ] ); + Qxtra = silk_min_int( Qxtra, silk_CLZ32( silk_MUL( D, silk_RSHIFT( silk_SMULWB( w_max, c_max ), 4 ) ) ) - 5 ); + Qxtra = silk_max_int( Qxtra, 0 ); + for( i = 0; i < D; i++ ) { + cn[ i ] = silk_LSHIFT( ( opus_int )c[ i ], Qxtra ); + silk_assert( silk_abs(cn[i]) <= ( silk_int16_MAX + 1 ) ); /* Check that silk_SMLAWB can be used */ + } + lshifts -= Qxtra; + + /* Compute wxx - 2 * wXx * c */ + tmp = 0; + for( i = 0; i < D; i++ ) { + tmp = silk_SMLAWB( tmp, wXx[ i ], cn[ i ] ); + } + nrg = silk_RSHIFT( wxx, 1 + lshifts ) - tmp; /* Q: -lshifts - 1 */ + + /* Add c' * wXX * c, assuming wXX is symmetric */ + tmp2 = 0; + for( i = 0; i < D; i++ ) { + tmp = 0; + pRow = &wXX[ i * D ]; + for( j = i + 1; j < D; j++ ) { + tmp = silk_SMLAWB( tmp, pRow[ j ], cn[ j ] ); + } + tmp = silk_SMLAWB( tmp, silk_RSHIFT( pRow[ i ], 1 ), cn[ i ] ); + tmp2 = silk_SMLAWB( tmp2, tmp, cn[ i ] ); + } + nrg = silk_ADD_LSHIFT32( nrg, tmp2, lshifts ); /* Q: -lshifts - 1 */ + + /* Keep one bit free always, because we add them for LSF interpolation */ + if( nrg < 1 ) { + nrg = 1; + } else if( nrg > silk_RSHIFT( silk_int32_MAX, lshifts + 2 ) ) { + nrg = silk_int32_MAX >> 1; + } else { + nrg = silk_LSHIFT( nrg, lshifts + 1 ); /* Q0 */ + } + return nrg; + +} diff --git a/lib/opus/silk/fixed/residual_energy_FIX.c b/lib/opus/silk/fixed/residual_energy_FIX.c new file mode 100644 index 00000000..6c7cade9 --- /dev/null +++ b/lib/opus/silk/fixed/residual_energy_FIX.c @@ -0,0 +1,98 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" +#include "stack_alloc.h" + +/* Calculates residual energies of input subframes where all subframes have LPC_order */ +/* of preceding samples */ +void silk_residual_energy_FIX( + opus_int32 nrgs[ MAX_NB_SUBFR ], /* O Residual energy per subframe */ + opus_int nrgsQ[ MAX_NB_SUBFR ], /* O Q value per subframe */ + const opus_int16 x[], /* I Input signal */ + opus_int16 a_Q12[ 2 ][ MAX_LPC_ORDER ], /* I AR coefs for each frame half */ + const opus_int32 gains[ MAX_NB_SUBFR ], /* I Quantization gains */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr, /* I Number of subframes */ + const opus_int LPC_order, /* I LPC order */ + int arch /* I Run-time architecture */ +) +{ + opus_int offset, i, j, rshift, lz1, lz2; + opus_int16 *LPC_res_ptr; + VARDECL( opus_int16, LPC_res ); + const opus_int16 *x_ptr; + opus_int32 tmp32; + SAVE_STACK; + + x_ptr = x; + offset = LPC_order + subfr_length; + + /* Filter input to create the LPC residual for each frame half, and measure subframe energies */ + ALLOC( LPC_res, ( MAX_NB_SUBFR >> 1 ) * offset, opus_int16 ); + celt_assert( ( nb_subfr >> 1 ) * ( MAX_NB_SUBFR >> 1 ) == nb_subfr ); + for( i = 0; i < nb_subfr >> 1; i++ ) { + /* Calculate half frame LPC residual signal including preceding samples */ + silk_LPC_analysis_filter( LPC_res, x_ptr, a_Q12[ i ], ( MAX_NB_SUBFR >> 1 ) * offset, LPC_order, arch ); + + /* Point to first subframe of the just calculated LPC residual signal */ + LPC_res_ptr = LPC_res + LPC_order; + for( j = 0; j < ( MAX_NB_SUBFR >> 1 ); j++ ) { + /* Measure subframe energy */ + silk_sum_sqr_shift( &nrgs[ i * ( MAX_NB_SUBFR >> 1 ) + j ], &rshift, LPC_res_ptr, subfr_length ); + + /* Set Q values for the measured energy */ + nrgsQ[ i * ( MAX_NB_SUBFR >> 1 ) + j ] = -rshift; + + /* Move to next subframe */ + LPC_res_ptr += offset; + } + /* Move to next frame half */ + x_ptr += ( MAX_NB_SUBFR >> 1 ) * offset; + } + + /* Apply the squared subframe gains */ + for( i = 0; i < nb_subfr; i++ ) { + /* Fully upscale gains and energies */ + lz1 = silk_CLZ32( nrgs[ i ] ) - 1; + lz2 = silk_CLZ32( gains[ i ] ) - 1; + + tmp32 = silk_LSHIFT32( gains[ i ], lz2 ); + + /* Find squared gains */ + tmp32 = silk_SMMUL( tmp32, tmp32 ); /* Q( 2 * lz2 - 32 )*/ + + /* Scale energies */ + nrgs[ i ] = silk_SMMUL( tmp32, silk_LSHIFT32( nrgs[ i ], lz1 ) ); /* Q( nrgsQ[ i ] + lz1 + 2 * lz2 - 32 - 32 )*/ + nrgsQ[ i ] += lz1 + 2 * lz2 - 32 - 32; + } + RESTORE_STACK; +} diff --git a/lib/opus/silk/fixed/schur64_FIX.c b/lib/opus/silk/fixed/schur64_FIX.c new file mode 100644 index 00000000..4b7e19ea --- /dev/null +++ b/lib/opus/silk/fixed/schur64_FIX.c @@ -0,0 +1,93 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Slower than schur(), but more accurate. */ +/* Uses SMULL(), available on armv4 */ +opus_int32 silk_schur64( /* O returns residual energy */ + opus_int32 rc_Q16[], /* O Reflection coefficients [order] Q16 */ + const opus_int32 c[], /* I Correlations [order+1] */ + opus_int32 order /* I Prediction order */ +) +{ + opus_int k, n; + opus_int32 C[ SILK_MAX_ORDER_LPC + 1 ][ 2 ]; + opus_int32 Ctmp1_Q30, Ctmp2_Q30, rc_tmp_Q31; + + celt_assert( order >= 0 && order <= SILK_MAX_ORDER_LPC ); + + /* Check for invalid input */ + if( c[ 0 ] <= 0 ) { + silk_memset( rc_Q16, 0, order * sizeof( opus_int32 ) ); + return 0; + } + + k = 0; + do { + C[ k ][ 0 ] = C[ k ][ 1 ] = c[ k ]; + } while( ++k <= order ); + + for( k = 0; k < order; k++ ) { + /* Check that we won't be getting an unstable rc, otherwise stop here. */ + if (silk_abs_int32(C[ k + 1 ][ 0 ]) >= C[ 0 ][ 1 ]) { + if ( C[ k + 1 ][ 0 ] > 0 ) { + rc_Q16[ k ] = -SILK_FIX_CONST( .99f, 16 ); + } else { + rc_Q16[ k ] = SILK_FIX_CONST( .99f, 16 ); + } + k++; + break; + } + + /* Get reflection coefficient: divide two Q30 values and get result in Q31 */ + rc_tmp_Q31 = silk_DIV32_varQ( -C[ k + 1 ][ 0 ], C[ 0 ][ 1 ], 31 ); + + /* Save the output */ + rc_Q16[ k ] = silk_RSHIFT_ROUND( rc_tmp_Q31, 15 ); + + /* Update correlations */ + for( n = 0; n < order - k; n++ ) { + Ctmp1_Q30 = C[ n + k + 1 ][ 0 ]; + Ctmp2_Q30 = C[ n ][ 1 ]; + + /* Multiply and add the highest int32 */ + C[ n + k + 1 ][ 0 ] = Ctmp1_Q30 + silk_SMMUL( silk_LSHIFT( Ctmp2_Q30, 1 ), rc_tmp_Q31 ); + C[ n ][ 1 ] = Ctmp2_Q30 + silk_SMMUL( silk_LSHIFT( Ctmp1_Q30, 1 ), rc_tmp_Q31 ); + } + } + + for(; k < order; k++ ) { + rc_Q16[ k ] = 0; + } + + return silk_max_32( 1, C[ 0 ][ 1 ] ); +} diff --git a/lib/opus/silk/fixed/schur_FIX.c b/lib/opus/silk/fixed/schur_FIX.c new file mode 100644 index 00000000..2840f6b1 --- /dev/null +++ b/lib/opus/silk/fixed/schur_FIX.c @@ -0,0 +1,107 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Faster than schur64(), but much less accurate. */ +/* uses SMLAWB(), requiring armv5E and higher. */ +opus_int32 silk_schur( /* O Returns residual energy */ + opus_int16 *rc_Q15, /* O reflection coefficients [order] Q15 */ + const opus_int32 *c, /* I correlations [order+1] */ + const opus_int32 order /* I prediction order */ +) +{ + opus_int k, n, lz; + opus_int32 C[ SILK_MAX_ORDER_LPC + 1 ][ 2 ]; + opus_int32 Ctmp1, Ctmp2, rc_tmp_Q15; + + celt_assert( order >= 0 && order <= SILK_MAX_ORDER_LPC ); + + /* Get number of leading zeros */ + lz = silk_CLZ32( c[ 0 ] ); + + /* Copy correlations and adjust level to Q30 */ + k = 0; + if( lz < 2 ) { + /* lz must be 1, so shift one to the right */ + do { + C[ k ][ 0 ] = C[ k ][ 1 ] = silk_RSHIFT( c[ k ], 1 ); + } while( ++k <= order ); + } else if( lz > 2 ) { + /* Shift to the left */ + lz -= 2; + do { + C[ k ][ 0 ] = C[ k ][ 1 ] = silk_LSHIFT( c[ k ], lz ); + } while( ++k <= order ); + } else { + /* No need to shift */ + do { + C[ k ][ 0 ] = C[ k ][ 1 ] = c[ k ]; + } while( ++k <= order ); + } + + for( k = 0; k < order; k++ ) { + /* Check that we won't be getting an unstable rc, otherwise stop here. */ + if (silk_abs_int32(C[ k + 1 ][ 0 ]) >= C[ 0 ][ 1 ]) { + if ( C[ k + 1 ][ 0 ] > 0 ) { + rc_Q15[ k ] = -SILK_FIX_CONST( .99f, 15 ); + } else { + rc_Q15[ k ] = SILK_FIX_CONST( .99f, 15 ); + } + k++; + break; + } + + /* Get reflection coefficient */ + rc_tmp_Q15 = -silk_DIV32_16( C[ k + 1 ][ 0 ], silk_max_32( silk_RSHIFT( C[ 0 ][ 1 ], 15 ), 1 ) ); + + /* Clip (shouldn't happen for properly conditioned inputs) */ + rc_tmp_Q15 = silk_SAT16( rc_tmp_Q15 ); + + /* Store */ + rc_Q15[ k ] = (opus_int16)rc_tmp_Q15; + + /* Update correlations */ + for( n = 0; n < order - k; n++ ) { + Ctmp1 = C[ n + k + 1 ][ 0 ]; + Ctmp2 = C[ n ][ 1 ]; + C[ n + k + 1 ][ 0 ] = silk_SMLAWB( Ctmp1, silk_LSHIFT( Ctmp2, 1 ), rc_tmp_Q15 ); + C[ n ][ 1 ] = silk_SMLAWB( Ctmp2, silk_LSHIFT( Ctmp1, 1 ), rc_tmp_Q15 ); + } + } + + for(; k < order; k++ ) { + rc_Q15[ k ] = 0; + } + + /* return residual energy */ + return silk_max_32( 1, C[ 0 ][ 1 ] ); +} diff --git a/lib/opus/silk/fixed/structs_FIX.h b/lib/opus/silk/fixed/structs_FIX.h new file mode 100644 index 00000000..2774a97b --- /dev/null +++ b/lib/opus/silk/fixed/structs_FIX.h @@ -0,0 +1,116 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_STRUCTS_FIX_H +#define SILK_STRUCTS_FIX_H + +#include "typedef.h" +#include "main.h" +#include "structs.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +/********************************/ +/* Noise shaping analysis state */ +/********************************/ +typedef struct { + opus_int8 LastGainIndex; + opus_int32 HarmBoost_smth_Q16; + opus_int32 HarmShapeGain_smth_Q16; + opus_int32 Tilt_smth_Q16; +} silk_shape_state_FIX; + +/********************************/ +/* Encoder state FIX */ +/********************************/ +typedef struct { + silk_encoder_state sCmn; /* Common struct, shared with floating-point code */ + silk_shape_state_FIX sShape; /* Shape state */ + + /* Buffer for find pitch and noise shape analysis */ + silk_DWORD_ALIGN opus_int16 x_buf[ 2 * MAX_FRAME_LENGTH + LA_SHAPE_MAX ];/* Buffer for find pitch and noise shape analysis */ + opus_int LTPCorr_Q15; /* Normalized correlation from pitch lag estimator */ + opus_int32 resNrgSmth; +} silk_encoder_state_FIX; + +/************************/ +/* Encoder control FIX */ +/************************/ +typedef struct { + /* Prediction and coding parameters */ + opus_int32 Gains_Q16[ MAX_NB_SUBFR ]; + silk_DWORD_ALIGN opus_int16 PredCoef_Q12[ 2 ][ MAX_LPC_ORDER ]; + opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ]; + opus_int LTP_scale_Q14; + opus_int pitchL[ MAX_NB_SUBFR ]; + + /* Noise shaping parameters */ + /* Testing */ + silk_DWORD_ALIGN opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ]; + opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ]; /* Packs two int16 coefficients per int32 value */ + opus_int Tilt_Q14[ MAX_NB_SUBFR ]; + opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ]; + opus_int Lambda_Q10; + opus_int input_quality_Q14; + opus_int coding_quality_Q14; + + /* measures */ + opus_int32 predGain_Q16; + opus_int LTPredCodGain_Q7; + opus_int32 ResNrg[ MAX_NB_SUBFR ]; /* Residual energy per subframe */ + opus_int ResNrgQ[ MAX_NB_SUBFR ]; /* Q domain for the residual energy > 0 */ + + /* Parameters for CBR mode */ + opus_int32 GainsUnq_Q16[ MAX_NB_SUBFR ]; + opus_int8 lastGainIndexPrev; +} silk_encoder_control_FIX; + +/************************/ +/* Encoder Super Struct */ +/************************/ +typedef struct { + silk_encoder_state_FIX state_Fxx[ ENCODER_NUM_CHANNELS ]; + stereo_enc_state sStereo; + opus_int32 nBitsUsedLBRR; + opus_int32 nBitsExceeded; + opus_int nChannelsAPI; + opus_int nChannelsInternal; + opus_int nPrevChannelsInternal; + opus_int timeSinceSwitchAllowed_ms; + opus_int allowBandwidthSwitch; + opus_int prev_decode_only_middle; +} silk_encoder; + + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/fixed/vector_ops_FIX.c b/lib/opus/silk/fixed/vector_ops_FIX.c new file mode 100644 index 00000000..dcf84070 --- /dev/null +++ b/lib/opus/silk/fixed/vector_ops_FIX.c @@ -0,0 +1,102 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "pitch.h" + +/* Copy and multiply a vector by a constant */ +void silk_scale_copy_vector16( + opus_int16 *data_out, + const opus_int16 *data_in, + opus_int32 gain_Q16, /* I Gain in Q16 */ + const opus_int dataSize /* I Length */ +) +{ + opus_int i; + opus_int32 tmp32; + + for( i = 0; i < dataSize; i++ ) { + tmp32 = silk_SMULWB( gain_Q16, data_in[ i ] ); + data_out[ i ] = (opus_int16)silk_CHECK_FIT16( tmp32 ); + } +} + +/* Multiply a vector by a constant */ +void silk_scale_vector32_Q26_lshift_18( + opus_int32 *data1, /* I/O Q0/Q18 */ + opus_int32 gain_Q26, /* I Q26 */ + opus_int dataSize /* I length */ +) +{ + opus_int i; + + for( i = 0; i < dataSize; i++ ) { + data1[ i ] = (opus_int32)silk_CHECK_FIT32( silk_RSHIFT64( silk_SMULL( data1[ i ], gain_Q26 ), 8 ) ); /* OUTPUT: Q18 */ + } +} + +/* sum = for(i=0;i6, memory access can be reduced by half. */ +opus_int32 silk_inner_prod_aligned( + const opus_int16 *const inVec1, /* I input vector 1 */ + const opus_int16 *const inVec2, /* I input vector 2 */ + const opus_int len, /* I vector lengths */ + int arch /* I Run-time architecture */ +) +{ +#ifdef FIXED_POINT + return celt_inner_prod(inVec1, inVec2, len, arch); +#else + opus_int i; + opus_int32 sum = 0; + for( i = 0; i < len; i++ ) { + sum = silk_SMLABB( sum, inVec1[ i ], inVec2[ i ] ); + } + return sum; +#endif +} + +opus_int64 silk_inner_prod16_c( + const opus_int16 *inVec1, /* I input vector 1 */ + const opus_int16 *inVec2, /* I input vector 2 */ + const opus_int len /* I vector lengths */ +) +{ + opus_int i; + opus_int64 sum = 0; + for( i = 0; i < len; i++ ) { + sum = silk_SMLALBB( sum, inVec1[ i ], inVec2[ i ] ); + } + return sum; +} diff --git a/lib/opus/silk/fixed/warped_autocorrelation_FIX.c b/lib/opus/silk/fixed/warped_autocorrelation_FIX.c new file mode 100644 index 00000000..5c79553b --- /dev/null +++ b/lib/opus/silk/fixed/warped_autocorrelation_FIX.c @@ -0,0 +1,92 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FIX.h" + +#if defined(MIPSr1_ASM) +#include "mips/warped_autocorrelation_FIX_mipsr1.h" +#endif + + +/* Autocorrelations for a warped frequency axis */ +#ifndef OVERRIDE_silk_warped_autocorrelation_FIX_c +void silk_warped_autocorrelation_FIX_c( + opus_int32 *corr, /* O Result [order + 1] */ + opus_int *scale, /* O Scaling of the correlation vector */ + const opus_int16 *input, /* I Input data to correlate */ + const opus_int warping_Q16, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +) +{ + opus_int n, i, lsh; + opus_int32 tmp1_QS, tmp2_QS; + opus_int32 state_QS[ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; + opus_int64 corr_QC[ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; + + /* Order must be even */ + celt_assert( ( order & 1 ) == 0 ); + silk_assert( 2 * QS - QC >= 0 ); + + /* Loop over samples */ + for( n = 0; n < length; n++ ) { + tmp1_QS = silk_LSHIFT32( (opus_int32)input[ n ], QS ); + /* Loop over allpass sections */ + for( i = 0; i < order; i += 2 ) { + /* Output of allpass section */ + tmp2_QS = silk_SMLAWB( state_QS[ i ], state_QS[ i + 1 ] - tmp1_QS, warping_Q16 ); + state_QS[ i ] = tmp1_QS; + corr_QC[ i ] += silk_RSHIFT64( silk_SMULL( tmp1_QS, state_QS[ 0 ] ), 2 * QS - QC ); + /* Output of allpass section */ + tmp1_QS = silk_SMLAWB( state_QS[ i + 1 ], state_QS[ i + 2 ] - tmp2_QS, warping_Q16 ); + state_QS[ i + 1 ] = tmp2_QS; + corr_QC[ i + 1 ] += silk_RSHIFT64( silk_SMULL( tmp2_QS, state_QS[ 0 ] ), 2 * QS - QC ); + } + state_QS[ order ] = tmp1_QS; + corr_QC[ order ] += silk_RSHIFT64( silk_SMULL( tmp1_QS, state_QS[ 0 ] ), 2 * QS - QC ); + } + + lsh = silk_CLZ64( corr_QC[ 0 ] ) - 35; + lsh = silk_LIMIT( lsh, -12 - QC, 30 - QC ); + *scale = -( QC + lsh ); + silk_assert( *scale >= -30 && *scale <= 12 ); + if( lsh >= 0 ) { + for( i = 0; i < order + 1; i++ ) { + corr[ i ] = (opus_int32)silk_CHECK_FIT32( silk_LSHIFT64( corr_QC[ i ], lsh ) ); + } + } else { + for( i = 0; i < order + 1; i++ ) { + corr[ i ] = (opus_int32)silk_CHECK_FIT32( silk_RSHIFT64( corr_QC[ i ], -lsh ) ); + } + } + silk_assert( corr_QC[ 0 ] >= 0 ); /* If breaking, decrease QC*/ +} +#endif /* OVERRIDE_silk_warped_autocorrelation_FIX_c */ diff --git a/lib/opus/silk/fixed/x86/burg_modified_FIX_sse4_1.c b/lib/opus/silk/fixed/x86/burg_modified_FIX_sse4_1.c new file mode 100644 index 00000000..e58bf079 --- /dev/null +++ b/lib/opus/silk/fixed/x86/burg_modified_FIX_sse4_1.c @@ -0,0 +1,400 @@ +/* Copyright (c) 2014-2020, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang FrancisQuiers + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include + +#include "SigProc_FIX.h" +#include "define.h" +#include "tuning_parameters.h" +#include "pitch.h" +#include "celt/x86/x86cpu.h" + +#define MAX_FRAME_SIZE 384 /* subfr_length * nb_subfr = ( 0.005 * 16000 + 16 ) * 4 = 384 */ + +#define QA 25 +#define N_BITS_HEAD_ROOM 3 +#define MIN_RSHIFTS -16 +#define MAX_RSHIFTS (32 - QA) + +/* Compute reflection coefficients from input signal */ +void silk_burg_modified_sse4_1( + opus_int32 *res_nrg, /* O Residual energy */ + opus_int *res_nrg_Q, /* O Residual energy Q value */ + opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ + const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ + const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ + const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I Number of subframes stacked in x */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */ +) +{ + opus_int k, n, s, lz, rshifts, reached_max_gain; + opus_int32 C0, num, nrg, rc_Q31, invGain_Q30, Atmp_QA, Atmp1, tmp1, tmp2, x1, x2; + const opus_int16 *x_ptr; + opus_int32 C_first_row[ SILK_MAX_ORDER_LPC ]; + opus_int32 C_last_row[ SILK_MAX_ORDER_LPC ]; + opus_int32 Af_QA[ SILK_MAX_ORDER_LPC ]; + opus_int32 CAf[ SILK_MAX_ORDER_LPC + 1 ]; + opus_int32 CAb[ SILK_MAX_ORDER_LPC + 1 ]; + opus_int32 xcorr[ SILK_MAX_ORDER_LPC ]; + opus_int64 C0_64; + + __m128i FIRST_3210, LAST_3210, ATMP_3210, TMP1_3210, TMP2_3210, T1_3210, T2_3210, PTR_3210, SUBFR_3210, X1_3210, X2_3210; + __m128i CONST1 = _mm_set1_epi32(1); + + celt_assert( subfr_length * nb_subfr <= MAX_FRAME_SIZE ); + + /* Compute autocorrelations, added over subframes */ + C0_64 = silk_inner_prod16( x, x, subfr_length*nb_subfr, arch ); + lz = silk_CLZ64(C0_64); + rshifts = 32 + 1 + N_BITS_HEAD_ROOM - lz; + if (rshifts > MAX_RSHIFTS) rshifts = MAX_RSHIFTS; + if (rshifts < MIN_RSHIFTS) rshifts = MIN_RSHIFTS; + + if (rshifts > 0) { + C0 = (opus_int32)silk_RSHIFT64(C0_64, rshifts ); + } else { + C0 = silk_LSHIFT32((opus_int32)C0_64, -rshifts ); + } + + CAb[ 0 ] = CAf[ 0 ] = C0 + silk_SMMUL( SILK_FIX_CONST( FIND_LPC_COND_FAC, 32 ), C0 ) + 1; /* Q(-rshifts) */ + silk_memset( C_first_row, 0, SILK_MAX_ORDER_LPC * sizeof( opus_int32 ) ); + if( rshifts > 0 ) { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + for( n = 1; n < D + 1; n++ ) { + C_first_row[ n - 1 ] += (opus_int32)silk_RSHIFT64( + silk_inner_prod16( x_ptr, x_ptr + n, subfr_length - n, arch ), rshifts ); + } + } + } else { + for( s = 0; s < nb_subfr; s++ ) { + int i; + opus_int32 d; + x_ptr = x + s * subfr_length; + celt_pitch_xcorr(x_ptr, x_ptr + 1, xcorr, subfr_length - D, D, arch ); + for( n = 1; n < D + 1; n++ ) { + for ( i = n + subfr_length - D, d = 0; i < subfr_length; i++ ) + d = MAC16_16( d, x_ptr[ i ], x_ptr[ i - n ] ); + xcorr[ n - 1 ] += d; + } + for( n = 1; n < D + 1; n++ ) { + C_first_row[ n - 1 ] += silk_LSHIFT32( xcorr[ n - 1 ], -rshifts ); + } + } + } + silk_memcpy( C_last_row, C_first_row, SILK_MAX_ORDER_LPC * sizeof( opus_int32 ) ); + + /* Initialize */ + CAb[ 0 ] = CAf[ 0 ] = C0 + silk_SMMUL( SILK_FIX_CONST( FIND_LPC_COND_FAC, 32 ), C0 ) + 1; /* Q(-rshifts) */ + + invGain_Q30 = (opus_int32)1 << 30; + reached_max_gain = 0; + for( n = 0; n < D; n++ ) { + /* Update first row of correlation matrix (without first element) */ + /* Update last row of correlation matrix (without last element, stored in reversed order) */ + /* Update C * Af */ + /* Update C * flipud(Af) (stored in reversed order) */ + if( rshifts > -2 ) { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + x1 = -silk_LSHIFT32( (opus_int32)x_ptr[ n ], 16 - rshifts ); /* Q(16-rshifts) */ + x2 = -silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], 16 - rshifts ); /* Q(16-rshifts) */ + tmp1 = silk_LSHIFT32( (opus_int32)x_ptr[ n ], QA - 16 ); /* Q(QA-16) */ + tmp2 = silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], QA - 16 ); /* Q(QA-16) */ + for( k = 0; k < n; k++ ) { + C_first_row[ k ] = silk_SMLAWB( C_first_row[ k ], x1, x_ptr[ n - k - 1 ] ); /* Q( -rshifts ) */ + C_last_row[ k ] = silk_SMLAWB( C_last_row[ k ], x2, x_ptr[ subfr_length - n + k ] ); /* Q( -rshifts ) */ + Atmp_QA = Af_QA[ k ]; + tmp1 = silk_SMLAWB( tmp1, Atmp_QA, x_ptr[ n - k - 1 ] ); /* Q(QA-16) */ + tmp2 = silk_SMLAWB( tmp2, Atmp_QA, x_ptr[ subfr_length - n + k ] ); /* Q(QA-16) */ + } + tmp1 = silk_LSHIFT32( -tmp1, 32 - QA - rshifts ); /* Q(16-rshifts) */ + tmp2 = silk_LSHIFT32( -tmp2, 32 - QA - rshifts ); /* Q(16-rshifts) */ + for( k = 0; k <= n; k++ ) { + CAf[ k ] = silk_SMLAWB( CAf[ k ], tmp1, x_ptr[ n - k ] ); /* Q( -rshift ) */ + CAb[ k ] = silk_SMLAWB( CAb[ k ], tmp2, x_ptr[ subfr_length - n + k - 1 ] ); /* Q( -rshift ) */ + } + } + } else { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + x1 = -silk_LSHIFT32( (opus_int32)x_ptr[ n ], -rshifts ); /* Q( -rshifts ) */ + x2 = -silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], -rshifts ); /* Q( -rshifts ) */ + tmp1 = silk_LSHIFT32( (opus_int32)x_ptr[ n ], 17 ); /* Q17 */ + tmp2 = silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n - 1 ], 17 ); /* Q17 */ + + X1_3210 = _mm_set1_epi32( x1 ); + X2_3210 = _mm_set1_epi32( x2 ); + TMP1_3210 = _mm_setzero_si128(); + TMP2_3210 = _mm_setzero_si128(); + for( k = 0; k < n - 3; k += 4 ) { + PTR_3210 = OP_CVTEPI16_EPI32_M64( &x_ptr[ n - k - 1 - 3 ] ); + SUBFR_3210 = OP_CVTEPI16_EPI32_M64( &x_ptr[ subfr_length - n + k ] ); + FIRST_3210 = _mm_loadu_si128( (__m128i *)&C_first_row[ k ] ); + PTR_3210 = _mm_shuffle_epi32( PTR_3210, _MM_SHUFFLE( 0, 1, 2, 3 ) ); + LAST_3210 = _mm_loadu_si128( (__m128i *)&C_last_row[ k ] ); + ATMP_3210 = _mm_loadu_si128( (__m128i *)&Af_QA[ k ] ); + + T1_3210 = _mm_mullo_epi32( PTR_3210, X1_3210 ); + T2_3210 = _mm_mullo_epi32( SUBFR_3210, X2_3210 ); + + ATMP_3210 = _mm_srai_epi32( ATMP_3210, 7 ); + ATMP_3210 = _mm_add_epi32( ATMP_3210, CONST1 ); + ATMP_3210 = _mm_srai_epi32( ATMP_3210, 1 ); + + FIRST_3210 = _mm_add_epi32( FIRST_3210, T1_3210 ); + LAST_3210 = _mm_add_epi32( LAST_3210, T2_3210 ); + + PTR_3210 = _mm_mullo_epi32( ATMP_3210, PTR_3210 ); + SUBFR_3210 = _mm_mullo_epi32( ATMP_3210, SUBFR_3210 ); + + _mm_storeu_si128( (__m128i *)&C_first_row[ k ], FIRST_3210 ); + _mm_storeu_si128( (__m128i *)&C_last_row[ k ], LAST_3210 ); + + TMP1_3210 = _mm_add_epi32( TMP1_3210, PTR_3210 ); + TMP2_3210 = _mm_add_epi32( TMP2_3210, SUBFR_3210 ); + } + + TMP1_3210 = _mm_add_epi32( TMP1_3210, _mm_unpackhi_epi64(TMP1_3210, TMP1_3210 ) ); + TMP2_3210 = _mm_add_epi32( TMP2_3210, _mm_unpackhi_epi64(TMP2_3210, TMP2_3210 ) ); + TMP1_3210 = _mm_add_epi32( TMP1_3210, _mm_shufflelo_epi16(TMP1_3210, 0x0E ) ); + TMP2_3210 = _mm_add_epi32( TMP2_3210, _mm_shufflelo_epi16(TMP2_3210, 0x0E ) ); + + tmp1 += _mm_cvtsi128_si32( TMP1_3210 ); + tmp2 += _mm_cvtsi128_si32( TMP2_3210 ); + + for( ; k < n; k++ ) { + C_first_row[ k ] = silk_MLA( C_first_row[ k ], x1, x_ptr[ n - k - 1 ] ); /* Q( -rshifts ) */ + C_last_row[ k ] = silk_MLA( C_last_row[ k ], x2, x_ptr[ subfr_length - n + k ] ); /* Q( -rshifts ) */ + Atmp1 = silk_RSHIFT_ROUND( Af_QA[ k ], QA - 17 ); /* Q17 */ + /* We sometimes get overflows in the multiplications (even beyond +/- 2^32), + but they cancel each other and the real result seems to always fit in a 32-bit + signed integer. This was determined experimentally, not theoretically (unfortunately). */ + tmp1 = silk_MLA_ovflw( tmp1, x_ptr[ n - k - 1 ], Atmp1 ); /* Q17 */ + tmp2 = silk_MLA_ovflw( tmp2, x_ptr[ subfr_length - n + k ], Atmp1 ); /* Q17 */ + } + + tmp1 = -tmp1; /* Q17 */ + tmp2 = -tmp2; /* Q17 */ + + { + __m128i xmm_tmp1, xmm_tmp2; + __m128i xmm_x_ptr_n_k_x2x0, xmm_x_ptr_n_k_x3x1; + __m128i xmm_x_ptr_sub_x2x0, xmm_x_ptr_sub_x3x1; + + xmm_tmp1 = _mm_set1_epi32( tmp1 ); + xmm_tmp2 = _mm_set1_epi32( tmp2 ); + + for( k = 0; k <= n - 3; k += 4 ) { + xmm_x_ptr_n_k_x2x0 = OP_CVTEPI16_EPI32_M64( &x_ptr[ n - k - 3 ] ); + xmm_x_ptr_sub_x2x0 = OP_CVTEPI16_EPI32_M64( &x_ptr[ subfr_length - n + k - 1 ] ); + + xmm_x_ptr_n_k_x2x0 = _mm_shuffle_epi32( xmm_x_ptr_n_k_x2x0, _MM_SHUFFLE( 0, 1, 2, 3 ) ); + + xmm_x_ptr_n_k_x2x0 = _mm_slli_epi32( xmm_x_ptr_n_k_x2x0, -rshifts - 1 ); + xmm_x_ptr_sub_x2x0 = _mm_slli_epi32( xmm_x_ptr_sub_x2x0, -rshifts - 1 ); + + /* equal shift right 4 bytes, xmm_x_ptr_n_k_x3x1 = _mm_srli_si128(xmm_x_ptr_n_k_x2x0, 4)*/ + xmm_x_ptr_n_k_x3x1 = _mm_shuffle_epi32( xmm_x_ptr_n_k_x2x0, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + xmm_x_ptr_sub_x3x1 = _mm_shuffle_epi32( xmm_x_ptr_sub_x2x0, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + + xmm_x_ptr_n_k_x2x0 = _mm_mul_epi32( xmm_x_ptr_n_k_x2x0, xmm_tmp1 ); + xmm_x_ptr_n_k_x3x1 = _mm_mul_epi32( xmm_x_ptr_n_k_x3x1, xmm_tmp1 ); + xmm_x_ptr_sub_x2x0 = _mm_mul_epi32( xmm_x_ptr_sub_x2x0, xmm_tmp2 ); + xmm_x_ptr_sub_x3x1 = _mm_mul_epi32( xmm_x_ptr_sub_x3x1, xmm_tmp2 ); + + xmm_x_ptr_n_k_x2x0 = _mm_srli_epi64( xmm_x_ptr_n_k_x2x0, 16 ); + xmm_x_ptr_n_k_x3x1 = _mm_slli_epi64( xmm_x_ptr_n_k_x3x1, 16 ); + xmm_x_ptr_sub_x2x0 = _mm_srli_epi64( xmm_x_ptr_sub_x2x0, 16 ); + xmm_x_ptr_sub_x3x1 = _mm_slli_epi64( xmm_x_ptr_sub_x3x1, 16 ); + + xmm_x_ptr_n_k_x2x0 = _mm_blend_epi16( xmm_x_ptr_n_k_x2x0, xmm_x_ptr_n_k_x3x1, 0xCC ); + xmm_x_ptr_sub_x2x0 = _mm_blend_epi16( xmm_x_ptr_sub_x2x0, xmm_x_ptr_sub_x3x1, 0xCC ); + + X1_3210 = _mm_loadu_si128( (__m128i *)&CAf[ k ] ); + PTR_3210 = _mm_loadu_si128( (__m128i *)&CAb[ k ] ); + + X1_3210 = _mm_add_epi32( X1_3210, xmm_x_ptr_n_k_x2x0 ); + PTR_3210 = _mm_add_epi32( PTR_3210, xmm_x_ptr_sub_x2x0 ); + + _mm_storeu_si128( (__m128i *)&CAf[ k ], X1_3210 ); + _mm_storeu_si128( (__m128i *)&CAb[ k ], PTR_3210 ); + } + + for( ; k <= n; k++ ) { + CAf[ k ] = silk_SMLAWW( CAf[ k ], tmp1, + silk_LSHIFT32( (opus_int32)x_ptr[ n - k ], -rshifts - 1 ) ); /* Q( -rshift ) */ + CAb[ k ] = silk_SMLAWW( CAb[ k ], tmp2, + silk_LSHIFT32( (opus_int32)x_ptr[ subfr_length - n + k - 1 ], -rshifts - 1 ) ); /* Q( -rshift ) */ + } + } + } + } + + /* Calculate nominator and denominator for the next order reflection (parcor) coefficient */ + tmp1 = C_first_row[ n ]; /* Q( -rshifts ) */ + tmp2 = C_last_row[ n ]; /* Q( -rshifts ) */ + num = 0; /* Q( -rshifts ) */ + nrg = silk_ADD32( CAb[ 0 ], CAf[ 0 ] ); /* Q( 1-rshifts ) */ + for( k = 0; k < n; k++ ) { + Atmp_QA = Af_QA[ k ]; + lz = silk_CLZ32( silk_abs( Atmp_QA ) ) - 1; + lz = silk_min( 32 - QA, lz ); + Atmp1 = silk_LSHIFT32( Atmp_QA, lz ); /* Q( QA + lz ) */ + + tmp1 = silk_ADD_LSHIFT32( tmp1, silk_SMMUL( C_last_row[ n - k - 1 ], Atmp1 ), 32 - QA - lz ); /* Q( -rshifts ) */ + tmp2 = silk_ADD_LSHIFT32( tmp2, silk_SMMUL( C_first_row[ n - k - 1 ], Atmp1 ), 32 - QA - lz ); /* Q( -rshifts ) */ + num = silk_ADD_LSHIFT32( num, silk_SMMUL( CAb[ n - k ], Atmp1 ), 32 - QA - lz ); /* Q( -rshifts ) */ + nrg = silk_ADD_LSHIFT32( nrg, silk_SMMUL( silk_ADD32( CAb[ k + 1 ], CAf[ k + 1 ] ), + Atmp1 ), 32 - QA - lz ); /* Q( 1-rshifts ) */ + } + CAf[ n + 1 ] = tmp1; /* Q( -rshifts ) */ + CAb[ n + 1 ] = tmp2; /* Q( -rshifts ) */ + num = silk_ADD32( num, tmp2 ); /* Q( -rshifts ) */ + num = silk_LSHIFT32( -num, 1 ); /* Q( 1-rshifts ) */ + + /* Calculate the next order reflection (parcor) coefficient */ + if( silk_abs( num ) < nrg ) { + rc_Q31 = silk_DIV32_varQ( num, nrg, 31 ); + } else { + rc_Q31 = ( num > 0 ) ? silk_int32_MAX : silk_int32_MIN; + } + + /* Update inverse prediction gain */ + tmp1 = ( (opus_int32)1 << 30 ) - silk_SMMUL( rc_Q31, rc_Q31 ); + tmp1 = silk_LSHIFT( silk_SMMUL( invGain_Q30, tmp1 ), 2 ); + if( tmp1 <= minInvGain_Q30 ) { + /* Max prediction gain exceeded; set reflection coefficient such that max prediction gain is exactly hit */ + tmp2 = ( (opus_int32)1 << 30 ) - silk_DIV32_varQ( minInvGain_Q30, invGain_Q30, 30 ); /* Q30 */ + rc_Q31 = silk_SQRT_APPROX( tmp2 ); /* Q15 */ + if( rc_Q31 > 0 ) { + /* Newton-Raphson iteration */ + rc_Q31 = silk_RSHIFT32( rc_Q31 + silk_DIV32( tmp2, rc_Q31 ), 1 ); /* Q15 */ + rc_Q31 = silk_LSHIFT32( rc_Q31, 16 ); /* Q31 */ + if( num < 0 ) { + /* Ensure adjusted reflection coefficients has the original sign */ + rc_Q31 = -rc_Q31; + } + } + invGain_Q30 = minInvGain_Q30; + reached_max_gain = 1; + } else { + invGain_Q30 = tmp1; + } + + /* Update the AR coefficients */ + for( k = 0; k < (n + 1) >> 1; k++ ) { + tmp1 = Af_QA[ k ]; /* QA */ + tmp2 = Af_QA[ n - k - 1 ]; /* QA */ + Af_QA[ k ] = silk_ADD_LSHIFT32( tmp1, silk_SMMUL( tmp2, rc_Q31 ), 1 ); /* QA */ + Af_QA[ n - k - 1 ] = silk_ADD_LSHIFT32( tmp2, silk_SMMUL( tmp1, rc_Q31 ), 1 ); /* QA */ + } + Af_QA[ n ] = silk_RSHIFT32( rc_Q31, 31 - QA ); /* QA */ + + if( reached_max_gain ) { + /* Reached max prediction gain; set remaining coefficients to zero and exit loop */ + for( k = n + 1; k < D; k++ ) { + Af_QA[ k ] = 0; + } + break; + } + + /* Update C * Af and C * Ab */ + for( k = 0; k <= n + 1; k++ ) { + tmp1 = CAf[ k ]; /* Q( -rshifts ) */ + tmp2 = CAb[ n - k + 1 ]; /* Q( -rshifts ) */ + CAf[ k ] = silk_ADD_LSHIFT32( tmp1, silk_SMMUL( tmp2, rc_Q31 ), 1 ); /* Q( -rshifts ) */ + CAb[ n - k + 1 ] = silk_ADD_LSHIFT32( tmp2, silk_SMMUL( tmp1, rc_Q31 ), 1 ); /* Q( -rshifts ) */ + } + } + + if( reached_max_gain ) { + for( k = 0; k < D; k++ ) { + /* Scale coefficients */ + A_Q16[ k ] = -silk_RSHIFT_ROUND( Af_QA[ k ], QA - 16 ); + } + /* Subtract energy of preceding samples from C0 */ + if( rshifts > 0 ) { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + C0 -= (opus_int32)silk_RSHIFT64( silk_inner_prod16( x_ptr, x_ptr, D, arch ), rshifts ); + } + } else { + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + C0 -= silk_LSHIFT32( silk_inner_prod_aligned( x_ptr, x_ptr, D, arch ), -rshifts ); + } + } + /* Approximate residual energy */ + *res_nrg = silk_LSHIFT( silk_SMMUL( invGain_Q30, C0 ), 2 ); + *res_nrg_Q = -rshifts; + } else { + /* Return residual energy */ + nrg = CAf[ 0 ]; /* Q( -rshifts ) */ + tmp1 = (opus_int32)1 << 16; /* Q16 */ + for( k = 0; k < D; k++ ) { + Atmp1 = silk_RSHIFT_ROUND( Af_QA[ k ], QA - 16 ); /* Q16 */ + nrg = silk_SMLAWW( nrg, CAf[ k + 1 ], Atmp1 ); /* Q( -rshifts ) */ + tmp1 = silk_SMLAWW( tmp1, Atmp1, Atmp1 ); /* Q16 */ + A_Q16[ k ] = -Atmp1; + } + *res_nrg = silk_SMLAWW( nrg, silk_SMMUL( SILK_FIX_CONST( FIND_LPC_COND_FAC, 32 ), C0 ), -tmp1 );/* Q( -rshifts ) */ + *res_nrg_Q = -rshifts; + } + +#ifdef OPUS_CHECK_ASM + { + opus_int32 res_nrg_c = 0; + opus_int res_nrg_Q_c = 0; + opus_int32 A_Q16_c[ MAX_LPC_ORDER ] = {0}; + + silk_burg_modified_c( + &res_nrg_c, + &res_nrg_Q_c, + A_Q16_c, + x, + minInvGain_Q30, + subfr_length, + nb_subfr, + D, + 0 + ); + + silk_assert( *res_nrg == res_nrg_c ); + silk_assert( *res_nrg_Q == res_nrg_Q_c ); + silk_assert( !memcmp( A_Q16, A_Q16_c, D * sizeof( *A_Q16 ) ) ); + } +#endif +} diff --git a/lib/opus/silk/fixed/x86/vector_ops_FIX_sse4_1.c b/lib/opus/silk/fixed/x86/vector_ops_FIX_sse4_1.c new file mode 100644 index 00000000..a46289bb --- /dev/null +++ b/lib/opus/silk/fixed/x86/vector_ops_FIX_sse4_1.c @@ -0,0 +1,93 @@ +/* Copyright (c) 2014, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include "main.h" + +#include "SigProc_FIX.h" +#include "pitch.h" +#include "celt/x86/x86cpu.h" + +opus_int64 silk_inner_prod16_sse4_1( + const opus_int16 *inVec1, /* I input vector 1 */ + const opus_int16 *inVec2, /* I input vector 2 */ + const opus_int len /* I vector lengths */ +) +{ + opus_int i, dataSize4; + opus_int64 sum; + + __m128i xmm_prod_20, xmm_prod_31; + __m128i inVec1_3210, acc1; + __m128i inVec2_3210, acc2; + + sum = 0; + dataSize4 = len & ~3; + + acc1 = _mm_setzero_si128(); + acc2 = _mm_setzero_si128(); + + for( i = 0; i < dataSize4; i += 4 ) { + inVec1_3210 = OP_CVTEPI16_EPI32_M64( &inVec1[i + 0] ); + inVec2_3210 = OP_CVTEPI16_EPI32_M64( &inVec2[i + 0] ); + xmm_prod_20 = _mm_mul_epi32( inVec1_3210, inVec2_3210 ); + + inVec1_3210 = _mm_shuffle_epi32( inVec1_3210, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + inVec2_3210 = _mm_shuffle_epi32( inVec2_3210, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + xmm_prod_31 = _mm_mul_epi32( inVec1_3210, inVec2_3210 ); + + acc1 = _mm_add_epi64( acc1, xmm_prod_20 ); + acc2 = _mm_add_epi64( acc2, xmm_prod_31 ); + } + + acc1 = _mm_add_epi64( acc1, acc2 ); + + /* equal shift right 8 bytes */ + acc2 = _mm_shuffle_epi32( acc1, _MM_SHUFFLE( 0, 0, 3, 2 ) ); + acc1 = _mm_add_epi64( acc1, acc2 ); + + _mm_storel_epi64( (__m128i *)&sum, acc1 ); + + for( ; i < len; i++ ) { + sum = silk_SMLALBB( sum, inVec1[ i ], inVec2[ i ] ); + } + +#ifdef OPUS_CHECK_ASM + { + opus_int64 sum_c = silk_inner_prod16_c( inVec1, inVec2, len ); + silk_assert( sum == sum_c ); + } +#endif + + return sum; +} diff --git a/lib/opus/silk/float/LPC_analysis_filter_FLP.c b/lib/opus/silk/float/LPC_analysis_filter_FLP.c new file mode 100644 index 00000000..0e1a1fed --- /dev/null +++ b/lib/opus/silk/float/LPC_analysis_filter_FLP.c @@ -0,0 +1,249 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "main_FLP.h" + +/************************************************/ +/* LPC analysis filter */ +/* NB! State is kept internally and the */ +/* filter always starts with zero state */ +/* first Order output samples are set to zero */ +/************************************************/ + +/* 16th order LPC analysis filter, does not write first 16 samples */ +static OPUS_INLINE void silk_LPC_analysis_filter16_FLP( + silk_float r_LPC[], /* O LPC residual signal */ + const silk_float PredCoef[], /* I LPC coefficients */ + const silk_float s[], /* I Input signal */ + const opus_int length /* I Length of input signal */ +) +{ + opus_int ix; + silk_float LPC_pred; + const silk_float *s_ptr; + + for( ix = 16; ix < length; ix++ ) { + s_ptr = &s[ix - 1]; + + /* short-term prediction */ + LPC_pred = s_ptr[ 0 ] * PredCoef[ 0 ] + + s_ptr[ -1 ] * PredCoef[ 1 ] + + s_ptr[ -2 ] * PredCoef[ 2 ] + + s_ptr[ -3 ] * PredCoef[ 3 ] + + s_ptr[ -4 ] * PredCoef[ 4 ] + + s_ptr[ -5 ] * PredCoef[ 5 ] + + s_ptr[ -6 ] * PredCoef[ 6 ] + + s_ptr[ -7 ] * PredCoef[ 7 ] + + s_ptr[ -8 ] * PredCoef[ 8 ] + + s_ptr[ -9 ] * PredCoef[ 9 ] + + s_ptr[ -10 ] * PredCoef[ 10 ] + + s_ptr[ -11 ] * PredCoef[ 11 ] + + s_ptr[ -12 ] * PredCoef[ 12 ] + + s_ptr[ -13 ] * PredCoef[ 13 ] + + s_ptr[ -14 ] * PredCoef[ 14 ] + + s_ptr[ -15 ] * PredCoef[ 15 ]; + + /* prediction error */ + r_LPC[ix] = s_ptr[ 1 ] - LPC_pred; + } +} + +/* 12th order LPC analysis filter, does not write first 12 samples */ +static OPUS_INLINE void silk_LPC_analysis_filter12_FLP( + silk_float r_LPC[], /* O LPC residual signal */ + const silk_float PredCoef[], /* I LPC coefficients */ + const silk_float s[], /* I Input signal */ + const opus_int length /* I Length of input signal */ +) +{ + opus_int ix; + silk_float LPC_pred; + const silk_float *s_ptr; + + for( ix = 12; ix < length; ix++ ) { + s_ptr = &s[ix - 1]; + + /* short-term prediction */ + LPC_pred = s_ptr[ 0 ] * PredCoef[ 0 ] + + s_ptr[ -1 ] * PredCoef[ 1 ] + + s_ptr[ -2 ] * PredCoef[ 2 ] + + s_ptr[ -3 ] * PredCoef[ 3 ] + + s_ptr[ -4 ] * PredCoef[ 4 ] + + s_ptr[ -5 ] * PredCoef[ 5 ] + + s_ptr[ -6 ] * PredCoef[ 6 ] + + s_ptr[ -7 ] * PredCoef[ 7 ] + + s_ptr[ -8 ] * PredCoef[ 8 ] + + s_ptr[ -9 ] * PredCoef[ 9 ] + + s_ptr[ -10 ] * PredCoef[ 10 ] + + s_ptr[ -11 ] * PredCoef[ 11 ]; + + /* prediction error */ + r_LPC[ix] = s_ptr[ 1 ] - LPC_pred; + } +} + +/* 10th order LPC analysis filter, does not write first 10 samples */ +static OPUS_INLINE void silk_LPC_analysis_filter10_FLP( + silk_float r_LPC[], /* O LPC residual signal */ + const silk_float PredCoef[], /* I LPC coefficients */ + const silk_float s[], /* I Input signal */ + const opus_int length /* I Length of input signal */ +) +{ + opus_int ix; + silk_float LPC_pred; + const silk_float *s_ptr; + + for( ix = 10; ix < length; ix++ ) { + s_ptr = &s[ix - 1]; + + /* short-term prediction */ + LPC_pred = s_ptr[ 0 ] * PredCoef[ 0 ] + + s_ptr[ -1 ] * PredCoef[ 1 ] + + s_ptr[ -2 ] * PredCoef[ 2 ] + + s_ptr[ -3 ] * PredCoef[ 3 ] + + s_ptr[ -4 ] * PredCoef[ 4 ] + + s_ptr[ -5 ] * PredCoef[ 5 ] + + s_ptr[ -6 ] * PredCoef[ 6 ] + + s_ptr[ -7 ] * PredCoef[ 7 ] + + s_ptr[ -8 ] * PredCoef[ 8 ] + + s_ptr[ -9 ] * PredCoef[ 9 ]; + + /* prediction error */ + r_LPC[ix] = s_ptr[ 1 ] - LPC_pred; + } +} + +/* 8th order LPC analysis filter, does not write first 8 samples */ +static OPUS_INLINE void silk_LPC_analysis_filter8_FLP( + silk_float r_LPC[], /* O LPC residual signal */ + const silk_float PredCoef[], /* I LPC coefficients */ + const silk_float s[], /* I Input signal */ + const opus_int length /* I Length of input signal */ +) +{ + opus_int ix; + silk_float LPC_pred; + const silk_float *s_ptr; + + for( ix = 8; ix < length; ix++ ) { + s_ptr = &s[ix - 1]; + + /* short-term prediction */ + LPC_pred = s_ptr[ 0 ] * PredCoef[ 0 ] + + s_ptr[ -1 ] * PredCoef[ 1 ] + + s_ptr[ -2 ] * PredCoef[ 2 ] + + s_ptr[ -3 ] * PredCoef[ 3 ] + + s_ptr[ -4 ] * PredCoef[ 4 ] + + s_ptr[ -5 ] * PredCoef[ 5 ] + + s_ptr[ -6 ] * PredCoef[ 6 ] + + s_ptr[ -7 ] * PredCoef[ 7 ]; + + /* prediction error */ + r_LPC[ix] = s_ptr[ 1 ] - LPC_pred; + } +} + +/* 6th order LPC analysis filter, does not write first 6 samples */ +static OPUS_INLINE void silk_LPC_analysis_filter6_FLP( + silk_float r_LPC[], /* O LPC residual signal */ + const silk_float PredCoef[], /* I LPC coefficients */ + const silk_float s[], /* I Input signal */ + const opus_int length /* I Length of input signal */ +) +{ + opus_int ix; + silk_float LPC_pred; + const silk_float *s_ptr; + + for( ix = 6; ix < length; ix++ ) { + s_ptr = &s[ix - 1]; + + /* short-term prediction */ + LPC_pred = s_ptr[ 0 ] * PredCoef[ 0 ] + + s_ptr[ -1 ] * PredCoef[ 1 ] + + s_ptr[ -2 ] * PredCoef[ 2 ] + + s_ptr[ -3 ] * PredCoef[ 3 ] + + s_ptr[ -4 ] * PredCoef[ 4 ] + + s_ptr[ -5 ] * PredCoef[ 5 ]; + + /* prediction error */ + r_LPC[ix] = s_ptr[ 1 ] - LPC_pred; + } +} + +/************************************************/ +/* LPC analysis filter */ +/* NB! State is kept internally and the */ +/* filter always starts with zero state */ +/* first Order output samples are set to zero */ +/************************************************/ +void silk_LPC_analysis_filter_FLP( + silk_float r_LPC[], /* O LPC residual signal */ + const silk_float PredCoef[], /* I LPC coefficients */ + const silk_float s[], /* I Input signal */ + const opus_int length, /* I Length of input signal */ + const opus_int Order /* I LPC order */ +) +{ + celt_assert( Order <= length ); + + switch( Order ) { + case 6: + silk_LPC_analysis_filter6_FLP( r_LPC, PredCoef, s, length ); + break; + + case 8: + silk_LPC_analysis_filter8_FLP( r_LPC, PredCoef, s, length ); + break; + + case 10: + silk_LPC_analysis_filter10_FLP( r_LPC, PredCoef, s, length ); + break; + + case 12: + silk_LPC_analysis_filter12_FLP( r_LPC, PredCoef, s, length ); + break; + + case 16: + silk_LPC_analysis_filter16_FLP( r_LPC, PredCoef, s, length ); + break; + + default: + celt_assert( 0 ); + break; + } + + /* Set first Order output samples to zero */ + silk_memset( r_LPC, 0, Order * sizeof( silk_float ) ); +} + diff --git a/lib/opus/silk/float/LPC_inv_pred_gain_FLP.c b/lib/opus/silk/float/LPC_inv_pred_gain_FLP.c new file mode 100644 index 00000000..2be2122d --- /dev/null +++ b/lib/opus/silk/float/LPC_inv_pred_gain_FLP.c @@ -0,0 +1,73 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "SigProc_FLP.h" +#include "define.h" + +/* compute inverse of LPC prediction gain, and */ +/* test if LPC coefficients are stable (all poles within unit circle) */ +/* this code is based on silk_a2k_FLP() */ +silk_float silk_LPC_inverse_pred_gain_FLP( /* O return inverse prediction gain, energy domain */ + const silk_float *A, /* I prediction coefficients [order] */ + opus_int32 order /* I prediction order */ +) +{ + opus_int k, n; + double invGain, rc, rc_mult1, rc_mult2, tmp1, tmp2; + silk_float Atmp[ SILK_MAX_ORDER_LPC ]; + + silk_memcpy( Atmp, A, order * sizeof(silk_float) ); + + invGain = 1.0; + for( k = order - 1; k > 0; k-- ) { + rc = -Atmp[ k ]; + rc_mult1 = 1.0f - rc * rc; + invGain *= rc_mult1; + if( invGain * MAX_PREDICTION_POWER_GAIN < 1.0f ) { + return 0.0f; + } + rc_mult2 = 1.0f / rc_mult1; + for( n = 0; n < (k + 1) >> 1; n++ ) { + tmp1 = Atmp[ n ]; + tmp2 = Atmp[ k - n - 1 ]; + Atmp[ n ] = (silk_float)( ( tmp1 - tmp2 * rc ) * rc_mult2 ); + Atmp[ k - n - 1 ] = (silk_float)( ( tmp2 - tmp1 * rc ) * rc_mult2 ); + } + } + rc = -Atmp[ 0 ]; + rc_mult1 = 1.0f - rc * rc; + invGain *= rc_mult1; + if( invGain * MAX_PREDICTION_POWER_GAIN < 1.0f ) { + return 0.0f; + } + return (silk_float)invGain; +} diff --git a/lib/opus/silk/float/LTP_analysis_filter_FLP.c b/lib/opus/silk/float/LTP_analysis_filter_FLP.c new file mode 100644 index 00000000..849b7c1c --- /dev/null +++ b/lib/opus/silk/float/LTP_analysis_filter_FLP.c @@ -0,0 +1,75 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +void silk_LTP_analysis_filter_FLP( + silk_float *LTP_res, /* O LTP res MAX_NB_SUBFR*(pre_lgth+subfr_lngth) */ + const silk_float *x, /* I Input signal, with preceding samples */ + const silk_float B[ LTP_ORDER * MAX_NB_SUBFR ], /* I LTP coefficients for each subframe */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const silk_float invGains[ MAX_NB_SUBFR ], /* I Inverse quantization gains */ + const opus_int subfr_length, /* I Length of each subframe */ + const opus_int nb_subfr, /* I number of subframes */ + const opus_int pre_length /* I Preceding samples for each subframe */ +) +{ + const silk_float *x_ptr, *x_lag_ptr; + silk_float Btmp[ LTP_ORDER ]; + silk_float *LTP_res_ptr; + silk_float inv_gain; + opus_int k, i, j; + + x_ptr = x; + LTP_res_ptr = LTP_res; + for( k = 0; k < nb_subfr; k++ ) { + x_lag_ptr = x_ptr - pitchL[ k ]; + inv_gain = invGains[ k ]; + for( i = 0; i < LTP_ORDER; i++ ) { + Btmp[ i ] = B[ k * LTP_ORDER + i ]; + } + + /* LTP analysis FIR filter */ + for( i = 0; i < subfr_length + pre_length; i++ ) { + LTP_res_ptr[ i ] = x_ptr[ i ]; + /* Subtract long-term prediction */ + for( j = 0; j < LTP_ORDER; j++ ) { + LTP_res_ptr[ i ] -= Btmp[ j ] * x_lag_ptr[ LTP_ORDER / 2 - j ]; + } + LTP_res_ptr[ i ] *= inv_gain; + x_lag_ptr++; + } + + /* Update pointers */ + LTP_res_ptr += subfr_length + pre_length; + x_ptr += subfr_length; + } +} diff --git a/lib/opus/silk/float/LTP_scale_ctrl_FLP.c b/lib/opus/silk/float/LTP_scale_ctrl_FLP.c new file mode 100644 index 00000000..6f30ff09 --- /dev/null +++ b/lib/opus/silk/float/LTP_scale_ctrl_FLP.c @@ -0,0 +1,58 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +void silk_LTP_scale_ctrl_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + opus_int round_loss; + + if( condCoding == CODE_INDEPENDENTLY ) { + /* Only scale if first frame in packet */ + round_loss = psEnc->sCmn.PacketLoss_perc * psEnc->sCmn.nFramesPerPacket; + if ( psEnc->sCmn.LBRR_flag ) { + /* LBRR reduces the effective loss. In practice, it does not square the loss because + losses aren't independent, but that still seems to work best. We also never go below 2%. */ + round_loss = 2 + silk_SMULBB( round_loss, round_loss) / 100; + } + psEnc->sCmn.indices.LTP_scaleIndex = silk_SMULBB( psEncCtrl->LTPredCodGain, round_loss ) > silk_log2lin( 2900 - psEnc->sCmn.SNR_dB_Q7 ); + psEnc->sCmn.indices.LTP_scaleIndex += silk_SMULBB( psEncCtrl->LTPredCodGain, round_loss ) > silk_log2lin( 3900 - psEnc->sCmn.SNR_dB_Q7 ); + } else { + /* Default is minimum scaling */ + psEnc->sCmn.indices.LTP_scaleIndex = 0; + } + + psEncCtrl->LTP_scale = (silk_float)silk_LTPScales_table_Q14[ psEnc->sCmn.indices.LTP_scaleIndex ] / 16384.0f; +} diff --git a/lib/opus/silk/float/SigProc_FLP.h b/lib/opus/silk/float/SigProc_FLP.h new file mode 100644 index 00000000..953de8b0 --- /dev/null +++ b/lib/opus/silk/float/SigProc_FLP.h @@ -0,0 +1,197 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_SIGPROC_FLP_H +#define SILK_SIGPROC_FLP_H + +#include "SigProc_FIX.h" +#include "float_cast.h" +#include + +#ifdef __cplusplus +extern "C" +{ +#endif + +/********************************************************************/ +/* SIGNAL PROCESSING FUNCTIONS */ +/********************************************************************/ + +/* Chirp (bw expand) LP AR filter */ +void silk_bwexpander_FLP( + silk_float *ar, /* I/O AR filter to be expanded (without leading 1) */ + const opus_int d, /* I length of ar */ + const silk_float chirp /* I chirp factor (typically in range (0..1) ) */ +); + +/* compute inverse of LPC prediction gain, and */ +/* test if LPC coefficients are stable (all poles within unit circle) */ +/* this code is based on silk_FLP_a2k() */ +silk_float silk_LPC_inverse_pred_gain_FLP( /* O return inverse prediction gain, energy domain */ + const silk_float *A, /* I prediction coefficients [order] */ + opus_int32 order /* I prediction order */ +); + +silk_float silk_schur_FLP( /* O returns residual energy */ + silk_float refl_coef[], /* O reflection coefficients (length order) */ + const silk_float auto_corr[], /* I autocorrelation sequence (length order+1) */ + opus_int order /* I order */ +); + +void silk_k2a_FLP( + silk_float *A, /* O prediction coefficients [order] */ + const silk_float *rc, /* I reflection coefficients [order] */ + opus_int32 order /* I prediction order */ +); + +/* compute autocorrelation */ +void silk_autocorrelation_FLP( + silk_float *results, /* O result (length correlationCount) */ + const silk_float *inputData, /* I input data to correlate */ + opus_int inputDataSize, /* I length of input */ + opus_int correlationCount /* I number of correlation taps to compute */ +); + +opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, 1 unvoiced */ + const silk_float *frame, /* I Signal of length PE_FRAME_LENGTH_MS*Fs_kHz */ + opus_int *pitch_out, /* O Pitch lag values [nb_subfr] */ + opus_int16 *lagIndex, /* O Lag Index */ + opus_int8 *contourIndex, /* O Pitch contour Index */ + silk_float *LTPCorr, /* I/O Normalized correlation; input: value from previous frame */ + opus_int prevLag, /* I Last lag of previous frame; set to zero is unvoiced */ + const silk_float search_thres1, /* I First stage threshold for lag candidates 0 - 1 */ + const silk_float search_thres2, /* I Final threshold for lag candidates 0 - 1 */ + const opus_int Fs_kHz, /* I sample frequency (kHz) */ + const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ + const opus_int nb_subfr, /* I Number of 5 ms subframes */ + int arch /* I Run-time architecture */ +); + +void silk_insertion_sort_decreasing_FLP( + silk_float *a, /* I/O Unsorted / Sorted vector */ + opus_int *idx, /* O Index vector for the sorted elements */ + const opus_int L, /* I Vector length */ + const opus_int K /* I Number of correctly sorted positions */ +); + +/* Compute reflection coefficients from input signal */ +silk_float silk_burg_modified_FLP( /* O returns residual energy */ + silk_float A[], /* O prediction coefficients (length order) */ + const silk_float x[], /* I input signal, length: nb_subfr*(D+L_sub) */ + const silk_float minInvGain, /* I minimum inverse prediction gain */ + const opus_int subfr_length, /* I input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I number of subframes stacked in x */ + const opus_int D /* I order */ +); + +/* multiply a vector by a constant */ +void silk_scale_vector_FLP( + silk_float *data1, + silk_float gain, + opus_int dataSize +); + +/* copy and multiply a vector by a constant */ +void silk_scale_copy_vector_FLP( + silk_float *data_out, + const silk_float *data_in, + silk_float gain, + opus_int dataSize +); + +/* inner product of two silk_float arrays, with result as double */ +double silk_inner_product_FLP( + const silk_float *data1, + const silk_float *data2, + opus_int dataSize +); + +/* sum of squares of a silk_float array, with result as double */ +double silk_energy_FLP( + const silk_float *data, + opus_int dataSize +); + +/********************************************************************/ +/* MACROS */ +/********************************************************************/ + +#define PI (3.1415926536f) + +#define silk_min_float( a, b ) (((a) < (b)) ? (a) : (b)) +#define silk_max_float( a, b ) (((a) > (b)) ? (a) : (b)) +#define silk_abs_float( a ) ((silk_float)fabs(a)) + +/* sigmoid function */ +static OPUS_INLINE silk_float silk_sigmoid( silk_float x ) +{ + return (silk_float)(1.0 / (1.0 + exp(-x))); +} + +/* floating-point to integer conversion (rounding) */ +static OPUS_INLINE opus_int32 silk_float2int( silk_float x ) +{ + return (opus_int32)float2int( x ); +} + +/* floating-point to integer conversion (rounding) */ +static OPUS_INLINE void silk_float2short_array( + opus_int16 *out, + const silk_float *in, + opus_int32 length +) +{ + opus_int32 k; + for( k = length - 1; k >= 0; k-- ) { + out[k] = silk_SAT16( (opus_int32)float2int( in[k] ) ); + } +} + +/* integer to floating-point conversion */ +static OPUS_INLINE void silk_short2float_array( + silk_float *out, + const opus_int16 *in, + opus_int32 length +) +{ + opus_int32 k; + for( k = length - 1; k >= 0; k-- ) { + out[k] = (silk_float)in[k]; + } +} + +/* using log2() helps the fixed-point conversion */ +static OPUS_INLINE silk_float silk_log2( double x ) +{ + return ( silk_float )( 3.32192809488736 * log10( x ) ); +} + +#ifdef __cplusplus +} +#endif + +#endif /* SILK_SIGPROC_FLP_H */ diff --git a/lib/opus/silk/float/apply_sine_window_FLP.c b/lib/opus/silk/float/apply_sine_window_FLP.c new file mode 100644 index 00000000..e49e7179 --- /dev/null +++ b/lib/opus/silk/float/apply_sine_window_FLP.c @@ -0,0 +1,81 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +/* Apply sine window to signal vector */ +/* Window types: */ +/* 1 -> sine window from 0 to pi/2 */ +/* 2 -> sine window from pi/2 to pi */ +void silk_apply_sine_window_FLP( + silk_float px_win[], /* O Pointer to windowed signal */ + const silk_float px[], /* I Pointer to input signal */ + const opus_int win_type, /* I Selects a window type */ + const opus_int length /* I Window length, multiple of 4 */ +) +{ + opus_int k; + silk_float freq, c, S0, S1; + + celt_assert( win_type == 1 || win_type == 2 ); + + /* Length must be multiple of 4 */ + celt_assert( ( length & 3 ) == 0 ); + + freq = PI / ( length + 1 ); + + /* Approximation of 2 * cos(f) */ + c = 2.0f - freq * freq; + + /* Initialize state */ + if( win_type < 2 ) { + /* Start from 0 */ + S0 = 0.0f; + /* Approximation of sin(f) */ + S1 = freq; + } else { + /* Start from 1 */ + S0 = 1.0f; + /* Approximation of cos(f) */ + S1 = 0.5f * c; + } + + /* Uses the recursive equation: sin(n*f) = 2 * cos(f) * sin((n-1)*f) - sin((n-2)*f) */ + /* 4 samples at a time */ + for( k = 0; k < length; k += 4 ) { + px_win[ k + 0 ] = px[ k + 0 ] * 0.5f * ( S0 + S1 ); + px_win[ k + 1 ] = px[ k + 1 ] * S1; + S0 = c * S1 - S0; + px_win[ k + 2 ] = px[ k + 2 ] * 0.5f * ( S1 + S0 ); + px_win[ k + 3 ] = px[ k + 3 ] * S0; + S1 = c * S0 - S1; + } +} diff --git a/lib/opus/silk/float/autocorrelation_FLP.c b/lib/opus/silk/float/autocorrelation_FLP.c new file mode 100644 index 00000000..8b8a9e65 --- /dev/null +++ b/lib/opus/silk/float/autocorrelation_FLP.c @@ -0,0 +1,52 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "typedef.h" +#include "SigProc_FLP.h" + +/* compute autocorrelation */ +void silk_autocorrelation_FLP( + silk_float *results, /* O result (length correlationCount) */ + const silk_float *inputData, /* I input data to correlate */ + opus_int inputDataSize, /* I length of input */ + opus_int correlationCount /* I number of correlation taps to compute */ +) +{ + opus_int i; + + if( correlationCount > inputDataSize ) { + correlationCount = inputDataSize; + } + + for( i = 0; i < correlationCount; i++ ) { + results[ i ] = (silk_float)silk_inner_product_FLP( inputData, inputData + i, inputDataSize - i ); + } +} diff --git a/lib/opus/silk/float/burg_modified_FLP.c b/lib/opus/silk/float/burg_modified_FLP.c new file mode 100644 index 00000000..756b76a3 --- /dev/null +++ b/lib/opus/silk/float/burg_modified_FLP.c @@ -0,0 +1,186 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" +#include "tuning_parameters.h" +#include "define.h" + +#define MAX_FRAME_SIZE 384 /* subfr_length * nb_subfr = ( 0.005 * 16000 + 16 ) * 4 = 384*/ + +/* Compute reflection coefficients from input signal */ +silk_float silk_burg_modified_FLP( /* O returns residual energy */ + silk_float A[], /* O prediction coefficients (length order) */ + const silk_float x[], /* I input signal, length: nb_subfr*(D+L_sub) */ + const silk_float minInvGain, /* I minimum inverse prediction gain */ + const opus_int subfr_length, /* I input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I number of subframes stacked in x */ + const opus_int D /* I order */ +) +{ + opus_int k, n, s, reached_max_gain; + double C0, invGain, num, nrg_f, nrg_b, rc, Atmp, tmp1, tmp2; + const silk_float *x_ptr; + double C_first_row[ SILK_MAX_ORDER_LPC ], C_last_row[ SILK_MAX_ORDER_LPC ]; + double CAf[ SILK_MAX_ORDER_LPC + 1 ], CAb[ SILK_MAX_ORDER_LPC + 1 ]; + double Af[ SILK_MAX_ORDER_LPC ]; + + celt_assert( subfr_length * nb_subfr <= MAX_FRAME_SIZE ); + + /* Compute autocorrelations, added over subframes */ + C0 = silk_energy_FLP( x, nb_subfr * subfr_length ); + silk_memset( C_first_row, 0, SILK_MAX_ORDER_LPC * sizeof( double ) ); + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + for( n = 1; n < D + 1; n++ ) { + C_first_row[ n - 1 ] += silk_inner_product_FLP( x_ptr, x_ptr + n, subfr_length - n ); + } + } + silk_memcpy( C_last_row, C_first_row, SILK_MAX_ORDER_LPC * sizeof( double ) ); + + /* Initialize */ + CAb[ 0 ] = CAf[ 0 ] = C0 + FIND_LPC_COND_FAC * C0 + 1e-9f; + invGain = 1.0f; + reached_max_gain = 0; + for( n = 0; n < D; n++ ) { + /* Update first row of correlation matrix (without first element) */ + /* Update last row of correlation matrix (without last element, stored in reversed order) */ + /* Update C * Af */ + /* Update C * flipud(Af) (stored in reversed order) */ + for( s = 0; s < nb_subfr; s++ ) { + x_ptr = x + s * subfr_length; + tmp1 = x_ptr[ n ]; + tmp2 = x_ptr[ subfr_length - n - 1 ]; + for( k = 0; k < n; k++ ) { + C_first_row[ k ] -= x_ptr[ n ] * x_ptr[ n - k - 1 ]; + C_last_row[ k ] -= x_ptr[ subfr_length - n - 1 ] * x_ptr[ subfr_length - n + k ]; + Atmp = Af[ k ]; + tmp1 += x_ptr[ n - k - 1 ] * Atmp; + tmp2 += x_ptr[ subfr_length - n + k ] * Atmp; + } + for( k = 0; k <= n; k++ ) { + CAf[ k ] -= tmp1 * x_ptr[ n - k ]; + CAb[ k ] -= tmp2 * x_ptr[ subfr_length - n + k - 1 ]; + } + } + tmp1 = C_first_row[ n ]; + tmp2 = C_last_row[ n ]; + for( k = 0; k < n; k++ ) { + Atmp = Af[ k ]; + tmp1 += C_last_row[ n - k - 1 ] * Atmp; + tmp2 += C_first_row[ n - k - 1 ] * Atmp; + } + CAf[ n + 1 ] = tmp1; + CAb[ n + 1 ] = tmp2; + + /* Calculate nominator and denominator for the next order reflection (parcor) coefficient */ + num = CAb[ n + 1 ]; + nrg_b = CAb[ 0 ]; + nrg_f = CAf[ 0 ]; + for( k = 0; k < n; k++ ) { + Atmp = Af[ k ]; + num += CAb[ n - k ] * Atmp; + nrg_b += CAb[ k + 1 ] * Atmp; + nrg_f += CAf[ k + 1 ] * Atmp; + } + silk_assert( nrg_f > 0.0 ); + silk_assert( nrg_b > 0.0 ); + + /* Calculate the next order reflection (parcor) coefficient */ + rc = -2.0 * num / ( nrg_f + nrg_b ); + silk_assert( rc > -1.0 && rc < 1.0 ); + + /* Update inverse prediction gain */ + tmp1 = invGain * ( 1.0 - rc * rc ); + if( tmp1 <= minInvGain ) { + /* Max prediction gain exceeded; set reflection coefficient such that max prediction gain is exactly hit */ + rc = sqrt( 1.0 - minInvGain / invGain ); + if( num > 0 ) { + /* Ensure adjusted reflection coefficients has the original sign */ + rc = -rc; + } + invGain = minInvGain; + reached_max_gain = 1; + } else { + invGain = tmp1; + } + + /* Update the AR coefficients */ + for( k = 0; k < (n + 1) >> 1; k++ ) { + tmp1 = Af[ k ]; + tmp2 = Af[ n - k - 1 ]; + Af[ k ] = tmp1 + rc * tmp2; + Af[ n - k - 1 ] = tmp2 + rc * tmp1; + } + Af[ n ] = rc; + + if( reached_max_gain ) { + /* Reached max prediction gain; set remaining coefficients to zero and exit loop */ + for( k = n + 1; k < D; k++ ) { + Af[ k ] = 0.0; + } + break; + } + + /* Update C * Af and C * Ab */ + for( k = 0; k <= n + 1; k++ ) { + tmp1 = CAf[ k ]; + CAf[ k ] += rc * CAb[ n - k + 1 ]; + CAb[ n - k + 1 ] += rc * tmp1; + } + } + + if( reached_max_gain ) { + /* Convert to silk_float */ + for( k = 0; k < D; k++ ) { + A[ k ] = (silk_float)( -Af[ k ] ); + } + /* Subtract energy of preceding samples from C0 */ + for( s = 0; s < nb_subfr; s++ ) { + C0 -= silk_energy_FLP( x + s * subfr_length, D ); + } + /* Approximate residual energy */ + nrg_f = C0 * invGain; + } else { + /* Compute residual energy and store coefficients as silk_float */ + nrg_f = CAf[ 0 ]; + tmp1 = 1.0; + for( k = 0; k < D; k++ ) { + Atmp = Af[ k ]; + nrg_f += CAf[ k + 1 ] * Atmp; + tmp1 += Atmp * Atmp; + A[ k ] = (silk_float)(-Atmp); + } + nrg_f -= FIND_LPC_COND_FAC * C0 * tmp1; + } + + /* Return residual energy */ + return (silk_float)nrg_f; +} diff --git a/lib/opus/silk/float/bwexpander_FLP.c b/lib/opus/silk/float/bwexpander_FLP.c new file mode 100644 index 00000000..d55a4d79 --- /dev/null +++ b/lib/opus/silk/float/bwexpander_FLP.c @@ -0,0 +1,49 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" + +/* Chirp (bw expand) LP AR filter */ +void silk_bwexpander_FLP( + silk_float *ar, /* I/O AR filter to be expanded (without leading 1) */ + const opus_int d, /* I length of ar */ + const silk_float chirp /* I chirp factor (typically in range (0..1) ) */ +) +{ + opus_int i; + silk_float cfac = chirp; + + for( i = 0; i < d - 1; i++ ) { + ar[ i ] *= cfac; + cfac *= chirp; + } + ar[ d - 1 ] *= cfac; +} diff --git a/lib/opus/silk/float/corrMatrix_FLP.c b/lib/opus/silk/float/corrMatrix_FLP.c new file mode 100644 index 00000000..eae6a1cf --- /dev/null +++ b/lib/opus/silk/float/corrMatrix_FLP.c @@ -0,0 +1,93 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/********************************************************************** + * Correlation matrix computations for LS estimate. + **********************************************************************/ + +#include "main_FLP.h" + +/* Calculates correlation vector X'*t */ +void silk_corrVector_FLP( + const silk_float *x, /* I x vector [L+order-1] used to create X */ + const silk_float *t, /* I Target vector [L] */ + const opus_int L, /* I Length of vecors */ + const opus_int Order, /* I Max lag for correlation */ + silk_float *Xt /* O X'*t correlation vector [order] */ +) +{ + opus_int lag; + const silk_float *ptr1; + + ptr1 = &x[ Order - 1 ]; /* Points to first sample of column 0 of X: X[:,0] */ + for( lag = 0; lag < Order; lag++ ) { + /* Calculate X[:,lag]'*t */ + Xt[ lag ] = (silk_float)silk_inner_product_FLP( ptr1, t, L ); + ptr1--; /* Next column of X */ + } +} + +/* Calculates correlation matrix X'*X */ +void silk_corrMatrix_FLP( + const silk_float *x, /* I x vector [ L+order-1 ] used to create X */ + const opus_int L, /* I Length of vectors */ + const opus_int Order, /* I Max lag for correlation */ + silk_float *XX /* O X'*X correlation matrix [order x order] */ +) +{ + opus_int j, lag; + double energy; + const silk_float *ptr1, *ptr2; + + ptr1 = &x[ Order - 1 ]; /* First sample of column 0 of X */ + energy = silk_energy_FLP( ptr1, L ); /* X[:,0]'*X[:,0] */ + matrix_ptr( XX, 0, 0, Order ) = ( silk_float )energy; + for( j = 1; j < Order; j++ ) { + /* Calculate X[:,j]'*X[:,j] */ + energy += ptr1[ -j ] * ptr1[ -j ] - ptr1[ L - j ] * ptr1[ L - j ]; + matrix_ptr( XX, j, j, Order ) = ( silk_float )energy; + } + + ptr2 = &x[ Order - 2 ]; /* First sample of column 1 of X */ + for( lag = 1; lag < Order; lag++ ) { + /* Calculate X[:,0]'*X[:,lag] */ + energy = silk_inner_product_FLP( ptr1, ptr2, L ); + matrix_ptr( XX, lag, 0, Order ) = ( silk_float )energy; + matrix_ptr( XX, 0, lag, Order ) = ( silk_float )energy; + /* Calculate X[:,j]'*X[:,j + lag] */ + for( j = 1; j < ( Order - lag ); j++ ) { + energy += ptr1[ -j ] * ptr2[ -j ] - ptr1[ L - j ] * ptr2[ L - j ]; + matrix_ptr( XX, lag + j, j, Order ) = ( silk_float )energy; + matrix_ptr( XX, j, lag + j, Order ) = ( silk_float )energy; + } + ptr2--; /* Next column of X */ + } +} diff --git a/lib/opus/silk/float/encode_frame_FLP.c b/lib/opus/silk/float/encode_frame_FLP.c new file mode 100644 index 00000000..b029c3f5 --- /dev/null +++ b/lib/opus/silk/float/encode_frame_FLP.c @@ -0,0 +1,435 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "main_FLP.h" +#include "tuning_parameters.h" + +/* Low Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode with lower bitrate */ +static OPUS_INLINE void silk_LBRR_encode_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + const silk_float xfw[], /* I Input signal */ + opus_int condCoding /* I The type of conditional coding used so far for this frame */ +); + +void silk_encode_do_VAD_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + opus_int activity /* I Decision of Opus voice activity detector */ +) +{ + const opus_int activity_threshold = SILK_FIX_CONST( SPEECH_ACTIVITY_DTX_THRES, 8 ); + + /****************************/ + /* Voice Activity Detection */ + /****************************/ + silk_VAD_GetSA_Q8( &psEnc->sCmn, psEnc->sCmn.inputBuf + 1, psEnc->sCmn.arch ); + /* If Opus VAD is inactive and Silk VAD is active: lower Silk VAD to just under the threshold */ + if( activity == VAD_NO_ACTIVITY && psEnc->sCmn.speech_activity_Q8 >= activity_threshold ) { + psEnc->sCmn.speech_activity_Q8 = activity_threshold - 1; + } + + /**************************************************/ + /* Convert speech activity into VAD and DTX flags */ + /**************************************************/ + if( psEnc->sCmn.speech_activity_Q8 < activity_threshold ) { + psEnc->sCmn.indices.signalType = TYPE_NO_VOICE_ACTIVITY; + psEnc->sCmn.noSpeechCounter++; + if( psEnc->sCmn.noSpeechCounter <= NB_SPEECH_FRAMES_BEFORE_DTX ) { + psEnc->sCmn.inDTX = 0; + } else if( psEnc->sCmn.noSpeechCounter > MAX_CONSECUTIVE_DTX + NB_SPEECH_FRAMES_BEFORE_DTX ) { + psEnc->sCmn.noSpeechCounter = NB_SPEECH_FRAMES_BEFORE_DTX; + psEnc->sCmn.inDTX = 0; + } + psEnc->sCmn.VAD_flags[ psEnc->sCmn.nFramesEncoded ] = 0; + } else { + psEnc->sCmn.noSpeechCounter = 0; + psEnc->sCmn.inDTX = 0; + psEnc->sCmn.indices.signalType = TYPE_UNVOICED; + psEnc->sCmn.VAD_flags[ psEnc->sCmn.nFramesEncoded ] = 1; + } +} + +/****************/ +/* Encode frame */ +/****************/ +opus_int silk_encode_frame_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + opus_int32 *pnBytesOut, /* O Number of payload bytes; */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding, /* I The type of conditional coding to use */ + opus_int maxBits, /* I If > 0: maximum number of output bits */ + opus_int useCBR /* I Flag to force constant-bitrate operation */ +) +{ + silk_encoder_control_FLP sEncCtrl; + opus_int i, iter, maxIter, found_upper, found_lower, ret = 0; + silk_float *x_frame, *res_pitch_frame; + silk_float res_pitch[ 2 * MAX_FRAME_LENGTH + LA_PITCH_MAX ]; + ec_enc sRangeEnc_copy, sRangeEnc_copy2; + silk_nsq_state sNSQ_copy, sNSQ_copy2; + opus_int32 seed_copy, nBits, nBits_lower, nBits_upper, gainMult_lower, gainMult_upper; + opus_int32 gainsID, gainsID_lower, gainsID_upper; + opus_int16 gainMult_Q8; + opus_int16 ec_prevLagIndex_copy; + opus_int ec_prevSignalType_copy; + opus_int8 LastGainIndex_copy2; + opus_int32 pGains_Q16[ MAX_NB_SUBFR ]; + opus_uint8 ec_buf_copy[ 1275 ]; + opus_int gain_lock[ MAX_NB_SUBFR ] = {0}; + opus_int16 best_gain_mult[ MAX_NB_SUBFR ]; + opus_int best_sum[ MAX_NB_SUBFR ]; + + /* This is totally unnecessary but many compilers (including gcc) are too dumb to realise it */ + LastGainIndex_copy2 = nBits_lower = nBits_upper = gainMult_lower = gainMult_upper = 0; + + psEnc->sCmn.indices.Seed = psEnc->sCmn.frameCounter++ & 3; + + /**************************************************************/ + /* Set up Input Pointers, and insert frame in input buffer */ + /**************************************************************/ + /* pointers aligned with start of frame to encode */ + x_frame = psEnc->x_buf + psEnc->sCmn.ltp_mem_length; /* start of frame to encode */ + res_pitch_frame = res_pitch + psEnc->sCmn.ltp_mem_length; /* start of pitch LPC residual frame */ + + /***************************************/ + /* Ensure smooth bandwidth transitions */ + /***************************************/ + silk_LP_variable_cutoff( &psEnc->sCmn.sLP, psEnc->sCmn.inputBuf + 1, psEnc->sCmn.frame_length ); + + /*******************************************/ + /* Copy new frame to front of input buffer */ + /*******************************************/ + silk_short2float_array( x_frame + LA_SHAPE_MS * psEnc->sCmn.fs_kHz, psEnc->sCmn.inputBuf + 1, psEnc->sCmn.frame_length ); + + /* Add tiny signal to avoid high CPU load from denormalized floating point numbers */ + for( i = 0; i < 8; i++ ) { + x_frame[ LA_SHAPE_MS * psEnc->sCmn.fs_kHz + i * ( psEnc->sCmn.frame_length >> 3 ) ] += ( 1 - ( i & 2 ) ) * 1e-6f; + } + + if( !psEnc->sCmn.prefillFlag ) { + /*****************************************/ + /* Find pitch lags, initial LPC analysis */ + /*****************************************/ + silk_find_pitch_lags_FLP( psEnc, &sEncCtrl, res_pitch, x_frame, psEnc->sCmn.arch ); + + /************************/ + /* Noise shape analysis */ + /************************/ + silk_noise_shape_analysis_FLP( psEnc, &sEncCtrl, res_pitch_frame, x_frame ); + + /***************************************************/ + /* Find linear prediction coefficients (LPC + LTP) */ + /***************************************************/ + silk_find_pred_coefs_FLP( psEnc, &sEncCtrl, res_pitch_frame, x_frame, condCoding ); + + /****************************************/ + /* Process gains */ + /****************************************/ + silk_process_gains_FLP( psEnc, &sEncCtrl, condCoding ); + + /****************************************/ + /* Low Bitrate Redundant Encoding */ + /****************************************/ + silk_LBRR_encode_FLP( psEnc, &sEncCtrl, x_frame, condCoding ); + + /* Loop over quantizer and entroy coding to control bitrate */ + maxIter = 6; + gainMult_Q8 = SILK_FIX_CONST( 1, 8 ); + found_lower = 0; + found_upper = 0; + gainsID = silk_gains_ID( psEnc->sCmn.indices.GainsIndices, psEnc->sCmn.nb_subfr ); + gainsID_lower = -1; + gainsID_upper = -1; + /* Copy part of the input state */ + silk_memcpy( &sRangeEnc_copy, psRangeEnc, sizeof( ec_enc ) ); + silk_memcpy( &sNSQ_copy, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) ); + seed_copy = psEnc->sCmn.indices.Seed; + ec_prevLagIndex_copy = psEnc->sCmn.ec_prevLagIndex; + ec_prevSignalType_copy = psEnc->sCmn.ec_prevSignalType; + for( iter = 0; ; iter++ ) { + if( gainsID == gainsID_lower ) { + nBits = nBits_lower; + } else if( gainsID == gainsID_upper ) { + nBits = nBits_upper; + } else { + /* Restore part of the input state */ + if( iter > 0 ) { + silk_memcpy( psRangeEnc, &sRangeEnc_copy, sizeof( ec_enc ) ); + silk_memcpy( &psEnc->sCmn.sNSQ, &sNSQ_copy, sizeof( silk_nsq_state ) ); + psEnc->sCmn.indices.Seed = seed_copy; + psEnc->sCmn.ec_prevLagIndex = ec_prevLagIndex_copy; + psEnc->sCmn.ec_prevSignalType = ec_prevSignalType_copy; + } + + /*****************************************/ + /* Noise shaping quantization */ + /*****************************************/ + silk_NSQ_wrapper_FLP( psEnc, &sEncCtrl, &psEnc->sCmn.indices, &psEnc->sCmn.sNSQ, psEnc->sCmn.pulses, x_frame ); + + if ( iter == maxIter && !found_lower ) { + silk_memcpy( &sRangeEnc_copy2, psRangeEnc, sizeof( ec_enc ) ); + } + + /****************************************/ + /* Encode Parameters */ + /****************************************/ + silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding ); + + /****************************************/ + /* Encode Excitation Signal */ + /****************************************/ + silk_encode_pulses( psRangeEnc, psEnc->sCmn.indices.signalType, psEnc->sCmn.indices.quantOffsetType, + psEnc->sCmn.pulses, psEnc->sCmn.frame_length ); + + nBits = ec_tell( psRangeEnc ); + + /* If we still bust after the last iteration, do some damage control. */ + if ( iter == maxIter && !found_lower && nBits > maxBits ) { + silk_memcpy( psRangeEnc, &sRangeEnc_copy2, sizeof( ec_enc ) ); + + /* Keep gains the same as the last frame. */ + psEnc->sShape.LastGainIndex = sEncCtrl.lastGainIndexPrev; + for ( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + psEnc->sCmn.indices.GainsIndices[ i ] = 4; + } + if (condCoding != CODE_CONDITIONALLY) { + psEnc->sCmn.indices.GainsIndices[ 0 ] = sEncCtrl.lastGainIndexPrev; + } + psEnc->sCmn.ec_prevLagIndex = ec_prevLagIndex_copy; + psEnc->sCmn.ec_prevSignalType = ec_prevSignalType_copy; + /* Clear all pulses. */ + for ( i = 0; i < psEnc->sCmn.frame_length; i++ ) { + psEnc->sCmn.pulses[ i ] = 0; + } + + silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding ); + + silk_encode_pulses( psRangeEnc, psEnc->sCmn.indices.signalType, psEnc->sCmn.indices.quantOffsetType, + psEnc->sCmn.pulses, psEnc->sCmn.frame_length ); + + nBits = ec_tell( psRangeEnc ); + } + + if( useCBR == 0 && iter == 0 && nBits <= maxBits ) { + break; + } + } + + if( iter == maxIter ) { + if( found_lower && ( gainsID == gainsID_lower || nBits > maxBits ) ) { + /* Restore output state from earlier iteration that did meet the bitrate budget */ + silk_memcpy( psRangeEnc, &sRangeEnc_copy2, sizeof( ec_enc ) ); + celt_assert( sRangeEnc_copy2.offs <= 1275 ); + silk_memcpy( psRangeEnc->buf, ec_buf_copy, sRangeEnc_copy2.offs ); + silk_memcpy( &psEnc->sCmn.sNSQ, &sNSQ_copy2, sizeof( silk_nsq_state ) ); + psEnc->sShape.LastGainIndex = LastGainIndex_copy2; + } + break; + } + + if( nBits > maxBits ) { + if( found_lower == 0 && iter >= 2 ) { + /* Adjust the quantizer's rate/distortion tradeoff and discard previous "upper" results */ + sEncCtrl.Lambda = silk_max_float(sEncCtrl.Lambda*1.5f, 1.5f); + /* Reducing dithering can help us hit the target. */ + psEnc->sCmn.indices.quantOffsetType = 0; + found_upper = 0; + gainsID_upper = -1; + } else { + found_upper = 1; + nBits_upper = nBits; + gainMult_upper = gainMult_Q8; + gainsID_upper = gainsID; + } + } else if( nBits < maxBits - 5 ) { + found_lower = 1; + nBits_lower = nBits; + gainMult_lower = gainMult_Q8; + if( gainsID != gainsID_lower ) { + gainsID_lower = gainsID; + /* Copy part of the output state */ + silk_memcpy( &sRangeEnc_copy2, psRangeEnc, sizeof( ec_enc ) ); + celt_assert( psRangeEnc->offs <= 1275 ); + silk_memcpy( ec_buf_copy, psRangeEnc->buf, psRangeEnc->offs ); + silk_memcpy( &sNSQ_copy2, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) ); + LastGainIndex_copy2 = psEnc->sShape.LastGainIndex; + } + } else { + /* Within 5 bits of budget: close enough */ + break; + } + + if ( !found_lower && nBits > maxBits ) { + int j; + for ( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + int sum=0; + for ( j = i*psEnc->sCmn.subfr_length; j < (i+1)*psEnc->sCmn.subfr_length; j++ ) { + sum += abs( psEnc->sCmn.pulses[j] ); + } + if ( iter == 0 || (sum < best_sum[i] && !gain_lock[i]) ) { + best_sum[i] = sum; + best_gain_mult[i] = gainMult_Q8; + } else { + gain_lock[i] = 1; + } + } + } + if( ( found_lower & found_upper ) == 0 ) { + /* Adjust gain according to high-rate rate/distortion curve */ + if( nBits > maxBits ) { + if (gainMult_Q8 < 16384) { + gainMult_Q8 *= 2; + } else { + gainMult_Q8 = 32767; + } + } else { + opus_int32 gain_factor_Q16; + gain_factor_Q16 = silk_log2lin( silk_LSHIFT( nBits - maxBits, 7 ) / psEnc->sCmn.frame_length + SILK_FIX_CONST( 16, 7 ) ); + gainMult_Q8 = silk_SMULWB( gain_factor_Q16, gainMult_Q8 ); + } + } else { + /* Adjust gain by interpolating */ + gainMult_Q8 = gainMult_lower + ( ( gainMult_upper - gainMult_lower ) * ( maxBits - nBits_lower ) ) / ( nBits_upper - nBits_lower ); + /* New gain multplier must be between 25% and 75% of old range (note that gainMult_upper < gainMult_lower) */ + if( gainMult_Q8 > silk_ADD_RSHIFT32( gainMult_lower, gainMult_upper - gainMult_lower, 2 ) ) { + gainMult_Q8 = silk_ADD_RSHIFT32( gainMult_lower, gainMult_upper - gainMult_lower, 2 ); + } else + if( gainMult_Q8 < silk_SUB_RSHIFT32( gainMult_upper, gainMult_upper - gainMult_lower, 2 ) ) { + gainMult_Q8 = silk_SUB_RSHIFT32( gainMult_upper, gainMult_upper - gainMult_lower, 2 ); + } + } + + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + opus_int16 tmp; + if ( gain_lock[i] ) { + tmp = best_gain_mult[i]; + } else { + tmp = gainMult_Q8; + } + pGains_Q16[ i ] = silk_LSHIFT_SAT32( silk_SMULWB( sEncCtrl.GainsUnq_Q16[ i ], tmp ), 8 ); + } + + /* Quantize gains */ + psEnc->sShape.LastGainIndex = sEncCtrl.lastGainIndexPrev; + silk_gains_quant( psEnc->sCmn.indices.GainsIndices, pGains_Q16, + &psEnc->sShape.LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); + + /* Unique identifier of gains vector */ + gainsID = silk_gains_ID( psEnc->sCmn.indices.GainsIndices, psEnc->sCmn.nb_subfr ); + + /* Overwrite unquantized gains with quantized gains and convert back to Q0 from Q16 */ + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + sEncCtrl.Gains[ i ] = pGains_Q16[ i ] / 65536.0f; + } + } + } + + /* Update input buffer */ + silk_memmove( psEnc->x_buf, &psEnc->x_buf[ psEnc->sCmn.frame_length ], + ( psEnc->sCmn.ltp_mem_length + LA_SHAPE_MS * psEnc->sCmn.fs_kHz ) * sizeof( silk_float ) ); + + /* Exit without entropy coding */ + if( psEnc->sCmn.prefillFlag ) { + /* No payload */ + *pnBytesOut = 0; + return ret; + } + + /* Parameters needed for next frame */ + psEnc->sCmn.prevLag = sEncCtrl.pitchL[ psEnc->sCmn.nb_subfr - 1 ]; + psEnc->sCmn.prevSignalType = psEnc->sCmn.indices.signalType; + + /****************************************/ + /* Finalize payload */ + /****************************************/ + psEnc->sCmn.first_frame_after_reset = 0; + /* Payload size */ + *pnBytesOut = silk_RSHIFT( ec_tell( psRangeEnc ) + 7, 3 ); + + return ret; +} + +/* Low-Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode excitation at lower bitrate */ +static OPUS_INLINE void silk_LBRR_encode_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + const silk_float xfw[], /* I Input signal */ + opus_int condCoding /* I The type of conditional coding used so far for this frame */ +) +{ + opus_int k; + opus_int32 Gains_Q16[ MAX_NB_SUBFR ]; + silk_float TempGains[ MAX_NB_SUBFR ]; + SideInfoIndices *psIndices_LBRR = &psEnc->sCmn.indices_LBRR[ psEnc->sCmn.nFramesEncoded ]; + silk_nsq_state sNSQ_LBRR; + + /*******************************************/ + /* Control use of inband LBRR */ + /*******************************************/ + if( psEnc->sCmn.LBRR_enabled && psEnc->sCmn.speech_activity_Q8 > SILK_FIX_CONST( LBRR_SPEECH_ACTIVITY_THRES, 8 ) ) { + psEnc->sCmn.LBRR_flags[ psEnc->sCmn.nFramesEncoded ] = 1; + + /* Copy noise shaping quantizer state and quantization indices from regular encoding */ + silk_memcpy( &sNSQ_LBRR, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) ); + silk_memcpy( psIndices_LBRR, &psEnc->sCmn.indices, sizeof( SideInfoIndices ) ); + + /* Save original gains */ + silk_memcpy( TempGains, psEncCtrl->Gains, psEnc->sCmn.nb_subfr * sizeof( silk_float ) ); + + if( psEnc->sCmn.nFramesEncoded == 0 || psEnc->sCmn.LBRR_flags[ psEnc->sCmn.nFramesEncoded - 1 ] == 0 ) { + /* First frame in packet or previous frame not LBRR coded */ + psEnc->sCmn.LBRRprevLastGainIndex = psEnc->sShape.LastGainIndex; + + /* Increase Gains to get target LBRR rate */ + psIndices_LBRR->GainsIndices[ 0 ] += psEnc->sCmn.LBRR_GainIncreases; + psIndices_LBRR->GainsIndices[ 0 ] = silk_min_int( psIndices_LBRR->GainsIndices[ 0 ], N_LEVELS_QGAIN - 1 ); + } + + /* Decode to get gains in sync with decoder */ + silk_gains_dequant( Gains_Q16, psIndices_LBRR->GainsIndices, + &psEnc->sCmn.LBRRprevLastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); + + /* Overwrite unquantized gains with quantized gains and convert back to Q0 from Q16 */ + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->Gains[ k ] = Gains_Q16[ k ] * ( 1.0f / 65536.0f ); + } + + /*****************************************/ + /* Noise shaping quantization */ + /*****************************************/ + silk_NSQ_wrapper_FLP( psEnc, psEncCtrl, psIndices_LBRR, &sNSQ_LBRR, + psEnc->sCmn.pulses_LBRR[ psEnc->sCmn.nFramesEncoded ], xfw ); + + /* Restore original gains */ + silk_memcpy( psEncCtrl->Gains, TempGains, psEnc->sCmn.nb_subfr * sizeof( silk_float ) ); + } +} diff --git a/lib/opus/silk/float/energy_FLP.c b/lib/opus/silk/float/energy_FLP.c new file mode 100644 index 00000000..7bc7173c --- /dev/null +++ b/lib/opus/silk/float/energy_FLP.c @@ -0,0 +1,59 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" + +/* sum of squares of a silk_float array, with result as double */ +double silk_energy_FLP( + const silk_float *data, + opus_int dataSize +) +{ + opus_int i; + double result; + + /* 4x unrolled loop */ + result = 0.0; + for( i = 0; i < dataSize - 3; i += 4 ) { + result += data[ i + 0 ] * (double)data[ i + 0 ] + + data[ i + 1 ] * (double)data[ i + 1 ] + + data[ i + 2 ] * (double)data[ i + 2 ] + + data[ i + 3 ] * (double)data[ i + 3 ]; + } + + /* add any remaining products */ + for( ; i < dataSize; i++ ) { + result += data[ i ] * (double)data[ i ]; + } + + silk_assert( result >= 0.0 ); + return result; +} diff --git a/lib/opus/silk/float/find_LPC_FLP.c b/lib/opus/silk/float/find_LPC_FLP.c new file mode 100644 index 00000000..fa3ffe7f --- /dev/null +++ b/lib/opus/silk/float/find_LPC_FLP.c @@ -0,0 +1,104 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "define.h" +#include "main_FLP.h" +#include "tuning_parameters.h" + +/* LPC analysis */ +void silk_find_LPC_FLP( + silk_encoder_state *psEncC, /* I/O Encoder state */ + opus_int16 NLSF_Q15[], /* O NLSFs */ + const silk_float x[], /* I Input signal */ + const silk_float minInvGain /* I Inverse of max prediction gain */ +) +{ + opus_int k, subfr_length; + silk_float a[ MAX_LPC_ORDER ]; + + /* Used only for NLSF interpolation */ + silk_float res_nrg, res_nrg_2nd, res_nrg_interp; + opus_int16 NLSF0_Q15[ MAX_LPC_ORDER ]; + silk_float a_tmp[ MAX_LPC_ORDER ]; + silk_float LPC_res[ MAX_FRAME_LENGTH + MAX_NB_SUBFR * MAX_LPC_ORDER ]; + + subfr_length = psEncC->subfr_length + psEncC->predictLPCOrder; + + /* Default: No interpolation */ + psEncC->indices.NLSFInterpCoef_Q2 = 4; + + /* Burg AR analysis for the full frame */ + res_nrg = silk_burg_modified_FLP( a, x, minInvGain, subfr_length, psEncC->nb_subfr, psEncC->predictLPCOrder ); + + if( psEncC->useInterpolatedNLSFs && !psEncC->first_frame_after_reset && psEncC->nb_subfr == MAX_NB_SUBFR ) { + /* Optimal solution for last 10 ms; subtract residual energy here, as that's easier than */ + /* adding it to the residual energy of the first 10 ms in each iteration of the search below */ + res_nrg -= silk_burg_modified_FLP( a_tmp, x + ( MAX_NB_SUBFR / 2 ) * subfr_length, minInvGain, subfr_length, MAX_NB_SUBFR / 2, psEncC->predictLPCOrder ); + + /* Convert to NLSFs */ + silk_A2NLSF_FLP( NLSF_Q15, a_tmp, psEncC->predictLPCOrder ); + + /* Search over interpolation indices to find the one with lowest residual energy */ + res_nrg_2nd = silk_float_MAX; + for( k = 3; k >= 0; k-- ) { + /* Interpolate NLSFs for first half */ + silk_interpolate( NLSF0_Q15, psEncC->prev_NLSFq_Q15, NLSF_Q15, k, psEncC->predictLPCOrder ); + + /* Convert to LPC for residual energy evaluation */ + silk_NLSF2A_FLP( a_tmp, NLSF0_Q15, psEncC->predictLPCOrder, psEncC->arch ); + + /* Calculate residual energy with LSF interpolation */ + silk_LPC_analysis_filter_FLP( LPC_res, a_tmp, x, 2 * subfr_length, psEncC->predictLPCOrder ); + res_nrg_interp = (silk_float)( + silk_energy_FLP( LPC_res + psEncC->predictLPCOrder, subfr_length - psEncC->predictLPCOrder ) + + silk_energy_FLP( LPC_res + psEncC->predictLPCOrder + subfr_length, subfr_length - psEncC->predictLPCOrder ) ); + + /* Determine whether current interpolated NLSFs are best so far */ + if( res_nrg_interp < res_nrg ) { + /* Interpolation has lower residual energy */ + res_nrg = res_nrg_interp; + psEncC->indices.NLSFInterpCoef_Q2 = (opus_int8)k; + } else if( res_nrg_interp > res_nrg_2nd ) { + /* No reason to continue iterating - residual energies will continue to climb */ + break; + } + res_nrg_2nd = res_nrg_interp; + } + } + + if( psEncC->indices.NLSFInterpCoef_Q2 == 4 ) { + /* NLSF interpolation is currently inactive, calculate NLSFs from full frame AR coefficients */ + silk_A2NLSF_FLP( NLSF_Q15, a, psEncC->predictLPCOrder ); + } + + celt_assert( psEncC->indices.NLSFInterpCoef_Q2 == 4 || + ( psEncC->useInterpolatedNLSFs && !psEncC->first_frame_after_reset && psEncC->nb_subfr == MAX_NB_SUBFR ) ); +} diff --git a/lib/opus/silk/float/find_LTP_FLP.c b/lib/opus/silk/float/find_LTP_FLP.c new file mode 100644 index 00000000..f9706493 --- /dev/null +++ b/lib/opus/silk/float/find_LTP_FLP.c @@ -0,0 +1,64 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" +#include "tuning_parameters.h" + +void silk_find_LTP_FLP( + silk_float XX[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* O Weight for LTP quantization */ + silk_float xX[ MAX_NB_SUBFR * LTP_ORDER ], /* O Weight for LTP quantization */ + const silk_float r_ptr[], /* I LPC residual */ + const opus_int lag[ MAX_NB_SUBFR ], /* I LTP lags */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr /* I number of subframes */ +) +{ + opus_int k; + silk_float *xX_ptr, *XX_ptr; + const silk_float *lag_ptr; + silk_float xx, temp; + + xX_ptr = xX; + XX_ptr = XX; + for( k = 0; k < nb_subfr; k++ ) { + lag_ptr = r_ptr - ( lag[ k ] + LTP_ORDER / 2 ); + silk_corrMatrix_FLP( lag_ptr, subfr_length, LTP_ORDER, XX_ptr ); + silk_corrVector_FLP( lag_ptr, r_ptr, subfr_length, LTP_ORDER, xX_ptr ); + xx = ( silk_float )silk_energy_FLP( r_ptr, subfr_length + LTP_ORDER ); + temp = 1.0f / silk_max( xx, LTP_CORR_INV_MAX * 0.5f * ( XX_ptr[ 0 ] + XX_ptr[ 24 ] ) + 1.0f ); + silk_scale_vector_FLP( XX_ptr, temp, LTP_ORDER * LTP_ORDER ); + silk_scale_vector_FLP( xX_ptr, temp, LTP_ORDER ); + + r_ptr += subfr_length; + XX_ptr += LTP_ORDER * LTP_ORDER; + xX_ptr += LTP_ORDER; + } +} diff --git a/lib/opus/silk/float/find_pitch_lags_FLP.c b/lib/opus/silk/float/find_pitch_lags_FLP.c new file mode 100644 index 00000000..dedbcd28 --- /dev/null +++ b/lib/opus/silk/float/find_pitch_lags_FLP.c @@ -0,0 +1,132 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "main_FLP.h" +#include "tuning_parameters.h" + +void silk_find_pitch_lags_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + silk_float res[], /* O Residual */ + const silk_float x[], /* I Speech signal */ + int arch /* I Run-time architecture */ +) +{ + opus_int buf_len; + silk_float thrhld, res_nrg; + const silk_float *x_buf_ptr, *x_buf; + silk_float auto_corr[ MAX_FIND_PITCH_LPC_ORDER + 1 ]; + silk_float A[ MAX_FIND_PITCH_LPC_ORDER ]; + silk_float refl_coef[ MAX_FIND_PITCH_LPC_ORDER ]; + silk_float Wsig[ FIND_PITCH_LPC_WIN_MAX ]; + silk_float *Wsig_ptr; + + /******************************************/ + /* Set up buffer lengths etc based on Fs */ + /******************************************/ + buf_len = psEnc->sCmn.la_pitch + psEnc->sCmn.frame_length + psEnc->sCmn.ltp_mem_length; + + /* Safety check */ + celt_assert( buf_len >= psEnc->sCmn.pitch_LPC_win_length ); + + x_buf = x - psEnc->sCmn.ltp_mem_length; + + /******************************************/ + /* Estimate LPC AR coeficients */ + /******************************************/ + + /* Calculate windowed signal */ + + /* First LA_LTP samples */ + x_buf_ptr = x_buf + buf_len - psEnc->sCmn.pitch_LPC_win_length; + Wsig_ptr = Wsig; + silk_apply_sine_window_FLP( Wsig_ptr, x_buf_ptr, 1, psEnc->sCmn.la_pitch ); + + /* Middle non-windowed samples */ + Wsig_ptr += psEnc->sCmn.la_pitch; + x_buf_ptr += psEnc->sCmn.la_pitch; + silk_memcpy( Wsig_ptr, x_buf_ptr, ( psEnc->sCmn.pitch_LPC_win_length - ( psEnc->sCmn.la_pitch << 1 ) ) * sizeof( silk_float ) ); + + /* Last LA_LTP samples */ + Wsig_ptr += psEnc->sCmn.pitch_LPC_win_length - ( psEnc->sCmn.la_pitch << 1 ); + x_buf_ptr += psEnc->sCmn.pitch_LPC_win_length - ( psEnc->sCmn.la_pitch << 1 ); + silk_apply_sine_window_FLP( Wsig_ptr, x_buf_ptr, 2, psEnc->sCmn.la_pitch ); + + /* Calculate autocorrelation sequence */ + silk_autocorrelation_FLP( auto_corr, Wsig, psEnc->sCmn.pitch_LPC_win_length, psEnc->sCmn.pitchEstimationLPCOrder + 1 ); + + /* Add white noise, as a fraction of the energy */ + auto_corr[ 0 ] += auto_corr[ 0 ] * FIND_PITCH_WHITE_NOISE_FRACTION + 1; + + /* Calculate the reflection coefficients using Schur */ + res_nrg = silk_schur_FLP( refl_coef, auto_corr, psEnc->sCmn.pitchEstimationLPCOrder ); + + /* Prediction gain */ + psEncCtrl->predGain = auto_corr[ 0 ] / silk_max_float( res_nrg, 1.0f ); + + /* Convert reflection coefficients to prediction coefficients */ + silk_k2a_FLP( A, refl_coef, psEnc->sCmn.pitchEstimationLPCOrder ); + + /* Bandwidth expansion */ + silk_bwexpander_FLP( A, psEnc->sCmn.pitchEstimationLPCOrder, FIND_PITCH_BANDWIDTH_EXPANSION ); + + /*****************************************/ + /* LPC analysis filtering */ + /*****************************************/ + silk_LPC_analysis_filter_FLP( res, A, x_buf, buf_len, psEnc->sCmn.pitchEstimationLPCOrder ); + + if( psEnc->sCmn.indices.signalType != TYPE_NO_VOICE_ACTIVITY && psEnc->sCmn.first_frame_after_reset == 0 ) { + /* Threshold for pitch estimator */ + thrhld = 0.6f; + thrhld -= 0.004f * psEnc->sCmn.pitchEstimationLPCOrder; + thrhld -= 0.1f * psEnc->sCmn.speech_activity_Q8 * ( 1.0f / 256.0f ); + thrhld -= 0.15f * (psEnc->sCmn.prevSignalType >> 1); + thrhld -= 0.1f * psEnc->sCmn.input_tilt_Q15 * ( 1.0f / 32768.0f ); + + /*****************************************/ + /* Call Pitch estimator */ + /*****************************************/ + if( silk_pitch_analysis_core_FLP( res, psEncCtrl->pitchL, &psEnc->sCmn.indices.lagIndex, + &psEnc->sCmn.indices.contourIndex, &psEnc->LTPCorr, psEnc->sCmn.prevLag, psEnc->sCmn.pitchEstimationThreshold_Q16 / 65536.0f, + thrhld, psEnc->sCmn.fs_kHz, psEnc->sCmn.pitchEstimationComplexity, psEnc->sCmn.nb_subfr, arch ) == 0 ) + { + psEnc->sCmn.indices.signalType = TYPE_VOICED; + } else { + psEnc->sCmn.indices.signalType = TYPE_UNVOICED; + } + } else { + silk_memset( psEncCtrl->pitchL, 0, sizeof( psEncCtrl->pitchL ) ); + psEnc->sCmn.indices.lagIndex = 0; + psEnc->sCmn.indices.contourIndex = 0; + psEnc->LTPCorr = 0; + } +} diff --git a/lib/opus/silk/float/find_pred_coefs_FLP.c b/lib/opus/silk/float/find_pred_coefs_FLP.c new file mode 100644 index 00000000..6f790788 --- /dev/null +++ b/lib/opus/silk/float/find_pred_coefs_FLP.c @@ -0,0 +1,117 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +/* Find LPC and LTP coefficients */ +void silk_find_pred_coefs_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + const silk_float res_pitch[], /* I Residual from pitch analysis */ + const silk_float x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + opus_int i; + silk_float XXLTP[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ]; + silk_float xXLTP[ MAX_NB_SUBFR * LTP_ORDER ]; + silk_float invGains[ MAX_NB_SUBFR ]; + /* Set to NLSF_Q15 to zero so we don't copy junk to the state. */ + opus_int16 NLSF_Q15[ MAX_LPC_ORDER ]={0}; + const silk_float *x_ptr; + silk_float *x_pre_ptr, LPC_in_pre[ MAX_NB_SUBFR * MAX_LPC_ORDER + MAX_FRAME_LENGTH ]; + silk_float minInvGain; + + /* Weighting for weighted least squares */ + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + silk_assert( psEncCtrl->Gains[ i ] > 0.0f ); + invGains[ i ] = 1.0f / psEncCtrl->Gains[ i ]; + } + + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /**********/ + /* VOICED */ + /**********/ + celt_assert( psEnc->sCmn.ltp_mem_length - psEnc->sCmn.predictLPCOrder >= psEncCtrl->pitchL[ 0 ] + LTP_ORDER / 2 ); + + /* LTP analysis */ + silk_find_LTP_FLP( XXLTP, xXLTP, res_pitch, psEncCtrl->pitchL, psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr ); + + /* Quantize LTP gain parameters */ + silk_quant_LTP_gains_FLP( psEncCtrl->LTPCoef, psEnc->sCmn.indices.LTPIndex, &psEnc->sCmn.indices.PERIndex, + &psEnc->sCmn.sum_log_gain_Q7, &psEncCtrl->LTPredCodGain, XXLTP, xXLTP, psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.arch ); + + /* Control LTP scaling */ + silk_LTP_scale_ctrl_FLP( psEnc, psEncCtrl, condCoding ); + + /* Create LTP residual */ + silk_LTP_analysis_filter_FLP( LPC_in_pre, x - psEnc->sCmn.predictLPCOrder, psEncCtrl->LTPCoef, + psEncCtrl->pitchL, invGains, psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.predictLPCOrder ); + } else { + /************/ + /* UNVOICED */ + /************/ + /* Create signal with prepended subframes, scaled by inverse gains */ + x_ptr = x - psEnc->sCmn.predictLPCOrder; + x_pre_ptr = LPC_in_pre; + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + silk_scale_copy_vector_FLP( x_pre_ptr, x_ptr, invGains[ i ], + psEnc->sCmn.subfr_length + psEnc->sCmn.predictLPCOrder ); + x_pre_ptr += psEnc->sCmn.subfr_length + psEnc->sCmn.predictLPCOrder; + x_ptr += psEnc->sCmn.subfr_length; + } + silk_memset( psEncCtrl->LTPCoef, 0, psEnc->sCmn.nb_subfr * LTP_ORDER * sizeof( silk_float ) ); + psEncCtrl->LTPredCodGain = 0.0f; + psEnc->sCmn.sum_log_gain_Q7 = 0; + } + + /* Limit on total predictive coding gain */ + if( psEnc->sCmn.first_frame_after_reset ) { + minInvGain = 1.0f / MAX_PREDICTION_POWER_GAIN_AFTER_RESET; + } else { + minInvGain = (silk_float)pow( 2, psEncCtrl->LTPredCodGain / 3 ) / MAX_PREDICTION_POWER_GAIN; + minInvGain /= 0.25f + 0.75f * psEncCtrl->coding_quality; + } + + /* LPC_in_pre contains the LTP-filtered input for voiced, and the unfiltered input for unvoiced */ + silk_find_LPC_FLP( &psEnc->sCmn, NLSF_Q15, LPC_in_pre, minInvGain ); + + /* Quantize LSFs */ + silk_process_NLSFs_FLP( &psEnc->sCmn, psEncCtrl->PredCoef, NLSF_Q15, psEnc->sCmn.prev_NLSFq_Q15 ); + + /* Calculate residual energy using quantized LPC coefficients */ + silk_residual_energy_FLP( psEncCtrl->ResNrg, LPC_in_pre, psEncCtrl->PredCoef, psEncCtrl->Gains, + psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.predictLPCOrder ); + + /* Copy to prediction struct for use in next frame for interpolation */ + silk_memcpy( psEnc->sCmn.prev_NLSFq_Q15, NLSF_Q15, sizeof( psEnc->sCmn.prev_NLSFq_Q15 ) ); +} + diff --git a/lib/opus/silk/float/inner_product_FLP.c b/lib/opus/silk/float/inner_product_FLP.c new file mode 100644 index 00000000..cdd39d24 --- /dev/null +++ b/lib/opus/silk/float/inner_product_FLP.c @@ -0,0 +1,59 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" + +/* inner product of two silk_float arrays, with result as double */ +double silk_inner_product_FLP( + const silk_float *data1, + const silk_float *data2, + opus_int dataSize +) +{ + opus_int i; + double result; + + /* 4x unrolled loop */ + result = 0.0; + for( i = 0; i < dataSize - 3; i += 4 ) { + result += data1[ i + 0 ] * (double)data2[ i + 0 ] + + data1[ i + 1 ] * (double)data2[ i + 1 ] + + data1[ i + 2 ] * (double)data2[ i + 2 ] + + data1[ i + 3 ] * (double)data2[ i + 3 ]; + } + + /* add any remaining products */ + for( ; i < dataSize; i++ ) { + result += data1[ i ] * (double)data2[ i ]; + } + + return result; +} diff --git a/lib/opus/silk/float/k2a_FLP.c b/lib/opus/silk/float/k2a_FLP.c new file mode 100644 index 00000000..1448008d --- /dev/null +++ b/lib/opus/silk/float/k2a_FLP.c @@ -0,0 +1,54 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" + +/* step up function, converts reflection coefficients to prediction coefficients */ +void silk_k2a_FLP( + silk_float *A, /* O prediction coefficients [order] */ + const silk_float *rc, /* I reflection coefficients [order] */ + opus_int32 order /* I prediction order */ +) +{ + opus_int k, n; + silk_float rck, tmp1, tmp2; + + for( k = 0; k < order; k++ ) { + rck = rc[ k ]; + for( n = 0; n < (k + 1) >> 1; n++ ) { + tmp1 = A[ n ]; + tmp2 = A[ k - n - 1 ]; + A[ n ] = tmp1 + tmp2 * rck; + A[ k - n - 1 ] = tmp2 + tmp1 * rck; + } + A[ k ] = -rck; + } +} diff --git a/lib/opus/silk/float/main_FLP.h b/lib/opus/silk/float/main_FLP.h new file mode 100644 index 00000000..5dc0ccf4 --- /dev/null +++ b/lib/opus/silk/float/main_FLP.h @@ -0,0 +1,286 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MAIN_FLP_H +#define SILK_MAIN_FLP_H + +#include "SigProc_FLP.h" +#include "SigProc_FIX.h" +#include "structs_FLP.h" +#include "main.h" +#include "define.h" +#include "debug.h" +#include "entenc.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +#define silk_encoder_state_Fxx silk_encoder_state_FLP +#define silk_encode_do_VAD_Fxx silk_encode_do_VAD_FLP +#define silk_encode_frame_Fxx silk_encode_frame_FLP + +/*********************/ +/* Encoder Functions */ +/*********************/ + +/* High-pass filter with cutoff frequency adaptation based on pitch lag statistics */ +void silk_HP_variable_cutoff( + silk_encoder_state_Fxx state_Fxx[] /* I/O Encoder states */ +); + +/* Encoder main function */ +void silk_encode_do_VAD_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + opus_int activity /* I Decision of Opus voice activity detector */ +); + +/* Encoder main function */ +opus_int silk_encode_frame_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + opus_int32 *pnBytesOut, /* O Number of payload bytes; */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding, /* I The type of conditional coding to use */ + opus_int maxBits, /* I If > 0: maximum number of output bits */ + opus_int useCBR /* I Flag to force constant-bitrate operation */ +); + +/* Initializes the Silk encoder state */ +opus_int silk_init_encoder( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + int arch /* I Run-tim architecture */ +); + +/* Control the Silk encoder */ +opus_int silk_control_encoder( + silk_encoder_state_FLP *psEnc, /* I/O Pointer to Silk encoder state FLP */ + silk_EncControlStruct *encControl, /* I Control structure */ + const opus_int allow_bw_switch, /* I Flag to allow switching audio bandwidth */ + const opus_int channelNb, /* I Channel number */ + const opus_int force_fs_kHz +); + +/**************************/ +/* Noise shaping analysis */ +/**************************/ +/* Compute noise shaping coefficients and initial gain values */ +void silk_noise_shape_analysis_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + const silk_float *pitch_res, /* I LPC residual from pitch analysis */ + const silk_float *x /* I Input signal [frame_length + la_shape] */ +); + +/* Autocorrelations for a warped frequency axis */ +void silk_warped_autocorrelation_FLP( + silk_float *corr, /* O Result [order + 1] */ + const silk_float *input, /* I Input data to correlate */ + const silk_float warping, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +); + +/* Calculation of LTP state scaling */ +void silk_LTP_scale_ctrl_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/**********************************************/ +/* Prediction Analysis */ +/**********************************************/ +/* Find pitch lags */ +void silk_find_pitch_lags_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + silk_float res[], /* O Residual */ + const silk_float x[], /* I Speech signal */ + int arch /* I Run-time architecture */ +); + +/* Find LPC and LTP coefficients */ +void silk_find_pred_coefs_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + const silk_float res_pitch[], /* I Residual from pitch analysis */ + const silk_float x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/* LPC analysis */ +void silk_find_LPC_FLP( + silk_encoder_state *psEncC, /* I/O Encoder state */ + opus_int16 NLSF_Q15[], /* O NLSFs */ + const silk_float x[], /* I Input signal */ + const silk_float minInvGain /* I Prediction gain from LTP (dB) */ +); + +/* LTP analysis */ +void silk_find_LTP_FLP( + silk_float XX[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* O Weight for LTP quantization */ + silk_float xX[ MAX_NB_SUBFR * LTP_ORDER ], /* O Weight for LTP quantization */ + const silk_float r_ptr[], /* I LPC residual */ + const opus_int lag[ MAX_NB_SUBFR ], /* I LTP lags */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr /* I number of subframes */ +); + +void silk_LTP_analysis_filter_FLP( + silk_float *LTP_res, /* O LTP res MAX_NB_SUBFR*(pre_lgth+subfr_lngth) */ + const silk_float *x, /* I Input signal, with preceding samples */ + const silk_float B[ LTP_ORDER * MAX_NB_SUBFR ], /* I LTP coefficients for each subframe */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const silk_float invGains[ MAX_NB_SUBFR ], /* I Inverse quantization gains */ + const opus_int subfr_length, /* I Length of each subframe */ + const opus_int nb_subfr, /* I number of subframes */ + const opus_int pre_length /* I Preceding samples for each subframe */ +); + +/* Calculates residual energies of input subframes where all subframes have LPC_order */ +/* of preceding samples */ +void silk_residual_energy_FLP( + silk_float nrgs[ MAX_NB_SUBFR ], /* O Residual energy per subframe */ + const silk_float x[], /* I Input signal */ + silk_float a[ 2 ][ MAX_LPC_ORDER ], /* I AR coefs for each frame half */ + const silk_float gains[], /* I Quantization gains */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr, /* I number of subframes */ + const opus_int LPC_order /* I LPC order */ +); + +/* 16th order LPC analysis filter */ +void silk_LPC_analysis_filter_FLP( + silk_float r_LPC[], /* O LPC residual signal */ + const silk_float PredCoef[], /* I LPC coefficients */ + const silk_float s[], /* I Input signal */ + const opus_int length, /* I Length of input signal */ + const opus_int Order /* I LPC order */ +); + +/* LTP tap quantizer */ +void silk_quant_LTP_gains_FLP( + silk_float B[ MAX_NB_SUBFR * LTP_ORDER ], /* O Quantized LTP gains */ + opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook index */ + opus_int8 *periodicity_index, /* O Periodicity index */ + opus_int32 *sum_log_gain_Q7, /* I/O Cumulative max prediction gain */ + silk_float *pred_gain_dB, /* O LTP prediction gain */ + const silk_float XX[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* I Correlation matrix */ + const silk_float xX[ MAX_NB_SUBFR * LTP_ORDER ], /* I Correlation vector */ + const opus_int subfr_len, /* I Number of samples per subframe */ + const opus_int nb_subfr, /* I Number of subframes */ + int arch /* I Run-time architecture */ +); + +/* Residual energy: nrg = wxx - 2 * wXx * c + c' * wXX * c */ +silk_float silk_residual_energy_covar_FLP( /* O Weighted residual energy */ + const silk_float *c, /* I Filter coefficients */ + silk_float *wXX, /* I/O Weighted correlation matrix, reg. out */ + const silk_float *wXx, /* I Weighted correlation vector */ + const silk_float wxx, /* I Weighted correlation value */ + const opus_int D /* I Dimension */ +); + +/* Processing of gains */ +void silk_process_gains_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/******************/ +/* Linear Algebra */ +/******************/ +/* Calculates correlation matrix X'*X */ +void silk_corrMatrix_FLP( + const silk_float *x, /* I x vector [ L+order-1 ] used to create X */ + const opus_int L, /* I Length of vectors */ + const opus_int Order, /* I Max lag for correlation */ + silk_float *XX /* O X'*X correlation matrix [order x order] */ +); + +/* Calculates correlation vector X'*t */ +void silk_corrVector_FLP( + const silk_float *x, /* I x vector [L+order-1] used to create X */ + const silk_float *t, /* I Target vector [L] */ + const opus_int L, /* I Length of vecors */ + const opus_int Order, /* I Max lag for correlation */ + silk_float *Xt /* O X'*t correlation vector [order] */ +); + +/* Apply sine window to signal vector. */ +/* Window types: */ +/* 1 -> sine window from 0 to pi/2 */ +/* 2 -> sine window from pi/2 to pi */ +void silk_apply_sine_window_FLP( + silk_float px_win[], /* O Pointer to windowed signal */ + const silk_float px[], /* I Pointer to input signal */ + const opus_int win_type, /* I Selects a window type */ + const opus_int length /* I Window length, multiple of 4 */ +); + +/* Wrapper functions. Call flp / fix code */ + +/* Convert AR filter coefficients to NLSF parameters */ +void silk_A2NLSF_FLP( + opus_int16 *NLSF_Q15, /* O NLSF vector [ LPC_order ] */ + const silk_float *pAR, /* I LPC coefficients [ LPC_order ] */ + const opus_int LPC_order /* I LPC order */ +); + +/* Convert NLSF parameters to AR prediction filter coefficients */ +void silk_NLSF2A_FLP( + silk_float *pAR, /* O LPC coefficients [ LPC_order ] */ + const opus_int16 *NLSF_Q15, /* I NLSF vector [ LPC_order ] */ + const opus_int LPC_order, /* I LPC order */ + int arch /* I Run-time architecture */ +); + +/* Limit, stabilize, and quantize NLSFs */ +void silk_process_NLSFs_FLP( + silk_encoder_state *psEncC, /* I/O Encoder state */ + silk_float PredCoef[ 2 ][ MAX_LPC_ORDER ], /* O Prediction coefficients */ + opus_int16 NLSF_Q15[ MAX_LPC_ORDER ], /* I/O Normalized LSFs (quant out) (0 - (2^15-1)) */ + const opus_int16 prev_NLSF_Q15[ MAX_LPC_ORDER ] /* I Previous Normalized LSFs (0 - (2^15-1)) */ +); + +/* Floating-point Silk NSQ wrapper */ +void silk_NSQ_wrapper_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + SideInfoIndices *psIndices, /* I/O Quantization indices */ + silk_nsq_state *psNSQ, /* I/O Noise Shaping Quantzation state */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const silk_float x[] /* I Prefiltered input signal */ +); + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/float/noise_shape_analysis_FLP.c b/lib/opus/silk/float/noise_shape_analysis_FLP.c new file mode 100644 index 00000000..cb3d8a50 --- /dev/null +++ b/lib/opus/silk/float/noise_shape_analysis_FLP.c @@ -0,0 +1,350 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" +#include "tuning_parameters.h" + +/* Compute gain to make warped filter coefficients have a zero mean log frequency response on a */ +/* non-warped frequency scale. (So that it can be implemented with a minimum-phase monic filter.) */ +/* Note: A monic filter is one with the first coefficient equal to 1.0. In Silk we omit the first */ +/* coefficient in an array of coefficients, for monic filters. */ +static OPUS_INLINE silk_float warped_gain( + const silk_float *coefs, + silk_float lambda, + opus_int order +) { + opus_int i; + silk_float gain; + + lambda = -lambda; + gain = coefs[ order - 1 ]; + for( i = order - 2; i >= 0; i-- ) { + gain = lambda * gain + coefs[ i ]; + } + return (silk_float)( 1.0f / ( 1.0f - lambda * gain ) ); +} + +/* Convert warped filter coefficients to monic pseudo-warped coefficients and limit maximum */ +/* amplitude of monic warped coefficients by using bandwidth expansion on the true coefficients */ +static OPUS_INLINE void warped_true2monic_coefs( + silk_float *coefs, + silk_float lambda, + silk_float limit, + opus_int order +) { + opus_int i, iter, ind = 0; + silk_float tmp, maxabs, chirp, gain; + + /* Convert to monic coefficients */ + for( i = order - 1; i > 0; i-- ) { + coefs[ i - 1 ] -= lambda * coefs[ i ]; + } + gain = ( 1.0f - lambda * lambda ) / ( 1.0f + lambda * coefs[ 0 ] ); + for( i = 0; i < order; i++ ) { + coefs[ i ] *= gain; + } + + /* Limit */ + for( iter = 0; iter < 10; iter++ ) { + /* Find maximum absolute value */ + maxabs = -1.0f; + for( i = 0; i < order; i++ ) { + tmp = silk_abs_float( coefs[ i ] ); + if( tmp > maxabs ) { + maxabs = tmp; + ind = i; + } + } + if( maxabs <= limit ) { + /* Coefficients are within range - done */ + return; + } + + /* Convert back to true warped coefficients */ + for( i = 1; i < order; i++ ) { + coefs[ i - 1 ] += lambda * coefs[ i ]; + } + gain = 1.0f / gain; + for( i = 0; i < order; i++ ) { + coefs[ i ] *= gain; + } + + /* Apply bandwidth expansion */ + chirp = 0.99f - ( 0.8f + 0.1f * iter ) * ( maxabs - limit ) / ( maxabs * ( ind + 1 ) ); + silk_bwexpander_FLP( coefs, order, chirp ); + + /* Convert to monic warped coefficients */ + for( i = order - 1; i > 0; i-- ) { + coefs[ i - 1 ] -= lambda * coefs[ i ]; + } + gain = ( 1.0f - lambda * lambda ) / ( 1.0f + lambda * coefs[ 0 ] ); + for( i = 0; i < order; i++ ) { + coefs[ i ] *= gain; + } + } + silk_assert( 0 ); +} + +static OPUS_INLINE void limit_coefs( + silk_float *coefs, + silk_float limit, + opus_int order +) { + opus_int i, iter, ind = 0; + silk_float tmp, maxabs, chirp; + + for( iter = 0; iter < 10; iter++ ) { + /* Find maximum absolute value */ + maxabs = -1.0f; + for( i = 0; i < order; i++ ) { + tmp = silk_abs_float( coefs[ i ] ); + if( tmp > maxabs ) { + maxabs = tmp; + ind = i; + } + } + if( maxabs <= limit ) { + /* Coefficients are within range - done */ + return; + } + + /* Apply bandwidth expansion */ + chirp = 0.99f - ( 0.8f + 0.1f * iter ) * ( maxabs - limit ) / ( maxabs * ( ind + 1 ) ); + silk_bwexpander_FLP( coefs, order, chirp ); + } + silk_assert( 0 ); +} + +/* Compute noise shaping coefficients and initial gain values */ +void silk_noise_shape_analysis_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + const silk_float *pitch_res, /* I LPC residual from pitch analysis */ + const silk_float *x /* I Input signal [frame_length + la_shape] */ +) +{ + silk_shape_state_FLP *psShapeSt = &psEnc->sShape; + opus_int k, nSamples, nSegs; + silk_float SNR_adj_dB, HarmShapeGain, Tilt; + silk_float nrg, log_energy, log_energy_prev, energy_variation; + silk_float BWExp, gain_mult, gain_add, strength, b, warping; + silk_float x_windowed[ SHAPE_LPC_WIN_MAX ]; + silk_float auto_corr[ MAX_SHAPE_LPC_ORDER + 1 ]; + silk_float rc[ MAX_SHAPE_LPC_ORDER + 1 ]; + const silk_float *x_ptr, *pitch_res_ptr; + + /* Point to start of first LPC analysis block */ + x_ptr = x - psEnc->sCmn.la_shape; + + /****************/ + /* GAIN CONTROL */ + /****************/ + SNR_adj_dB = psEnc->sCmn.SNR_dB_Q7 * ( 1 / 128.0f ); + + /* Input quality is the average of the quality in the lowest two VAD bands */ + psEncCtrl->input_quality = 0.5f * ( psEnc->sCmn.input_quality_bands_Q15[ 0 ] + psEnc->sCmn.input_quality_bands_Q15[ 1 ] ) * ( 1.0f / 32768.0f ); + + /* Coding quality level, between 0.0 and 1.0 */ + psEncCtrl->coding_quality = silk_sigmoid( 0.25f * ( SNR_adj_dB - 20.0f ) ); + + if( psEnc->sCmn.useCBR == 0 ) { + /* Reduce coding SNR during low speech activity */ + b = 1.0f - psEnc->sCmn.speech_activity_Q8 * ( 1.0f / 256.0f ); + SNR_adj_dB -= BG_SNR_DECR_dB * psEncCtrl->coding_quality * ( 0.5f + 0.5f * psEncCtrl->input_quality ) * b * b; + } + + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Reduce gains for periodic signals */ + SNR_adj_dB += HARM_SNR_INCR_dB * psEnc->LTPCorr; + } else { + /* For unvoiced signals and low-quality input, adjust the quality slower than SNR_dB setting */ + SNR_adj_dB += ( -0.4f * psEnc->sCmn.SNR_dB_Q7 * ( 1 / 128.0f ) + 6.0f ) * ( 1.0f - psEncCtrl->input_quality ); + } + + /*************************/ + /* SPARSENESS PROCESSING */ + /*************************/ + /* Set quantizer offset */ + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Initially set to 0; may be overruled in process_gains(..) */ + psEnc->sCmn.indices.quantOffsetType = 0; + } else { + /* Sparseness measure, based on relative fluctuations of energy per 2 milliseconds */ + nSamples = 2 * psEnc->sCmn.fs_kHz; + energy_variation = 0.0f; + log_energy_prev = 0.0f; + pitch_res_ptr = pitch_res; + nSegs = silk_SMULBB( SUB_FRAME_LENGTH_MS, psEnc->sCmn.nb_subfr ) / 2; + for( k = 0; k < nSegs; k++ ) { + nrg = ( silk_float )nSamples + ( silk_float )silk_energy_FLP( pitch_res_ptr, nSamples ); + log_energy = silk_log2( nrg ); + if( k > 0 ) { + energy_variation += silk_abs_float( log_energy - log_energy_prev ); + } + log_energy_prev = log_energy; + pitch_res_ptr += nSamples; + } + + /* Set quantization offset depending on sparseness measure */ + if( energy_variation > ENERGY_VARIATION_THRESHOLD_QNT_OFFSET * (nSegs-1) ) { + psEnc->sCmn.indices.quantOffsetType = 0; + } else { + psEnc->sCmn.indices.quantOffsetType = 1; + } + } + + /*******************************/ + /* Control bandwidth expansion */ + /*******************************/ + /* More BWE for signals with high prediction gain */ + strength = FIND_PITCH_WHITE_NOISE_FRACTION * psEncCtrl->predGain; /* between 0.0 and 1.0 */ + BWExp = BANDWIDTH_EXPANSION / ( 1.0f + strength * strength ); + + /* Slightly more warping in analysis will move quantization noise up in frequency, where it's better masked */ + warping = (silk_float)psEnc->sCmn.warping_Q16 / 65536.0f + 0.01f * psEncCtrl->coding_quality; + + /********************************************/ + /* Compute noise shaping AR coefs and gains */ + /********************************************/ + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + /* Apply window: sine slope followed by flat part followed by cosine slope */ + opus_int shift, slope_part, flat_part; + flat_part = psEnc->sCmn.fs_kHz * 3; + slope_part = ( psEnc->sCmn.shapeWinLength - flat_part ) / 2; + + silk_apply_sine_window_FLP( x_windowed, x_ptr, 1, slope_part ); + shift = slope_part; + silk_memcpy( x_windowed + shift, x_ptr + shift, flat_part * sizeof(silk_float) ); + shift += flat_part; + silk_apply_sine_window_FLP( x_windowed + shift, x_ptr + shift, 2, slope_part ); + + /* Update pointer: next LPC analysis block */ + x_ptr += psEnc->sCmn.subfr_length; + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Calculate warped auto correlation */ + silk_warped_autocorrelation_FLP( auto_corr, x_windowed, warping, + psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder ); + } else { + /* Calculate regular auto correlation */ + silk_autocorrelation_FLP( auto_corr, x_windowed, psEnc->sCmn.shapeWinLength, psEnc->sCmn.shapingLPCOrder + 1 ); + } + + /* Add white noise, as a fraction of energy */ + auto_corr[ 0 ] += auto_corr[ 0 ] * SHAPE_WHITE_NOISE_FRACTION + 1.0f; + + /* Convert correlations to prediction coefficients, and compute residual energy */ + nrg = silk_schur_FLP( rc, auto_corr, psEnc->sCmn.shapingLPCOrder ); + silk_k2a_FLP( &psEncCtrl->AR[ k * MAX_SHAPE_LPC_ORDER ], rc, psEnc->sCmn.shapingLPCOrder ); + psEncCtrl->Gains[ k ] = ( silk_float )sqrt( nrg ); + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Adjust gain for warping */ + psEncCtrl->Gains[ k ] *= warped_gain( &psEncCtrl->AR[ k * MAX_SHAPE_LPC_ORDER ], warping, psEnc->sCmn.shapingLPCOrder ); + } + + /* Bandwidth expansion for synthesis filter shaping */ + silk_bwexpander_FLP( &psEncCtrl->AR[ k * MAX_SHAPE_LPC_ORDER ], psEnc->sCmn.shapingLPCOrder, BWExp ); + + if( psEnc->sCmn.warping_Q16 > 0 ) { + /* Convert to monic warped prediction coefficients and limit absolute values */ + warped_true2monic_coefs( &psEncCtrl->AR[ k * MAX_SHAPE_LPC_ORDER ], warping, 3.999f, psEnc->sCmn.shapingLPCOrder ); + } else { + /* Limit absolute values */ + limit_coefs( &psEncCtrl->AR[ k * MAX_SHAPE_LPC_ORDER ], 3.999f, psEnc->sCmn.shapingLPCOrder ); + } + } + + /*****************/ + /* Gain tweaking */ + /*****************/ + /* Increase gains during low speech activity */ + gain_mult = (silk_float)pow( 2.0f, -0.16f * SNR_adj_dB ); + gain_add = (silk_float)pow( 2.0f, 0.16f * MIN_QGAIN_DB ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->Gains[ k ] *= gain_mult; + psEncCtrl->Gains[ k ] += gain_add; + } + + /************************************************/ + /* Control low-frequency shaping and noise tilt */ + /************************************************/ + /* Less low frequency shaping for noisy inputs */ + strength = LOW_FREQ_SHAPING * ( 1.0f + LOW_QUALITY_LOW_FREQ_SHAPING_DECR * ( psEnc->sCmn.input_quality_bands_Q15[ 0 ] * ( 1.0f / 32768.0f ) - 1.0f ) ); + strength *= psEnc->sCmn.speech_activity_Q8 * ( 1.0f / 256.0f ); + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Reduce low frequencies quantization noise for periodic signals, depending on pitch lag */ + /*f = 400; freqz([1, -0.98 + 2e-4 * f], [1, -0.97 + 7e-4 * f], 2^12, Fs); axis([0, 1000, -10, 1])*/ + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + b = 0.2f / psEnc->sCmn.fs_kHz + 3.0f / psEncCtrl->pitchL[ k ]; + psEncCtrl->LF_MA_shp[ k ] = -1.0f + b; + psEncCtrl->LF_AR_shp[ k ] = 1.0f - b - b * strength; + } + Tilt = - HP_NOISE_COEF - + (1 - HP_NOISE_COEF) * HARM_HP_NOISE_COEF * psEnc->sCmn.speech_activity_Q8 * ( 1.0f / 256.0f ); + } else { + b = 1.3f / psEnc->sCmn.fs_kHz; + psEncCtrl->LF_MA_shp[ 0 ] = -1.0f + b; + psEncCtrl->LF_AR_shp[ 0 ] = 1.0f - b - b * strength * 0.6f; + for( k = 1; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->LF_MA_shp[ k ] = psEncCtrl->LF_MA_shp[ 0 ]; + psEncCtrl->LF_AR_shp[ k ] = psEncCtrl->LF_AR_shp[ 0 ]; + } + Tilt = -HP_NOISE_COEF; + } + + /****************************/ + /* HARMONIC SHAPING CONTROL */ + /****************************/ + if( USE_HARM_SHAPING && psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + /* Harmonic noise shaping */ + HarmShapeGain = HARMONIC_SHAPING; + + /* More harmonic noise shaping for high bitrates or noisy input */ + HarmShapeGain += HIGH_RATE_OR_LOW_QUALITY_HARMONIC_SHAPING * + ( 1.0f - ( 1.0f - psEncCtrl->coding_quality ) * psEncCtrl->input_quality ); + + /* Less harmonic noise shaping for less periodic signals */ + HarmShapeGain *= ( silk_float )sqrt( psEnc->LTPCorr ); + } else { + HarmShapeGain = 0.0f; + } + + /*************************/ + /* Smooth over subframes */ + /*************************/ + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psShapeSt->HarmShapeGain_smth += SUBFR_SMTH_COEF * ( HarmShapeGain - psShapeSt->HarmShapeGain_smth ); + psEncCtrl->HarmShapeGain[ k ] = psShapeSt->HarmShapeGain_smth; + psShapeSt->Tilt_smth += SUBFR_SMTH_COEF * ( Tilt - psShapeSt->Tilt_smth ); + psEncCtrl->Tilt[ k ] = psShapeSt->Tilt_smth; + } +} diff --git a/lib/opus/silk/float/pitch_analysis_core_FLP.c b/lib/opus/silk/float/pitch_analysis_core_FLP.c new file mode 100644 index 00000000..f351bc37 --- /dev/null +++ b/lib/opus/silk/float/pitch_analysis_core_FLP.c @@ -0,0 +1,630 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/***************************************************************************** +* Pitch analyser function +******************************************************************************/ +#include "SigProc_FLP.h" +#include "SigProc_FIX.h" +#include "pitch_est_defines.h" +#include "pitch.h" + +#define SCRATCH_SIZE 22 + +/************************************************************/ +/* Internally used functions */ +/************************************************************/ +static void silk_P_Ana_calc_corr_st3( + silk_float cross_corr_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ], /* O 3 DIM correlation array */ + const silk_float frame[], /* I vector to correlate */ + opus_int start_lag, /* I start lag */ + opus_int sf_length, /* I sub frame length */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ +); + +static void silk_P_Ana_calc_energy_st3( + silk_float energies_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ], /* O 3 DIM correlation array */ + const silk_float frame[], /* I vector to correlate */ + opus_int start_lag, /* I start lag */ + opus_int sf_length, /* I sub frame length */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity /* I Complexity setting */ +); + +/************************************************************/ +/* CORE PITCH ANALYSIS FUNCTION */ +/************************************************************/ +opus_int silk_pitch_analysis_core_FLP( /* O Voicing estimate: 0 voiced, 1 unvoiced */ + const silk_float *frame, /* I Signal of length PE_FRAME_LENGTH_MS*Fs_kHz */ + opus_int *pitch_out, /* O Pitch lag values [nb_subfr] */ + opus_int16 *lagIndex, /* O Lag Index */ + opus_int8 *contourIndex, /* O Pitch contour Index */ + silk_float *LTPCorr, /* I/O Normalized correlation; input: value from previous frame */ + opus_int prevLag, /* I Last lag of previous frame; set to zero is unvoiced */ + const silk_float search_thres1, /* I First stage threshold for lag candidates 0 - 1 */ + const silk_float search_thres2, /* I Final threshold for lag candidates 0 - 1 */ + const opus_int Fs_kHz, /* I sample frequency (kHz) */ + const opus_int complexity, /* I Complexity setting, 0-2, where 2 is highest */ + const opus_int nb_subfr, /* I Number of 5 ms subframes */ + int arch /* I Run-time architecture */ +) +{ + opus_int i, k, d, j; + silk_float frame_8kHz[ PE_MAX_FRAME_LENGTH_MS * 8 ]; + silk_float frame_4kHz[ PE_MAX_FRAME_LENGTH_MS * 4 ]; + opus_int16 frame_8_FIX[ PE_MAX_FRAME_LENGTH_MS * 8 ]; + opus_int16 frame_4_FIX[ PE_MAX_FRAME_LENGTH_MS * 4 ]; + opus_int32 filt_state[ 6 ]; + silk_float threshold, contour_bias; + silk_float C[ PE_MAX_NB_SUBFR][ (PE_MAX_LAG >> 1) + 5 ]; + opus_val32 xcorr[ PE_MAX_LAG_MS * 4 - PE_MIN_LAG_MS * 4 + 1 ]; + silk_float CC[ PE_NB_CBKS_STAGE2_EXT ]; + const silk_float *target_ptr, *basis_ptr; + double cross_corr, normalizer, energy, energy_tmp; + opus_int d_srch[ PE_D_SRCH_LENGTH ]; + opus_int16 d_comp[ (PE_MAX_LAG >> 1) + 5 ]; + opus_int length_d_srch, length_d_comp; + silk_float Cmax, CCmax, CCmax_b, CCmax_new_b, CCmax_new; + opus_int CBimax, CBimax_new, lag, start_lag, end_lag, lag_new; + opus_int cbk_size; + silk_float lag_log2, prevLag_log2, delta_lag_log2_sqr; + silk_float energies_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ]; + silk_float cross_corr_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ]; + opus_int lag_counter; + opus_int frame_length, frame_length_8kHz, frame_length_4kHz; + opus_int sf_length, sf_length_8kHz, sf_length_4kHz; + opus_int min_lag, min_lag_8kHz, min_lag_4kHz; + opus_int max_lag, max_lag_8kHz, max_lag_4kHz; + opus_int nb_cbk_search; + const opus_int8 *Lag_CB_ptr; + + /* Check for valid sampling frequency */ + celt_assert( Fs_kHz == 8 || Fs_kHz == 12 || Fs_kHz == 16 ); + + /* Check for valid complexity setting */ + celt_assert( complexity >= SILK_PE_MIN_COMPLEX ); + celt_assert( complexity <= SILK_PE_MAX_COMPLEX ); + + silk_assert( search_thres1 >= 0.0f && search_thres1 <= 1.0f ); + silk_assert( search_thres2 >= 0.0f && search_thres2 <= 1.0f ); + + /* Set up frame lengths max / min lag for the sampling frequency */ + frame_length = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * Fs_kHz; + frame_length_4kHz = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * 4; + frame_length_8kHz = ( PE_LTP_MEM_LENGTH_MS + nb_subfr * PE_SUBFR_LENGTH_MS ) * 8; + sf_length = PE_SUBFR_LENGTH_MS * Fs_kHz; + sf_length_4kHz = PE_SUBFR_LENGTH_MS * 4; + sf_length_8kHz = PE_SUBFR_LENGTH_MS * 8; + min_lag = PE_MIN_LAG_MS * Fs_kHz; + min_lag_4kHz = PE_MIN_LAG_MS * 4; + min_lag_8kHz = PE_MIN_LAG_MS * 8; + max_lag = PE_MAX_LAG_MS * Fs_kHz - 1; + max_lag_4kHz = PE_MAX_LAG_MS * 4; + max_lag_8kHz = PE_MAX_LAG_MS * 8 - 1; + + /* Resample from input sampled at Fs_kHz to 8 kHz */ + if( Fs_kHz == 16 ) { + /* Resample to 16 -> 8 khz */ + opus_int16 frame_16_FIX[ 16 * PE_MAX_FRAME_LENGTH_MS ]; + silk_float2short_array( frame_16_FIX, frame, frame_length ); + silk_memset( filt_state, 0, 2 * sizeof( opus_int32 ) ); + silk_resampler_down2( filt_state, frame_8_FIX, frame_16_FIX, frame_length ); + silk_short2float_array( frame_8kHz, frame_8_FIX, frame_length_8kHz ); + } else if( Fs_kHz == 12 ) { + /* Resample to 12 -> 8 khz */ + opus_int16 frame_12_FIX[ 12 * PE_MAX_FRAME_LENGTH_MS ]; + silk_float2short_array( frame_12_FIX, frame, frame_length ); + silk_memset( filt_state, 0, 6 * sizeof( opus_int32 ) ); + silk_resampler_down2_3( filt_state, frame_8_FIX, frame_12_FIX, frame_length ); + silk_short2float_array( frame_8kHz, frame_8_FIX, frame_length_8kHz ); + } else { + celt_assert( Fs_kHz == 8 ); + silk_float2short_array( frame_8_FIX, frame, frame_length_8kHz ); + } + + /* Decimate again to 4 kHz */ + silk_memset( filt_state, 0, 2 * sizeof( opus_int32 ) ); + silk_resampler_down2( filt_state, frame_4_FIX, frame_8_FIX, frame_length_8kHz ); + silk_short2float_array( frame_4kHz, frame_4_FIX, frame_length_4kHz ); + + /* Low-pass filter */ + for( i = frame_length_4kHz - 1; i > 0; i-- ) { + frame_4kHz[ i ] = silk_ADD_SAT16( frame_4kHz[ i ], frame_4kHz[ i - 1 ] ); + } + + /****************************************************************************** + * FIRST STAGE, operating in 4 khz + ******************************************************************************/ + silk_memset(C, 0, sizeof(silk_float) * nb_subfr * ((PE_MAX_LAG >> 1) + 5)); + target_ptr = &frame_4kHz[ silk_LSHIFT( sf_length_4kHz, 2 ) ]; + for( k = 0; k < nb_subfr >> 1; k++ ) { + /* Check that we are within range of the array */ + celt_assert( target_ptr >= frame_4kHz ); + celt_assert( target_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); + + basis_ptr = target_ptr - min_lag_4kHz; + + /* Check that we are within range of the array */ + celt_assert( basis_ptr >= frame_4kHz ); + celt_assert( basis_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); + + celt_pitch_xcorr( target_ptr, target_ptr-max_lag_4kHz, xcorr, sf_length_8kHz, max_lag_4kHz - min_lag_4kHz + 1, arch ); + + /* Calculate first vector products before loop */ + cross_corr = xcorr[ max_lag_4kHz - min_lag_4kHz ]; + normalizer = silk_energy_FLP( target_ptr, sf_length_8kHz ) + + silk_energy_FLP( basis_ptr, sf_length_8kHz ) + + sf_length_8kHz * 4000.0f; + + C[ 0 ][ min_lag_4kHz ] += (silk_float)( 2 * cross_corr / normalizer ); + + /* From now on normalizer is computed recursively */ + for( d = min_lag_4kHz + 1; d <= max_lag_4kHz; d++ ) { + basis_ptr--; + + /* Check that we are within range of the array */ + silk_assert( basis_ptr >= frame_4kHz ); + silk_assert( basis_ptr + sf_length_8kHz <= frame_4kHz + frame_length_4kHz ); + + cross_corr = xcorr[ max_lag_4kHz - d ]; + + /* Add contribution of new sample and remove contribution from oldest sample */ + normalizer += + basis_ptr[ 0 ] * (double)basis_ptr[ 0 ] - + basis_ptr[ sf_length_8kHz ] * (double)basis_ptr[ sf_length_8kHz ]; + C[ 0 ][ d ] += (silk_float)( 2 * cross_corr / normalizer ); + } + /* Update target pointer */ + target_ptr += sf_length_8kHz; + } + + /* Apply short-lag bias */ + for( i = max_lag_4kHz; i >= min_lag_4kHz; i-- ) { + C[ 0 ][ i ] -= C[ 0 ][ i ] * i / 4096.0f; + } + + /* Sort */ + length_d_srch = 4 + 2 * complexity; + celt_assert( 3 * length_d_srch <= PE_D_SRCH_LENGTH ); + silk_insertion_sort_decreasing_FLP( &C[ 0 ][ min_lag_4kHz ], d_srch, max_lag_4kHz - min_lag_4kHz + 1, length_d_srch ); + + /* Escape if correlation is very low already here */ + Cmax = C[ 0 ][ min_lag_4kHz ]; + if( Cmax < 0.2f ) { + silk_memset( pitch_out, 0, nb_subfr * sizeof( opus_int ) ); + *LTPCorr = 0.0f; + *lagIndex = 0; + *contourIndex = 0; + return 1; + } + + threshold = search_thres1 * Cmax; + for( i = 0; i < length_d_srch; i++ ) { + /* Convert to 8 kHz indices for the sorted correlation that exceeds the threshold */ + if( C[ 0 ][ min_lag_4kHz + i ] > threshold ) { + d_srch[ i ] = silk_LSHIFT( d_srch[ i ] + min_lag_4kHz, 1 ); + } else { + length_d_srch = i; + break; + } + } + celt_assert( length_d_srch > 0 ); + + for( i = min_lag_8kHz - 5; i < max_lag_8kHz + 5; i++ ) { + d_comp[ i ] = 0; + } + for( i = 0; i < length_d_srch; i++ ) { + d_comp[ d_srch[ i ] ] = 1; + } + + /* Convolution */ + for( i = max_lag_8kHz + 3; i >= min_lag_8kHz; i-- ) { + d_comp[ i ] += d_comp[ i - 1 ] + d_comp[ i - 2 ]; + } + + length_d_srch = 0; + for( i = min_lag_8kHz; i < max_lag_8kHz + 1; i++ ) { + if( d_comp[ i + 1 ] > 0 ) { + d_srch[ length_d_srch ] = i; + length_d_srch++; + } + } + + /* Convolution */ + for( i = max_lag_8kHz + 3; i >= min_lag_8kHz; i-- ) { + d_comp[ i ] += d_comp[ i - 1 ] + d_comp[ i - 2 ] + d_comp[ i - 3 ]; + } + + length_d_comp = 0; + for( i = min_lag_8kHz; i < max_lag_8kHz + 4; i++ ) { + if( d_comp[ i ] > 0 ) { + d_comp[ length_d_comp ] = (opus_int16)( i - 2 ); + length_d_comp++; + } + } + + /********************************************************************************** + ** SECOND STAGE, operating at 8 kHz, on lag sections with high correlation + *************************************************************************************/ + /********************************************************************************* + * Find energy of each subframe projected onto its history, for a range of delays + *********************************************************************************/ + silk_memset( C, 0, PE_MAX_NB_SUBFR*((PE_MAX_LAG >> 1) + 5) * sizeof(silk_float)); + + if( Fs_kHz == 8 ) { + target_ptr = &frame[ PE_LTP_MEM_LENGTH_MS * 8 ]; + } else { + target_ptr = &frame_8kHz[ PE_LTP_MEM_LENGTH_MS * 8 ]; + } + for( k = 0; k < nb_subfr; k++ ) { + energy_tmp = silk_energy_FLP( target_ptr, sf_length_8kHz ) + 1.0; + for( j = 0; j < length_d_comp; j++ ) { + d = d_comp[ j ]; + basis_ptr = target_ptr - d; + cross_corr = silk_inner_product_FLP( basis_ptr, target_ptr, sf_length_8kHz ); + if( cross_corr > 0.0f ) { + energy = silk_energy_FLP( basis_ptr, sf_length_8kHz ); + C[ k ][ d ] = (silk_float)( 2 * cross_corr / ( energy + energy_tmp ) ); + } else { + C[ k ][ d ] = 0.0f; + } + } + target_ptr += sf_length_8kHz; + } + + /* search over lag range and lags codebook */ + /* scale factor for lag codebook, as a function of center lag */ + + CCmax = 0.0f; /* This value doesn't matter */ + CCmax_b = -1000.0f; + + CBimax = 0; /* To avoid returning undefined lag values */ + lag = -1; /* To check if lag with strong enough correlation has been found */ + + if( prevLag > 0 ) { + if( Fs_kHz == 12 ) { + prevLag = silk_LSHIFT( prevLag, 1 ) / 3; + } else if( Fs_kHz == 16 ) { + prevLag = silk_RSHIFT( prevLag, 1 ); + } + prevLag_log2 = silk_log2( (silk_float)prevLag ); + } else { + prevLag_log2 = 0; + } + + /* Set up stage 2 codebook based on number of subframes */ + if( nb_subfr == PE_MAX_NB_SUBFR ) { + cbk_size = PE_NB_CBKS_STAGE2_EXT; + Lag_CB_ptr = &silk_CB_lags_stage2[ 0 ][ 0 ]; + if( Fs_kHz == 8 && complexity > SILK_PE_MIN_COMPLEX ) { + /* If input is 8 khz use a larger codebook here because it is last stage */ + nb_cbk_search = PE_NB_CBKS_STAGE2_EXT; + } else { + nb_cbk_search = PE_NB_CBKS_STAGE2; + } + } else { + cbk_size = PE_NB_CBKS_STAGE2_10MS; + Lag_CB_ptr = &silk_CB_lags_stage2_10_ms[ 0 ][ 0 ]; + nb_cbk_search = PE_NB_CBKS_STAGE2_10MS; + } + + for( k = 0; k < length_d_srch; k++ ) { + d = d_srch[ k ]; + for( j = 0; j < nb_cbk_search; j++ ) { + CC[j] = 0.0f; + for( i = 0; i < nb_subfr; i++ ) { + /* Try all codebooks */ + CC[ j ] += C[ i ][ d + matrix_ptr( Lag_CB_ptr, i, j, cbk_size )]; + } + } + /* Find best codebook */ + CCmax_new = -1000.0f; + CBimax_new = 0; + for( i = 0; i < nb_cbk_search; i++ ) { + if( CC[ i ] > CCmax_new ) { + CCmax_new = CC[ i ]; + CBimax_new = i; + } + } + + /* Bias towards shorter lags */ + lag_log2 = silk_log2( (silk_float)d ); + CCmax_new_b = CCmax_new - PE_SHORTLAG_BIAS * nb_subfr * lag_log2; + + /* Bias towards previous lag */ + if( prevLag > 0 ) { + delta_lag_log2_sqr = lag_log2 - prevLag_log2; + delta_lag_log2_sqr *= delta_lag_log2_sqr; + CCmax_new_b -= PE_PREVLAG_BIAS * nb_subfr * (*LTPCorr) * delta_lag_log2_sqr / ( delta_lag_log2_sqr + 0.5f ); + } + + if( CCmax_new_b > CCmax_b && /* Find maximum biased correlation */ + CCmax_new > nb_subfr * search_thres2 /* Correlation needs to be high enough to be voiced */ + ) { + CCmax_b = CCmax_new_b; + CCmax = CCmax_new; + lag = d; + CBimax = CBimax_new; + } + } + + if( lag == -1 ) { + /* No suitable candidate found */ + silk_memset( pitch_out, 0, PE_MAX_NB_SUBFR * sizeof(opus_int) ); + *LTPCorr = 0.0f; + *lagIndex = 0; + *contourIndex = 0; + return 1; + } + + /* Output normalized correlation */ + *LTPCorr = (silk_float)( CCmax / nb_subfr ); + silk_assert( *LTPCorr >= 0.0f ); + + if( Fs_kHz > 8 ) { + /* Search in original signal */ + + /* Compensate for decimation */ + silk_assert( lag == silk_SAT16( lag ) ); + if( Fs_kHz == 12 ) { + lag = silk_RSHIFT_ROUND( silk_SMULBB( lag, 3 ), 1 ); + } else { /* Fs_kHz == 16 */ + lag = silk_LSHIFT( lag, 1 ); + } + + lag = silk_LIMIT_int( lag, min_lag, max_lag ); + start_lag = silk_max_int( lag - 2, min_lag ); + end_lag = silk_min_int( lag + 2, max_lag ); + lag_new = lag; /* to avoid undefined lag */ + CBimax = 0; /* to avoid undefined lag */ + + CCmax = -1000.0f; + + /* Calculate the correlations and energies needed in stage 3 */ + silk_P_Ana_calc_corr_st3( cross_corr_st3, frame, start_lag, sf_length, nb_subfr, complexity, arch ); + silk_P_Ana_calc_energy_st3( energies_st3, frame, start_lag, sf_length, nb_subfr, complexity ); + + lag_counter = 0; + silk_assert( lag == silk_SAT16( lag ) ); + contour_bias = PE_FLATCONTOUR_BIAS / lag; + + /* Set up cbk parameters according to complexity setting and frame length */ + if( nb_subfr == PE_MAX_NB_SUBFR ) { + nb_cbk_search = (opus_int)silk_nb_cbk_searchs_stage3[ complexity ]; + cbk_size = PE_NB_CBKS_STAGE3_MAX; + Lag_CB_ptr = &silk_CB_lags_stage3[ 0 ][ 0 ]; + } else { + nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; + cbk_size = PE_NB_CBKS_STAGE3_10MS; + Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; + } + + target_ptr = &frame[ PE_LTP_MEM_LENGTH_MS * Fs_kHz ]; + energy_tmp = silk_energy_FLP( target_ptr, nb_subfr * sf_length ) + 1.0; + for( d = start_lag; d <= end_lag; d++ ) { + for( j = 0; j < nb_cbk_search; j++ ) { + cross_corr = 0.0; + energy = energy_tmp; + for( k = 0; k < nb_subfr; k++ ) { + cross_corr += cross_corr_st3[ k ][ j ][ lag_counter ]; + energy += energies_st3[ k ][ j ][ lag_counter ]; + } + if( cross_corr > 0.0 ) { + CCmax_new = (silk_float)( 2 * cross_corr / energy ); + /* Reduce depending on flatness of contour */ + CCmax_new *= 1.0f - contour_bias * j; + } else { + CCmax_new = 0.0f; + } + + if( CCmax_new > CCmax && ( d + (opus_int)silk_CB_lags_stage3[ 0 ][ j ] ) <= max_lag ) { + CCmax = CCmax_new; + lag_new = d; + CBimax = j; + } + } + lag_counter++; + } + + for( k = 0; k < nb_subfr; k++ ) { + pitch_out[ k ] = lag_new + matrix_ptr( Lag_CB_ptr, k, CBimax, cbk_size ); + pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], min_lag, PE_MAX_LAG_MS * Fs_kHz ); + } + *lagIndex = (opus_int16)( lag_new - min_lag ); + *contourIndex = (opus_int8)CBimax; + } else { /* Fs_kHz == 8 */ + /* Save Lags */ + for( k = 0; k < nb_subfr; k++ ) { + pitch_out[ k ] = lag + matrix_ptr( Lag_CB_ptr, k, CBimax, cbk_size ); + pitch_out[ k ] = silk_LIMIT( pitch_out[ k ], min_lag_8kHz, PE_MAX_LAG_MS * 8 ); + } + *lagIndex = (opus_int16)( lag - min_lag_8kHz ); + *contourIndex = (opus_int8)CBimax; + } + celt_assert( *lagIndex >= 0 ); + /* return as voiced */ + return 0; +} + +/*********************************************************************** + * Calculates the correlations used in stage 3 search. In order to cover + * the whole lag codebook for all the searched offset lags (lag +- 2), + * the following correlations are needed in each sub frame: + * + * sf1: lag range [-8,...,7] total 16 correlations + * sf2: lag range [-4,...,4] total 9 correlations + * sf3: lag range [-3,....4] total 8 correltions + * sf4: lag range [-6,....8] total 15 correlations + * + * In total 48 correlations. The direct implementation computed in worst + * case 4*12*5 = 240 correlations, but more likely around 120. + ***********************************************************************/ +static void silk_P_Ana_calc_corr_st3( + silk_float cross_corr_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ], /* O 3 DIM correlation array */ + const silk_float frame[], /* I vector to correlate */ + opus_int start_lag, /* I start lag */ + opus_int sf_length, /* I sub frame length */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity, /* I Complexity setting */ + int arch /* I Run-time architecture */ +) +{ + const silk_float *target_ptr; + opus_int i, j, k, lag_counter, lag_low, lag_high; + opus_int nb_cbk_search, delta, idx, cbk_size; + silk_float scratch_mem[ SCRATCH_SIZE ]; + opus_val32 xcorr[ SCRATCH_SIZE ]; + const opus_int8 *Lag_range_ptr, *Lag_CB_ptr; + + celt_assert( complexity >= SILK_PE_MIN_COMPLEX ); + celt_assert( complexity <= SILK_PE_MAX_COMPLEX ); + + if( nb_subfr == PE_MAX_NB_SUBFR ) { + Lag_range_ptr = &silk_Lag_range_stage3[ complexity ][ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3[ 0 ][ 0 ]; + nb_cbk_search = silk_nb_cbk_searchs_stage3[ complexity ]; + cbk_size = PE_NB_CBKS_STAGE3_MAX; + } else { + celt_assert( nb_subfr == PE_MAX_NB_SUBFR >> 1); + Lag_range_ptr = &silk_Lag_range_stage3_10_ms[ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; + nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; + cbk_size = PE_NB_CBKS_STAGE3_10MS; + } + + target_ptr = &frame[ silk_LSHIFT( sf_length, 2 ) ]; /* Pointer to middle of frame */ + for( k = 0; k < nb_subfr; k++ ) { + lag_counter = 0; + + /* Calculate the correlations for each subframe */ + lag_low = matrix_ptr( Lag_range_ptr, k, 0, 2 ); + lag_high = matrix_ptr( Lag_range_ptr, k, 1, 2 ); + silk_assert(lag_high-lag_low+1 <= SCRATCH_SIZE); + celt_pitch_xcorr( target_ptr, target_ptr - start_lag - lag_high, xcorr, sf_length, lag_high - lag_low + 1, arch ); + for( j = lag_low; j <= lag_high; j++ ) { + silk_assert( lag_counter < SCRATCH_SIZE ); + scratch_mem[ lag_counter ] = xcorr[ lag_high - j ]; + lag_counter++; + } + + delta = matrix_ptr( Lag_range_ptr, k, 0, 2 ); + for( i = 0; i < nb_cbk_search; i++ ) { + /* Fill out the 3 dim array that stores the correlations for */ + /* each code_book vector for each start lag */ + idx = matrix_ptr( Lag_CB_ptr, k, i, cbk_size ) - delta; + for( j = 0; j < PE_NB_STAGE3_LAGS; j++ ) { + silk_assert( idx + j < SCRATCH_SIZE ); + silk_assert( idx + j < lag_counter ); + cross_corr_st3[ k ][ i ][ j ] = scratch_mem[ idx + j ]; + } + } + target_ptr += sf_length; + } +} + +/********************************************************************/ +/* Calculate the energies for first two subframes. The energies are */ +/* calculated recursively. */ +/********************************************************************/ +static void silk_P_Ana_calc_energy_st3( + silk_float energies_st3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ][ PE_NB_STAGE3_LAGS ], /* O 3 DIM correlation array */ + const silk_float frame[], /* I vector to correlate */ + opus_int start_lag, /* I start lag */ + opus_int sf_length, /* I sub frame length */ + opus_int nb_subfr, /* I number of subframes */ + opus_int complexity /* I Complexity setting */ +) +{ + const silk_float *target_ptr, *basis_ptr; + double energy; + opus_int k, i, j, lag_counter; + opus_int nb_cbk_search, delta, idx, cbk_size, lag_diff; + silk_float scratch_mem[ SCRATCH_SIZE ]; + const opus_int8 *Lag_range_ptr, *Lag_CB_ptr; + + celt_assert( complexity >= SILK_PE_MIN_COMPLEX ); + celt_assert( complexity <= SILK_PE_MAX_COMPLEX ); + + if( nb_subfr == PE_MAX_NB_SUBFR ) { + Lag_range_ptr = &silk_Lag_range_stage3[ complexity ][ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3[ 0 ][ 0 ]; + nb_cbk_search = silk_nb_cbk_searchs_stage3[ complexity ]; + cbk_size = PE_NB_CBKS_STAGE3_MAX; + } else { + celt_assert( nb_subfr == PE_MAX_NB_SUBFR >> 1); + Lag_range_ptr = &silk_Lag_range_stage3_10_ms[ 0 ][ 0 ]; + Lag_CB_ptr = &silk_CB_lags_stage3_10_ms[ 0 ][ 0 ]; + nb_cbk_search = PE_NB_CBKS_STAGE3_10MS; + cbk_size = PE_NB_CBKS_STAGE3_10MS; + } + + target_ptr = &frame[ silk_LSHIFT( sf_length, 2 ) ]; + for( k = 0; k < nb_subfr; k++ ) { + lag_counter = 0; + + /* Calculate the energy for first lag */ + basis_ptr = target_ptr - ( start_lag + matrix_ptr( Lag_range_ptr, k, 0, 2 ) ); + energy = silk_energy_FLP( basis_ptr, sf_length ) + 1e-3; + silk_assert( energy >= 0.0 ); + scratch_mem[lag_counter] = (silk_float)energy; + lag_counter++; + + lag_diff = ( matrix_ptr( Lag_range_ptr, k, 1, 2 ) - matrix_ptr( Lag_range_ptr, k, 0, 2 ) + 1 ); + for( i = 1; i < lag_diff; i++ ) { + /* remove part outside new window */ + energy -= basis_ptr[sf_length - i] * (double)basis_ptr[sf_length - i]; + silk_assert( energy >= 0.0 ); + + /* add part that comes into window */ + energy += basis_ptr[ -i ] * (double)basis_ptr[ -i ]; + silk_assert( energy >= 0.0 ); + silk_assert( lag_counter < SCRATCH_SIZE ); + scratch_mem[lag_counter] = (silk_float)energy; + lag_counter++; + } + + delta = matrix_ptr( Lag_range_ptr, k, 0, 2 ); + for( i = 0; i < nb_cbk_search; i++ ) { + /* Fill out the 3 dim array that stores the correlations for */ + /* each code_book vector for each start lag */ + idx = matrix_ptr( Lag_CB_ptr, k, i, cbk_size ) - delta; + for( j = 0; j < PE_NB_STAGE3_LAGS; j++ ) { + silk_assert( idx + j < SCRATCH_SIZE ); + silk_assert( idx + j < lag_counter ); + energies_st3[ k ][ i ][ j ] = scratch_mem[ idx + j ]; + silk_assert( energies_st3[ k ][ i ][ j ] >= 0.0f ); + } + } + target_ptr += sf_length; + } +} diff --git a/lib/opus/silk/float/process_gains_FLP.c b/lib/opus/silk/float/process_gains_FLP.c new file mode 100644 index 00000000..c0da0dae --- /dev/null +++ b/lib/opus/silk/float/process_gains_FLP.c @@ -0,0 +1,103 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" +#include "tuning_parameters.h" + +/* Processing of gains */ +void silk_process_gains_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ +) +{ + silk_shape_state_FLP *psShapeSt = &psEnc->sShape; + opus_int k; + opus_int32 pGains_Q16[ MAX_NB_SUBFR ]; + silk_float s, InvMaxSqrVal, gain, quant_offset; + + /* Gain reduction when LTP coding gain is high */ + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + s = 1.0f - 0.5f * silk_sigmoid( 0.25f * ( psEncCtrl->LTPredCodGain - 12.0f ) ); + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->Gains[ k ] *= s; + } + } + + /* Limit the quantized signal */ + InvMaxSqrVal = ( silk_float )( pow( 2.0f, 0.33f * ( 21.0f - psEnc->sCmn.SNR_dB_Q7 * ( 1 / 128.0f ) ) ) / psEnc->sCmn.subfr_length ); + + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + /* Soft limit on ratio residual energy and squared gains */ + gain = psEncCtrl->Gains[ k ]; + gain = ( silk_float )sqrt( gain * gain + psEncCtrl->ResNrg[ k ] * InvMaxSqrVal ); + psEncCtrl->Gains[ k ] = silk_min_float( gain, 32767.0f ); + } + + /* Prepare gains for noise shaping quantization */ + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + pGains_Q16[ k ] = (opus_int32)( psEncCtrl->Gains[ k ] * 65536.0f ); + } + + /* Save unquantized gains and gain Index */ + silk_memcpy( psEncCtrl->GainsUnq_Q16, pGains_Q16, psEnc->sCmn.nb_subfr * sizeof( opus_int32 ) ); + psEncCtrl->lastGainIndexPrev = psShapeSt->LastGainIndex; + + /* Quantize gains */ + silk_gains_quant( psEnc->sCmn.indices.GainsIndices, pGains_Q16, + &psShapeSt->LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); + + /* Overwrite unquantized gains with quantized gains and convert back to Q0 from Q16 */ + for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { + psEncCtrl->Gains[ k ] = pGains_Q16[ k ] / 65536.0f; + } + + /* Set quantizer offset for voiced signals. Larger offset when LTP coding gain is low or tilt is high (ie low-pass) */ + if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { + if( psEncCtrl->LTPredCodGain + psEnc->sCmn.input_tilt_Q15 * ( 1.0f / 32768.0f ) > 1.0f ) { + psEnc->sCmn.indices.quantOffsetType = 0; + } else { + psEnc->sCmn.indices.quantOffsetType = 1; + } + } + + /* Quantizer boundary adjustment */ + quant_offset = silk_Quantization_Offsets_Q10[ psEnc->sCmn.indices.signalType >> 1 ][ psEnc->sCmn.indices.quantOffsetType ] / 1024.0f; + psEncCtrl->Lambda = LAMBDA_OFFSET + + LAMBDA_DELAYED_DECISIONS * psEnc->sCmn.nStatesDelayedDecision + + LAMBDA_SPEECH_ACT * psEnc->sCmn.speech_activity_Q8 * ( 1.0f / 256.0f ) + + LAMBDA_INPUT_QUALITY * psEncCtrl->input_quality + + LAMBDA_CODING_QUALITY * psEncCtrl->coding_quality + + LAMBDA_QUANT_OFFSET * quant_offset; + + silk_assert( psEncCtrl->Lambda > 0.0f ); + silk_assert( psEncCtrl->Lambda < 2.0f ); +} diff --git a/lib/opus/silk/float/regularize_correlations_FLP.c b/lib/opus/silk/float/regularize_correlations_FLP.c new file mode 100644 index 00000000..df461260 --- /dev/null +++ b/lib/opus/silk/float/regularize_correlations_FLP.c @@ -0,0 +1,48 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +/* Add noise to matrix diagonal */ +void silk_regularize_correlations_FLP( + silk_float *XX, /* I/O Correlation matrices */ + silk_float *xx, /* I/O Correlation values */ + const silk_float noise, /* I Noise energy to add */ + const opus_int D /* I Dimension of XX */ +) +{ + opus_int i; + + for( i = 0; i < D; i++ ) { + matrix_ptr( &XX[ 0 ], i, i, D ) += noise; + } + xx[ 0 ] += noise; +} diff --git a/lib/opus/silk/float/residual_energy_FLP.c b/lib/opus/silk/float/residual_energy_FLP.c new file mode 100644 index 00000000..1bd07b33 --- /dev/null +++ b/lib/opus/silk/float/residual_energy_FLP.c @@ -0,0 +1,117 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +#define MAX_ITERATIONS_RESIDUAL_NRG 10 +#define REGULARIZATION_FACTOR 1e-8f + +/* Residual energy: nrg = wxx - 2 * wXx * c + c' * wXX * c */ +silk_float silk_residual_energy_covar_FLP( /* O Weighted residual energy */ + const silk_float *c, /* I Filter coefficients */ + silk_float *wXX, /* I/O Weighted correlation matrix, reg. out */ + const silk_float *wXx, /* I Weighted correlation vector */ + const silk_float wxx, /* I Weighted correlation value */ + const opus_int D /* I Dimension */ +) +{ + opus_int i, j, k; + silk_float tmp, nrg = 0.0f, regularization; + + /* Safety checks */ + celt_assert( D >= 0 ); + + regularization = REGULARIZATION_FACTOR * ( wXX[ 0 ] + wXX[ D * D - 1 ] ); + for( k = 0; k < MAX_ITERATIONS_RESIDUAL_NRG; k++ ) { + nrg = wxx; + + tmp = 0.0f; + for( i = 0; i < D; i++ ) { + tmp += wXx[ i ] * c[ i ]; + } + nrg -= 2.0f * tmp; + + /* compute c' * wXX * c, assuming wXX is symmetric */ + for( i = 0; i < D; i++ ) { + tmp = 0.0f; + for( j = i + 1; j < D; j++ ) { + tmp += matrix_c_ptr( wXX, i, j, D ) * c[ j ]; + } + nrg += c[ i ] * ( 2.0f * tmp + matrix_c_ptr( wXX, i, i, D ) * c[ i ] ); + } + if( nrg > 0 ) { + break; + } else { + /* Add white noise */ + for( i = 0; i < D; i++ ) { + matrix_c_ptr( wXX, i, i, D ) += regularization; + } + /* Increase noise for next run */ + regularization *= 2.0f; + } + } + if( k == MAX_ITERATIONS_RESIDUAL_NRG ) { + silk_assert( nrg == 0 ); + nrg = 1.0f; + } + + return nrg; +} + +/* Calculates residual energies of input subframes where all subframes have LPC_order */ +/* of preceding samples */ +void silk_residual_energy_FLP( + silk_float nrgs[ MAX_NB_SUBFR ], /* O Residual energy per subframe */ + const silk_float x[], /* I Input signal */ + silk_float a[ 2 ][ MAX_LPC_ORDER ], /* I AR coefs for each frame half */ + const silk_float gains[], /* I Quantization gains */ + const opus_int subfr_length, /* I Subframe length */ + const opus_int nb_subfr, /* I number of subframes */ + const opus_int LPC_order /* I LPC order */ +) +{ + opus_int shift; + silk_float *LPC_res_ptr, LPC_res[ ( MAX_FRAME_LENGTH + MAX_NB_SUBFR * MAX_LPC_ORDER ) / 2 ]; + + LPC_res_ptr = LPC_res + LPC_order; + shift = LPC_order + subfr_length; + + /* Filter input to create the LPC residual for each frame half, and measure subframe energies */ + silk_LPC_analysis_filter_FLP( LPC_res, a[ 0 ], x + 0 * shift, 2 * shift, LPC_order ); + nrgs[ 0 ] = ( silk_float )( gains[ 0 ] * gains[ 0 ] * silk_energy_FLP( LPC_res_ptr + 0 * shift, subfr_length ) ); + nrgs[ 1 ] = ( silk_float )( gains[ 1 ] * gains[ 1 ] * silk_energy_FLP( LPC_res_ptr + 1 * shift, subfr_length ) ); + + if( nb_subfr == MAX_NB_SUBFR ) { + silk_LPC_analysis_filter_FLP( LPC_res, a[ 1 ], x + 2 * shift, 2 * shift, LPC_order ); + nrgs[ 2 ] = ( silk_float )( gains[ 2 ] * gains[ 2 ] * silk_energy_FLP( LPC_res_ptr + 0 * shift, subfr_length ) ); + nrgs[ 3 ] = ( silk_float )( gains[ 3 ] * gains[ 3 ] * silk_energy_FLP( LPC_res_ptr + 1 * shift, subfr_length ) ); + } +} diff --git a/lib/opus/silk/float/scale_copy_vector_FLP.c b/lib/opus/silk/float/scale_copy_vector_FLP.c new file mode 100644 index 00000000..20db32b3 --- /dev/null +++ b/lib/opus/silk/float/scale_copy_vector_FLP.c @@ -0,0 +1,57 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" + +/* copy and multiply a vector by a constant */ +void silk_scale_copy_vector_FLP( + silk_float *data_out, + const silk_float *data_in, + silk_float gain, + opus_int dataSize +) +{ + opus_int i, dataSize4; + + /* 4x unrolled loop */ + dataSize4 = dataSize & 0xFFFC; + for( i = 0; i < dataSize4; i += 4 ) { + data_out[ i + 0 ] = gain * data_in[ i + 0 ]; + data_out[ i + 1 ] = gain * data_in[ i + 1 ]; + data_out[ i + 2 ] = gain * data_in[ i + 2 ]; + data_out[ i + 3 ] = gain * data_in[ i + 3 ]; + } + + /* any remaining elements */ + for( ; i < dataSize; i++ ) { + data_out[ i ] = gain * data_in[ i ]; + } +} diff --git a/lib/opus/silk/float/scale_vector_FLP.c b/lib/opus/silk/float/scale_vector_FLP.c new file mode 100644 index 00000000..108fdcbe --- /dev/null +++ b/lib/opus/silk/float/scale_vector_FLP.c @@ -0,0 +1,56 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" + +/* multiply a vector by a constant */ +void silk_scale_vector_FLP( + silk_float *data1, + silk_float gain, + opus_int dataSize +) +{ + opus_int i, dataSize4; + + /* 4x unrolled loop */ + dataSize4 = dataSize & 0xFFFC; + for( i = 0; i < dataSize4; i += 4 ) { + data1[ i + 0 ] *= gain; + data1[ i + 1 ] *= gain; + data1[ i + 2 ] *= gain; + data1[ i + 3 ] *= gain; + } + + /* any remaining elements */ + for( ; i < dataSize; i++ ) { + data1[ i ] *= gain; + } +} diff --git a/lib/opus/silk/float/schur_FLP.c b/lib/opus/silk/float/schur_FLP.c new file mode 100644 index 00000000..8526c748 --- /dev/null +++ b/lib/opus/silk/float/schur_FLP.c @@ -0,0 +1,70 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FLP.h" + +silk_float silk_schur_FLP( /* O returns residual energy */ + silk_float refl_coef[], /* O reflection coefficients (length order) */ + const silk_float auto_corr[], /* I autocorrelation sequence (length order+1) */ + opus_int order /* I order */ +) +{ + opus_int k, n; + double C[ SILK_MAX_ORDER_LPC + 1 ][ 2 ]; + double Ctmp1, Ctmp2, rc_tmp; + + celt_assert( order >= 0 && order <= SILK_MAX_ORDER_LPC ); + + /* Copy correlations */ + k = 0; + do { + C[ k ][ 0 ] = C[ k ][ 1 ] = auto_corr[ k ]; + } while( ++k <= order ); + + for( k = 0; k < order; k++ ) { + /* Get reflection coefficient */ + rc_tmp = -C[ k + 1 ][ 0 ] / silk_max_float( C[ 0 ][ 1 ], 1e-9f ); + + /* Save the output */ + refl_coef[ k ] = (silk_float)rc_tmp; + + /* Update correlations */ + for( n = 0; n < order - k; n++ ) { + Ctmp1 = C[ n + k + 1 ][ 0 ]; + Ctmp2 = C[ n ][ 1 ]; + C[ n + k + 1 ][ 0 ] = Ctmp1 + Ctmp2 * rc_tmp; + C[ n ][ 1 ] = Ctmp2 + Ctmp1 * rc_tmp; + } + } + + /* Return residual energy */ + return (silk_float)C[ 0 ][ 1 ]; +} diff --git a/lib/opus/silk/float/sort_FLP.c b/lib/opus/silk/float/sort_FLP.c new file mode 100644 index 00000000..0e18f319 --- /dev/null +++ b/lib/opus/silk/float/sort_FLP.c @@ -0,0 +1,83 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* Insertion sort (fast for already almost sorted arrays): */ +/* Best case: O(n) for an already sorted array */ +/* Worst case: O(n^2) for an inversely sorted array */ + +#include "typedef.h" +#include "SigProc_FLP.h" + +void silk_insertion_sort_decreasing_FLP( + silk_float *a, /* I/O Unsorted / Sorted vector */ + opus_int *idx, /* O Index vector for the sorted elements */ + const opus_int L, /* I Vector length */ + const opus_int K /* I Number of correctly sorted positions */ +) +{ + silk_float value; + opus_int i, j; + + /* Safety checks */ + celt_assert( K > 0 ); + celt_assert( L > 0 ); + celt_assert( L >= K ); + + /* Write start indices in index vector */ + for( i = 0; i < K; i++ ) { + idx[ i ] = i; + } + + /* Sort vector elements by value, decreasing order */ + for( i = 1; i < K; i++ ) { + value = a[ i ]; + for( j = i - 1; ( j >= 0 ) && ( value > a[ j ] ); j-- ) { + a[ j + 1 ] = a[ j ]; /* Shift value */ + idx[ j + 1 ] = idx[ j ]; /* Shift index */ + } + a[ j + 1 ] = value; /* Write value */ + idx[ j + 1 ] = i; /* Write index */ + } + + /* If less than L values are asked check the remaining values, */ + /* but only spend CPU to ensure that the K first values are correct */ + for( i = K; i < L; i++ ) { + value = a[ i ]; + if( value > a[ K - 1 ] ) { + for( j = K - 2; ( j >= 0 ) && ( value > a[ j ] ); j-- ) { + a[ j + 1 ] = a[ j ]; /* Shift value */ + idx[ j + 1 ] = idx[ j ]; /* Shift index */ + } + a[ j + 1 ] = value; /* Write value */ + idx[ j + 1 ] = i; /* Write index */ + } + } +} diff --git a/lib/opus/silk/float/structs_FLP.h b/lib/opus/silk/float/structs_FLP.h new file mode 100644 index 00000000..3150b386 --- /dev/null +++ b/lib/opus/silk/float/structs_FLP.h @@ -0,0 +1,112 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_STRUCTS_FLP_H +#define SILK_STRUCTS_FLP_H + +#include "typedef.h" +#include "main.h" +#include "structs.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +/********************************/ +/* Noise shaping analysis state */ +/********************************/ +typedef struct { + opus_int8 LastGainIndex; + silk_float HarmShapeGain_smth; + silk_float Tilt_smth; +} silk_shape_state_FLP; + +/********************************/ +/* Encoder state FLP */ +/********************************/ +typedef struct { + silk_encoder_state sCmn; /* Common struct, shared with fixed-point code */ + silk_shape_state_FLP sShape; /* Noise shaping state */ + + /* Buffer for find pitch and noise shape analysis */ + silk_float x_buf[ 2 * MAX_FRAME_LENGTH + LA_SHAPE_MAX ];/* Buffer for find pitch and noise shape analysis */ + silk_float LTPCorr; /* Normalized correlation from pitch lag estimator */ +} silk_encoder_state_FLP; + +/************************/ +/* Encoder control FLP */ +/************************/ +typedef struct { + /* Prediction and coding parameters */ + silk_float Gains[ MAX_NB_SUBFR ]; + silk_float PredCoef[ 2 ][ MAX_LPC_ORDER ]; /* holds interpolated and final coefficients */ + silk_float LTPCoef[LTP_ORDER * MAX_NB_SUBFR]; + silk_float LTP_scale; + opus_int pitchL[ MAX_NB_SUBFR ]; + + /* Noise shaping parameters */ + silk_float AR[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ]; + silk_float LF_MA_shp[ MAX_NB_SUBFR ]; + silk_float LF_AR_shp[ MAX_NB_SUBFR ]; + silk_float Tilt[ MAX_NB_SUBFR ]; + silk_float HarmShapeGain[ MAX_NB_SUBFR ]; + silk_float Lambda; + silk_float input_quality; + silk_float coding_quality; + + /* Measures */ + silk_float predGain; + silk_float LTPredCodGain; + silk_float ResNrg[ MAX_NB_SUBFR ]; /* Residual energy per subframe */ + + /* Parameters for CBR mode */ + opus_int32 GainsUnq_Q16[ MAX_NB_SUBFR ]; + opus_int8 lastGainIndexPrev; +} silk_encoder_control_FLP; + +/************************/ +/* Encoder Super Struct */ +/************************/ +typedef struct { + silk_encoder_state_FLP state_Fxx[ ENCODER_NUM_CHANNELS ]; + stereo_enc_state sStereo; + opus_int32 nBitsUsedLBRR; + opus_int32 nBitsExceeded; + opus_int nChannelsAPI; + opus_int nChannelsInternal; + opus_int nPrevChannelsInternal; + opus_int timeSinceSwitchAllowed_ms; + opus_int allowBandwidthSwitch; + opus_int prev_decode_only_middle; +} silk_encoder; + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/float/warped_autocorrelation_FLP.c b/lib/opus/silk/float/warped_autocorrelation_FLP.c new file mode 100644 index 00000000..09186e73 --- /dev/null +++ b/lib/opus/silk/float/warped_autocorrelation_FLP.c @@ -0,0 +1,73 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +/* Autocorrelations for a warped frequency axis */ +void silk_warped_autocorrelation_FLP( + silk_float *corr, /* O Result [order + 1] */ + const silk_float *input, /* I Input data to correlate */ + const silk_float warping, /* I Warping coefficient */ + const opus_int length, /* I Length of input */ + const opus_int order /* I Correlation order (even) */ +) +{ + opus_int n, i; + double tmp1, tmp2; + double state[ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; + double C[ MAX_SHAPE_LPC_ORDER + 1 ] = { 0 }; + + /* Order must be even */ + celt_assert( ( order & 1 ) == 0 ); + + /* Loop over samples */ + for( n = 0; n < length; n++ ) { + tmp1 = input[ n ]; + /* Loop over allpass sections */ + for( i = 0; i < order; i += 2 ) { + /* Output of allpass section */ + tmp2 = state[ i ] + warping * ( state[ i + 1 ] - tmp1 ); + state[ i ] = tmp1; + C[ i ] += state[ 0 ] * tmp1; + /* Output of allpass section */ + tmp1 = state[ i + 1 ] + warping * ( state[ i + 2 ] - tmp2 ); + state[ i + 1 ] = tmp2; + C[ i + 1 ] += state[ 0 ] * tmp2; + } + state[ order ] = tmp1; + C[ order ] += state[ 0 ] * tmp1; + } + + /* Copy correlations in silk_float output format */ + for( i = 0; i < order + 1; i++ ) { + corr[ i ] = ( silk_float )C[ i ]; + } +} diff --git a/lib/opus/silk/float/wrappers_FLP.c b/lib/opus/silk/float/wrappers_FLP.c new file mode 100644 index 00000000..c0c183e3 --- /dev/null +++ b/lib/opus/silk/float/wrappers_FLP.c @@ -0,0 +1,209 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main_FLP.h" + +/* Wrappers. Calls flp / fix code */ + +/* Convert AR filter coefficients to NLSF parameters */ +void silk_A2NLSF_FLP( + opus_int16 *NLSF_Q15, /* O NLSF vector [ LPC_order ] */ + const silk_float *pAR, /* I LPC coefficients [ LPC_order ] */ + const opus_int LPC_order /* I LPC order */ +) +{ + opus_int i; + opus_int32 a_fix_Q16[ MAX_LPC_ORDER ]; + + for( i = 0; i < LPC_order; i++ ) { + a_fix_Q16[ i ] = silk_float2int( pAR[ i ] * 65536.0f ); + } + + silk_A2NLSF( NLSF_Q15, a_fix_Q16, LPC_order ); +} + +/* Convert LSF parameters to AR prediction filter coefficients */ +void silk_NLSF2A_FLP( + silk_float *pAR, /* O LPC coefficients [ LPC_order ] */ + const opus_int16 *NLSF_Q15, /* I NLSF vector [ LPC_order ] */ + const opus_int LPC_order, /* I LPC order */ + int arch /* I Run-time architecture */ +) +{ + opus_int i; + opus_int16 a_fix_Q12[ MAX_LPC_ORDER ]; + + silk_NLSF2A( a_fix_Q12, NLSF_Q15, LPC_order, arch ); + + for( i = 0; i < LPC_order; i++ ) { + pAR[ i ] = ( silk_float )a_fix_Q12[ i ] * ( 1.0f / 4096.0f ); + } +} + +/******************************************/ +/* Floating-point NLSF processing wrapper */ +/******************************************/ +void silk_process_NLSFs_FLP( + silk_encoder_state *psEncC, /* I/O Encoder state */ + silk_float PredCoef[ 2 ][ MAX_LPC_ORDER ], /* O Prediction coefficients */ + opus_int16 NLSF_Q15[ MAX_LPC_ORDER ], /* I/O Normalized LSFs (quant out) (0 - (2^15-1)) */ + const opus_int16 prev_NLSF_Q15[ MAX_LPC_ORDER ] /* I Previous Normalized LSFs (0 - (2^15-1)) */ +) +{ + opus_int i, j; + opus_int16 PredCoef_Q12[ 2 ][ MAX_LPC_ORDER ]; + + silk_process_NLSFs( psEncC, PredCoef_Q12, NLSF_Q15, prev_NLSF_Q15); + + for( j = 0; j < 2; j++ ) { + for( i = 0; i < psEncC->predictLPCOrder; i++ ) { + PredCoef[ j ][ i ] = ( silk_float )PredCoef_Q12[ j ][ i ] * ( 1.0f / 4096.0f ); + } + } +} + +/****************************************/ +/* Floating-point Silk NSQ wrapper */ +/****************************************/ +void silk_NSQ_wrapper_FLP( + silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + SideInfoIndices *psIndices, /* I/O Quantization indices */ + silk_nsq_state *psNSQ, /* I/O Noise Shaping Quantzation state */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const silk_float x[] /* I Prefiltered input signal */ +) +{ + opus_int i, j; + opus_int16 x16[ MAX_FRAME_LENGTH ]; + opus_int32 Gains_Q16[ MAX_NB_SUBFR ]; + silk_DWORD_ALIGN opus_int16 PredCoef_Q12[ 2 ][ MAX_LPC_ORDER ]; + opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ]; + opus_int LTP_scale_Q14; + + /* Noise shaping parameters */ + opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ]; + opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ]; /* Packs two int16 coefficients per int32 value */ + opus_int Lambda_Q10; + opus_int Tilt_Q14[ MAX_NB_SUBFR ]; + opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ]; + + /* Convert control struct to fix control struct */ + /* Noise shape parameters */ + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + for( j = 0; j < psEnc->sCmn.shapingLPCOrder; j++ ) { + AR_Q13[ i * MAX_SHAPE_LPC_ORDER + j ] = silk_float2int( psEncCtrl->AR[ i * MAX_SHAPE_LPC_ORDER + j ] * 8192.0f ); + } + } + + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + LF_shp_Q14[ i ] = silk_LSHIFT32( silk_float2int( psEncCtrl->LF_AR_shp[ i ] * 16384.0f ), 16 ) | + (opus_uint16)silk_float2int( psEncCtrl->LF_MA_shp[ i ] * 16384.0f ); + Tilt_Q14[ i ] = (opus_int)silk_float2int( psEncCtrl->Tilt[ i ] * 16384.0f ); + HarmShapeGain_Q14[ i ] = (opus_int)silk_float2int( psEncCtrl->HarmShapeGain[ i ] * 16384.0f ); + } + Lambda_Q10 = ( opus_int )silk_float2int( psEncCtrl->Lambda * 1024.0f ); + + /* prediction and coding parameters */ + for( i = 0; i < psEnc->sCmn.nb_subfr * LTP_ORDER; i++ ) { + LTPCoef_Q14[ i ] = (opus_int16)silk_float2int( psEncCtrl->LTPCoef[ i ] * 16384.0f ); + } + + for( j = 0; j < 2; j++ ) { + for( i = 0; i < psEnc->sCmn.predictLPCOrder; i++ ) { + PredCoef_Q12[ j ][ i ] = (opus_int16)silk_float2int( psEncCtrl->PredCoef[ j ][ i ] * 4096.0f ); + } + } + + for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) { + Gains_Q16[ i ] = silk_float2int( psEncCtrl->Gains[ i ] * 65536.0f ); + silk_assert( Gains_Q16[ i ] > 0 ); + } + + if( psIndices->signalType == TYPE_VOICED ) { + LTP_scale_Q14 = silk_LTPScales_table_Q14[ psIndices->LTP_scaleIndex ]; + } else { + LTP_scale_Q14 = 0; + } + + /* Convert input to fix */ + for( i = 0; i < psEnc->sCmn.frame_length; i++ ) { + x16[ i ] = silk_float2int( x[ i ] ); + } + + /* Call NSQ */ + if( psEnc->sCmn.nStatesDelayedDecision > 1 || psEnc->sCmn.warping_Q16 > 0 ) { + silk_NSQ_del_dec( &psEnc->sCmn, psNSQ, psIndices, x16, pulses, PredCoef_Q12[ 0 ], LTPCoef_Q14, + AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, psEncCtrl->pitchL, Lambda_Q10, LTP_scale_Q14, psEnc->sCmn.arch ); + } else { + silk_NSQ( &psEnc->sCmn, psNSQ, psIndices, x16, pulses, PredCoef_Q12[ 0 ], LTPCoef_Q14, + AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, psEncCtrl->pitchL, Lambda_Q10, LTP_scale_Q14, psEnc->sCmn.arch ); + } +} + +/***********************************************/ +/* Floating-point Silk LTP quantiation wrapper */ +/***********************************************/ +void silk_quant_LTP_gains_FLP( + silk_float B[ MAX_NB_SUBFR * LTP_ORDER ], /* O Quantized LTP gains */ + opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook index */ + opus_int8 *periodicity_index, /* O Periodicity index */ + opus_int32 *sum_log_gain_Q7, /* I/O Cumulative max prediction gain */ + silk_float *pred_gain_dB, /* O LTP prediction gain */ + const silk_float XX[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ], /* I Correlation matrix */ + const silk_float xX[ MAX_NB_SUBFR * LTP_ORDER ], /* I Correlation vector */ + const opus_int subfr_len, /* I Number of samples per subframe */ + const opus_int nb_subfr, /* I Number of subframes */ + int arch /* I Run-time architecture */ +) +{ + opus_int i, pred_gain_dB_Q7; + opus_int16 B_Q14[ MAX_NB_SUBFR * LTP_ORDER ]; + opus_int32 XX_Q17[ MAX_NB_SUBFR * LTP_ORDER * LTP_ORDER ]; + opus_int32 xX_Q17[ MAX_NB_SUBFR * LTP_ORDER ]; + + i = 0; + do { + XX_Q17[ i ] = (opus_int32)silk_float2int( XX[ i ] * 131072.0f ); + } while ( ++i < nb_subfr * LTP_ORDER * LTP_ORDER ); + i = 0; + do { + xX_Q17[ i ] = (opus_int32)silk_float2int( xX[ i ] * 131072.0f ); + } while ( ++i < nb_subfr * LTP_ORDER ); + + silk_quant_LTP_gains( B_Q14, cbk_index, periodicity_index, sum_log_gain_Q7, &pred_gain_dB_Q7, XX_Q17, xX_Q17, subfr_len, nb_subfr, arch ); + + for( i = 0; i < nb_subfr * LTP_ORDER; i++ ) { + B[ i ] = (silk_float)B_Q14[ i ] * ( 1.0f / 16384.0f ); + } + + *pred_gain_dB = (silk_float)pred_gain_dB_Q7 * ( 1.0f / 128.0f ); +} diff --git a/lib/opus/silk/gain_quant.c b/lib/opus/silk/gain_quant.c new file mode 100644 index 00000000..ee65245a --- /dev/null +++ b/lib/opus/silk/gain_quant.c @@ -0,0 +1,142 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +#define OFFSET ( ( MIN_QGAIN_DB * 128 ) / 6 + 16 * 128 ) +#define SCALE_Q16 ( ( 65536 * ( N_LEVELS_QGAIN - 1 ) ) / ( ( ( MAX_QGAIN_DB - MIN_QGAIN_DB ) * 128 ) / 6 ) ) +#define INV_SCALE_Q16 ( ( 65536 * ( ( ( MAX_QGAIN_DB - MIN_QGAIN_DB ) * 128 ) / 6 ) ) / ( N_LEVELS_QGAIN - 1 ) ) + +/* Gain scalar quantization with hysteresis, uniform on log scale */ +void silk_gains_quant( + opus_int8 ind[ MAX_NB_SUBFR ], /* O gain indices */ + opus_int32 gain_Q16[ MAX_NB_SUBFR ], /* I/O gains (quantized out) */ + opus_int8 *prev_ind, /* I/O last index in previous frame */ + const opus_int conditional, /* I first gain is delta coded if 1 */ + const opus_int nb_subfr /* I number of subframes */ +) +{ + opus_int k, double_step_size_threshold; + + for( k = 0; k < nb_subfr; k++ ) { + /* Convert to log scale, scale, floor() */ + ind[ k ] = silk_SMULWB( SCALE_Q16, silk_lin2log( gain_Q16[ k ] ) - OFFSET ); + + /* Round towards previous quantized gain (hysteresis) */ + if( ind[ k ] < *prev_ind ) { + ind[ k ]++; + } + ind[ k ] = silk_LIMIT_int( ind[ k ], 0, N_LEVELS_QGAIN - 1 ); + + /* Compute delta indices and limit */ + if( k == 0 && conditional == 0 ) { + /* Full index */ + ind[ k ] = silk_LIMIT_int( ind[ k ], *prev_ind + MIN_DELTA_GAIN_QUANT, N_LEVELS_QGAIN - 1 ); + *prev_ind = ind[ k ]; + } else { + /* Delta index */ + ind[ k ] = ind[ k ] - *prev_ind; + + /* Double the quantization step size for large gain increases, so that the max gain level can be reached */ + double_step_size_threshold = 2 * MAX_DELTA_GAIN_QUANT - N_LEVELS_QGAIN + *prev_ind; + if( ind[ k ] > double_step_size_threshold ) { + ind[ k ] = double_step_size_threshold + silk_RSHIFT( ind[ k ] - double_step_size_threshold + 1, 1 ); + } + + ind[ k ] = silk_LIMIT_int( ind[ k ], MIN_DELTA_GAIN_QUANT, MAX_DELTA_GAIN_QUANT ); + + /* Accumulate deltas */ + if( ind[ k ] > double_step_size_threshold ) { + *prev_ind += silk_LSHIFT( ind[ k ], 1 ) - double_step_size_threshold; + *prev_ind = silk_min_int( *prev_ind, N_LEVELS_QGAIN - 1 ); + } else { + *prev_ind += ind[ k ]; + } + + /* Shift to make non-negative */ + ind[ k ] -= MIN_DELTA_GAIN_QUANT; + } + + /* Scale and convert to linear scale */ + gain_Q16[ k ] = silk_log2lin( silk_min_32( silk_SMULWB( INV_SCALE_Q16, *prev_ind ) + OFFSET, 3967 ) ); /* 3967 = 31 in Q7 */ + } +} + +/* Gains scalar dequantization, uniform on log scale */ +void silk_gains_dequant( + opus_int32 gain_Q16[ MAX_NB_SUBFR ], /* O quantized gains */ + const opus_int8 ind[ MAX_NB_SUBFR ], /* I gain indices */ + opus_int8 *prev_ind, /* I/O last index in previous frame */ + const opus_int conditional, /* I first gain is delta coded if 1 */ + const opus_int nb_subfr /* I number of subframes */ +) +{ + opus_int k, ind_tmp, double_step_size_threshold; + + for( k = 0; k < nb_subfr; k++ ) { + if( k == 0 && conditional == 0 ) { + /* Gain index is not allowed to go down more than 16 steps (~21.8 dB) */ + *prev_ind = silk_max_int( ind[ k ], *prev_ind - 16 ); + } else { + /* Delta index */ + ind_tmp = ind[ k ] + MIN_DELTA_GAIN_QUANT; + + /* Accumulate deltas */ + double_step_size_threshold = 2 * MAX_DELTA_GAIN_QUANT - N_LEVELS_QGAIN + *prev_ind; + if( ind_tmp > double_step_size_threshold ) { + *prev_ind += silk_LSHIFT( ind_tmp, 1 ) - double_step_size_threshold; + } else { + *prev_ind += ind_tmp; + } + } + *prev_ind = silk_LIMIT_int( *prev_ind, 0, N_LEVELS_QGAIN - 1 ); + + /* Scale and convert to linear scale */ + gain_Q16[ k ] = silk_log2lin( silk_min_32( silk_SMULWB( INV_SCALE_Q16, *prev_ind ) + OFFSET, 3967 ) ); /* 3967 = 31 in Q7 */ + } +} + +/* Compute unique identifier of gain indices vector */ +opus_int32 silk_gains_ID( /* O returns unique identifier of gains */ + const opus_int8 ind[ MAX_NB_SUBFR ], /* I gain indices */ + const opus_int nb_subfr /* I number of subframes */ +) +{ + opus_int k; + opus_int32 gainsID; + + gainsID = 0; + for( k = 0; k < nb_subfr; k++ ) { + gainsID = silk_ADD_LSHIFT32( ind[ k ], gainsID, 8 ); + } + + return gainsID; +} diff --git a/lib/opus/silk/init_decoder.c b/lib/opus/silk/init_decoder.c new file mode 100644 index 00000000..16c03dcd --- /dev/null +++ b/lib/opus/silk/init_decoder.c @@ -0,0 +1,57 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/************************/ +/* Init Decoder State */ +/************************/ +opus_int silk_init_decoder( + silk_decoder_state *psDec /* I/O Decoder state pointer */ +) +{ + /* Clear the entire encoder state, except anything copied */ + silk_memset( psDec, 0, sizeof( silk_decoder_state ) ); + + /* Used to deactivate LSF interpolation */ + psDec->first_frame_after_reset = 1; + psDec->prev_gain_Q16 = 65536; + psDec->arch = opus_select_arch(); + + /* Reset CNG state */ + silk_CNG_Reset( psDec ); + + /* Reset PLC state */ + silk_PLC_Reset( psDec ); + + return(0); +} + diff --git a/lib/opus/silk/init_encoder.c b/lib/opus/silk/init_encoder.c new file mode 100644 index 00000000..65995c33 --- /dev/null +++ b/lib/opus/silk/init_encoder.c @@ -0,0 +1,64 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif +#ifdef FIXED_POINT +#include "main_FIX.h" +#else +#include "main_FLP.h" +#endif +#include "tuning_parameters.h" +#include "cpu_support.h" + +/*********************************/ +/* Initialize Silk Encoder state */ +/*********************************/ +opus_int silk_init_encoder( + silk_encoder_state_Fxx *psEnc, /* I/O Pointer to Silk FIX encoder state */ + int arch /* I Run-time architecture */ +) +{ + opus_int ret = 0; + + /* Clear the entire encoder state */ + silk_memset( psEnc, 0, sizeof( silk_encoder_state_Fxx ) ); + + psEnc->sCmn.arch = arch; + + psEnc->sCmn.variable_HP_smth1_Q15 = silk_LSHIFT( silk_lin2log( SILK_FIX_CONST( VARIABLE_HP_MIN_CUTOFF_HZ, 16 ) ) - ( 16 << 7 ), 8 ); + psEnc->sCmn.variable_HP_smth2_Q15 = psEnc->sCmn.variable_HP_smth1_Q15; + + /* Used to deactivate LSF interpolation, pitch prediction */ + psEnc->sCmn.first_frame_after_reset = 1; + + /* Initialize Silk VAD */ + ret += silk_VAD_Init( &psEnc->sCmn.sVAD ); + + return ret; +} diff --git a/lib/opus/silk/inner_prod_aligned.c b/lib/opus/silk/inner_prod_aligned.c new file mode 100644 index 00000000..257ae9e0 --- /dev/null +++ b/lib/opus/silk/inner_prod_aligned.c @@ -0,0 +1,47 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +opus_int32 silk_inner_prod_aligned_scale( + const opus_int16 *const inVec1, /* I input vector 1 */ + const opus_int16 *const inVec2, /* I input vector 2 */ + const opus_int scale, /* I number of bits to shift */ + const opus_int len /* I vector lengths */ +) +{ + opus_int i; + opus_int32 sum = 0; + for( i = 0; i < len; i++ ) { + sum = silk_ADD_RSHIFT32( sum, silk_SMULBB( inVec1[ i ], inVec2[ i ] ), scale ); + } + return sum; +} diff --git a/lib/opus/silk/interpolate.c b/lib/opus/silk/interpolate.c new file mode 100644 index 00000000..833c28ef --- /dev/null +++ b/lib/opus/silk/interpolate.c @@ -0,0 +1,51 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Interpolate two vectors */ +void silk_interpolate( + opus_int16 xi[ MAX_LPC_ORDER ], /* O interpolated vector */ + const opus_int16 x0[ MAX_LPC_ORDER ], /* I first vector */ + const opus_int16 x1[ MAX_LPC_ORDER ], /* I second vector */ + const opus_int ifact_Q2, /* I interp. factor, weight on 2nd vector */ + const opus_int d /* I number of parameters */ +) +{ + opus_int i; + + celt_assert( ifact_Q2 >= 0 ); + celt_assert( ifact_Q2 <= 4 ); + + for( i = 0; i < d; i++ ) { + xi[ i ] = (opus_int16)silk_ADD_RSHIFT( x0[ i ], silk_SMULBB( x1[ i ] - x0[ i ], ifact_Q2 ), 2 ); + } +} diff --git a/lib/opus/silk/lin2log.c b/lib/opus/silk/lin2log.c new file mode 100644 index 00000000..0d5155aa --- /dev/null +++ b/lib/opus/silk/lin2log.c @@ -0,0 +1,46 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +/* Approximation of 128 * log2() (very close inverse of silk_log2lin()) */ +/* Convert input to a log scale */ +opus_int32 silk_lin2log( + const opus_int32 inLin /* I input in linear scale */ +) +{ + opus_int32 lz, frac_Q7; + + silk_CLZ_FRAC( inLin, &lz, &frac_Q7 ); + + /* Piece-wise parabolic approximation */ + return silk_ADD_LSHIFT32( silk_SMLAWB( frac_Q7, silk_MUL( frac_Q7, 128 - frac_Q7 ), 179 ), 31 - lz, 7 ); +} + diff --git a/lib/opus/silk/log2lin.c b/lib/opus/silk/log2lin.c new file mode 100644 index 00000000..b7c48e47 --- /dev/null +++ b/lib/opus/silk/log2lin.c @@ -0,0 +1,58 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Approximation of 2^() (very close inverse of silk_lin2log()) */ +/* Convert input to a linear scale */ +opus_int32 silk_log2lin( + const opus_int32 inLog_Q7 /* I input on log scale */ +) +{ + opus_int32 out, frac_Q7; + + if( inLog_Q7 < 0 ) { + return 0; + } else if ( inLog_Q7 >= 3967 ) { + return silk_int32_MAX; + } + + out = silk_LSHIFT( 1, silk_RSHIFT( inLog_Q7, 7 ) ); + frac_Q7 = inLog_Q7 & 0x7F; + if( inLog_Q7 < 2048 ) { + /* Piece-wise parabolic approximation */ + out = silk_ADD_RSHIFT32( out, silk_MUL( out, silk_SMLAWB( frac_Q7, silk_SMULBB( frac_Q7, 128 - frac_Q7 ), -174 ) ), 7 ); + } else { + /* Piece-wise parabolic approximation */ + out = silk_MLA( out, silk_RSHIFT( out, 7 ), silk_SMLAWB( frac_Q7, silk_SMULBB( frac_Q7, 128 - frac_Q7 ), -174 ) ); + } + return out; +} diff --git a/lib/opus/silk/macros.h b/lib/opus/silk/macros.h new file mode 100644 index 00000000..3c67b6e5 --- /dev/null +++ b/lib/opus/silk/macros.h @@ -0,0 +1,151 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MACROS_H +#define SILK_MACROS_H + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_types.h" +#include "opus_defines.h" +#include "arch.h" + +/* This is an OPUS_INLINE header file for general platform. */ + +/* (a32 * (opus_int32)((opus_int16)(b32))) >> 16 output have to be 32bit int */ +#if OPUS_FAST_INT64 +#define silk_SMULWB(a32, b32) ((opus_int32)(((a32) * (opus_int64)((opus_int16)(b32))) >> 16)) +#else +#define silk_SMULWB(a32, b32) ((((a32) >> 16) * (opus_int32)((opus_int16)(b32))) + ((((a32) & 0x0000FFFF) * (opus_int32)((opus_int16)(b32))) >> 16)) +#endif + +/* a32 + (b32 * (opus_int32)((opus_int16)(c32))) >> 16 output have to be 32bit int */ +#if OPUS_FAST_INT64 +#define silk_SMLAWB(a32, b32, c32) ((opus_int32)((a32) + (((b32) * (opus_int64)((opus_int16)(c32))) >> 16))) +#else +#define silk_SMLAWB(a32, b32, c32) ((a32) + ((((b32) >> 16) * (opus_int32)((opus_int16)(c32))) + ((((b32) & 0x0000FFFF) * (opus_int32)((opus_int16)(c32))) >> 16))) +#endif + +/* (a32 * (b32 >> 16)) >> 16 */ +#if OPUS_FAST_INT64 +#define silk_SMULWT(a32, b32) ((opus_int32)(((a32) * (opus_int64)((b32) >> 16)) >> 16)) +#else +#define silk_SMULWT(a32, b32) (((a32) >> 16) * ((b32) >> 16) + ((((a32) & 0x0000FFFF) * ((b32) >> 16)) >> 16)) +#endif + +/* a32 + (b32 * (c32 >> 16)) >> 16 */ +#if OPUS_FAST_INT64 +#define silk_SMLAWT(a32, b32, c32) ((opus_int32)((a32) + (((b32) * ((opus_int64)(c32) >> 16)) >> 16))) +#else +#define silk_SMLAWT(a32, b32, c32) ((a32) + (((b32) >> 16) * ((c32) >> 16)) + ((((b32) & 0x0000FFFF) * ((c32) >> 16)) >> 16)) +#endif + +/* (opus_int32)((opus_int16)(a3))) * (opus_int32)((opus_int16)(b32)) output have to be 32bit int */ +#define silk_SMULBB(a32, b32) ((opus_int32)((opus_int16)(a32)) * (opus_int32)((opus_int16)(b32))) + +/* a32 + (opus_int32)((opus_int16)(b32)) * (opus_int32)((opus_int16)(c32)) output have to be 32bit int */ +#define silk_SMLABB(a32, b32, c32) ((a32) + ((opus_int32)((opus_int16)(b32))) * (opus_int32)((opus_int16)(c32))) + +/* (opus_int32)((opus_int16)(a32)) * (b32 >> 16) */ +#define silk_SMULBT(a32, b32) ((opus_int32)((opus_int16)(a32)) * ((b32) >> 16)) + +/* a32 + (opus_int32)((opus_int16)(b32)) * (c32 >> 16) */ +#define silk_SMLABT(a32, b32, c32) ((a32) + ((opus_int32)((opus_int16)(b32))) * ((c32) >> 16)) + +/* a64 + (b32 * c32) */ +#define silk_SMLAL(a64, b32, c32) (silk_ADD64((a64), ((opus_int64)(b32) * (opus_int64)(c32)))) + +/* (a32 * b32) >> 16 */ +#if OPUS_FAST_INT64 +#define silk_SMULWW(a32, b32) ((opus_int32)(((opus_int64)(a32) * (b32)) >> 16)) +#else +#define silk_SMULWW(a32, b32) silk_MLA(silk_SMULWB((a32), (b32)), (a32), silk_RSHIFT_ROUND((b32), 16)) +#endif + +/* a32 + ((b32 * c32) >> 16) */ +#if OPUS_FAST_INT64 +#define silk_SMLAWW(a32, b32, c32) ((opus_int32)((a32) + (((opus_int64)(b32) * (c32)) >> 16))) +#else +#define silk_SMLAWW(a32, b32, c32) silk_MLA(silk_SMLAWB((a32), (b32), (c32)), (b32), silk_RSHIFT_ROUND((c32), 16)) +#endif + +/* add/subtract with output saturated */ +#define silk_ADD_SAT32(a, b) ((((opus_uint32)(a) + (opus_uint32)(b)) & 0x80000000) == 0 ? \ + ((((a) & (b)) & 0x80000000) != 0 ? silk_int32_MIN : (a)+(b)) : \ + ((((a) | (b)) & 0x80000000) == 0 ? silk_int32_MAX : (a)+(b)) ) + +#define silk_SUB_SAT32(a, b) ((((opus_uint32)(a)-(opus_uint32)(b)) & 0x80000000) == 0 ? \ + (( (a) & ((b)^0x80000000) & 0x80000000) ? silk_int32_MIN : (a)-(b)) : \ + ((((a)^0x80000000) & (b) & 0x80000000) ? silk_int32_MAX : (a)-(b)) ) + +#if defined(MIPSr1_ASM) +#include "mips/macros_mipsr1.h" +#endif + +#include "ecintrin.h" +#ifndef OVERRIDE_silk_CLZ16 +static OPUS_INLINE opus_int32 silk_CLZ16(opus_int16 in16) +{ + return 32 - EC_ILOG(in16<<16|0x8000); +} +#endif + +#ifndef OVERRIDE_silk_CLZ32 +static OPUS_INLINE opus_int32 silk_CLZ32(opus_int32 in32) +{ + return in32 ? 32 - EC_ILOG(in32) : 32; +} +#endif + +/* Row based */ +#define matrix_ptr(Matrix_base_adr, row, column, N) \ + (*((Matrix_base_adr) + ((row)*(N)+(column)))) +#define matrix_adr(Matrix_base_adr, row, column, N) \ + ((Matrix_base_adr) + ((row)*(N)+(column))) + +/* Column based */ +#ifndef matrix_c_ptr +# define matrix_c_ptr(Matrix_base_adr, row, column, M) \ + (*((Matrix_base_adr) + ((row)+(M)*(column)))) +#endif + +#ifdef OPUS_ARM_INLINE_ASM +#include "arm/macros_armv4.h" +#endif + +#ifdef OPUS_ARM_INLINE_EDSP +#include "arm/macros_armv5e.h" +#endif + +#ifdef OPUS_ARM_PRESUME_AARCH64_NEON_INTR +#include "arm/macros_arm64.h" +#endif + +#endif /* SILK_MACROS_H */ + diff --git a/lib/opus/silk/main.h b/lib/opus/silk/main.h new file mode 100644 index 00000000..a5f56875 --- /dev/null +++ b/lib/opus/silk/main.h @@ -0,0 +1,476 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_MAIN_H +#define SILK_MAIN_H + +#include "SigProc_FIX.h" +#include "define.h" +#include "structs.h" +#include "tables.h" +#include "PLC.h" +#include "control.h" +#include "debug.h" +#include "entenc.h" +#include "entdec.h" + +#if defined(OPUS_X86_MAY_HAVE_SSE4_1) +#include "x86/main_sse.h" +#endif + +#if (defined(OPUS_ARM_ASM) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)) +#include "arm/NSQ_del_dec_arm.h" +#endif + +/* Convert Left/Right stereo signal to adaptive Mid/Side representation */ +void silk_stereo_LR_to_MS( + stereo_enc_state *state, /* I/O State */ + opus_int16 x1[], /* I/O Left input signal, becomes mid signal */ + opus_int16 x2[], /* I/O Right input signal, becomes side signal */ + opus_int8 ix[ 2 ][ 3 ], /* O Quantization indices */ + opus_int8 *mid_only_flag, /* O Flag: only mid signal coded */ + opus_int32 mid_side_rates_bps[], /* O Bitrates for mid and side signals */ + opus_int32 total_rate_bps, /* I Total bitrate */ + opus_int prev_speech_act_Q8, /* I Speech activity level in previous frame */ + opus_int toMono, /* I Last frame before a stereo->mono transition */ + opus_int fs_kHz, /* I Sample rate (kHz) */ + opus_int frame_length /* I Number of samples */ +); + +/* Convert adaptive Mid/Side representation to Left/Right stereo signal */ +void silk_stereo_MS_to_LR( + stereo_dec_state *state, /* I/O State */ + opus_int16 x1[], /* I/O Left input signal, becomes mid signal */ + opus_int16 x2[], /* I/O Right input signal, becomes side signal */ + const opus_int32 pred_Q13[], /* I Predictors */ + opus_int fs_kHz, /* I Samples rate (kHz) */ + opus_int frame_length /* I Number of samples */ +); + +/* Find least-squares prediction gain for one signal based on another and quantize it */ +opus_int32 silk_stereo_find_predictor( /* O Returns predictor in Q13 */ + opus_int32 *ratio_Q14, /* O Ratio of residual and mid energies */ + const opus_int16 x[], /* I Basis signal */ + const opus_int16 y[], /* I Target signal */ + opus_int32 mid_res_amp_Q0[], /* I/O Smoothed mid, residual norms */ + opus_int length, /* I Number of samples */ + opus_int smooth_coef_Q16 /* I Smoothing coefficient */ +); + +/* Quantize mid/side predictors */ +void silk_stereo_quant_pred( + opus_int32 pred_Q13[], /* I/O Predictors (out: quantized) */ + opus_int8 ix[ 2 ][ 3 ] /* O Quantization indices */ +); + +/* Entropy code the mid/side quantization indices */ +void silk_stereo_encode_pred( + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int8 ix[ 2 ][ 3 ] /* I Quantization indices */ +); + +/* Entropy code the mid-only flag */ +void silk_stereo_encode_mid_only( + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int8 mid_only_flag +); + +/* Decode mid/side predictors */ +void silk_stereo_decode_pred( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int32 pred_Q13[] /* O Predictors */ +); + +/* Decode mid-only flag */ +void silk_stereo_decode_mid_only( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int *decode_only_mid /* O Flag that only mid channel has been coded */ +); + +/* Encodes signs of excitation */ +void silk_encode_signs( + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + const opus_int8 pulses[], /* I pulse signal */ + opus_int length, /* I length of input */ + const opus_int signalType, /* I Signal type */ + const opus_int quantOffsetType, /* I Quantization offset type */ + const opus_int sum_pulses[ MAX_NB_SHELL_BLOCKS ] /* I Sum of absolute pulses per block */ +); + +/* Decodes signs of excitation */ +void silk_decode_signs( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 pulses[], /* I/O pulse signal */ + opus_int length, /* I length of input */ + const opus_int signalType, /* I Signal type */ + const opus_int quantOffsetType, /* I Quantization offset type */ + const opus_int sum_pulses[ MAX_NB_SHELL_BLOCKS ] /* I Sum of absolute pulses per block */ +); + +/* Check encoder control struct */ +opus_int check_control_input( + silk_EncControlStruct *encControl /* I Control structure */ +); + +/* Control internal sampling rate */ +opus_int silk_control_audio_bandwidth( + silk_encoder_state *psEncC, /* I/O Pointer to Silk encoder state */ + silk_EncControlStruct *encControl /* I Control structure */ +); + +/* Control SNR of redidual quantizer */ +opus_int silk_control_SNR( + silk_encoder_state *psEncC, /* I/O Pointer to Silk encoder state */ + opus_int32 TargetRate_bps /* I Target max bitrate (bps) */ +); + +/***************/ +/* Shell coder */ +/***************/ + +/* Encode quantization indices of excitation */ +void silk_encode_pulses( + ec_enc *psRangeEnc, /* I/O compressor data structure */ + const opus_int signalType, /* I Signal type */ + const opus_int quantOffsetType, /* I quantOffsetType */ + opus_int8 pulses[], /* I quantization indices */ + const opus_int frame_length /* I Frame length */ +); + +/* Shell encoder, operates on one shell code frame of 16 pulses */ +void silk_shell_encoder( + ec_enc *psRangeEnc, /* I/O compressor data structure */ + const opus_int *pulses0 /* I data: nonnegative pulse amplitudes */ +); + +/* Shell decoder, operates on one shell code frame of 16 pulses */ +void silk_shell_decoder( + opus_int16 *pulses0, /* O data: nonnegative pulse amplitudes */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + const opus_int pulses4 /* I number of pulses per pulse-subframe */ +); + +/* Gain scalar quantization with hysteresis, uniform on log scale */ +void silk_gains_quant( + opus_int8 ind[ MAX_NB_SUBFR ], /* O gain indices */ + opus_int32 gain_Q16[ MAX_NB_SUBFR ], /* I/O gains (quantized out) */ + opus_int8 *prev_ind, /* I/O last index in previous frame */ + const opus_int conditional, /* I first gain is delta coded if 1 */ + const opus_int nb_subfr /* I number of subframes */ +); + +/* Gains scalar dequantization, uniform on log scale */ +void silk_gains_dequant( + opus_int32 gain_Q16[ MAX_NB_SUBFR ], /* O quantized gains */ + const opus_int8 ind[ MAX_NB_SUBFR ], /* I gain indices */ + opus_int8 *prev_ind, /* I/O last index in previous frame */ + const opus_int conditional, /* I first gain is delta coded if 1 */ + const opus_int nb_subfr /* I number of subframes */ +); + +/* Compute unique identifier of gain indices vector */ +opus_int32 silk_gains_ID( /* O returns unique identifier of gains */ + const opus_int8 ind[ MAX_NB_SUBFR ], /* I gain indices */ + const opus_int nb_subfr /* I number of subframes */ +); + +/* Interpolate two vectors */ +void silk_interpolate( + opus_int16 xi[ MAX_LPC_ORDER ], /* O interpolated vector */ + const opus_int16 x0[ MAX_LPC_ORDER ], /* I first vector */ + const opus_int16 x1[ MAX_LPC_ORDER ], /* I second vector */ + const opus_int ifact_Q2, /* I interp. factor, weight on 2nd vector */ + const opus_int d /* I number of parameters */ +); + +/* LTP tap quantizer */ +void silk_quant_LTP_gains( + opus_int16 B_Q14[ MAX_NB_SUBFR * LTP_ORDER ], /* O Quantized LTP gains */ + opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook Index */ + opus_int8 *periodicity_index, /* O Periodicity Index */ + opus_int32 *sum_gain_dB_Q7, /* I/O Cumulative max prediction gain */ + opus_int *pred_gain_dB_Q7, /* O LTP prediction gain */ + const opus_int32 XX_Q17[ MAX_NB_SUBFR*LTP_ORDER*LTP_ORDER ], /* I Correlation matrix in Q18 */ + const opus_int32 xX_Q17[ MAX_NB_SUBFR*LTP_ORDER ], /* I Correlation vector in Q18 */ + const opus_int subfr_len, /* I Number of samples per subframe */ + const opus_int nb_subfr, /* I Number of subframes */ + int arch /* I Run-time architecture */ +); + +/* Entropy constrained matrix-weighted VQ, for a single input data vector */ +void silk_VQ_WMat_EC_c( + opus_int8 *ind, /* O index of best codebook vector */ + opus_int32 *res_nrg_Q15, /* O best residual energy */ + opus_int32 *rate_dist_Q8, /* O best total bitrate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ + const opus_int32 *XX_Q17, /* I correlation matrix */ + const opus_int32 *xX_Q17, /* I correlation vector */ + const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ + const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ + const opus_int subfr_len, /* I number of samples per subframe */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ + const opus_int L /* I number of vectors in codebook */ +); + +#if !defined(OVERRIDE_silk_VQ_WMat_EC) +#define silk_VQ_WMat_EC(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, subfr_len, max_gain_Q7, L, arch) \ + ((void)(arch),silk_VQ_WMat_EC_c(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, subfr_len, max_gain_Q7, L)) +#endif + +/************************************/ +/* Noise shaping quantization (NSQ) */ +/************************************/ + +void silk_NSQ_c( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +); + +#if !defined(OVERRIDE_silk_NSQ) +#define silk_NSQ(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14, arch) \ + ((void)(arch),silk_NSQ_c(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14)) +#endif + +/* Noise shaping using delayed decision */ +void silk_NSQ_del_dec_c( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +); + +#if !defined(OVERRIDE_silk_NSQ_del_dec) +#define silk_NSQ_del_dec(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14, arch) \ + ((void)(arch),silk_NSQ_del_dec_c(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14)) +#endif + +/************/ +/* Silk VAD */ +/************/ +/* Initialize the Silk VAD */ +opus_int silk_VAD_Init( /* O Return value, 0 if success */ + silk_VAD_state *psSilk_VAD /* I/O Pointer to Silk VAD state */ +); + +/* Get speech activity level in Q8 */ +opus_int silk_VAD_GetSA_Q8_c( /* O Return value, 0 if success */ + silk_encoder_state *psEncC, /* I/O Encoder state */ + const opus_int16 pIn[] /* I PCM input */ +); + +#if !defined(OVERRIDE_silk_VAD_GetSA_Q8) +#define silk_VAD_GetSA_Q8(psEnC, pIn, arch) ((void)(arch),silk_VAD_GetSA_Q8_c(psEnC, pIn)) +#endif + +/* Low-pass filter with variable cutoff frequency based on */ +/* piece-wise linear interpolation between elliptic filters */ +/* Start by setting transition_frame_no = 1; */ +void silk_LP_variable_cutoff( + silk_LP_state *psLP, /* I/O LP filter state */ + opus_int16 *frame, /* I/O Low-pass filtered output signal */ + const opus_int frame_length /* I Frame length */ +); + +/******************/ +/* NLSF Quantizer */ +/******************/ +/* Limit, stabilize, convert and quantize NLSFs */ +void silk_process_NLSFs( + silk_encoder_state *psEncC, /* I/O Encoder state */ + opus_int16 PredCoef_Q12[ 2 ][ MAX_LPC_ORDER ], /* O Prediction coefficients */ + opus_int16 pNLSF_Q15[ MAX_LPC_ORDER ], /* I/O Normalized LSFs (quant out) (0 - (2^15-1)) */ + const opus_int16 prev_NLSFq_Q15[ MAX_LPC_ORDER ] /* I Previous Normalized LSFs (0 - (2^15-1)) */ +); + +opus_int32 silk_NLSF_encode( /* O Returns RD value in Q25 */ + opus_int8 *NLSFIndices, /* I Codebook path vector [ LPC_ORDER + 1 ] */ + opus_int16 *pNLSF_Q15, /* I/O Quantized NLSF vector [ LPC_ORDER ] */ + const silk_NLSF_CB_struct *psNLSF_CB, /* I Codebook object */ + const opus_int16 *pW_QW, /* I NLSF weight vector [ LPC_ORDER ] */ + const opus_int NLSF_mu_Q20, /* I Rate weight for the RD optimization */ + const opus_int nSurvivors, /* I Max survivors after first stage */ + const opus_int signalType /* I Signal type: 0/1/2 */ +); + +/* Compute quantization errors for an LPC_order element input vector for a VQ codebook */ +void silk_NLSF_VQ( + opus_int32 err_Q26[], /* O Quantization errors [K] */ + const opus_int16 in_Q15[], /* I Input vectors to be quantized [LPC_order] */ + const opus_uint8 pCB_Q8[], /* I Codebook vectors [K*LPC_order] */ + const opus_int16 pWght_Q9[], /* I Codebook weights [K*LPC_order] */ + const opus_int K, /* I Number of codebook vectors */ + const opus_int LPC_order /* I Number of LPCs */ +); + +/* Delayed-decision quantizer for NLSF residuals */ +opus_int32 silk_NLSF_del_dec_quant( /* O Returns RD value in Q25 */ + opus_int8 indices[], /* O Quantization indices [ order ] */ + const opus_int16 x_Q10[], /* I Input [ order ] */ + const opus_int16 w_Q5[], /* I Weights [ order ] */ + const opus_uint8 pred_coef_Q8[], /* I Backward predictor coefs [ order ] */ + const opus_int16 ec_ix[], /* I Indices to entropy coding tables [ order ] */ + const opus_uint8 ec_rates_Q5[], /* I Rates [] */ + const opus_int quant_step_size_Q16, /* I Quantization step size */ + const opus_int16 inv_quant_step_size_Q6, /* I Inverse quantization step size */ + const opus_int32 mu_Q20, /* I R/D tradeoff */ + const opus_int16 order /* I Number of input values */ +); + +/* Unpack predictor values and indices for entropy coding tables */ +void silk_NLSF_unpack( + opus_int16 ec_ix[], /* O Indices to entropy tables [ LPC_ORDER ] */ + opus_uint8 pred_Q8[], /* O LSF predictor [ LPC_ORDER ] */ + const silk_NLSF_CB_struct *psNLSF_CB, /* I Codebook object */ + const opus_int CB1_index /* I Index of vector in first LSF codebook */ +); + +/***********************/ +/* NLSF vector decoder */ +/***********************/ +void silk_NLSF_decode( + opus_int16 *pNLSF_Q15, /* O Quantized NLSF vector [ LPC_ORDER ] */ + opus_int8 *NLSFIndices, /* I Codebook path vector [ LPC_ORDER + 1 ] */ + const silk_NLSF_CB_struct *psNLSF_CB /* I Codebook object */ +); + +/****************************************************/ +/* Decoder Functions */ +/****************************************************/ +opus_int silk_init_decoder( + silk_decoder_state *psDec /* I/O Decoder state pointer */ +); + +/* Set decoder sampling rate */ +opus_int silk_decoder_set_fs( + silk_decoder_state *psDec, /* I/O Decoder state pointer */ + opus_int fs_kHz, /* I Sampling frequency (kHz) */ + opus_int32 fs_API_Hz /* I API Sampling frequency (Hz) */ +); + +/****************/ +/* Decode frame */ +/****************/ +opus_int silk_decode_frame( + silk_decoder_state *psDec, /* I/O Pointer to Silk decoder state */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 pOut[], /* O Pointer to output speech frame */ + opus_int32 *pN, /* O Pointer to size of output frame */ + opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int condCoding, /* I The type of conditional coding to use */ + int arch /* I Run-time architecture */ +); + +/* Decode indices from bitstream */ +void silk_decode_indices( + silk_decoder_state *psDec, /* I/O State */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int FrameIndex, /* I Frame number */ + opus_int decode_LBRR, /* I Flag indicating LBRR data is being decoded */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/* Decode parameters from payload */ +void silk_decode_parameters( + silk_decoder_state *psDec, /* I/O State */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +/* Core decoder. Performs inverse NSQ operation LTP + LPC */ +void silk_decode_core( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I Decoder control */ + opus_int16 xq[], /* O Decoded speech */ + const opus_int16 pulses[ MAX_FRAME_LENGTH ], /* I Pulse signal */ + int arch /* I Run-time architecture */ +); + +/* Decode quantization indices of excitation (Shell coding) */ +void silk_decode_pulses( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int16 pulses[], /* O Excitation signal */ + const opus_int signalType, /* I Sigtype */ + const opus_int quantOffsetType, /* I quantOffsetType */ + const opus_int frame_length /* I Frame length */ +); + +/******************/ +/* CNG */ +/******************/ + +/* Reset CNG */ +void silk_CNG_Reset( + silk_decoder_state *psDec /* I/O Decoder state */ +); + +/* Updates CNG estimate, and applies the CNG when packet was lost */ +void silk_CNG( + silk_decoder_state *psDec, /* I/O Decoder state */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int16 frame[], /* I/O Signal */ + opus_int length /* I Length of residual */ +); + +/* Encoding of various parameters */ +void silk_encode_indices( + silk_encoder_state *psEncC, /* I/O Encoder state */ + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int FrameIndex, /* I Frame number */ + opus_int encode_LBRR, /* I Flag indicating LBRR data is being encoded */ + opus_int condCoding /* I The type of conditional coding to use */ +); + +#endif diff --git a/lib/opus/silk/meson.build b/lib/opus/silk/meson.build new file mode 100644 index 00000000..917048b2 --- /dev/null +++ b/lib/opus/silk/meson.build @@ -0,0 +1,63 @@ +silk_sources = sources['SILK_SOURCES'] + +silk_sources_sse4_1 = sources['SILK_SOURCES_SSE4_1'] + +silk_sources_neon_intr = sources['SILK_SOURCES_ARM_NEON_INTR'] + +silk_sources_fixed_neon_intr = sources['SILK_SOURCES_FIXED_ARM_NEON_INTR'] + +silk_sources_fixed = sources['SILK_SOURCES_FIXED'] + +silk_sources_fixed_sse4_1 = sources['SILK_SOURCES_FIXED_SSE4_1'] + +silk_sources_float = sources['SILK_SOURCES_FLOAT'] + +if opt_fixed_point + silk_sources += silk_sources_fixed +else + silk_sources += silk_sources_float +endif + +silk_includes = [opus_includes, include_directories('float', 'fixed')] +silk_static_libs = [] + +if host_cpu_family in ['x86', 'x86_64'] and opus_conf.has('OPUS_HAVE_RTCD') + silk_sources += sources['SILK_SOURCES_X86_RTCD'] +endif + +if host_cpu_family in ['arm', 'aarch64'] and have_arm_intrinsics_or_asm + if opus_conf.has('OPUS_HAVE_RTCD') + silk_sources += sources['SILK_SOURCES_ARM_RTCD'] + endif +endif + +foreach intr_name : ['sse4_1', 'neon_intr'] + have_intr = get_variable('have_' + intr_name) + if not have_intr + continue + endif + + intr_sources = get_variable('silk_sources_' + intr_name) + if opt_fixed_point + intr_sources += get_variable('silk_sources_fixed_' + intr_name) + endif + + intr_args = get_variable('opus_@0@_args'.format(intr_name), []) + silk_static_libs += static_library('silk_' + intr_name, intr_sources, + c_args: intr_args, + include_directories: silk_includes, + install: false) +endforeach + +silk_c_args = [] +if host_machine.system() == 'windows' + silk_c_args += ['-DDLL_EXPORT'] +endif + +silk_lib = static_library('opus-silk', + silk_sources, + c_args: silk_c_args, + include_directories: silk_includes, + link_whole: silk_static_libs, + dependencies: libm, + install: false) diff --git a/lib/opus/silk/mips/NSQ_del_dec_mipsr1.h b/lib/opus/silk/mips/NSQ_del_dec_mipsr1.h new file mode 100644 index 00000000..cd70713a --- /dev/null +++ b/lib/opus/silk/mips/NSQ_del_dec_mipsr1.h @@ -0,0 +1,410 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef __NSQ_DEL_DEC_MIPSR1_H__ +#define __NSQ_DEL_DEC_MIPSR1_H__ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" + +#define OVERRIDE_silk_noise_shape_quantizer_del_dec +static inline void silk_noise_shape_quantizer_del_dec( + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP filter state */ + opus_int32 delayedGain_Q10[], /* I/O Gain delay buffer */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int subfr, /* I Subframe number */ + opus_int shapingLPCOrder, /* I Shaping LPC filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + opus_int warping_Q16, /* I */ + opus_int nStatesDelayedDecision, /* I Number of states in decision tree */ + opus_int *smpl_buf_idx, /* I/O Index to newest samples in buffers */ + opus_int decisionDelay, /* I */ + int arch /* I */ +) +{ + opus_int i, j, k, Winner_ind, RDmin_ind, RDmax_ind, last_smple_idx; + opus_int32 Winner_rand_state; + opus_int32 LTP_pred_Q14, LPC_pred_Q14, n_AR_Q14, n_LTP_Q14; + opus_int32 n_LF_Q14, r_Q10, rr_Q10, rd1_Q10, rd2_Q10, RDmin_Q10, RDmax_Q10; + opus_int32 q1_Q0, q1_Q10, q2_Q10, exc_Q14, LPC_exc_Q14, xq_Q14, Gain_Q10; + opus_int32 tmp1, tmp2, sLF_AR_shp_Q14; + opus_int32 *pred_lag_ptr, *shp_lag_ptr, *psLPC_Q14; + NSQ_sample_struct psSampleState[ MAX_DEL_DEC_STATES ][ 2 ]; + NSQ_del_dec_struct *psDD; + NSQ_sample_struct *psSS; + opus_int16 b_Q14_0, b_Q14_1, b_Q14_2, b_Q14_3, b_Q14_4; + opus_int16 a_Q12_0, a_Q12_1, a_Q12_2, a_Q12_3, a_Q12_4, a_Q12_5, a_Q12_6; + opus_int16 a_Q12_7, a_Q12_8, a_Q12_9, a_Q12_10, a_Q12_11, a_Q12_12, a_Q12_13; + opus_int16 a_Q12_14, a_Q12_15; + + opus_int32 cur, prev, next; + + /*Unused.*/ + (void)arch; + + //Intialize b_Q14 variables + b_Q14_0 = b_Q14[ 0 ]; + b_Q14_1 = b_Q14[ 1 ]; + b_Q14_2 = b_Q14[ 2 ]; + b_Q14_3 = b_Q14[ 3 ]; + b_Q14_4 = b_Q14[ 4 ]; + + //Intialize a_Q12 variables + a_Q12_0 = a_Q12[0]; + a_Q12_1 = a_Q12[1]; + a_Q12_2 = a_Q12[2]; + a_Q12_3 = a_Q12[3]; + a_Q12_4 = a_Q12[4]; + a_Q12_5 = a_Q12[5]; + a_Q12_6 = a_Q12[6]; + a_Q12_7 = a_Q12[7]; + a_Q12_8 = a_Q12[8]; + a_Q12_9 = a_Q12[9]; + a_Q12_10 = a_Q12[10]; + a_Q12_11 = a_Q12[11]; + a_Q12_12 = a_Q12[12]; + a_Q12_13 = a_Q12[13]; + a_Q12_14 = a_Q12[14]; + a_Q12_15 = a_Q12[15]; + + long long temp64; + + silk_assert( nStatesDelayedDecision > 0 ); + + shp_lag_ptr = &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - lag + HARM_SHAPE_FIR_TAPS / 2 ]; + pred_lag_ptr = &sLTP_Q15[ NSQ->sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + Gain_Q10 = silk_RSHIFT( Gain_Q16, 6 ); + + for( i = 0; i < length; i++ ) { + /* Perform common calculations used in all states */ + + /* Long-term prediction */ + if( signalType == TYPE_VOICED ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + temp64 = __builtin_mips_mult(pred_lag_ptr[ 0 ], b_Q14_0 ); + temp64 = __builtin_mips_madd( temp64, pred_lag_ptr[ -1 ], b_Q14_1 ); + temp64 = __builtin_mips_madd( temp64, pred_lag_ptr[ -2 ], b_Q14_2 ); + temp64 = __builtin_mips_madd( temp64, pred_lag_ptr[ -3 ], b_Q14_3 ); + temp64 = __builtin_mips_madd( temp64, pred_lag_ptr[ -4 ], b_Q14_4 ); + temp64 += 32768; + LTP_pred_Q14 = __builtin_mips_extr_w(temp64, 16); + LTP_pred_Q14 = silk_LSHIFT( LTP_pred_Q14, 1 ); /* Q13 -> Q14 */ + pred_lag_ptr++; + } else { + LTP_pred_Q14 = 0; + } + + /* Long-term shaping */ + if( lag > 0 ) { + /* Symmetric, packed FIR coefficients */ + n_LTP_Q14 = silk_SMULWB( silk_ADD32( shp_lag_ptr[ 0 ], shp_lag_ptr[ -2 ] ), HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SMLAWT( n_LTP_Q14, shp_lag_ptr[ -1 ], HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SUB_LSHIFT32( LTP_pred_Q14, n_LTP_Q14, 2 ); /* Q12 -> Q14 */ + shp_lag_ptr++; + } else { + n_LTP_Q14 = 0; + } + + for( k = 0; k < nStatesDelayedDecision; k++ ) { + /* Delayed decision state */ + psDD = &psDelDec[ k ]; + + /* Sample state */ + psSS = psSampleState[ k ]; + + /* Generate dither */ + psDD->Seed = silk_RAND( psDD->Seed ); + + /* Pointer used in short term prediction and shaping */ + psLPC_Q14 = &psDD->sLPC_Q14[ NSQ_LPC_BUF_LENGTH - 1 + i ]; + /* Short-term prediction */ + silk_assert( predictLPCOrder == 10 || predictLPCOrder == 16 ); + temp64 = __builtin_mips_mult(psLPC_Q14[ 0 ], a_Q12_0 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -1 ], a_Q12_1 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -2 ], a_Q12_2 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -3 ], a_Q12_3 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -4 ], a_Q12_4 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -5 ], a_Q12_5 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -6 ], a_Q12_6 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -7 ], a_Q12_7 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -8 ], a_Q12_8 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -9 ], a_Q12_9 ); + if( predictLPCOrder == 16 ) { + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -10 ], a_Q12_10 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -11 ], a_Q12_11 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -12 ], a_Q12_12 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -13 ], a_Q12_13 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -14 ], a_Q12_14 ); + temp64 = __builtin_mips_madd( temp64, psLPC_Q14[ -15 ], a_Q12_15 ); + } + temp64 += 32768; + LPC_pred_Q14 = __builtin_mips_extr_w(temp64, 16); + + LPC_pred_Q14 = silk_LSHIFT( LPC_pred_Q14, 4 ); /* Q10 -> Q14 */ + + /* Noise shape feedback */ + silk_assert( ( shapingLPCOrder & 1 ) == 0 ); /* check that order is even */ + /* Output of lowpass section */ + tmp2 = silk_SMLAWB( psLPC_Q14[ 0 ], psDD->sAR2_Q14[ 0 ], warping_Q16 ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( psDD->sAR2_Q14[ 0 ], psDD->sAR2_Q14[ 1 ] - tmp2, warping_Q16 ); + psDD->sAR2_Q14[ 0 ] = tmp2; + + temp64 = __builtin_mips_mult(tmp2, AR_shp_Q13[ 0 ] ); + + prev = psDD->sAR2_Q14[ 1 ]; + + /* Loop over allpass sections */ + for( j = 2; j < shapingLPCOrder; j += 2 ) { + cur = psDD->sAR2_Q14[ j ]; + next = psDD->sAR2_Q14[ j+1 ]; + /* Output of allpass section */ + tmp2 = silk_SMLAWB( prev, cur - tmp1, warping_Q16 ); + psDD->sAR2_Q14[ j - 1 ] = tmp1; + temp64 = __builtin_mips_madd( temp64, tmp1, AR_shp_Q13[ j - 1 ] ); + temp64 = __builtin_mips_madd( temp64, tmp2, AR_shp_Q13[ j ] ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( cur, next - tmp2, warping_Q16 ); + psDD->sAR2_Q14[ j + 0 ] = tmp2; + prev = next; + } + psDD->sAR2_Q14[ shapingLPCOrder - 1 ] = tmp1; + temp64 = __builtin_mips_madd( temp64, tmp1, AR_shp_Q13[ shapingLPCOrder - 1 ] ); + temp64 += 32768; + n_AR_Q14 = __builtin_mips_extr_w(temp64, 16); + n_AR_Q14 = silk_LSHIFT( n_AR_Q14, 1 ); /* Q11 -> Q12 */ + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, psDD->LF_AR_Q14, Tilt_Q14 ); /* Q12 */ + n_AR_Q14 = silk_LSHIFT( n_AR_Q14, 2 ); /* Q12 -> Q14 */ + + n_LF_Q14 = silk_SMULWB( psDD->Shape_Q14[ *smpl_buf_idx ], LF_shp_Q14 ); /* Q12 */ + n_LF_Q14 = silk_SMLAWT( n_LF_Q14, psDD->LF_AR_Q14, LF_shp_Q14 ); /* Q12 */ + n_LF_Q14 = silk_LSHIFT( n_LF_Q14, 2 ); /* Q12 -> Q14 */ + + /* Input minus prediction plus noise feedback */ + /* r = x[ i ] - LTP_pred - LPC_pred + n_AR + n_Tilt + n_LF + n_LTP */ + tmp1 = silk_ADD32( n_AR_Q14, n_LF_Q14 ); /* Q14 */ + tmp2 = silk_ADD32( n_LTP_Q14, LPC_pred_Q14 ); /* Q13 */ + tmp1 = silk_SUB32( tmp2, tmp1 ); /* Q13 */ + tmp1 = silk_RSHIFT_ROUND( tmp1, 4 ); /* Q10 */ + + r_Q10 = silk_SUB32( x_Q10[ i ], tmp1 ); /* residual error Q10 */ + + /* Flip sign depending on dither */ + if ( psDD->Seed < 0 ) { + r_Q10 = -r_Q10; + } + r_Q10 = silk_LIMIT_32( r_Q10, -(31 << 10), 30 << 10 ); + + /* Find two quantization level candidates and measure their rate-distortion */ + q1_Q10 = silk_SUB32( r_Q10, offset_Q10 ); + q1_Q0 = silk_RSHIFT( q1_Q10, 10 ); + if( q1_Q0 > 0 ) { + q1_Q10 = silk_SUB32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q10 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == 0 ) { + q1_Q10 = offset_Q10; + q2_Q10 = silk_ADD32( q1_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q10 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == -1 ) { + q2_Q10 = offset_Q10; + q1_Q10 = silk_SUB32( q2_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q10 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else { /* q1_Q0 < -1 */ + q1_Q10 = silk_ADD32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q10 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( -q2_Q10, Lambda_Q10 ); + } + rr_Q10 = silk_SUB32( r_Q10, q1_Q10 ); + rd1_Q10 = silk_RSHIFT( silk_SMLABB( rd1_Q10, rr_Q10, rr_Q10 ), 10 ); + rr_Q10 = silk_SUB32( r_Q10, q2_Q10 ); + rd2_Q10 = silk_RSHIFT( silk_SMLABB( rd2_Q10, rr_Q10, rr_Q10 ), 10 ); + + if( rd1_Q10 < rd2_Q10 ) { + psSS[ 0 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd1_Q10 ); + psSS[ 1 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd2_Q10 ); + psSS[ 0 ].Q_Q10 = q1_Q10; + psSS[ 1 ].Q_Q10 = q2_Q10; + } else { + psSS[ 0 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd2_Q10 ); + psSS[ 1 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd1_Q10 ); + psSS[ 0 ].Q_Q10 = q2_Q10; + psSS[ 1 ].Q_Q10 = q1_Q10; + } + + /* Update states for best quantization */ + + /* Quantized excitation */ + exc_Q14 = silk_LSHIFT32( psSS[ 0 ].Q_Q10, 4 ); + if ( psDD->Seed < 0 ) { + exc_Q14 = -exc_Q14; + } + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD32( exc_Q14, LTP_pred_Q14 ); + xq_Q14 = silk_ADD32( LPC_exc_Q14, LPC_pred_Q14 ); + + /* Update states */ + sLF_AR_shp_Q14 = silk_SUB32( xq_Q14, n_AR_Q14 ); + psSS[ 0 ].sLTP_shp_Q14 = silk_SUB32( sLF_AR_shp_Q14, n_LF_Q14 ); + psSS[ 0 ].LF_AR_Q14 = sLF_AR_shp_Q14; + psSS[ 0 ].LPC_exc_Q14 = LPC_exc_Q14; + psSS[ 0 ].xq_Q14 = xq_Q14; + + /* Update states for second best quantization */ + + /* Quantized excitation */ + exc_Q14 = silk_LSHIFT32( psSS[ 1 ].Q_Q10, 4 ); + if ( psDD->Seed < 0 ) { + exc_Q14 = -exc_Q14; + } + + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD32( exc_Q14, LTP_pred_Q14 ); + xq_Q14 = silk_ADD32( LPC_exc_Q14, LPC_pred_Q14 ); + + /* Update states */ + sLF_AR_shp_Q14 = silk_SUB32( xq_Q14, n_AR_Q14 ); + psSS[ 1 ].sLTP_shp_Q14 = silk_SUB32( sLF_AR_shp_Q14, n_LF_Q14 ); + psSS[ 1 ].LF_AR_Q14 = sLF_AR_shp_Q14; + psSS[ 1 ].LPC_exc_Q14 = LPC_exc_Q14; + psSS[ 1 ].xq_Q14 = xq_Q14; + } + + *smpl_buf_idx = ( *smpl_buf_idx - 1 ) % DECISION_DELAY; + if( *smpl_buf_idx < 0 ) *smpl_buf_idx += DECISION_DELAY; + last_smple_idx = ( *smpl_buf_idx + decisionDelay ) % DECISION_DELAY; + + /* Find winner */ + RDmin_Q10 = psSampleState[ 0 ][ 0 ].RD_Q10; + Winner_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + if( psSampleState[ k ][ 0 ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ k ][ 0 ].RD_Q10; + Winner_ind = k; + } + } + + /* Increase RD values of expired states */ + Winner_rand_state = psDelDec[ Winner_ind ].RandState[ last_smple_idx ]; + for( k = 0; k < nStatesDelayedDecision; k++ ) { + if( psDelDec[ k ].RandState[ last_smple_idx ] != Winner_rand_state ) { + psSampleState[ k ][ 0 ].RD_Q10 = silk_ADD32( psSampleState[ k ][ 0 ].RD_Q10, silk_int32_MAX >> 4 ); + psSampleState[ k ][ 1 ].RD_Q10 = silk_ADD32( psSampleState[ k ][ 1 ].RD_Q10, silk_int32_MAX >> 4 ); + silk_assert( psSampleState[ k ][ 0 ].RD_Q10 >= 0 ); + } + } + + /* Find worst in first set and best in second set */ + RDmax_Q10 = psSampleState[ 0 ][ 0 ].RD_Q10; + RDmin_Q10 = psSampleState[ 0 ][ 1 ].RD_Q10; + RDmax_ind = 0; + RDmin_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + /* find worst in first set */ + if( psSampleState[ k ][ 0 ].RD_Q10 > RDmax_Q10 ) { + RDmax_Q10 = psSampleState[ k ][ 0 ].RD_Q10; + RDmax_ind = k; + } + /* find best in second set */ + if( psSampleState[ k ][ 1 ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ k ][ 1 ].RD_Q10; + RDmin_ind = k; + } + } + + /* Replace a state if best from second set outperforms worst in first set */ + if( RDmin_Q10 < RDmax_Q10 ) { + silk_memcpy( ( (opus_int32 *)&psDelDec[ RDmax_ind ] ) + i, + ( (opus_int32 *)&psDelDec[ RDmin_ind ] ) + i, sizeof( NSQ_del_dec_struct ) - i * sizeof( opus_int32) ); + silk_memcpy( &psSampleState[ RDmax_ind ][ 0 ], &psSampleState[ RDmin_ind ][ 1 ], sizeof( NSQ_sample_struct ) ); + } + + /* Write samples from winner to output and long-term filter states */ + psDD = &psDelDec[ Winner_ind ]; + if( subfr > 0 || i >= decisionDelay ) { + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDD->Q_Q10[ last_smple_idx ], 10 ); + xq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDD->Xq_Q14[ last_smple_idx ], delayedGain_Q10[ last_smple_idx ] ), 8 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay ] = psDD->Shape_Q14[ last_smple_idx ]; + sLTP_Q15[ NSQ->sLTP_buf_idx - decisionDelay ] = psDD->Pred_Q15[ last_smple_idx ]; + } + NSQ->sLTP_shp_buf_idx++; + NSQ->sLTP_buf_idx++; + + /* Update states */ + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + psSS = &psSampleState[ k ][ 0 ]; + psDD->LF_AR_Q14 = psSS->LF_AR_Q14; + psDD->sLPC_Q14[ NSQ_LPC_BUF_LENGTH + i ] = psSS->xq_Q14; + psDD->Xq_Q14[ *smpl_buf_idx ] = psSS->xq_Q14; + psDD->Q_Q10[ *smpl_buf_idx ] = psSS->Q_Q10; + psDD->Pred_Q15[ *smpl_buf_idx ] = silk_LSHIFT32( psSS->LPC_exc_Q14, 1 ); + psDD->Shape_Q14[ *smpl_buf_idx ] = psSS->sLTP_shp_Q14; + psDD->Seed = silk_ADD32_ovflw( psDD->Seed, silk_RSHIFT_ROUND( psSS->Q_Q10, 10 ) ); + psDD->RandState[ *smpl_buf_idx ] = psDD->Seed; + psDD->RD_Q10 = psSS->RD_Q10; + } + delayedGain_Q10[ *smpl_buf_idx ] = Gain_Q10; + } + /* Update LPC states */ + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + silk_memcpy( psDD->sLPC_Q14, &psDD->sLPC_Q14[ length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + } +} + +#endif /* __NSQ_DEL_DEC_MIPSR1_H__ */ diff --git a/lib/opus/silk/mips/macros_mipsr1.h b/lib/opus/silk/mips/macros_mipsr1.h new file mode 100644 index 00000000..12ed981a --- /dev/null +++ b/lib/opus/silk/mips/macros_mipsr1.h @@ -0,0 +1,92 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + + +#ifndef __SILK_MACROS_MIPSR1_H__ +#define __SILK_MACROS_MIPSR1_H__ + +#define mips_clz(x) __builtin_clz(x) + +#undef silk_SMULWB +static inline int silk_SMULWB(int a, int b) +{ + long long ac; + int c; + + ac = __builtin_mips_mult(a, (opus_int32)(opus_int16)b); + c = __builtin_mips_extr_w(ac, 16); + + return c; +} + +#undef silk_SMLAWB +#define silk_SMLAWB(a32, b32, c32) ((a32) + silk_SMULWB(b32, c32)) + +#undef silk_SMULWW +static inline int silk_SMULWW(int a, int b) +{ + long long ac; + int c; + + ac = __builtin_mips_mult(a, b); + c = __builtin_mips_extr_w(ac, 16); + + return c; +} + +#undef silk_SMLAWW +static inline int silk_SMLAWW(int a, int b, int c) +{ + long long ac; + int res; + + ac = __builtin_mips_mult(b, c); + res = __builtin_mips_extr_w(ac, 16); + res += a; + + return res; +} + +#define OVERRIDE_silk_CLZ16 +static inline opus_int32 silk_CLZ16(opus_int16 in16) +{ + int re32; + opus_int32 in32 = (opus_int32 )in16; + re32 = mips_clz(in32); + re32-=16; + return re32; +} + +#define OVERRIDE_silk_CLZ32 +static inline opus_int32 silk_CLZ32(opus_int32 in32) +{ + int re32; + re32 = mips_clz(in32); + return re32; +} + +#endif /* __SILK_MACROS_MIPSR1_H__ */ diff --git a/lib/opus/silk/mips/sigproc_fix_mipsr1.h b/lib/opus/silk/mips/sigproc_fix_mipsr1.h new file mode 100644 index 00000000..51520c0a --- /dev/null +++ b/lib/opus/silk/mips/sigproc_fix_mipsr1.h @@ -0,0 +1,60 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_SIGPROC_FIX_MIPSR1_H +#define SILK_SIGPROC_FIX_MIPSR1_H + +#undef silk_SAT16 +static inline short int silk_SAT16(int a) +{ + int c; + c = __builtin_mips_shll_s_w(a, 16); + c = c>>16; + + return c; +} + +#undef silk_LSHIFT_SAT32 +static inline int silk_LSHIFT_SAT32(int a, int shift) +{ + int r; + + r = __builtin_mips_shll_s_w(a, shift); + + return r; +} + +#undef silk_RSHIFT_ROUND +static inline int silk_RSHIFT_ROUND(int a, int shift) +{ + int r; + + r = __builtin_mips_shra_r_w(a, shift); + return r; +} + +#endif /* SILK_SIGPROC_FIX_MIPSR1_H */ diff --git a/lib/opus/silk/pitch_est_defines.h b/lib/opus/silk/pitch_est_defines.h new file mode 100644 index 00000000..e1e4b5d7 --- /dev/null +++ b/lib/opus/silk/pitch_est_defines.h @@ -0,0 +1,88 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_PE_DEFINES_H +#define SILK_PE_DEFINES_H + +#include "SigProc_FIX.h" + +/********************************************************/ +/* Definitions for pitch estimator */ +/********************************************************/ + +#define PE_MAX_FS_KHZ 16 /* Maximum sampling frequency used */ + +#define PE_MAX_NB_SUBFR 4 +#define PE_SUBFR_LENGTH_MS 5 /* 5 ms */ + +#define PE_LTP_MEM_LENGTH_MS ( 4 * PE_SUBFR_LENGTH_MS ) + +#define PE_MAX_FRAME_LENGTH_MS ( PE_LTP_MEM_LENGTH_MS + PE_MAX_NB_SUBFR * PE_SUBFR_LENGTH_MS ) +#define PE_MAX_FRAME_LENGTH ( PE_MAX_FRAME_LENGTH_MS * PE_MAX_FS_KHZ ) +#define PE_MAX_FRAME_LENGTH_ST_1 ( PE_MAX_FRAME_LENGTH >> 2 ) +#define PE_MAX_FRAME_LENGTH_ST_2 ( PE_MAX_FRAME_LENGTH >> 1 ) + +#define PE_MAX_LAG_MS 18 /* 18 ms -> 56 Hz */ +#define PE_MIN_LAG_MS 2 /* 2 ms -> 500 Hz */ +#define PE_MAX_LAG ( PE_MAX_LAG_MS * PE_MAX_FS_KHZ ) +#define PE_MIN_LAG ( PE_MIN_LAG_MS * PE_MAX_FS_KHZ ) + +#define PE_D_SRCH_LENGTH 24 + +#define PE_NB_STAGE3_LAGS 5 + +#define PE_NB_CBKS_STAGE2 3 +#define PE_NB_CBKS_STAGE2_EXT 11 + +#define PE_NB_CBKS_STAGE3_MAX 34 +#define PE_NB_CBKS_STAGE3_MID 24 +#define PE_NB_CBKS_STAGE3_MIN 16 + +#define PE_NB_CBKS_STAGE3_10MS 12 +#define PE_NB_CBKS_STAGE2_10MS 3 + +#define PE_SHORTLAG_BIAS 0.2f /* for logarithmic weighting */ +#define PE_PREVLAG_BIAS 0.2f /* for logarithmic weighting */ +#define PE_FLATCONTOUR_BIAS 0.05f + +#define SILK_PE_MIN_COMPLEX 0 +#define SILK_PE_MID_COMPLEX 1 +#define SILK_PE_MAX_COMPLEX 2 + +/* Tables for 20 ms frames */ +extern const opus_int8 silk_CB_lags_stage2[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE2_EXT ]; +extern const opus_int8 silk_CB_lags_stage3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ]; +extern const opus_int8 silk_Lag_range_stage3[ SILK_PE_MAX_COMPLEX + 1 ] [ PE_MAX_NB_SUBFR ][ 2 ]; +extern const opus_int8 silk_nb_cbk_searchs_stage3[ SILK_PE_MAX_COMPLEX + 1 ]; + +/* Tables for 10 ms frames */ +extern const opus_int8 silk_CB_lags_stage2_10_ms[ PE_MAX_NB_SUBFR >> 1][ 3 ]; +extern const opus_int8 silk_CB_lags_stage3_10_ms[ PE_MAX_NB_SUBFR >> 1 ][ 12 ]; +extern const opus_int8 silk_Lag_range_stage3_10_ms[ PE_MAX_NB_SUBFR >> 1 ][ 2 ]; + +#endif + diff --git a/lib/opus/silk/pitch_est_tables.c b/lib/opus/silk/pitch_est_tables.c new file mode 100644 index 00000000..81a8baca --- /dev/null +++ b/lib/opus/silk/pitch_est_tables.c @@ -0,0 +1,99 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "typedef.h" +#include "pitch_est_defines.h" + +const opus_int8 silk_CB_lags_stage2_10_ms[ PE_MAX_NB_SUBFR >> 1][ PE_NB_CBKS_STAGE2_10MS ] = +{ + {0, 1, 0}, + {0, 0, 1} +}; + +const opus_int8 silk_CB_lags_stage3_10_ms[ PE_MAX_NB_SUBFR >> 1 ][ PE_NB_CBKS_STAGE3_10MS ] = +{ + { 0, 0, 1,-1, 1,-1, 2,-2, 2,-2, 3,-3}, + { 0, 1, 0, 1,-1, 2,-1, 2,-2, 3,-2, 3} +}; + +const opus_int8 silk_Lag_range_stage3_10_ms[ PE_MAX_NB_SUBFR >> 1 ][ 2 ] = +{ + {-3, 7}, + {-2, 7} +}; + +const opus_int8 silk_CB_lags_stage2[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE2_EXT ] = +{ + {0, 2,-1,-1,-1, 0, 0, 1, 1, 0, 1}, + {0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0}, + {0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0}, + {0,-1, 2, 1, 0, 1, 1, 0, 0,-1,-1} +}; + +const opus_int8 silk_CB_lags_stage3[ PE_MAX_NB_SUBFR ][ PE_NB_CBKS_STAGE3_MAX ] = +{ + {0, 0, 1,-1, 0, 1,-1, 0,-1, 1,-2, 2,-2,-2, 2,-3, 2, 3,-3,-4, 3,-4, 4, 4,-5, 5,-6,-5, 6,-7, 6, 5, 8,-9}, + {0, 0, 1, 0, 0, 0, 0, 0, 0, 0,-1, 1, 0, 0, 1,-1, 0, 1,-1,-1, 1,-1, 2, 1,-1, 2,-2,-2, 2,-2, 2, 2, 3,-3}, + {0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1,-1, 1, 0, 0, 2, 1,-1, 2,-1,-1, 2,-1, 2, 2,-1, 3,-2,-2,-2, 3}, + {0, 1, 0, 0, 1, 0, 1,-1, 2,-1, 2,-1, 2, 3,-2, 3,-2,-2, 4, 4,-3, 5,-3,-4, 6,-4, 6, 5,-5, 8,-6,-5,-7, 9} +}; + +const opus_int8 silk_Lag_range_stage3[ SILK_PE_MAX_COMPLEX + 1 ] [ PE_MAX_NB_SUBFR ][ 2 ] = +{ + /* Lags to search for low number of stage3 cbks */ + { + {-5,8}, + {-1,6}, + {-1,6}, + {-4,10} + }, + /* Lags to search for middle number of stage3 cbks */ + { + {-6,10}, + {-2,6}, + {-1,6}, + {-5,10} + }, + /* Lags to search for max number of stage3 cbks */ + { + {-9,12}, + {-3,7}, + {-2,7}, + {-7,13} + } +}; + +const opus_int8 silk_nb_cbk_searchs_stage3[ SILK_PE_MAX_COMPLEX + 1 ] = +{ + PE_NB_CBKS_STAGE3_MIN, + PE_NB_CBKS_STAGE3_MID, + PE_NB_CBKS_STAGE3_MAX +}; diff --git a/lib/opus/silk/process_NLSFs.c b/lib/opus/silk/process_NLSFs.c new file mode 100644 index 00000000..d1308095 --- /dev/null +++ b/lib/opus/silk/process_NLSFs.c @@ -0,0 +1,107 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Limit, stabilize, convert and quantize NLSFs */ +void silk_process_NLSFs( + silk_encoder_state *psEncC, /* I/O Encoder state */ + opus_int16 PredCoef_Q12[ 2 ][ MAX_LPC_ORDER ], /* O Prediction coefficients */ + opus_int16 pNLSF_Q15[ MAX_LPC_ORDER ], /* I/O Normalized LSFs (quant out) (0 - (2^15-1)) */ + const opus_int16 prev_NLSFq_Q15[ MAX_LPC_ORDER ] /* I Previous Normalized LSFs (0 - (2^15-1)) */ +) +{ + opus_int i, doInterpolate; + opus_int NLSF_mu_Q20; + opus_int16 i_sqr_Q15; + opus_int16 pNLSF0_temp_Q15[ MAX_LPC_ORDER ]; + opus_int16 pNLSFW_QW[ MAX_LPC_ORDER ]; + opus_int16 pNLSFW0_temp_QW[ MAX_LPC_ORDER ]; + + silk_assert( psEncC->speech_activity_Q8 >= 0 ); + silk_assert( psEncC->speech_activity_Q8 <= SILK_FIX_CONST( 1.0, 8 ) ); + celt_assert( psEncC->useInterpolatedNLSFs == 1 || psEncC->indices.NLSFInterpCoef_Q2 == ( 1 << 2 ) ); + + /***********************/ + /* Calculate mu values */ + /***********************/ + /* NLSF_mu = 0.003 - 0.0015 * psEnc->speech_activity; */ + NLSF_mu_Q20 = silk_SMLAWB( SILK_FIX_CONST( 0.003, 20 ), SILK_FIX_CONST( -0.001, 28 ), psEncC->speech_activity_Q8 ); + if( psEncC->nb_subfr == 2 ) { + /* Multiply by 1.5 for 10 ms packets */ + NLSF_mu_Q20 = silk_ADD_RSHIFT( NLSF_mu_Q20, NLSF_mu_Q20, 1 ); + } + + celt_assert( NLSF_mu_Q20 > 0 ); + silk_assert( NLSF_mu_Q20 <= SILK_FIX_CONST( 0.005, 20 ) ); + + /* Calculate NLSF weights */ + silk_NLSF_VQ_weights_laroia( pNLSFW_QW, pNLSF_Q15, psEncC->predictLPCOrder ); + + /* Update NLSF weights for interpolated NLSFs */ + doInterpolate = ( psEncC->useInterpolatedNLSFs == 1 ) && ( psEncC->indices.NLSFInterpCoef_Q2 < 4 ); + if( doInterpolate ) { + /* Calculate the interpolated NLSF vector for the first half */ + silk_interpolate( pNLSF0_temp_Q15, prev_NLSFq_Q15, pNLSF_Q15, + psEncC->indices.NLSFInterpCoef_Q2, psEncC->predictLPCOrder ); + + /* Calculate first half NLSF weights for the interpolated NLSFs */ + silk_NLSF_VQ_weights_laroia( pNLSFW0_temp_QW, pNLSF0_temp_Q15, psEncC->predictLPCOrder ); + + /* Update NLSF weights with contribution from first half */ + i_sqr_Q15 = silk_LSHIFT( silk_SMULBB( psEncC->indices.NLSFInterpCoef_Q2, psEncC->indices.NLSFInterpCoef_Q2 ), 11 ); + for( i = 0; i < psEncC->predictLPCOrder; i++ ) { + pNLSFW_QW[ i ] = silk_ADD16( silk_RSHIFT( pNLSFW_QW[ i ], 1 ), silk_RSHIFT( + silk_SMULBB( pNLSFW0_temp_QW[ i ], i_sqr_Q15 ), 16) ); + silk_assert( pNLSFW_QW[ i ] >= 1 ); + } + } + + silk_NLSF_encode( psEncC->indices.NLSFIndices, pNLSF_Q15, psEncC->psNLSF_CB, pNLSFW_QW, + NLSF_mu_Q20, psEncC->NLSF_MSVQ_Survivors, psEncC->indices.signalType ); + + /* Convert quantized NLSFs back to LPC coefficients */ + silk_NLSF2A( PredCoef_Q12[ 1 ], pNLSF_Q15, psEncC->predictLPCOrder, psEncC->arch ); + + if( doInterpolate ) { + /* Calculate the interpolated, quantized LSF vector for the first half */ + silk_interpolate( pNLSF0_temp_Q15, prev_NLSFq_Q15, pNLSF_Q15, + psEncC->indices.NLSFInterpCoef_Q2, psEncC->predictLPCOrder ); + + /* Convert back to LPC coefficients */ + silk_NLSF2A( PredCoef_Q12[ 0 ], pNLSF0_temp_Q15, psEncC->predictLPCOrder, psEncC->arch ); + + } else { + /* Copy LPC coefficients for first half from second half */ + celt_assert( psEncC->predictLPCOrder <= MAX_LPC_ORDER ); + silk_memcpy( PredCoef_Q12[ 0 ], PredCoef_Q12[ 1 ], psEncC->predictLPCOrder * sizeof( opus_int16 ) ); + } +} diff --git a/lib/opus/silk/quant_LTP_gains.c b/lib/opus/silk/quant_LTP_gains.c new file mode 100644 index 00000000..d6b8eff8 --- /dev/null +++ b/lib/opus/silk/quant_LTP_gains.c @@ -0,0 +1,132 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "tuning_parameters.h" + +void silk_quant_LTP_gains( + opus_int16 B_Q14[ MAX_NB_SUBFR * LTP_ORDER ], /* O Quantized LTP gains */ + opus_int8 cbk_index[ MAX_NB_SUBFR ], /* O Codebook Index */ + opus_int8 *periodicity_index, /* O Periodicity Index */ + opus_int32 *sum_log_gain_Q7, /* I/O Cumulative max prediction gain */ + opus_int *pred_gain_dB_Q7, /* O LTP prediction gain */ + const opus_int32 XX_Q17[ MAX_NB_SUBFR*LTP_ORDER*LTP_ORDER ], /* I Correlation matrix in Q18 */ + const opus_int32 xX_Q17[ MAX_NB_SUBFR*LTP_ORDER ], /* I Correlation vector in Q18 */ + const opus_int subfr_len, /* I Number of samples per subframe */ + const opus_int nb_subfr, /* I Number of subframes */ + int arch /* I Run-time architecture */ +) +{ + opus_int j, k, cbk_size; + opus_int8 temp_idx[ MAX_NB_SUBFR ]; + const opus_uint8 *cl_ptr_Q5; + const opus_int8 *cbk_ptr_Q7; + const opus_uint8 *cbk_gain_ptr_Q7; + const opus_int32 *XX_Q17_ptr, *xX_Q17_ptr; + opus_int32 res_nrg_Q15_subfr, res_nrg_Q15, rate_dist_Q7_subfr, rate_dist_Q7, min_rate_dist_Q7; + opus_int32 sum_log_gain_tmp_Q7, best_sum_log_gain_Q7, max_gain_Q7; + opus_int gain_Q7; + + /***************************************************/ + /* iterate over different codebooks with different */ + /* rates/distortions, and choose best */ + /***************************************************/ + min_rate_dist_Q7 = silk_int32_MAX; + best_sum_log_gain_Q7 = 0; + for( k = 0; k < 3; k++ ) { + /* Safety margin for pitch gain control, to take into account factors + such as state rescaling/rewhitening. */ + opus_int32 gain_safety = SILK_FIX_CONST( 0.4, 7 ); + + cl_ptr_Q5 = silk_LTP_gain_BITS_Q5_ptrs[ k ]; + cbk_ptr_Q7 = silk_LTP_vq_ptrs_Q7[ k ]; + cbk_gain_ptr_Q7 = silk_LTP_vq_gain_ptrs_Q7[ k ]; + cbk_size = silk_LTP_vq_sizes[ k ]; + + /* Set up pointers to first subframe */ + XX_Q17_ptr = XX_Q17; + xX_Q17_ptr = xX_Q17; + + res_nrg_Q15 = 0; + rate_dist_Q7 = 0; + sum_log_gain_tmp_Q7 = *sum_log_gain_Q7; + for( j = 0; j < nb_subfr; j++ ) { + max_gain_Q7 = silk_log2lin( ( SILK_FIX_CONST( MAX_SUM_LOG_GAIN_DB / 6.0, 7 ) - sum_log_gain_tmp_Q7 ) + + SILK_FIX_CONST( 7, 7 ) ) - gain_safety; + silk_VQ_WMat_EC( + &temp_idx[ j ], /* O index of best codebook vector */ + &res_nrg_Q15_subfr, /* O residual energy */ + &rate_dist_Q7_subfr, /* O best weighted quantization error + mu * rate */ + &gain_Q7, /* O sum of absolute LTP coefficients */ + XX_Q17_ptr, /* I correlation matrix */ + xX_Q17_ptr, /* I correlation vector */ + cbk_ptr_Q7, /* I codebook */ + cbk_gain_ptr_Q7, /* I codebook effective gains */ + cl_ptr_Q5, /* I code length for each codebook vector */ + subfr_len, /* I number of samples per subframe */ + max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ + cbk_size, /* I number of vectors in codebook */ + arch /* I Run-time architecture */ + ); + + res_nrg_Q15 = silk_ADD_POS_SAT32( res_nrg_Q15, res_nrg_Q15_subfr ); + rate_dist_Q7 = silk_ADD_POS_SAT32( rate_dist_Q7, rate_dist_Q7_subfr ); + sum_log_gain_tmp_Q7 = silk_max(0, sum_log_gain_tmp_Q7 + + silk_lin2log( gain_safety + gain_Q7 ) - SILK_FIX_CONST( 7, 7 )); + + XX_Q17_ptr += LTP_ORDER * LTP_ORDER; + xX_Q17_ptr += LTP_ORDER; + } + + if( rate_dist_Q7 <= min_rate_dist_Q7 ) { + min_rate_dist_Q7 = rate_dist_Q7; + *periodicity_index = (opus_int8)k; + silk_memcpy( cbk_index, temp_idx, nb_subfr * sizeof( opus_int8 ) ); + best_sum_log_gain_Q7 = sum_log_gain_tmp_Q7; + } + } + + cbk_ptr_Q7 = silk_LTP_vq_ptrs_Q7[ *periodicity_index ]; + for( j = 0; j < nb_subfr; j++ ) { + for( k = 0; k < LTP_ORDER; k++ ) { + B_Q14[ j * LTP_ORDER + k ] = silk_LSHIFT( cbk_ptr_Q7[ cbk_index[ j ] * LTP_ORDER + k ], 7 ); + } + } + + if( nb_subfr == 2 ) { + res_nrg_Q15 = silk_RSHIFT32( res_nrg_Q15, 1 ); + } else { + res_nrg_Q15 = silk_RSHIFT32( res_nrg_Q15, 2 ); + } + + *sum_log_gain_Q7 = best_sum_log_gain_Q7; + *pred_gain_dB_Q7 = (opus_int)silk_SMULBB( -3, silk_lin2log( res_nrg_Q15 ) - ( 15 << 7 ) ); +} diff --git a/lib/opus/silk/resampler.c b/lib/opus/silk/resampler.c new file mode 100644 index 00000000..1f11e508 --- /dev/null +++ b/lib/opus/silk/resampler.c @@ -0,0 +1,215 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* + * Matrix of resampling methods used: + * Fs_out (kHz) + * 8 12 16 24 48 + * + * 8 C UF U UF UF + * 12 AF C UF U UF + * Fs_in (kHz) 16 D AF C UF UF + * 24 AF D AF C U + * 48 AF AF AF D C + * + * C -> Copy (no resampling) + * D -> Allpass-based 2x downsampling + * U -> Allpass-based 2x upsampling + * UF -> Allpass-based 2x upsampling followed by FIR interpolation + * AF -> AR2 filter followed by FIR interpolation + */ + +#include "resampler_private.h" + +/* Tables with delay compensation values to equalize total delay for different modes */ +static const opus_int8 delay_matrix_enc[ 5 ][ 3 ] = { +/* in \ out 8 12 16 */ +/* 8 */ { 6, 0, 3 }, +/* 12 */ { 0, 7, 3 }, +/* 16 */ { 0, 1, 10 }, +/* 24 */ { 0, 2, 6 }, +/* 48 */ { 18, 10, 12 } +}; + +static const opus_int8 delay_matrix_dec[ 3 ][ 5 ] = { +/* in \ out 8 12 16 24 48 */ +/* 8 */ { 4, 0, 2, 0, 0 }, +/* 12 */ { 0, 9, 4, 7, 4 }, +/* 16 */ { 0, 3, 12, 7, 7 } +}; + +/* Simple way to make [8000, 12000, 16000, 24000, 48000] to [0, 1, 2, 3, 4] */ +#define rateID(R) ( ( ( ((R)>>12) - ((R)>16000) ) >> ((R)>24000) ) - 1 ) + +#define USE_silk_resampler_copy (0) +#define USE_silk_resampler_private_up2_HQ_wrapper (1) +#define USE_silk_resampler_private_IIR_FIR (2) +#define USE_silk_resampler_private_down_FIR (3) + +/* Initialize/reset the resampler state for a given pair of input/output sampling rates */ +opus_int silk_resampler_init( + silk_resampler_state_struct *S, /* I/O Resampler state */ + opus_int32 Fs_Hz_in, /* I Input sampling rate (Hz) */ + opus_int32 Fs_Hz_out, /* I Output sampling rate (Hz) */ + opus_int forEnc /* I If 1: encoder; if 0: decoder */ +) +{ + opus_int up2x; + + /* Clear state */ + silk_memset( S, 0, sizeof( silk_resampler_state_struct ) ); + + /* Input checking */ + if( forEnc ) { + if( ( Fs_Hz_in != 8000 && Fs_Hz_in != 12000 && Fs_Hz_in != 16000 && Fs_Hz_in != 24000 && Fs_Hz_in != 48000 ) || + ( Fs_Hz_out != 8000 && Fs_Hz_out != 12000 && Fs_Hz_out != 16000 ) ) { + celt_assert( 0 ); + return -1; + } + S->inputDelay = delay_matrix_enc[ rateID( Fs_Hz_in ) ][ rateID( Fs_Hz_out ) ]; + } else { + if( ( Fs_Hz_in != 8000 && Fs_Hz_in != 12000 && Fs_Hz_in != 16000 ) || + ( Fs_Hz_out != 8000 && Fs_Hz_out != 12000 && Fs_Hz_out != 16000 && Fs_Hz_out != 24000 && Fs_Hz_out != 48000 ) ) { + celt_assert( 0 ); + return -1; + } + S->inputDelay = delay_matrix_dec[ rateID( Fs_Hz_in ) ][ rateID( Fs_Hz_out ) ]; + } + + S->Fs_in_kHz = silk_DIV32_16( Fs_Hz_in, 1000 ); + S->Fs_out_kHz = silk_DIV32_16( Fs_Hz_out, 1000 ); + + /* Number of samples processed per batch */ + S->batchSize = S->Fs_in_kHz * RESAMPLER_MAX_BATCH_SIZE_MS; + + /* Find resampler with the right sampling ratio */ + up2x = 0; + if( Fs_Hz_out > Fs_Hz_in ) { + /* Upsample */ + if( Fs_Hz_out == silk_MUL( Fs_Hz_in, 2 ) ) { /* Fs_out : Fs_in = 2 : 1 */ + /* Special case: directly use 2x upsampler */ + S->resampler_function = USE_silk_resampler_private_up2_HQ_wrapper; + } else { + /* Default resampler */ + S->resampler_function = USE_silk_resampler_private_IIR_FIR; + up2x = 1; + } + } else if ( Fs_Hz_out < Fs_Hz_in ) { + /* Downsample */ + S->resampler_function = USE_silk_resampler_private_down_FIR; + if( silk_MUL( Fs_Hz_out, 4 ) == silk_MUL( Fs_Hz_in, 3 ) ) { /* Fs_out : Fs_in = 3 : 4 */ + S->FIR_Fracs = 3; + S->FIR_Order = RESAMPLER_DOWN_ORDER_FIR0; + S->Coefs = silk_Resampler_3_4_COEFS; + } else if( silk_MUL( Fs_Hz_out, 3 ) == silk_MUL( Fs_Hz_in, 2 ) ) { /* Fs_out : Fs_in = 2 : 3 */ + S->FIR_Fracs = 2; + S->FIR_Order = RESAMPLER_DOWN_ORDER_FIR0; + S->Coefs = silk_Resampler_2_3_COEFS; + } else if( silk_MUL( Fs_Hz_out, 2 ) == Fs_Hz_in ) { /* Fs_out : Fs_in = 1 : 2 */ + S->FIR_Fracs = 1; + S->FIR_Order = RESAMPLER_DOWN_ORDER_FIR1; + S->Coefs = silk_Resampler_1_2_COEFS; + } else if( silk_MUL( Fs_Hz_out, 3 ) == Fs_Hz_in ) { /* Fs_out : Fs_in = 1 : 3 */ + S->FIR_Fracs = 1; + S->FIR_Order = RESAMPLER_DOWN_ORDER_FIR2; + S->Coefs = silk_Resampler_1_3_COEFS; + } else if( silk_MUL( Fs_Hz_out, 4 ) == Fs_Hz_in ) { /* Fs_out : Fs_in = 1 : 4 */ + S->FIR_Fracs = 1; + S->FIR_Order = RESAMPLER_DOWN_ORDER_FIR2; + S->Coefs = silk_Resampler_1_4_COEFS; + } else if( silk_MUL( Fs_Hz_out, 6 ) == Fs_Hz_in ) { /* Fs_out : Fs_in = 1 : 6 */ + S->FIR_Fracs = 1; + S->FIR_Order = RESAMPLER_DOWN_ORDER_FIR2; + S->Coefs = silk_Resampler_1_6_COEFS; + } else { + /* None available */ + celt_assert( 0 ); + return -1; + } + } else { + /* Input and output sampling rates are equal: copy */ + S->resampler_function = USE_silk_resampler_copy; + } + + /* Ratio of input/output samples */ + S->invRatio_Q16 = silk_LSHIFT32( silk_DIV32( silk_LSHIFT32( Fs_Hz_in, 14 + up2x ), Fs_Hz_out ), 2 ); + /* Make sure the ratio is rounded up */ + while( silk_SMULWW( S->invRatio_Q16, Fs_Hz_out ) < silk_LSHIFT32( Fs_Hz_in, up2x ) ) { + S->invRatio_Q16++; + } + + return 0; +} + +/* Resampler: convert from one sampling rate to another */ +/* Input and output sampling rate are at most 48000 Hz */ +opus_int silk_resampler( + silk_resampler_state_struct *S, /* I/O Resampler state */ + opus_int16 out[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + opus_int32 inLen /* I Number of input samples */ +) +{ + opus_int nSamples; + + /* Need at least 1 ms of input data */ + celt_assert( inLen >= S->Fs_in_kHz ); + /* Delay can't exceed the 1 ms of buffering */ + celt_assert( S->inputDelay <= S->Fs_in_kHz ); + + nSamples = S->Fs_in_kHz - S->inputDelay; + + /* Copy to delay buffer */ + silk_memcpy( &S->delayBuf[ S->inputDelay ], in, nSamples * sizeof( opus_int16 ) ); + + switch( S->resampler_function ) { + case USE_silk_resampler_private_up2_HQ_wrapper: + silk_resampler_private_up2_HQ_wrapper( S, out, S->delayBuf, S->Fs_in_kHz ); + silk_resampler_private_up2_HQ_wrapper( S, &out[ S->Fs_out_kHz ], &in[ nSamples ], inLen - S->Fs_in_kHz ); + break; + case USE_silk_resampler_private_IIR_FIR: + silk_resampler_private_IIR_FIR( S, out, S->delayBuf, S->Fs_in_kHz ); + silk_resampler_private_IIR_FIR( S, &out[ S->Fs_out_kHz ], &in[ nSamples ], inLen - S->Fs_in_kHz ); + break; + case USE_silk_resampler_private_down_FIR: + silk_resampler_private_down_FIR( S, out, S->delayBuf, S->Fs_in_kHz ); + silk_resampler_private_down_FIR( S, &out[ S->Fs_out_kHz ], &in[ nSamples ], inLen - S->Fs_in_kHz ); + break; + default: + silk_memcpy( out, S->delayBuf, S->Fs_in_kHz * sizeof( opus_int16 ) ); + silk_memcpy( &out[ S->Fs_out_kHz ], &in[ nSamples ], ( inLen - S->Fs_in_kHz ) * sizeof( opus_int16 ) ); + } + + /* Copy to delay buffer */ + silk_memcpy( S->delayBuf, &in[ inLen - S->inputDelay ], S->inputDelay * sizeof( opus_int16 ) ); + + return 0; +} diff --git a/lib/opus/silk/resampler_down2.c b/lib/opus/silk/resampler_down2.c new file mode 100644 index 00000000..971d7bfd --- /dev/null +++ b/lib/opus/silk/resampler_down2.c @@ -0,0 +1,74 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "resampler_rom.h" + +/* Downsample by a factor 2 */ +void silk_resampler_down2( + opus_int32 *S, /* I/O State vector [ 2 ] */ + opus_int16 *out, /* O Output signal [ floor(len/2) ] */ + const opus_int16 *in, /* I Input signal [ len ] */ + opus_int32 inLen /* I Number of input samples */ +) +{ + opus_int32 k, len2 = silk_RSHIFT32( inLen, 1 ); + opus_int32 in32, out32, Y, X; + + celt_assert( silk_resampler_down2_0 > 0 ); + celt_assert( silk_resampler_down2_1 < 0 ); + + /* Internal variables and state are in Q10 format */ + for( k = 0; k < len2; k++ ) { + /* Convert to Q10 */ + in32 = silk_LSHIFT( (opus_int32)in[ 2 * k ], 10 ); + + /* All-pass section for even input sample */ + Y = silk_SUB32( in32, S[ 0 ] ); + X = silk_SMLAWB( Y, Y, silk_resampler_down2_1 ); + out32 = silk_ADD32( S[ 0 ], X ); + S[ 0 ] = silk_ADD32( in32, X ); + + /* Convert to Q10 */ + in32 = silk_LSHIFT( (opus_int32)in[ 2 * k + 1 ], 10 ); + + /* All-pass section for odd input sample, and add to output of previous section */ + Y = silk_SUB32( in32, S[ 1 ] ); + X = silk_SMULWB( Y, silk_resampler_down2_0 ); + out32 = silk_ADD32( out32, S[ 1 ] ); + out32 = silk_ADD32( out32, X ); + S[ 1 ] = silk_ADD32( in32, X ); + + /* Add, convert back to int16 and store to output */ + out[ k ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( out32, 11 ) ); + } +} + diff --git a/lib/opus/silk/resampler_down2_3.c b/lib/opus/silk/resampler_down2_3.c new file mode 100644 index 00000000..4342614d --- /dev/null +++ b/lib/opus/silk/resampler_down2_3.c @@ -0,0 +1,103 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "resampler_private.h" +#include "stack_alloc.h" + +#define ORDER_FIR 4 + +/* Downsample by a factor 2/3, low quality */ +void silk_resampler_down2_3( + opus_int32 *S, /* I/O State vector [ 6 ] */ + opus_int16 *out, /* O Output signal [ floor(2*inLen/3) ] */ + const opus_int16 *in, /* I Input signal [ inLen ] */ + opus_int32 inLen /* I Number of input samples */ +) +{ + opus_int32 nSamplesIn, counter, res_Q6; + VARDECL( opus_int32, buf ); + opus_int32 *buf_ptr; + SAVE_STACK; + + ALLOC( buf, RESAMPLER_MAX_BATCH_SIZE_IN + ORDER_FIR, opus_int32 ); + + /* Copy buffered samples to start of buffer */ + silk_memcpy( buf, S, ORDER_FIR * sizeof( opus_int32 ) ); + + /* Iterate over blocks of frameSizeIn input samples */ + while( 1 ) { + nSamplesIn = silk_min( inLen, RESAMPLER_MAX_BATCH_SIZE_IN ); + + /* Second-order AR filter (output in Q8) */ + silk_resampler_private_AR2( &S[ ORDER_FIR ], &buf[ ORDER_FIR ], in, + silk_Resampler_2_3_COEFS_LQ, nSamplesIn ); + + /* Interpolate filtered signal */ + buf_ptr = buf; + counter = nSamplesIn; + while( counter > 2 ) { + /* Inner product */ + res_Q6 = silk_SMULWB( buf_ptr[ 0 ], silk_Resampler_2_3_COEFS_LQ[ 2 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 1 ], silk_Resampler_2_3_COEFS_LQ[ 3 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 2 ], silk_Resampler_2_3_COEFS_LQ[ 5 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 3 ], silk_Resampler_2_3_COEFS_LQ[ 4 ] ); + + /* Scale down, saturate and store in output array */ + *out++ = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( res_Q6, 6 ) ); + + res_Q6 = silk_SMULWB( buf_ptr[ 1 ], silk_Resampler_2_3_COEFS_LQ[ 4 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 2 ], silk_Resampler_2_3_COEFS_LQ[ 5 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 3 ], silk_Resampler_2_3_COEFS_LQ[ 3 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 4 ], silk_Resampler_2_3_COEFS_LQ[ 2 ] ); + + /* Scale down, saturate and store in output array */ + *out++ = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( res_Q6, 6 ) ); + + buf_ptr += 3; + counter -= 3; + } + + in += nSamplesIn; + inLen -= nSamplesIn; + + if( inLen > 0 ) { + /* More iterations to do; copy last part of filtered signal to beginning of buffer */ + silk_memcpy( buf, &buf[ nSamplesIn ], ORDER_FIR * sizeof( opus_int32 ) ); + } else { + break; + } + } + + /* Copy last part of filtered signal to the state for the next call */ + silk_memcpy( S, &buf[ nSamplesIn ], ORDER_FIR * sizeof( opus_int32 ) ); + RESTORE_STACK; +} diff --git a/lib/opus/silk/resampler_private.h b/lib/opus/silk/resampler_private.h new file mode 100644 index 00000000..422a7d9d --- /dev/null +++ b/lib/opus/silk/resampler_private.h @@ -0,0 +1,88 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_RESAMPLER_PRIVATE_H +#define SILK_RESAMPLER_PRIVATE_H + +#ifdef __cplusplus +extern "C" { +#endif + +#include "SigProc_FIX.h" +#include "resampler_structs.h" +#include "resampler_rom.h" + +/* Number of input samples to process in the inner loop */ +#define RESAMPLER_MAX_BATCH_SIZE_MS 10 +#define RESAMPLER_MAX_FS_KHZ 48 +#define RESAMPLER_MAX_BATCH_SIZE_IN ( RESAMPLER_MAX_BATCH_SIZE_MS * RESAMPLER_MAX_FS_KHZ ) + +/* Description: Hybrid IIR/FIR polyphase implementation of resampling */ +void silk_resampler_private_IIR_FIR( + void *SS, /* I/O Resampler state */ + opus_int16 out[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + opus_int32 inLen /* I Number of input samples */ +); + +/* Description: Hybrid IIR/FIR polyphase implementation of resampling */ +void silk_resampler_private_down_FIR( + void *SS, /* I/O Resampler state */ + opus_int16 out[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + opus_int32 inLen /* I Number of input samples */ +); + +/* Upsample by a factor 2, high quality */ +void silk_resampler_private_up2_HQ_wrapper( + void *SS, /* I/O Resampler state (unused) */ + opus_int16 *out, /* O Output signal [ 2 * len ] */ + const opus_int16 *in, /* I Input signal [ len ] */ + opus_int32 len /* I Number of input samples */ +); + +/* Upsample by a factor 2, high quality */ +void silk_resampler_private_up2_HQ( + opus_int32 *S, /* I/O Resampler state [ 6 ] */ + opus_int16 *out, /* O Output signal [ 2 * len ] */ + const opus_int16 *in, /* I Input signal [ len ] */ + opus_int32 len /* I Number of input samples */ +); + +/* Second order AR filter */ +void silk_resampler_private_AR2( + opus_int32 S[], /* I/O State vector [ 2 ] */ + opus_int32 out_Q8[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + const opus_int16 A_Q14[], /* I AR coefficients, Q14 */ + opus_int32 len /* I Signal length */ +); + +#ifdef __cplusplus +} +#endif +#endif /* SILK_RESAMPLER_PRIVATE_H */ diff --git a/lib/opus/silk/resampler_private_AR2.c b/lib/opus/silk/resampler_private_AR2.c new file mode 100644 index 00000000..5fff2371 --- /dev/null +++ b/lib/opus/silk/resampler_private_AR2.c @@ -0,0 +1,55 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "resampler_private.h" + +/* Second order AR filter with single delay elements */ +void silk_resampler_private_AR2( + opus_int32 S[], /* I/O State vector [ 2 ] */ + opus_int32 out_Q8[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + const opus_int16 A_Q14[], /* I AR coefficients, Q14 */ + opus_int32 len /* I Signal length */ +) +{ + opus_int32 k; + opus_int32 out32; + + for( k = 0; k < len; k++ ) { + out32 = silk_ADD_LSHIFT32( S[ 0 ], (opus_int32)in[ k ], 8 ); + out_Q8[ k ] = out32; + out32 = silk_LSHIFT( out32, 2 ); + S[ 0 ] = silk_SMLAWB( S[ 1 ], out32, A_Q14[ 0 ] ); + S[ 1 ] = silk_SMULWB( out32, A_Q14[ 1 ] ); + } +} + diff --git a/lib/opus/silk/resampler_private_IIR_FIR.c b/lib/opus/silk/resampler_private_IIR_FIR.c new file mode 100644 index 00000000..6b2b3a2e --- /dev/null +++ b/lib/opus/silk/resampler_private_IIR_FIR.c @@ -0,0 +1,107 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "resampler_private.h" +#include "stack_alloc.h" + +static OPUS_INLINE opus_int16 *silk_resampler_private_IIR_FIR_INTERPOL( + opus_int16 *out, + opus_int16 *buf, + opus_int32 max_index_Q16, + opus_int32 index_increment_Q16 +) +{ + opus_int32 index_Q16, res_Q15; + opus_int16 *buf_ptr; + opus_int32 table_index; + + /* Interpolate upsampled signal and store in output array */ + for( index_Q16 = 0; index_Q16 < max_index_Q16; index_Q16 += index_increment_Q16 ) { + table_index = silk_SMULWB( index_Q16 & 0xFFFF, 12 ); + buf_ptr = &buf[ index_Q16 >> 16 ]; + + res_Q15 = silk_SMULBB( buf_ptr[ 0 ], silk_resampler_frac_FIR_12[ table_index ][ 0 ] ); + res_Q15 = silk_SMLABB( res_Q15, buf_ptr[ 1 ], silk_resampler_frac_FIR_12[ table_index ][ 1 ] ); + res_Q15 = silk_SMLABB( res_Q15, buf_ptr[ 2 ], silk_resampler_frac_FIR_12[ table_index ][ 2 ] ); + res_Q15 = silk_SMLABB( res_Q15, buf_ptr[ 3 ], silk_resampler_frac_FIR_12[ table_index ][ 3 ] ); + res_Q15 = silk_SMLABB( res_Q15, buf_ptr[ 4 ], silk_resampler_frac_FIR_12[ 11 - table_index ][ 3 ] ); + res_Q15 = silk_SMLABB( res_Q15, buf_ptr[ 5 ], silk_resampler_frac_FIR_12[ 11 - table_index ][ 2 ] ); + res_Q15 = silk_SMLABB( res_Q15, buf_ptr[ 6 ], silk_resampler_frac_FIR_12[ 11 - table_index ][ 1 ] ); + res_Q15 = silk_SMLABB( res_Q15, buf_ptr[ 7 ], silk_resampler_frac_FIR_12[ 11 - table_index ][ 0 ] ); + *out++ = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( res_Q15, 15 ) ); + } + return out; +} +/* Upsample using a combination of allpass-based 2x upsampling and FIR interpolation */ +void silk_resampler_private_IIR_FIR( + void *SS, /* I/O Resampler state */ + opus_int16 out[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + opus_int32 inLen /* I Number of input samples */ +) +{ + silk_resampler_state_struct *S = (silk_resampler_state_struct *)SS; + opus_int32 nSamplesIn; + opus_int32 max_index_Q16, index_increment_Q16; + VARDECL( opus_int16, buf ); + SAVE_STACK; + + ALLOC( buf, 2 * S->batchSize + RESAMPLER_ORDER_FIR_12, opus_int16 ); + + /* Copy buffered samples to start of buffer */ + silk_memcpy( buf, S->sFIR.i16, RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + + /* Iterate over blocks of frameSizeIn input samples */ + index_increment_Q16 = S->invRatio_Q16; + while( 1 ) { + nSamplesIn = silk_min( inLen, S->batchSize ); + + /* Upsample 2x */ + silk_resampler_private_up2_HQ( S->sIIR, &buf[ RESAMPLER_ORDER_FIR_12 ], in, nSamplesIn ); + + max_index_Q16 = silk_LSHIFT32( nSamplesIn, 16 + 1 ); /* + 1 because 2x upsampling */ + out = silk_resampler_private_IIR_FIR_INTERPOL( out, buf, max_index_Q16, index_increment_Q16 ); + in += nSamplesIn; + inLen -= nSamplesIn; + + if( inLen > 0 ) { + /* More iterations to do; copy last part of filtered signal to beginning of buffer */ + silk_memcpy( buf, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + } else { + break; + } + } + + /* Copy last part of filtered signal to the state for the next call */ + silk_memcpy( S->sFIR.i16, &buf[ nSamplesIn << 1 ], RESAMPLER_ORDER_FIR_12 * sizeof( opus_int16 ) ); + RESTORE_STACK; +} diff --git a/lib/opus/silk/resampler_private_down_FIR.c b/lib/opus/silk/resampler_private_down_FIR.c new file mode 100644 index 00000000..3e8735a3 --- /dev/null +++ b/lib/opus/silk/resampler_private_down_FIR.c @@ -0,0 +1,194 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "resampler_private.h" +#include "stack_alloc.h" + +static OPUS_INLINE opus_int16 *silk_resampler_private_down_FIR_INTERPOL( + opus_int16 *out, + opus_int32 *buf, + const opus_int16 *FIR_Coefs, + opus_int FIR_Order, + opus_int FIR_Fracs, + opus_int32 max_index_Q16, + opus_int32 index_increment_Q16 +) +{ + opus_int32 index_Q16, res_Q6; + opus_int32 *buf_ptr; + opus_int32 interpol_ind; + const opus_int16 *interpol_ptr; + + switch( FIR_Order ) { + case RESAMPLER_DOWN_ORDER_FIR0: + for( index_Q16 = 0; index_Q16 < max_index_Q16; index_Q16 += index_increment_Q16 ) { + /* Integer part gives pointer to buffered input */ + buf_ptr = buf + silk_RSHIFT( index_Q16, 16 ); + + /* Fractional part gives interpolation coefficients */ + interpol_ind = silk_SMULWB( index_Q16 & 0xFFFF, FIR_Fracs ); + + /* Inner product */ + interpol_ptr = &FIR_Coefs[ RESAMPLER_DOWN_ORDER_FIR0 / 2 * interpol_ind ]; + res_Q6 = silk_SMULWB( buf_ptr[ 0 ], interpol_ptr[ 0 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 1 ], interpol_ptr[ 1 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 2 ], interpol_ptr[ 2 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 3 ], interpol_ptr[ 3 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 4 ], interpol_ptr[ 4 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 5 ], interpol_ptr[ 5 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 6 ], interpol_ptr[ 6 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 7 ], interpol_ptr[ 7 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 8 ], interpol_ptr[ 8 ] ); + interpol_ptr = &FIR_Coefs[ RESAMPLER_DOWN_ORDER_FIR0 / 2 * ( FIR_Fracs - 1 - interpol_ind ) ]; + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 17 ], interpol_ptr[ 0 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 16 ], interpol_ptr[ 1 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 15 ], interpol_ptr[ 2 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 14 ], interpol_ptr[ 3 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 13 ], interpol_ptr[ 4 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 12 ], interpol_ptr[ 5 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 11 ], interpol_ptr[ 6 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 10 ], interpol_ptr[ 7 ] ); + res_Q6 = silk_SMLAWB( res_Q6, buf_ptr[ 9 ], interpol_ptr[ 8 ] ); + + /* Scale down, saturate and store in output array */ + *out++ = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( res_Q6, 6 ) ); + } + break; + case RESAMPLER_DOWN_ORDER_FIR1: + for( index_Q16 = 0; index_Q16 < max_index_Q16; index_Q16 += index_increment_Q16 ) { + /* Integer part gives pointer to buffered input */ + buf_ptr = buf + silk_RSHIFT( index_Q16, 16 ); + + /* Inner product */ + res_Q6 = silk_SMULWB( silk_ADD32( buf_ptr[ 0 ], buf_ptr[ 23 ] ), FIR_Coefs[ 0 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 1 ], buf_ptr[ 22 ] ), FIR_Coefs[ 1 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 2 ], buf_ptr[ 21 ] ), FIR_Coefs[ 2 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 3 ], buf_ptr[ 20 ] ), FIR_Coefs[ 3 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 4 ], buf_ptr[ 19 ] ), FIR_Coefs[ 4 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 5 ], buf_ptr[ 18 ] ), FIR_Coefs[ 5 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 6 ], buf_ptr[ 17 ] ), FIR_Coefs[ 6 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 7 ], buf_ptr[ 16 ] ), FIR_Coefs[ 7 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 8 ], buf_ptr[ 15 ] ), FIR_Coefs[ 8 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 9 ], buf_ptr[ 14 ] ), FIR_Coefs[ 9 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 10 ], buf_ptr[ 13 ] ), FIR_Coefs[ 10 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 11 ], buf_ptr[ 12 ] ), FIR_Coefs[ 11 ] ); + + /* Scale down, saturate and store in output array */ + *out++ = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( res_Q6, 6 ) ); + } + break; + case RESAMPLER_DOWN_ORDER_FIR2: + for( index_Q16 = 0; index_Q16 < max_index_Q16; index_Q16 += index_increment_Q16 ) { + /* Integer part gives pointer to buffered input */ + buf_ptr = buf + silk_RSHIFT( index_Q16, 16 ); + + /* Inner product */ + res_Q6 = silk_SMULWB( silk_ADD32( buf_ptr[ 0 ], buf_ptr[ 35 ] ), FIR_Coefs[ 0 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 1 ], buf_ptr[ 34 ] ), FIR_Coefs[ 1 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 2 ], buf_ptr[ 33 ] ), FIR_Coefs[ 2 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 3 ], buf_ptr[ 32 ] ), FIR_Coefs[ 3 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 4 ], buf_ptr[ 31 ] ), FIR_Coefs[ 4 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 5 ], buf_ptr[ 30 ] ), FIR_Coefs[ 5 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 6 ], buf_ptr[ 29 ] ), FIR_Coefs[ 6 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 7 ], buf_ptr[ 28 ] ), FIR_Coefs[ 7 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 8 ], buf_ptr[ 27 ] ), FIR_Coefs[ 8 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 9 ], buf_ptr[ 26 ] ), FIR_Coefs[ 9 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 10 ], buf_ptr[ 25 ] ), FIR_Coefs[ 10 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 11 ], buf_ptr[ 24 ] ), FIR_Coefs[ 11 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 12 ], buf_ptr[ 23 ] ), FIR_Coefs[ 12 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 13 ], buf_ptr[ 22 ] ), FIR_Coefs[ 13 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 14 ], buf_ptr[ 21 ] ), FIR_Coefs[ 14 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 15 ], buf_ptr[ 20 ] ), FIR_Coefs[ 15 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 16 ], buf_ptr[ 19 ] ), FIR_Coefs[ 16 ] ); + res_Q6 = silk_SMLAWB( res_Q6, silk_ADD32( buf_ptr[ 17 ], buf_ptr[ 18 ] ), FIR_Coefs[ 17 ] ); + + /* Scale down, saturate and store in output array */ + *out++ = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( res_Q6, 6 ) ); + } + break; + default: + celt_assert( 0 ); + } + return out; +} + +/* Resample with a 2nd order AR filter followed by FIR interpolation */ +void silk_resampler_private_down_FIR( + void *SS, /* I/O Resampler state */ + opus_int16 out[], /* O Output signal */ + const opus_int16 in[], /* I Input signal */ + opus_int32 inLen /* I Number of input samples */ +) +{ + silk_resampler_state_struct *S = (silk_resampler_state_struct *)SS; + opus_int32 nSamplesIn; + opus_int32 max_index_Q16, index_increment_Q16; + VARDECL( opus_int32, buf ); + const opus_int16 *FIR_Coefs; + SAVE_STACK; + + ALLOC( buf, S->batchSize + S->FIR_Order, opus_int32 ); + + /* Copy buffered samples to start of buffer */ + silk_memcpy( buf, S->sFIR.i32, S->FIR_Order * sizeof( opus_int32 ) ); + + FIR_Coefs = &S->Coefs[ 2 ]; + + /* Iterate over blocks of frameSizeIn input samples */ + index_increment_Q16 = S->invRatio_Q16; + while( 1 ) { + nSamplesIn = silk_min( inLen, S->batchSize ); + + /* Second-order AR filter (output in Q8) */ + silk_resampler_private_AR2( S->sIIR, &buf[ S->FIR_Order ], in, S->Coefs, nSamplesIn ); + + max_index_Q16 = silk_LSHIFT32( nSamplesIn, 16 ); + + /* Interpolate filtered signal */ + out = silk_resampler_private_down_FIR_INTERPOL( out, buf, FIR_Coefs, S->FIR_Order, + S->FIR_Fracs, max_index_Q16, index_increment_Q16 ); + + in += nSamplesIn; + inLen -= nSamplesIn; + + if( inLen > 1 ) { + /* More iterations to do; copy last part of filtered signal to beginning of buffer */ + silk_memcpy( buf, &buf[ nSamplesIn ], S->FIR_Order * sizeof( opus_int32 ) ); + } else { + break; + } + } + + /* Copy last part of filtered signal to the state for the next call */ + silk_memcpy( S->sFIR.i32, &buf[ nSamplesIn ], S->FIR_Order * sizeof( opus_int32 ) ); + RESTORE_STACK; +} diff --git a/lib/opus/silk/resampler_private_up2_HQ.c b/lib/opus/silk/resampler_private_up2_HQ.c new file mode 100644 index 00000000..c7ec8de3 --- /dev/null +++ b/lib/opus/silk/resampler_private_up2_HQ.c @@ -0,0 +1,113 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" +#include "resampler_private.h" + +/* Upsample by a factor 2, high quality */ +/* Uses 2nd order allpass filters for the 2x upsampling, followed by a */ +/* notch filter just above Nyquist. */ +void silk_resampler_private_up2_HQ( + opus_int32 *S, /* I/O Resampler state [ 6 ] */ + opus_int16 *out, /* O Output signal [ 2 * len ] */ + const opus_int16 *in, /* I Input signal [ len ] */ + opus_int32 len /* I Number of input samples */ +) +{ + opus_int32 k; + opus_int32 in32, out32_1, out32_2, Y, X; + + silk_assert( silk_resampler_up2_hq_0[ 0 ] > 0 ); + silk_assert( silk_resampler_up2_hq_0[ 1 ] > 0 ); + silk_assert( silk_resampler_up2_hq_0[ 2 ] < 0 ); + silk_assert( silk_resampler_up2_hq_1[ 0 ] > 0 ); + silk_assert( silk_resampler_up2_hq_1[ 1 ] > 0 ); + silk_assert( silk_resampler_up2_hq_1[ 2 ] < 0 ); + + /* Internal variables and state are in Q10 format */ + for( k = 0; k < len; k++ ) { + /* Convert to Q10 */ + in32 = silk_LSHIFT( (opus_int32)in[ k ], 10 ); + + /* First all-pass section for even output sample */ + Y = silk_SUB32( in32, S[ 0 ] ); + X = silk_SMULWB( Y, silk_resampler_up2_hq_0[ 0 ] ); + out32_1 = silk_ADD32( S[ 0 ], X ); + S[ 0 ] = silk_ADD32( in32, X ); + + /* Second all-pass section for even output sample */ + Y = silk_SUB32( out32_1, S[ 1 ] ); + X = silk_SMULWB( Y, silk_resampler_up2_hq_0[ 1 ] ); + out32_2 = silk_ADD32( S[ 1 ], X ); + S[ 1 ] = silk_ADD32( out32_1, X ); + + /* Third all-pass section for even output sample */ + Y = silk_SUB32( out32_2, S[ 2 ] ); + X = silk_SMLAWB( Y, Y, silk_resampler_up2_hq_0[ 2 ] ); + out32_1 = silk_ADD32( S[ 2 ], X ); + S[ 2 ] = silk_ADD32( out32_2, X ); + + /* Apply gain in Q15, convert back to int16 and store to output */ + out[ 2 * k ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( out32_1, 10 ) ); + + /* First all-pass section for odd output sample */ + Y = silk_SUB32( in32, S[ 3 ] ); + X = silk_SMULWB( Y, silk_resampler_up2_hq_1[ 0 ] ); + out32_1 = silk_ADD32( S[ 3 ], X ); + S[ 3 ] = silk_ADD32( in32, X ); + + /* Second all-pass section for odd output sample */ + Y = silk_SUB32( out32_1, S[ 4 ] ); + X = silk_SMULWB( Y, silk_resampler_up2_hq_1[ 1 ] ); + out32_2 = silk_ADD32( S[ 4 ], X ); + S[ 4 ] = silk_ADD32( out32_1, X ); + + /* Third all-pass section for odd output sample */ + Y = silk_SUB32( out32_2, S[ 5 ] ); + X = silk_SMLAWB( Y, Y, silk_resampler_up2_hq_1[ 2 ] ); + out32_1 = silk_ADD32( S[ 5 ], X ); + S[ 5 ] = silk_ADD32( out32_2, X ); + + /* Apply gain in Q15, convert back to int16 and store to output */ + out[ 2 * k + 1 ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( out32_1, 10 ) ); + } +} + +void silk_resampler_private_up2_HQ_wrapper( + void *SS, /* I/O Resampler state (unused) */ + opus_int16 *out, /* O Output signal [ 2 * len ] */ + const opus_int16 *in, /* I Input signal [ len ] */ + opus_int32 len /* I Number of input samples */ +) +{ + silk_resampler_state_struct *S = (silk_resampler_state_struct *)SS; + silk_resampler_private_up2_HQ( S->sIIR, out, in, len ); +} diff --git a/lib/opus/silk/resampler_rom.c b/lib/opus/silk/resampler_rom.c new file mode 100644 index 00000000..5e6b0447 --- /dev/null +++ b/lib/opus/silk/resampler_rom.c @@ -0,0 +1,96 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* Filter coefficients for IIR/FIR polyphase resampling * + * Total size: 179 Words (358 Bytes) */ + +#include "resampler_private.h" + +/* Matlab code for the notch filter coefficients: */ +/* B = [1, 0.147, 1]; A = [1, 0.107, 0.89]; G = 0.93; freqz(G * B, A, 2^14, 16e3); axis([0, 8000, -10, 1]) */ +/* fprintf('\t%6d, %6d, %6d, %6d\n', round(B(2)*2^16), round(-A(2)*2^16), round((1-A(3))*2^16), round(G*2^15)) */ +/* const opus_int16 silk_resampler_up2_hq_notch[ 4 ] = { 9634, -7012, 7209, 30474 }; */ + +/* Tables with IIR and FIR coefficients for fractional downsamplers (123 Words) */ +silk_DWORD_ALIGN const opus_int16 silk_Resampler_3_4_COEFS[ 2 + 3 * RESAMPLER_DOWN_ORDER_FIR0 / 2 ] = { + -20694, -13867, + -49, 64, 17, -157, 353, -496, 163, 11047, 22205, + -39, 6, 91, -170, 186, 23, -896, 6336, 19928, + -19, -36, 102, -89, -24, 328, -951, 2568, 15909, +}; + +silk_DWORD_ALIGN const opus_int16 silk_Resampler_2_3_COEFS[ 2 + 2 * RESAMPLER_DOWN_ORDER_FIR0 / 2 ] = { + -14457, -14019, + 64, 128, -122, 36, 310, -768, 584, 9267, 17733, + 12, 128, 18, -142, 288, -117, -865, 4123, 14459, +}; + +silk_DWORD_ALIGN const opus_int16 silk_Resampler_1_2_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR1 / 2 ] = { + 616, -14323, + -10, 39, 58, -46, -84, 120, 184, -315, -541, 1284, 5380, 9024, +}; + +silk_DWORD_ALIGN const opus_int16 silk_Resampler_1_3_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR2 / 2 ] = { + 16102, -15162, + -13, 0, 20, 26, 5, -31, -43, -4, 65, 90, 7, -157, -248, -44, 593, 1583, 2612, 3271, +}; + +silk_DWORD_ALIGN const opus_int16 silk_Resampler_1_4_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR2 / 2 ] = { + 22500, -15099, + 3, -14, -20, -15, 2, 25, 37, 25, -16, -71, -107, -79, 50, 292, 623, 982, 1288, 1464, +}; + +silk_DWORD_ALIGN const opus_int16 silk_Resampler_1_6_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR2 / 2 ] = { + 27540, -15257, + 17, 12, 8, 1, -10, -22, -30, -32, -22, 3, 44, 100, 168, 243, 317, 381, 429, 455, +}; + +silk_DWORD_ALIGN const opus_int16 silk_Resampler_2_3_COEFS_LQ[ 2 + 2 * 2 ] = { + -2797, -6507, + 4697, 10739, + 1567, 8276, +}; + +/* Table with interplation fractions of 1/24, 3/24, 5/24, ... , 23/24 : 23/24 (46 Words) */ +silk_DWORD_ALIGN const opus_int16 silk_resampler_frac_FIR_12[ 12 ][ RESAMPLER_ORDER_FIR_12 / 2 ] = { + { 189, -600, 617, 30567 }, + { 117, -159, -1070, 29704 }, + { 52, 221, -2392, 28276 }, + { -4, 529, -3350, 26341 }, + { -48, 758, -3956, 23973 }, + { -80, 905, -4235, 21254 }, + { -99, 972, -4222, 18278 }, + { -107, 967, -3957, 15143 }, + { -103, 896, -3487, 11950 }, + { -91, 773, -2865, 8798 }, + { -71, 611, -2143, 5784 }, + { -46, 425, -1375, 2996 }, +}; diff --git a/lib/opus/silk/resampler_rom.h b/lib/opus/silk/resampler_rom.h new file mode 100644 index 00000000..490b3388 --- /dev/null +++ b/lib/opus/silk/resampler_rom.h @@ -0,0 +1,68 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_FIX_RESAMPLER_ROM_H +#define SILK_FIX_RESAMPLER_ROM_H + +#ifdef __cplusplus +extern "C" +{ +#endif + +#include "typedef.h" +#include "resampler_structs.h" + +#define RESAMPLER_DOWN_ORDER_FIR0 18 +#define RESAMPLER_DOWN_ORDER_FIR1 24 +#define RESAMPLER_DOWN_ORDER_FIR2 36 +#define RESAMPLER_ORDER_FIR_12 8 + +/* Tables for 2x downsampler */ +static const opus_int16 silk_resampler_down2_0 = 9872; +static const opus_int16 silk_resampler_down2_1 = 39809 - 65536; + +/* Tables for 2x upsampler, high quality */ +static const opus_int16 silk_resampler_up2_hq_0[ 3 ] = { 1746, 14986, 39083 - 65536 }; +static const opus_int16 silk_resampler_up2_hq_1[ 3 ] = { 6854, 25769, 55542 - 65536 }; + +/* Tables with IIR and FIR coefficients for fractional downsamplers */ +extern const opus_int16 silk_Resampler_3_4_COEFS[ 2 + 3 * RESAMPLER_DOWN_ORDER_FIR0 / 2 ]; +extern const opus_int16 silk_Resampler_2_3_COEFS[ 2 + 2 * RESAMPLER_DOWN_ORDER_FIR0 / 2 ]; +extern const opus_int16 silk_Resampler_1_2_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR1 / 2 ]; +extern const opus_int16 silk_Resampler_1_3_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR2 / 2 ]; +extern const opus_int16 silk_Resampler_1_4_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR2 / 2 ]; +extern const opus_int16 silk_Resampler_1_6_COEFS[ 2 + RESAMPLER_DOWN_ORDER_FIR2 / 2 ]; +extern const opus_int16 silk_Resampler_2_3_COEFS_LQ[ 2 + 2 * 2 ]; + +/* Table with interplation fractions of 1/24, 3/24, ..., 23/24 */ +extern const opus_int16 silk_resampler_frac_FIR_12[ 12 ][ RESAMPLER_ORDER_FIR_12 / 2 ]; + +#ifdef __cplusplus +} +#endif + +#endif /* SILK_FIX_RESAMPLER_ROM_H */ diff --git a/lib/opus/silk/resampler_structs.h b/lib/opus/silk/resampler_structs.h new file mode 100644 index 00000000..9e9457d1 --- /dev/null +++ b/lib/opus/silk/resampler_structs.h @@ -0,0 +1,60 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_RESAMPLER_STRUCTS_H +#define SILK_RESAMPLER_STRUCTS_H + +#ifdef __cplusplus +extern "C" { +#endif + +#define SILK_RESAMPLER_MAX_FIR_ORDER 36 +#define SILK_RESAMPLER_MAX_IIR_ORDER 6 + +typedef struct _silk_resampler_state_struct{ + opus_int32 sIIR[ SILK_RESAMPLER_MAX_IIR_ORDER ]; /* this must be the first element of this struct */ + union{ + opus_int32 i32[ SILK_RESAMPLER_MAX_FIR_ORDER ]; + opus_int16 i16[ SILK_RESAMPLER_MAX_FIR_ORDER ]; + } sFIR; + opus_int16 delayBuf[ 48 ]; + opus_int resampler_function; + opus_int batchSize; + opus_int32 invRatio_Q16; + opus_int FIR_Order; + opus_int FIR_Fracs; + opus_int Fs_in_kHz; + opus_int Fs_out_kHz; + opus_int inputDelay; + const opus_int16 *Coefs; +} silk_resampler_state_struct; + +#ifdef __cplusplus +} +#endif +#endif /* SILK_RESAMPLER_STRUCTS_H */ + diff --git a/lib/opus/silk/shell_coder.c b/lib/opus/silk/shell_coder.c new file mode 100644 index 00000000..4af34147 --- /dev/null +++ b/lib/opus/silk/shell_coder.c @@ -0,0 +1,151 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* shell coder; pulse-subframe length is hardcoded */ + +static OPUS_INLINE void combine_pulses( + opus_int *out, /* O combined pulses vector [len] */ + const opus_int *in, /* I input vector [2 * len] */ + const opus_int len /* I number of OUTPUT samples */ +) +{ + opus_int k; + for( k = 0; k < len; k++ ) { + out[ k ] = in[ 2 * k ] + in[ 2 * k + 1 ]; + } +} + +static OPUS_INLINE void encode_split( + ec_enc *psRangeEnc, /* I/O compressor data structure */ + const opus_int p_child1, /* I pulse amplitude of first child subframe */ + const opus_int p, /* I pulse amplitude of current subframe */ + const opus_uint8 *shell_table /* I table of shell cdfs */ +) +{ + if( p > 0 ) { + ec_enc_icdf( psRangeEnc, p_child1, &shell_table[ silk_shell_code_table_offsets[ p ] ], 8 ); + } +} + +static OPUS_INLINE void decode_split( + opus_int16 *p_child1, /* O pulse amplitude of first child subframe */ + opus_int16 *p_child2, /* O pulse amplitude of second child subframe */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + const opus_int p, /* I pulse amplitude of current subframe */ + const opus_uint8 *shell_table /* I table of shell cdfs */ +) +{ + if( p > 0 ) { + p_child1[ 0 ] = ec_dec_icdf( psRangeDec, &shell_table[ silk_shell_code_table_offsets[ p ] ], 8 ); + p_child2[ 0 ] = p - p_child1[ 0 ]; + } else { + p_child1[ 0 ] = 0; + p_child2[ 0 ] = 0; + } +} + +/* Shell encoder, operates on one shell code frame of 16 pulses */ +void silk_shell_encoder( + ec_enc *psRangeEnc, /* I/O compressor data structure */ + const opus_int *pulses0 /* I data: nonnegative pulse amplitudes */ +) +{ + opus_int pulses1[ 8 ], pulses2[ 4 ], pulses3[ 2 ], pulses4[ 1 ]; + + /* this function operates on one shell code frame of 16 pulses */ + silk_assert( SHELL_CODEC_FRAME_LENGTH == 16 ); + + /* tree representation per pulse-subframe */ + combine_pulses( pulses1, pulses0, 8 ); + combine_pulses( pulses2, pulses1, 4 ); + combine_pulses( pulses3, pulses2, 2 ); + combine_pulses( pulses4, pulses3, 1 ); + + encode_split( psRangeEnc, pulses3[ 0 ], pulses4[ 0 ], silk_shell_code_table3 ); + + encode_split( psRangeEnc, pulses2[ 0 ], pulses3[ 0 ], silk_shell_code_table2 ); + + encode_split( psRangeEnc, pulses1[ 0 ], pulses2[ 0 ], silk_shell_code_table1 ); + encode_split( psRangeEnc, pulses0[ 0 ], pulses1[ 0 ], silk_shell_code_table0 ); + encode_split( psRangeEnc, pulses0[ 2 ], pulses1[ 1 ], silk_shell_code_table0 ); + + encode_split( psRangeEnc, pulses1[ 2 ], pulses2[ 1 ], silk_shell_code_table1 ); + encode_split( psRangeEnc, pulses0[ 4 ], pulses1[ 2 ], silk_shell_code_table0 ); + encode_split( psRangeEnc, pulses0[ 6 ], pulses1[ 3 ], silk_shell_code_table0 ); + + encode_split( psRangeEnc, pulses2[ 2 ], pulses3[ 1 ], silk_shell_code_table2 ); + + encode_split( psRangeEnc, pulses1[ 4 ], pulses2[ 2 ], silk_shell_code_table1 ); + encode_split( psRangeEnc, pulses0[ 8 ], pulses1[ 4 ], silk_shell_code_table0 ); + encode_split( psRangeEnc, pulses0[ 10 ], pulses1[ 5 ], silk_shell_code_table0 ); + + encode_split( psRangeEnc, pulses1[ 6 ], pulses2[ 3 ], silk_shell_code_table1 ); + encode_split( psRangeEnc, pulses0[ 12 ], pulses1[ 6 ], silk_shell_code_table0 ); + encode_split( psRangeEnc, pulses0[ 14 ], pulses1[ 7 ], silk_shell_code_table0 ); +} + + +/* Shell decoder, operates on one shell code frame of 16 pulses */ +void silk_shell_decoder( + opus_int16 *pulses0, /* O data: nonnegative pulse amplitudes */ + ec_dec *psRangeDec, /* I/O Compressor data structure */ + const opus_int pulses4 /* I number of pulses per pulse-subframe */ +) +{ + opus_int16 pulses3[ 2 ], pulses2[ 4 ], pulses1[ 8 ]; + + /* this function operates on one shell code frame of 16 pulses */ + silk_assert( SHELL_CODEC_FRAME_LENGTH == 16 ); + + decode_split( &pulses3[ 0 ], &pulses3[ 1 ], psRangeDec, pulses4, silk_shell_code_table3 ); + + decode_split( &pulses2[ 0 ], &pulses2[ 1 ], psRangeDec, pulses3[ 0 ], silk_shell_code_table2 ); + + decode_split( &pulses1[ 0 ], &pulses1[ 1 ], psRangeDec, pulses2[ 0 ], silk_shell_code_table1 ); + decode_split( &pulses0[ 0 ], &pulses0[ 1 ], psRangeDec, pulses1[ 0 ], silk_shell_code_table0 ); + decode_split( &pulses0[ 2 ], &pulses0[ 3 ], psRangeDec, pulses1[ 1 ], silk_shell_code_table0 ); + + decode_split( &pulses1[ 2 ], &pulses1[ 3 ], psRangeDec, pulses2[ 1 ], silk_shell_code_table1 ); + decode_split( &pulses0[ 4 ], &pulses0[ 5 ], psRangeDec, pulses1[ 2 ], silk_shell_code_table0 ); + decode_split( &pulses0[ 6 ], &pulses0[ 7 ], psRangeDec, pulses1[ 3 ], silk_shell_code_table0 ); + + decode_split( &pulses2[ 2 ], &pulses2[ 3 ], psRangeDec, pulses3[ 1 ], silk_shell_code_table2 ); + + decode_split( &pulses1[ 4 ], &pulses1[ 5 ], psRangeDec, pulses2[ 2 ], silk_shell_code_table1 ); + decode_split( &pulses0[ 8 ], &pulses0[ 9 ], psRangeDec, pulses1[ 4 ], silk_shell_code_table0 ); + decode_split( &pulses0[ 10 ], &pulses0[ 11 ], psRangeDec, pulses1[ 5 ], silk_shell_code_table0 ); + + decode_split( &pulses1[ 6 ], &pulses1[ 7 ], psRangeDec, pulses2[ 3 ], silk_shell_code_table1 ); + decode_split( &pulses0[ 12 ], &pulses0[ 13 ], psRangeDec, pulses1[ 6 ], silk_shell_code_table0 ); + decode_split( &pulses0[ 14 ], &pulses0[ 15 ], psRangeDec, pulses1[ 7 ], silk_shell_code_table0 ); +} diff --git a/lib/opus/silk/sigm_Q15.c b/lib/opus/silk/sigm_Q15.c new file mode 100644 index 00000000..3c507d25 --- /dev/null +++ b/lib/opus/silk/sigm_Q15.c @@ -0,0 +1,76 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* Approximate sigmoid function */ + +#include "SigProc_FIX.h" + +/* fprintf(1, '%d, ', round(1024 * ([1 ./ (1 + exp(-(1:5))), 1] - 1 ./ (1 + exp(-(0:5)))))); */ +static const opus_int32 sigm_LUT_slope_Q10[ 6 ] = { + 237, 153, 73, 30, 12, 7 +}; +/* fprintf(1, '%d, ', round(32767 * 1 ./ (1 + exp(-(0:5))))); */ +static const opus_int32 sigm_LUT_pos_Q15[ 6 ] = { + 16384, 23955, 28861, 31213, 32178, 32548 +}; +/* fprintf(1, '%d, ', round(32767 * 1 ./ (1 + exp((0:5))))); */ +static const opus_int32 sigm_LUT_neg_Q15[ 6 ] = { + 16384, 8812, 3906, 1554, 589, 219 +}; + +opus_int silk_sigm_Q15( + opus_int in_Q5 /* I */ +) +{ + opus_int ind; + + if( in_Q5 < 0 ) { + /* Negative input */ + in_Q5 = -in_Q5; + if( in_Q5 >= 6 * 32 ) { + return 0; /* Clip */ + } else { + /* Linear interpolation of look up table */ + ind = silk_RSHIFT( in_Q5, 5 ); + return( sigm_LUT_neg_Q15[ ind ] - silk_SMULBB( sigm_LUT_slope_Q10[ ind ], in_Q5 & 0x1F ) ); + } + } else { + /* Positive input */ + if( in_Q5 >= 6 * 32 ) { + return 32767; /* clip */ + } else { + /* Linear interpolation of look up table */ + ind = silk_RSHIFT( in_Q5, 5 ); + return( sigm_LUT_pos_Q15[ ind ] + silk_SMULBB( sigm_LUT_slope_Q10[ ind ], in_Q5 & 0x1F ) ); + } + } +} + diff --git a/lib/opus/silk/sort.c b/lib/opus/silk/sort.c new file mode 100644 index 00000000..4fba16f8 --- /dev/null +++ b/lib/opus/silk/sort.c @@ -0,0 +1,154 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* Insertion sort (fast for already almost sorted arrays): */ +/* Best case: O(n) for an already sorted array */ +/* Worst case: O(n^2) for an inversely sorted array */ +/* */ +/* Shell short: https://en.wikipedia.org/wiki/Shell_sort */ + +#include "SigProc_FIX.h" + +void silk_insertion_sort_increasing( + opus_int32 *a, /* I/O Unsorted / Sorted vector */ + opus_int *idx, /* O Index vector for the sorted elements */ + const opus_int L, /* I Vector length */ + const opus_int K /* I Number of correctly sorted positions */ +) +{ + opus_int32 value; + opus_int i, j; + + /* Safety checks */ + celt_assert( K > 0 ); + celt_assert( L > 0 ); + celt_assert( L >= K ); + + /* Write start indices in index vector */ + for( i = 0; i < K; i++ ) { + idx[ i ] = i; + } + + /* Sort vector elements by value, increasing order */ + for( i = 1; i < K; i++ ) { + value = a[ i ]; + for( j = i - 1; ( j >= 0 ) && ( value < a[ j ] ); j-- ) { + a[ j + 1 ] = a[ j ]; /* Shift value */ + idx[ j + 1 ] = idx[ j ]; /* Shift index */ + } + a[ j + 1 ] = value; /* Write value */ + idx[ j + 1 ] = i; /* Write index */ + } + + /* If less than L values are asked for, check the remaining values, */ + /* but only spend CPU to ensure that the K first values are correct */ + for( i = K; i < L; i++ ) { + value = a[ i ]; + if( value < a[ K - 1 ] ) { + for( j = K - 2; ( j >= 0 ) && ( value < a[ j ] ); j-- ) { + a[ j + 1 ] = a[ j ]; /* Shift value */ + idx[ j + 1 ] = idx[ j ]; /* Shift index */ + } + a[ j + 1 ] = value; /* Write value */ + idx[ j + 1 ] = i; /* Write index */ + } + } +} + +#ifdef FIXED_POINT +/* This function is only used by the fixed-point build */ +void silk_insertion_sort_decreasing_int16( + opus_int16 *a, /* I/O Unsorted / Sorted vector */ + opus_int *idx, /* O Index vector for the sorted elements */ + const opus_int L, /* I Vector length */ + const opus_int K /* I Number of correctly sorted positions */ +) +{ + opus_int i, j; + opus_int value; + + /* Safety checks */ + celt_assert( K > 0 ); + celt_assert( L > 0 ); + celt_assert( L >= K ); + + /* Write start indices in index vector */ + for( i = 0; i < K; i++ ) { + idx[ i ] = i; + } + + /* Sort vector elements by value, decreasing order */ + for( i = 1; i < K; i++ ) { + value = a[ i ]; + for( j = i - 1; ( j >= 0 ) && ( value > a[ j ] ); j-- ) { + a[ j + 1 ] = a[ j ]; /* Shift value */ + idx[ j + 1 ] = idx[ j ]; /* Shift index */ + } + a[ j + 1 ] = value; /* Write value */ + idx[ j + 1 ] = i; /* Write index */ + } + + /* If less than L values are asked for, check the remaining values, */ + /* but only spend CPU to ensure that the K first values are correct */ + for( i = K; i < L; i++ ) { + value = a[ i ]; + if( value > a[ K - 1 ] ) { + for( j = K - 2; ( j >= 0 ) && ( value > a[ j ] ); j-- ) { + a[ j + 1 ] = a[ j ]; /* Shift value */ + idx[ j + 1 ] = idx[ j ]; /* Shift index */ + } + a[ j + 1 ] = value; /* Write value */ + idx[ j + 1 ] = i; /* Write index */ + } + } +} +#endif + +void silk_insertion_sort_increasing_all_values_int16( + opus_int16 *a, /* I/O Unsorted / Sorted vector */ + const opus_int L /* I Vector length */ +) +{ + opus_int value; + opus_int i, j; + + /* Safety checks */ + celt_assert( L > 0 ); + + /* Sort vector elements by value, increasing order */ + for( i = 1; i < L; i++ ) { + value = a[ i ]; + for( j = i - 1; ( j >= 0 ) && ( value < a[ j ] ); j-- ) { + a[ j + 1 ] = a[ j ]; /* Shift value */ + } + a[ j + 1 ] = value; /* Write value */ + } +} diff --git a/lib/opus/silk/stereo_LR_to_MS.c b/lib/opus/silk/stereo_LR_to_MS.c new file mode 100644 index 00000000..751452cb --- /dev/null +++ b/lib/opus/silk/stereo_LR_to_MS.c @@ -0,0 +1,229 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" +#include "stack_alloc.h" + +/* Convert Left/Right stereo signal to adaptive Mid/Side representation */ +void silk_stereo_LR_to_MS( + stereo_enc_state *state, /* I/O State */ + opus_int16 x1[], /* I/O Left input signal, becomes mid signal */ + opus_int16 x2[], /* I/O Right input signal, becomes side signal */ + opus_int8 ix[ 2 ][ 3 ], /* O Quantization indices */ + opus_int8 *mid_only_flag, /* O Flag: only mid signal coded */ + opus_int32 mid_side_rates_bps[], /* O Bitrates for mid and side signals */ + opus_int32 total_rate_bps, /* I Total bitrate */ + opus_int prev_speech_act_Q8, /* I Speech activity level in previous frame */ + opus_int toMono, /* I Last frame before a stereo->mono transition */ + opus_int fs_kHz, /* I Sample rate (kHz) */ + opus_int frame_length /* I Number of samples */ +) +{ + opus_int n, is10msFrame, denom_Q16, delta0_Q13, delta1_Q13; + opus_int32 sum, diff, smooth_coef_Q16, pred_Q13[ 2 ], pred0_Q13, pred1_Q13; + opus_int32 LP_ratio_Q14, HP_ratio_Q14, frac_Q16, frac_3_Q16, min_mid_rate_bps, width_Q14, w_Q24, deltaw_Q24; + VARDECL( opus_int16, side ); + VARDECL( opus_int16, LP_mid ); + VARDECL( opus_int16, HP_mid ); + VARDECL( opus_int16, LP_side ); + VARDECL( opus_int16, HP_side ); + opus_int16 *mid = &x1[ -2 ]; + SAVE_STACK; + + ALLOC( side, frame_length + 2, opus_int16 ); + /* Convert to basic mid/side signals */ + for( n = 0; n < frame_length + 2; n++ ) { + sum = x1[ n - 2 ] + (opus_int32)x2[ n - 2 ]; + diff = x1[ n - 2 ] - (opus_int32)x2[ n - 2 ]; + mid[ n ] = (opus_int16)silk_RSHIFT_ROUND( sum, 1 ); + side[ n ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( diff, 1 ) ); + } + + /* Buffering */ + silk_memcpy( mid, state->sMid, 2 * sizeof( opus_int16 ) ); + silk_memcpy( side, state->sSide, 2 * sizeof( opus_int16 ) ); + silk_memcpy( state->sMid, &mid[ frame_length ], 2 * sizeof( opus_int16 ) ); + silk_memcpy( state->sSide, &side[ frame_length ], 2 * sizeof( opus_int16 ) ); + + /* LP and HP filter mid signal */ + ALLOC( LP_mid, frame_length, opus_int16 ); + ALLOC( HP_mid, frame_length, opus_int16 ); + for( n = 0; n < frame_length; n++ ) { + sum = silk_RSHIFT_ROUND( silk_ADD_LSHIFT32( mid[ n ] + (opus_int32)mid[ n + 2 ], mid[ n + 1 ], 1 ), 2 ); + LP_mid[ n ] = sum; + HP_mid[ n ] = mid[ n + 1 ] - sum; + } + + /* LP and HP filter side signal */ + ALLOC( LP_side, frame_length, opus_int16 ); + ALLOC( HP_side, frame_length, opus_int16 ); + for( n = 0; n < frame_length; n++ ) { + sum = silk_RSHIFT_ROUND( silk_ADD_LSHIFT32( side[ n ] + (opus_int32)side[ n + 2 ], side[ n + 1 ], 1 ), 2 ); + LP_side[ n ] = sum; + HP_side[ n ] = side[ n + 1 ] - sum; + } + + /* Find energies and predictors */ + is10msFrame = frame_length == 10 * fs_kHz; + smooth_coef_Q16 = is10msFrame ? + SILK_FIX_CONST( STEREO_RATIO_SMOOTH_COEF / 2, 16 ) : + SILK_FIX_CONST( STEREO_RATIO_SMOOTH_COEF, 16 ); + smooth_coef_Q16 = silk_SMULWB( silk_SMULBB( prev_speech_act_Q8, prev_speech_act_Q8 ), smooth_coef_Q16 ); + + pred_Q13[ 0 ] = silk_stereo_find_predictor( &LP_ratio_Q14, LP_mid, LP_side, &state->mid_side_amp_Q0[ 0 ], frame_length, smooth_coef_Q16 ); + pred_Q13[ 1 ] = silk_stereo_find_predictor( &HP_ratio_Q14, HP_mid, HP_side, &state->mid_side_amp_Q0[ 2 ], frame_length, smooth_coef_Q16 ); + /* Ratio of the norms of residual and mid signals */ + frac_Q16 = silk_SMLABB( HP_ratio_Q14, LP_ratio_Q14, 3 ); + frac_Q16 = silk_min( frac_Q16, SILK_FIX_CONST( 1, 16 ) ); + + /* Determine bitrate distribution between mid and side, and possibly reduce stereo width */ + total_rate_bps -= is10msFrame ? 1200 : 600; /* Subtract approximate bitrate for coding stereo parameters */ + if( total_rate_bps < 1 ) { + total_rate_bps = 1; + } + min_mid_rate_bps = silk_SMLABB( 2000, fs_kHz, 600 ); + silk_assert( min_mid_rate_bps < 32767 ); + /* Default bitrate distribution: 8 parts for Mid and (5+3*frac) parts for Side. so: mid_rate = ( 8 / ( 13 + 3 * frac ) ) * total_ rate */ + frac_3_Q16 = silk_MUL( 3, frac_Q16 ); + mid_side_rates_bps[ 0 ] = silk_DIV32_varQ( total_rate_bps, SILK_FIX_CONST( 8 + 5, 16 ) + frac_3_Q16, 16+3 ); + /* If Mid bitrate below minimum, reduce stereo width */ + if( mid_side_rates_bps[ 0 ] < min_mid_rate_bps ) { + mid_side_rates_bps[ 0 ] = min_mid_rate_bps; + mid_side_rates_bps[ 1 ] = total_rate_bps - mid_side_rates_bps[ 0 ]; + /* width = 4 * ( 2 * side_rate - min_rate ) / ( ( 1 + 3 * frac ) * min_rate ) */ + width_Q14 = silk_DIV32_varQ( silk_LSHIFT( mid_side_rates_bps[ 1 ], 1 ) - min_mid_rate_bps, + silk_SMULWB( SILK_FIX_CONST( 1, 16 ) + frac_3_Q16, min_mid_rate_bps ), 14+2 ); + width_Q14 = silk_LIMIT( width_Q14, 0, SILK_FIX_CONST( 1, 14 ) ); + } else { + mid_side_rates_bps[ 1 ] = total_rate_bps - mid_side_rates_bps[ 0 ]; + width_Q14 = SILK_FIX_CONST( 1, 14 ); + } + + /* Smoother */ + state->smth_width_Q14 = (opus_int16)silk_SMLAWB( state->smth_width_Q14, width_Q14 - state->smth_width_Q14, smooth_coef_Q16 ); + + /* At very low bitrates or for inputs that are nearly amplitude panned, switch to panned-mono coding */ + *mid_only_flag = 0; + if( toMono ) { + /* Last frame before stereo->mono transition; collapse stereo width */ + width_Q14 = 0; + pred_Q13[ 0 ] = 0; + pred_Q13[ 1 ] = 0; + silk_stereo_quant_pred( pred_Q13, ix ); + } else if( state->width_prev_Q14 == 0 && + ( 8 * total_rate_bps < 13 * min_mid_rate_bps || silk_SMULWB( frac_Q16, state->smth_width_Q14 ) < SILK_FIX_CONST( 0.05, 14 ) ) ) + { + /* Code as panned-mono; previous frame already had zero width */ + /* Scale down and quantize predictors */ + pred_Q13[ 0 ] = silk_RSHIFT( silk_SMULBB( state->smth_width_Q14, pred_Q13[ 0 ] ), 14 ); + pred_Q13[ 1 ] = silk_RSHIFT( silk_SMULBB( state->smth_width_Q14, pred_Q13[ 1 ] ), 14 ); + silk_stereo_quant_pred( pred_Q13, ix ); + /* Collapse stereo width */ + width_Q14 = 0; + pred_Q13[ 0 ] = 0; + pred_Q13[ 1 ] = 0; + mid_side_rates_bps[ 0 ] = total_rate_bps; + mid_side_rates_bps[ 1 ] = 0; + *mid_only_flag = 1; + } else if( state->width_prev_Q14 != 0 && + ( 8 * total_rate_bps < 11 * min_mid_rate_bps || silk_SMULWB( frac_Q16, state->smth_width_Q14 ) < SILK_FIX_CONST( 0.02, 14 ) ) ) + { + /* Transition to zero-width stereo */ + /* Scale down and quantize predictors */ + pred_Q13[ 0 ] = silk_RSHIFT( silk_SMULBB( state->smth_width_Q14, pred_Q13[ 0 ] ), 14 ); + pred_Q13[ 1 ] = silk_RSHIFT( silk_SMULBB( state->smth_width_Q14, pred_Q13[ 1 ] ), 14 ); + silk_stereo_quant_pred( pred_Q13, ix ); + /* Collapse stereo width */ + width_Q14 = 0; + pred_Q13[ 0 ] = 0; + pred_Q13[ 1 ] = 0; + } else if( state->smth_width_Q14 > SILK_FIX_CONST( 0.95, 14 ) ) { + /* Full-width stereo coding */ + silk_stereo_quant_pred( pred_Q13, ix ); + width_Q14 = SILK_FIX_CONST( 1, 14 ); + } else { + /* Reduced-width stereo coding; scale down and quantize predictors */ + pred_Q13[ 0 ] = silk_RSHIFT( silk_SMULBB( state->smth_width_Q14, pred_Q13[ 0 ] ), 14 ); + pred_Q13[ 1 ] = silk_RSHIFT( silk_SMULBB( state->smth_width_Q14, pred_Q13[ 1 ] ), 14 ); + silk_stereo_quant_pred( pred_Q13, ix ); + width_Q14 = state->smth_width_Q14; + } + + /* Make sure to keep on encoding until the tapered output has been transmitted */ + if( *mid_only_flag == 1 ) { + state->silent_side_len += frame_length - STEREO_INTERP_LEN_MS * fs_kHz; + if( state->silent_side_len < LA_SHAPE_MS * fs_kHz ) { + *mid_only_flag = 0; + } else { + /* Limit to avoid wrapping around */ + state->silent_side_len = 10000; + } + } else { + state->silent_side_len = 0; + } + + if( *mid_only_flag == 0 && mid_side_rates_bps[ 1 ] < 1 ) { + mid_side_rates_bps[ 1 ] = 1; + mid_side_rates_bps[ 0 ] = silk_max_int( 1, total_rate_bps - mid_side_rates_bps[ 1 ]); + } + + /* Interpolate predictors and subtract prediction from side channel */ + pred0_Q13 = -state->pred_prev_Q13[ 0 ]; + pred1_Q13 = -state->pred_prev_Q13[ 1 ]; + w_Q24 = silk_LSHIFT( state->width_prev_Q14, 10 ); + denom_Q16 = silk_DIV32_16( (opus_int32)1 << 16, STEREO_INTERP_LEN_MS * fs_kHz ); + delta0_Q13 = -silk_RSHIFT_ROUND( silk_SMULBB( pred_Q13[ 0 ] - state->pred_prev_Q13[ 0 ], denom_Q16 ), 16 ); + delta1_Q13 = -silk_RSHIFT_ROUND( silk_SMULBB( pred_Q13[ 1 ] - state->pred_prev_Q13[ 1 ], denom_Q16 ), 16 ); + deltaw_Q24 = silk_LSHIFT( silk_SMULWB( width_Q14 - state->width_prev_Q14, denom_Q16 ), 10 ); + for( n = 0; n < STEREO_INTERP_LEN_MS * fs_kHz; n++ ) { + pred0_Q13 += delta0_Q13; + pred1_Q13 += delta1_Q13; + w_Q24 += deltaw_Q24; + sum = silk_LSHIFT( silk_ADD_LSHIFT32( mid[ n ] + (opus_int32)mid[ n + 2 ], mid[ n + 1 ], 1 ), 9 ); /* Q11 */ + sum = silk_SMLAWB( silk_SMULWB( w_Q24, side[ n + 1 ] ), sum, pred0_Q13 ); /* Q8 */ + sum = silk_SMLAWB( sum, silk_LSHIFT( (opus_int32)mid[ n + 1 ], 11 ), pred1_Q13 ); /* Q8 */ + x2[ n - 1 ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( sum, 8 ) ); + } + + pred0_Q13 = -pred_Q13[ 0 ]; + pred1_Q13 = -pred_Q13[ 1 ]; + w_Q24 = silk_LSHIFT( width_Q14, 10 ); + for( n = STEREO_INTERP_LEN_MS * fs_kHz; n < frame_length; n++ ) { + sum = silk_LSHIFT( silk_ADD_LSHIFT32( mid[ n ] + (opus_int32)mid[ n + 2 ], mid[ n + 1 ], 1 ), 9 ); /* Q11 */ + sum = silk_SMLAWB( silk_SMULWB( w_Q24, side[ n + 1 ] ), sum, pred0_Q13 ); /* Q8 */ + sum = silk_SMLAWB( sum, silk_LSHIFT( (opus_int32)mid[ n + 1 ], 11 ), pred1_Q13 ); /* Q8 */ + x2[ n - 1 ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( sum, 8 ) ); + } + state->pred_prev_Q13[ 0 ] = (opus_int16)pred_Q13[ 0 ]; + state->pred_prev_Q13[ 1 ] = (opus_int16)pred_Q13[ 1 ]; + state->width_prev_Q14 = (opus_int16)width_Q14; + RESTORE_STACK; +} diff --git a/lib/opus/silk/stereo_MS_to_LR.c b/lib/opus/silk/stereo_MS_to_LR.c new file mode 100644 index 00000000..1e01bb6e --- /dev/null +++ b/lib/opus/silk/stereo_MS_to_LR.c @@ -0,0 +1,85 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Convert adaptive Mid/Side representation to Left/Right stereo signal */ +void silk_stereo_MS_to_LR( + stereo_dec_state *state, /* I/O State */ + opus_int16 x1[], /* I/O Left input signal, becomes mid signal */ + opus_int16 x2[], /* I/O Right input signal, becomes side signal */ + const opus_int32 pred_Q13[], /* I Predictors */ + opus_int fs_kHz, /* I Samples rate (kHz) */ + opus_int frame_length /* I Number of samples */ +) +{ + opus_int n, denom_Q16, delta0_Q13, delta1_Q13; + opus_int32 sum, diff, pred0_Q13, pred1_Q13; + + /* Buffering */ + silk_memcpy( x1, state->sMid, 2 * sizeof( opus_int16 ) ); + silk_memcpy( x2, state->sSide, 2 * sizeof( opus_int16 ) ); + silk_memcpy( state->sMid, &x1[ frame_length ], 2 * sizeof( opus_int16 ) ); + silk_memcpy( state->sSide, &x2[ frame_length ], 2 * sizeof( opus_int16 ) ); + + /* Interpolate predictors and add prediction to side channel */ + pred0_Q13 = state->pred_prev_Q13[ 0 ]; + pred1_Q13 = state->pred_prev_Q13[ 1 ]; + denom_Q16 = silk_DIV32_16( (opus_int32)1 << 16, STEREO_INTERP_LEN_MS * fs_kHz ); + delta0_Q13 = silk_RSHIFT_ROUND( silk_SMULBB( pred_Q13[ 0 ] - state->pred_prev_Q13[ 0 ], denom_Q16 ), 16 ); + delta1_Q13 = silk_RSHIFT_ROUND( silk_SMULBB( pred_Q13[ 1 ] - state->pred_prev_Q13[ 1 ], denom_Q16 ), 16 ); + for( n = 0; n < STEREO_INTERP_LEN_MS * fs_kHz; n++ ) { + pred0_Q13 += delta0_Q13; + pred1_Q13 += delta1_Q13; + sum = silk_LSHIFT( silk_ADD_LSHIFT32( x1[ n ] + (opus_int32)x1[ n + 2 ], x1[ n + 1 ], 1 ), 9 ); /* Q11 */ + sum = silk_SMLAWB( silk_LSHIFT( (opus_int32)x2[ n + 1 ], 8 ), sum, pred0_Q13 ); /* Q8 */ + sum = silk_SMLAWB( sum, silk_LSHIFT( (opus_int32)x1[ n + 1 ], 11 ), pred1_Q13 ); /* Q8 */ + x2[ n + 1 ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( sum, 8 ) ); + } + pred0_Q13 = pred_Q13[ 0 ]; + pred1_Q13 = pred_Q13[ 1 ]; + for( n = STEREO_INTERP_LEN_MS * fs_kHz; n < frame_length; n++ ) { + sum = silk_LSHIFT( silk_ADD_LSHIFT32( x1[ n ] + (opus_int32)x1[ n + 2 ], x1[ n + 1 ], 1 ), 9 ); /* Q11 */ + sum = silk_SMLAWB( silk_LSHIFT( (opus_int32)x2[ n + 1 ], 8 ), sum, pred0_Q13 ); /* Q8 */ + sum = silk_SMLAWB( sum, silk_LSHIFT( (opus_int32)x1[ n + 1 ], 11 ), pred1_Q13 ); /* Q8 */ + x2[ n + 1 ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( sum, 8 ) ); + } + state->pred_prev_Q13[ 0 ] = pred_Q13[ 0 ]; + state->pred_prev_Q13[ 1 ] = pred_Q13[ 1 ]; + + /* Convert to left/right signals */ + for( n = 0; n < frame_length; n++ ) { + sum = x1[ n + 1 ] + (opus_int32)x2[ n + 1 ]; + diff = x1[ n + 1 ] - (opus_int32)x2[ n + 1 ]; + x1[ n + 1 ] = (opus_int16)silk_SAT16( sum ); + x2[ n + 1 ] = (opus_int16)silk_SAT16( diff ); + } +} diff --git a/lib/opus/silk/stereo_decode_pred.c b/lib/opus/silk/stereo_decode_pred.c new file mode 100644 index 00000000..56ba3925 --- /dev/null +++ b/lib/opus/silk/stereo_decode_pred.c @@ -0,0 +1,73 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Decode mid/side predictors */ +void silk_stereo_decode_pred( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int32 pred_Q13[] /* O Predictors */ +) +{ + opus_int n, ix[ 2 ][ 3 ]; + opus_int32 low_Q13, step_Q13; + + /* Entropy decoding */ + n = ec_dec_icdf( psRangeDec, silk_stereo_pred_joint_iCDF, 8 ); + ix[ 0 ][ 2 ] = silk_DIV32_16( n, 5 ); + ix[ 1 ][ 2 ] = n - 5 * ix[ 0 ][ 2 ]; + for( n = 0; n < 2; n++ ) { + ix[ n ][ 0 ] = ec_dec_icdf( psRangeDec, silk_uniform3_iCDF, 8 ); + ix[ n ][ 1 ] = ec_dec_icdf( psRangeDec, silk_uniform5_iCDF, 8 ); + } + + /* Dequantize */ + for( n = 0; n < 2; n++ ) { + ix[ n ][ 0 ] += 3 * ix[ n ][ 2 ]; + low_Q13 = silk_stereo_pred_quant_Q13[ ix[ n ][ 0 ] ]; + step_Q13 = silk_SMULWB( silk_stereo_pred_quant_Q13[ ix[ n ][ 0 ] + 1 ] - low_Q13, + SILK_FIX_CONST( 0.5 / STEREO_QUANT_SUB_STEPS, 16 ) ); + pred_Q13[ n ] = silk_SMLABB( low_Q13, step_Q13, 2 * ix[ n ][ 1 ] + 1 ); + } + + /* Subtract second from first predictor (helps when actually applying these) */ + pred_Q13[ 0 ] -= pred_Q13[ 1 ]; +} + +/* Decode mid-only flag */ +void silk_stereo_decode_mid_only( + ec_dec *psRangeDec, /* I/O Compressor data structure */ + opus_int *decode_only_mid /* O Flag that only mid channel has been coded */ +) +{ + /* Decode flag that only mid channel is coded */ + *decode_only_mid = ec_dec_icdf( psRangeDec, silk_stereo_only_code_mid_iCDF, 8 ); +} diff --git a/lib/opus/silk/stereo_encode_pred.c b/lib/opus/silk/stereo_encode_pred.c new file mode 100644 index 00000000..03becb67 --- /dev/null +++ b/lib/opus/silk/stereo_encode_pred.c @@ -0,0 +1,62 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Entropy code the mid/side quantization indices */ +void silk_stereo_encode_pred( + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int8 ix[ 2 ][ 3 ] /* I Quantization indices */ +) +{ + opus_int n; + + /* Entropy coding */ + n = 5 * ix[ 0 ][ 2 ] + ix[ 1 ][ 2 ]; + celt_assert( n < 25 ); + ec_enc_icdf( psRangeEnc, n, silk_stereo_pred_joint_iCDF, 8 ); + for( n = 0; n < 2; n++ ) { + celt_assert( ix[ n ][ 0 ] < 3 ); + celt_assert( ix[ n ][ 1 ] < STEREO_QUANT_SUB_STEPS ); + ec_enc_icdf( psRangeEnc, ix[ n ][ 0 ], silk_uniform3_iCDF, 8 ); + ec_enc_icdf( psRangeEnc, ix[ n ][ 1 ], silk_uniform5_iCDF, 8 ); + } +} + +/* Entropy code the mid-only flag */ +void silk_stereo_encode_mid_only( + ec_enc *psRangeEnc, /* I/O Compressor data structure */ + opus_int8 mid_only_flag +) +{ + /* Encode flag that only mid channel is coded */ + ec_enc_icdf( psRangeEnc, mid_only_flag, silk_stereo_only_code_mid_iCDF, 8 ); +} diff --git a/lib/opus/silk/stereo_find_predictor.c b/lib/opus/silk/stereo_find_predictor.c new file mode 100644 index 00000000..e30e90bd --- /dev/null +++ b/lib/opus/silk/stereo_find_predictor.c @@ -0,0 +1,79 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Find least-squares prediction gain for one signal based on another and quantize it */ +opus_int32 silk_stereo_find_predictor( /* O Returns predictor in Q13 */ + opus_int32 *ratio_Q14, /* O Ratio of residual and mid energies */ + const opus_int16 x[], /* I Basis signal */ + const opus_int16 y[], /* I Target signal */ + opus_int32 mid_res_amp_Q0[], /* I/O Smoothed mid, residual norms */ + opus_int length, /* I Number of samples */ + opus_int smooth_coef_Q16 /* I Smoothing coefficient */ +) +{ + opus_int scale, scale1, scale2; + opus_int32 nrgx, nrgy, corr, pred_Q13, pred2_Q10; + + /* Find predictor */ + silk_sum_sqr_shift( &nrgx, &scale1, x, length ); + silk_sum_sqr_shift( &nrgy, &scale2, y, length ); + scale = silk_max_int( scale1, scale2 ); + scale = scale + ( scale & 1 ); /* make even */ + nrgy = silk_RSHIFT32( nrgy, scale - scale2 ); + nrgx = silk_RSHIFT32( nrgx, scale - scale1 ); + nrgx = silk_max_int( nrgx, 1 ); + corr = silk_inner_prod_aligned_scale( x, y, scale, length ); + pred_Q13 = silk_DIV32_varQ( corr, nrgx, 13 ); + pred_Q13 = silk_LIMIT( pred_Q13, -(1 << 14), 1 << 14 ); + pred2_Q10 = silk_SMULWB( pred_Q13, pred_Q13 ); + + /* Faster update for signals with large prediction parameters */ + smooth_coef_Q16 = (opus_int)silk_max_int( smooth_coef_Q16, silk_abs( pred2_Q10 ) ); + + /* Smoothed mid and residual norms */ + silk_assert( smooth_coef_Q16 < 32768 ); + scale = silk_RSHIFT( scale, 1 ); + mid_res_amp_Q0[ 0 ] = silk_SMLAWB( mid_res_amp_Q0[ 0 ], silk_LSHIFT( silk_SQRT_APPROX( nrgx ), scale ) - mid_res_amp_Q0[ 0 ], + smooth_coef_Q16 ); + /* Residual energy = nrgy - 2 * pred * corr + pred^2 * nrgx */ + nrgy = silk_SUB_LSHIFT32( nrgy, silk_SMULWB( corr, pred_Q13 ), 3 + 1 ); + nrgy = silk_ADD_LSHIFT32( nrgy, silk_SMULWB( nrgx, pred2_Q10 ), 6 ); + mid_res_amp_Q0[ 1 ] = silk_SMLAWB( mid_res_amp_Q0[ 1 ], silk_LSHIFT( silk_SQRT_APPROX( nrgy ), scale ) - mid_res_amp_Q0[ 1 ], + smooth_coef_Q16 ); + + /* Ratio of smoothed residual and mid norms */ + *ratio_Q14 = silk_DIV32_varQ( mid_res_amp_Q0[ 1 ], silk_max( mid_res_amp_Q0[ 0 ], 1 ), 14 ); + *ratio_Q14 = silk_LIMIT( *ratio_Q14, 0, 32767 ); + + return pred_Q13; +} diff --git a/lib/opus/silk/stereo_quant_pred.c b/lib/opus/silk/stereo_quant_pred.c new file mode 100644 index 00000000..d4ced6c3 --- /dev/null +++ b/lib/opus/silk/stereo_quant_pred.c @@ -0,0 +1,73 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "main.h" + +/* Quantize mid/side predictors */ +void silk_stereo_quant_pred( + opus_int32 pred_Q13[], /* I/O Predictors (out: quantized) */ + opus_int8 ix[ 2 ][ 3 ] /* O Quantization indices */ +) +{ + opus_int i, j, n; + opus_int32 low_Q13, step_Q13, lvl_Q13, err_min_Q13, err_Q13, quant_pred_Q13 = 0; + + /* Quantize */ + for( n = 0; n < 2; n++ ) { + /* Brute-force search over quantization levels */ + err_min_Q13 = silk_int32_MAX; + for( i = 0; i < STEREO_QUANT_TAB_SIZE - 1; i++ ) { + low_Q13 = silk_stereo_pred_quant_Q13[ i ]; + step_Q13 = silk_SMULWB( silk_stereo_pred_quant_Q13[ i + 1 ] - low_Q13, + SILK_FIX_CONST( 0.5 / STEREO_QUANT_SUB_STEPS, 16 ) ); + for( j = 0; j < STEREO_QUANT_SUB_STEPS; j++ ) { + lvl_Q13 = silk_SMLABB( low_Q13, step_Q13, 2 * j + 1 ); + err_Q13 = silk_abs( pred_Q13[ n ] - lvl_Q13 ); + if( err_Q13 < err_min_Q13 ) { + err_min_Q13 = err_Q13; + quant_pred_Q13 = lvl_Q13; + ix[ n ][ 0 ] = i; + ix[ n ][ 1 ] = j; + } else { + /* Error increasing, so we're past the optimum */ + goto done; + } + } + } + done: + ix[ n ][ 2 ] = silk_DIV32_16( ix[ n ][ 0 ], 3 ); + ix[ n ][ 0 ] -= ix[ n ][ 2 ] * 3; + pred_Q13[ n ] = quant_pred_Q13; + } + + /* Subtract second from first predictor (helps when actually applying these) */ + pred_Q13[ 0 ] -= pred_Q13[ 1 ]; +} diff --git a/lib/opus/silk/structs.h b/lib/opus/silk/structs.h new file mode 100644 index 00000000..3380c757 --- /dev/null +++ b/lib/opus/silk/structs.h @@ -0,0 +1,329 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_STRUCTS_H +#define SILK_STRUCTS_H + +#include "typedef.h" +#include "SigProc_FIX.h" +#include "define.h" +#include "entenc.h" +#include "entdec.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +/************************************/ +/* Noise shaping quantization state */ +/************************************/ +typedef struct { + opus_int16 xq[ 2 * MAX_FRAME_LENGTH ]; /* Buffer for quantized output signal */ + opus_int32 sLTP_shp_Q14[ 2 * MAX_FRAME_LENGTH ]; + opus_int32 sLPC_Q14[ MAX_SUB_FRAME_LENGTH + NSQ_LPC_BUF_LENGTH ]; + opus_int32 sAR2_Q14[ MAX_SHAPE_LPC_ORDER ]; + opus_int32 sLF_AR_shp_Q14; + opus_int32 sDiff_shp_Q14; + opus_int lagPrev; + opus_int sLTP_buf_idx; + opus_int sLTP_shp_buf_idx; + opus_int32 rand_seed; + opus_int32 prev_gain_Q16; + opus_int rewhite_flag; +} silk_nsq_state; + +/********************************/ +/* VAD state */ +/********************************/ +typedef struct { + opus_int32 AnaState[ 2 ]; /* Analysis filterbank state: 0-8 kHz */ + opus_int32 AnaState1[ 2 ]; /* Analysis filterbank state: 0-4 kHz */ + opus_int32 AnaState2[ 2 ]; /* Analysis filterbank state: 0-2 kHz */ + opus_int32 XnrgSubfr[ VAD_N_BANDS ]; /* Subframe energies */ + opus_int32 NrgRatioSmth_Q8[ VAD_N_BANDS ]; /* Smoothed energy level in each band */ + opus_int16 HPstate; /* State of differentiator in the lowest band */ + opus_int32 NL[ VAD_N_BANDS ]; /* Noise energy level in each band */ + opus_int32 inv_NL[ VAD_N_BANDS ]; /* Inverse noise energy level in each band */ + opus_int32 NoiseLevelBias[ VAD_N_BANDS ]; /* Noise level estimator bias/offset */ + opus_int32 counter; /* Frame counter used in the initial phase */ +} silk_VAD_state; + +/* Variable cut-off low-pass filter state */ +typedef struct { + opus_int32 In_LP_State[ 2 ]; /* Low pass filter state */ + opus_int32 transition_frame_no; /* Counter which is mapped to a cut-off frequency */ + opus_int mode; /* Operating mode, <0: switch down, >0: switch up; 0: do nothing */ + opus_int32 saved_fs_kHz; /* If non-zero, holds the last sampling rate before a bandwidth switching reset. */ +} silk_LP_state; + +/* Structure containing NLSF codebook */ +typedef struct { + const opus_int16 nVectors; + const opus_int16 order; + const opus_int16 quantStepSize_Q16; + const opus_int16 invQuantStepSize_Q6; + const opus_uint8 *CB1_NLSF_Q8; + const opus_int16 *CB1_Wght_Q9; + const opus_uint8 *CB1_iCDF; + const opus_uint8 *pred_Q8; + const opus_uint8 *ec_sel; + const opus_uint8 *ec_iCDF; + const opus_uint8 *ec_Rates_Q5; + const opus_int16 *deltaMin_Q15; +} silk_NLSF_CB_struct; + +typedef struct { + opus_int16 pred_prev_Q13[ 2 ]; + opus_int16 sMid[ 2 ]; + opus_int16 sSide[ 2 ]; + opus_int32 mid_side_amp_Q0[ 4 ]; + opus_int16 smth_width_Q14; + opus_int16 width_prev_Q14; + opus_int16 silent_side_len; + opus_int8 predIx[ MAX_FRAMES_PER_PACKET ][ 2 ][ 3 ]; + opus_int8 mid_only_flags[ MAX_FRAMES_PER_PACKET ]; +} stereo_enc_state; + +typedef struct { + opus_int16 pred_prev_Q13[ 2 ]; + opus_int16 sMid[ 2 ]; + opus_int16 sSide[ 2 ]; +} stereo_dec_state; + +typedef struct { + opus_int8 GainsIndices[ MAX_NB_SUBFR ]; + opus_int8 LTPIndex[ MAX_NB_SUBFR ]; + opus_int8 NLSFIndices[ MAX_LPC_ORDER + 1 ]; + opus_int16 lagIndex; + opus_int8 contourIndex; + opus_int8 signalType; + opus_int8 quantOffsetType; + opus_int8 NLSFInterpCoef_Q2; + opus_int8 PERIndex; + opus_int8 LTP_scaleIndex; + opus_int8 Seed; +} SideInfoIndices; + +/********************************/ +/* Encoder state */ +/********************************/ +typedef struct { + opus_int32 In_HP_State[ 2 ]; /* High pass filter state */ + opus_int32 variable_HP_smth1_Q15; /* State of first smoother */ + opus_int32 variable_HP_smth2_Q15; /* State of second smoother */ + silk_LP_state sLP; /* Low pass filter state */ + silk_VAD_state sVAD; /* Voice activity detector state */ + silk_nsq_state sNSQ; /* Noise Shape Quantizer State */ + opus_int16 prev_NLSFq_Q15[ MAX_LPC_ORDER ]; /* Previously quantized NLSF vector */ + opus_int speech_activity_Q8; /* Speech activity */ + opus_int allow_bandwidth_switch; /* Flag indicating that switching of internal bandwidth is allowed */ + opus_int8 LBRRprevLastGainIndex; + opus_int8 prevSignalType; + opus_int prevLag; + opus_int pitch_LPC_win_length; + opus_int max_pitch_lag; /* Highest possible pitch lag (samples) */ + opus_int32 API_fs_Hz; /* API sampling frequency (Hz) */ + opus_int32 prev_API_fs_Hz; /* Previous API sampling frequency (Hz) */ + opus_int maxInternal_fs_Hz; /* Maximum internal sampling frequency (Hz) */ + opus_int minInternal_fs_Hz; /* Minimum internal sampling frequency (Hz) */ + opus_int desiredInternal_fs_Hz; /* Soft request for internal sampling frequency (Hz) */ + opus_int fs_kHz; /* Internal sampling frequency (kHz) */ + opus_int nb_subfr; /* Number of 5 ms subframes in a frame */ + opus_int frame_length; /* Frame length (samples) */ + opus_int subfr_length; /* Subframe length (samples) */ + opus_int ltp_mem_length; /* Length of LTP memory */ + opus_int la_pitch; /* Look-ahead for pitch analysis (samples) */ + opus_int la_shape; /* Look-ahead for noise shape analysis (samples) */ + opus_int shapeWinLength; /* Window length for noise shape analysis (samples) */ + opus_int32 TargetRate_bps; /* Target bitrate (bps) */ + opus_int PacketSize_ms; /* Number of milliseconds to put in each packet */ + opus_int PacketLoss_perc; /* Packet loss rate measured by farend */ + opus_int32 frameCounter; + opus_int Complexity; /* Complexity setting */ + opus_int nStatesDelayedDecision; /* Number of states in delayed decision quantization */ + opus_int useInterpolatedNLSFs; /* Flag for using NLSF interpolation */ + opus_int shapingLPCOrder; /* Filter order for noise shaping filters */ + opus_int predictLPCOrder; /* Filter order for prediction filters */ + opus_int pitchEstimationComplexity; /* Complexity level for pitch estimator */ + opus_int pitchEstimationLPCOrder; /* Whitening filter order for pitch estimator */ + opus_int32 pitchEstimationThreshold_Q16; /* Threshold for pitch estimator */ + opus_int32 sum_log_gain_Q7; /* Cumulative max prediction gain */ + opus_int NLSF_MSVQ_Survivors; /* Number of survivors in NLSF MSVQ */ + opus_int first_frame_after_reset; /* Flag for deactivating NLSF interpolation, pitch prediction */ + opus_int controlled_since_last_payload; /* Flag for ensuring codec_control only runs once per packet */ + opus_int warping_Q16; /* Warping parameter for warped noise shaping */ + opus_int useCBR; /* Flag to enable constant bitrate */ + opus_int prefillFlag; /* Flag to indicate that only buffers are prefilled, no coding */ + const opus_uint8 *pitch_lag_low_bits_iCDF; /* Pointer to iCDF table for low bits of pitch lag index */ + const opus_uint8 *pitch_contour_iCDF; /* Pointer to iCDF table for pitch contour index */ + const silk_NLSF_CB_struct *psNLSF_CB; /* Pointer to NLSF codebook */ + opus_int input_quality_bands_Q15[ VAD_N_BANDS ]; + opus_int input_tilt_Q15; + opus_int SNR_dB_Q7; /* Quality setting */ + + opus_int8 VAD_flags[ MAX_FRAMES_PER_PACKET ]; + opus_int8 LBRR_flag; + opus_int LBRR_flags[ MAX_FRAMES_PER_PACKET ]; + + SideInfoIndices indices; + opus_int8 pulses[ MAX_FRAME_LENGTH ]; + + int arch; + + /* Input/output buffering */ + opus_int16 inputBuf[ MAX_FRAME_LENGTH + 2 ]; /* Buffer containing input signal */ + opus_int inputBufIx; + opus_int nFramesPerPacket; + opus_int nFramesEncoded; /* Number of frames analyzed in current packet */ + + opus_int nChannelsAPI; + opus_int nChannelsInternal; + opus_int channelNb; + + /* Parameters For LTP scaling Control */ + opus_int frames_since_onset; + + /* Specifically for entropy coding */ + opus_int ec_prevSignalType; + opus_int16 ec_prevLagIndex; + + silk_resampler_state_struct resampler_state; + + /* DTX */ + opus_int useDTX; /* Flag to enable DTX */ + opus_int inDTX; /* Flag to signal DTX period */ + opus_int noSpeechCounter; /* Counts concecutive nonactive frames, used by DTX */ + + /* Inband Low Bitrate Redundancy (LBRR) data */ + opus_int useInBandFEC; /* Saves the API setting for query */ + opus_int LBRR_enabled; /* Depends on useInBandFRC, bitrate and packet loss rate */ + opus_int LBRR_GainIncreases; /* Gains increment for coding LBRR frames */ + SideInfoIndices indices_LBRR[ MAX_FRAMES_PER_PACKET ]; + opus_int8 pulses_LBRR[ MAX_FRAMES_PER_PACKET ][ MAX_FRAME_LENGTH ]; +} silk_encoder_state; + + +/* Struct for Packet Loss Concealment */ +typedef struct { + opus_int32 pitchL_Q8; /* Pitch lag to use for voiced concealment */ + opus_int16 LTPCoef_Q14[ LTP_ORDER ]; /* LTP coeficients to use for voiced concealment */ + opus_int16 prevLPC_Q12[ MAX_LPC_ORDER ]; + opus_int last_frame_lost; /* Was previous frame lost */ + opus_int32 rand_seed; /* Seed for unvoiced signal generation */ + opus_int16 randScale_Q14; /* Scaling of unvoiced random signal */ + opus_int32 conc_energy; + opus_int conc_energy_shift; + opus_int16 prevLTP_scale_Q14; + opus_int32 prevGain_Q16[ 2 ]; + opus_int fs_kHz; + opus_int nb_subfr; + opus_int subfr_length; +} silk_PLC_struct; + +/* Struct for CNG */ +typedef struct { + opus_int32 CNG_exc_buf_Q14[ MAX_FRAME_LENGTH ]; + opus_int16 CNG_smth_NLSF_Q15[ MAX_LPC_ORDER ]; + opus_int32 CNG_synth_state[ MAX_LPC_ORDER ]; + opus_int32 CNG_smth_Gain_Q16; + opus_int32 rand_seed; + opus_int fs_kHz; +} silk_CNG_struct; + +/********************************/ +/* Decoder state */ +/********************************/ +typedef struct { + opus_int32 prev_gain_Q16; + opus_int32 exc_Q14[ MAX_FRAME_LENGTH ]; + opus_int32 sLPC_Q14_buf[ MAX_LPC_ORDER ]; + opus_int16 outBuf[ MAX_FRAME_LENGTH + 2 * MAX_SUB_FRAME_LENGTH ]; /* Buffer for output signal */ + opus_int lagPrev; /* Previous Lag */ + opus_int8 LastGainIndex; /* Previous gain index */ + opus_int fs_kHz; /* Sampling frequency in kHz */ + opus_int32 fs_API_hz; /* API sample frequency (Hz) */ + opus_int nb_subfr; /* Number of 5 ms subframes in a frame */ + opus_int frame_length; /* Frame length (samples) */ + opus_int subfr_length; /* Subframe length (samples) */ + opus_int ltp_mem_length; /* Length of LTP memory */ + opus_int LPC_order; /* LPC order */ + opus_int16 prevNLSF_Q15[ MAX_LPC_ORDER ]; /* Used to interpolate LSFs */ + opus_int first_frame_after_reset; /* Flag for deactivating NLSF interpolation */ + const opus_uint8 *pitch_lag_low_bits_iCDF; /* Pointer to iCDF table for low bits of pitch lag index */ + const opus_uint8 *pitch_contour_iCDF; /* Pointer to iCDF table for pitch contour index */ + + /* For buffering payload in case of more frames per packet */ + opus_int nFramesDecoded; + opus_int nFramesPerPacket; + + /* Specifically for entropy coding */ + opus_int ec_prevSignalType; + opus_int16 ec_prevLagIndex; + + opus_int VAD_flags[ MAX_FRAMES_PER_PACKET ]; + opus_int LBRR_flag; + opus_int LBRR_flags[ MAX_FRAMES_PER_PACKET ]; + + silk_resampler_state_struct resampler_state; + + const silk_NLSF_CB_struct *psNLSF_CB; /* Pointer to NLSF codebook */ + + /* Quantization indices */ + SideInfoIndices indices; + + /* CNG state */ + silk_CNG_struct sCNG; + + /* Stuff used for PLC */ + opus_int lossCnt; + opus_int prevSignalType; + int arch; + + silk_PLC_struct sPLC; + +} silk_decoder_state; + +/************************/ +/* Decoder control */ +/************************/ +typedef struct { + /* Prediction and coding parameters */ + opus_int pitchL[ MAX_NB_SUBFR ]; + opus_int32 Gains_Q16[ MAX_NB_SUBFR ]; + /* Holds interpolated and final coefficients, 4-byte aligned */ + silk_DWORD_ALIGN opus_int16 PredCoef_Q12[ 2 ][ MAX_LPC_ORDER ]; + opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ]; + opus_int LTP_scale_Q14; +} silk_decoder_control; + + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/sum_sqr_shift.c b/lib/opus/silk/sum_sqr_shift.c new file mode 100644 index 00000000..4fd0c3d7 --- /dev/null +++ b/lib/opus/silk/sum_sqr_shift.c @@ -0,0 +1,83 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "SigProc_FIX.h" + +/* Compute number of bits to right shift the sum of squares of a vector */ +/* of int16s to make it fit in an int32 */ +void silk_sum_sqr_shift( + opus_int32 *energy, /* O Energy of x, after shifting to the right */ + opus_int *shift, /* O Number of bits right shift applied to energy */ + const opus_int16 *x, /* I Input vector */ + opus_int len /* I Length of input vector */ +) +{ + opus_int i, shft; + opus_uint32 nrg_tmp; + opus_int32 nrg; + + /* Do a first run with the maximum shift we could have. */ + shft = 31-silk_CLZ32(len); + /* Let's be conservative with rounding and start with nrg=len. */ + nrg = len; + for( i = 0; i < len - 1; i += 2 ) { + nrg_tmp = silk_SMULBB( x[ i ], x[ i ] ); + nrg_tmp = silk_SMLABB_ovflw( nrg_tmp, x[ i + 1 ], x[ i + 1 ] ); + nrg = (opus_int32)silk_ADD_RSHIFT_uint( nrg, nrg_tmp, shft ); + } + if( i < len ) { + /* One sample left to process */ + nrg_tmp = silk_SMULBB( x[ i ], x[ i ] ); + nrg = (opus_int32)silk_ADD_RSHIFT_uint( nrg, nrg_tmp, shft ); + } + silk_assert( nrg >= 0 ); + /* Make sure the result will fit in a 32-bit signed integer with two bits + of headroom. */ + shft = silk_max_32(0, shft+3 - silk_CLZ32(nrg)); + nrg = 0; + for( i = 0 ; i < len - 1; i += 2 ) { + nrg_tmp = silk_SMULBB( x[ i ], x[ i ] ); + nrg_tmp = silk_SMLABB_ovflw( nrg_tmp, x[ i + 1 ], x[ i + 1 ] ); + nrg = (opus_int32)silk_ADD_RSHIFT_uint( nrg, nrg_tmp, shft ); + } + if( i < len ) { + /* One sample left to process */ + nrg_tmp = silk_SMULBB( x[ i ], x[ i ] ); + nrg = (opus_int32)silk_ADD_RSHIFT_uint( nrg, nrg_tmp, shft ); + } + + silk_assert( nrg >= 0 ); + + /* Output arguments */ + *shift = shft; + *energy = nrg; +} + diff --git a/lib/opus/silk/table_LSF_cos.c b/lib/opus/silk/table_LSF_cos.c new file mode 100644 index 00000000..ec9dc639 --- /dev/null +++ b/lib/opus/silk/table_LSF_cos.c @@ -0,0 +1,70 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "tables.h" + +/* Cosine approximation table for LSF conversion */ +/* Q12 values (even) */ +const opus_int16 silk_LSFCosTab_FIX_Q12[ LSF_COS_TAB_SZ_FIX + 1 ] = { + 8192, 8190, 8182, 8170, + 8152, 8130, 8104, 8072, + 8034, 7994, 7946, 7896, + 7840, 7778, 7714, 7644, + 7568, 7490, 7406, 7318, + 7226, 7128, 7026, 6922, + 6812, 6698, 6580, 6458, + 6332, 6204, 6070, 5934, + 5792, 5648, 5502, 5352, + 5198, 5040, 4880, 4718, + 4552, 4382, 4212, 4038, + 3862, 3684, 3502, 3320, + 3136, 2948, 2760, 2570, + 2378, 2186, 1990, 1794, + 1598, 1400, 1202, 1002, + 802, 602, 402, 202, + 0, -202, -402, -602, + -802, -1002, -1202, -1400, + -1598, -1794, -1990, -2186, + -2378, -2570, -2760, -2948, + -3136, -3320, -3502, -3684, + -3862, -4038, -4212, -4382, + -4552, -4718, -4880, -5040, + -5198, -5352, -5502, -5648, + -5792, -5934, -6070, -6204, + -6332, -6458, -6580, -6698, + -6812, -6922, -7026, -7128, + -7226, -7318, -7406, -7490, + -7568, -7644, -7714, -7778, + -7840, -7896, -7946, -7994, + -8034, -8072, -8104, -8130, + -8152, -8170, -8182, -8190, + -8192 +}; diff --git a/lib/opus/silk/tables.h b/lib/opus/silk/tables.h new file mode 100644 index 00000000..95230c45 --- /dev/null +++ b/lib/opus/silk/tables.h @@ -0,0 +1,114 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_TABLES_H +#define SILK_TABLES_H + +#include "define.h" +#include "structs.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +/* Entropy coding tables (with size in bytes indicated) */ +extern const opus_uint8 silk_gain_iCDF[ 3 ][ N_LEVELS_QGAIN / 8 ]; /* 24 */ +extern const opus_uint8 silk_delta_gain_iCDF[ MAX_DELTA_GAIN_QUANT - MIN_DELTA_GAIN_QUANT + 1 ]; /* 41 */ + +extern const opus_uint8 silk_pitch_lag_iCDF[ 2 * ( PITCH_EST_MAX_LAG_MS - PITCH_EST_MIN_LAG_MS ) ];/* 32 */ +extern const opus_uint8 silk_pitch_delta_iCDF[ 21 ]; /* 21 */ +extern const opus_uint8 silk_pitch_contour_iCDF[ 34 ]; /* 34 */ +extern const opus_uint8 silk_pitch_contour_NB_iCDF[ 11 ]; /* 11 */ +extern const opus_uint8 silk_pitch_contour_10_ms_iCDF[ 12 ]; /* 12 */ +extern const opus_uint8 silk_pitch_contour_10_ms_NB_iCDF[ 3 ]; /* 3 */ + +extern const opus_uint8 silk_pulses_per_block_iCDF[ N_RATE_LEVELS ][ SILK_MAX_PULSES + 2 ]; /* 180 */ +extern const opus_uint8 silk_pulses_per_block_BITS_Q5[ N_RATE_LEVELS - 1 ][ SILK_MAX_PULSES + 2 ]; /* 162 */ + +extern const opus_uint8 silk_rate_levels_iCDF[ 2 ][ N_RATE_LEVELS - 1 ]; /* 18 */ +extern const opus_uint8 silk_rate_levels_BITS_Q5[ 2 ][ N_RATE_LEVELS - 1 ]; /* 18 */ + +extern const opus_uint8 silk_max_pulses_table[ 4 ]; /* 4 */ + +extern const opus_uint8 silk_shell_code_table0[ 152 ]; /* 152 */ +extern const opus_uint8 silk_shell_code_table1[ 152 ]; /* 152 */ +extern const opus_uint8 silk_shell_code_table2[ 152 ]; /* 152 */ +extern const opus_uint8 silk_shell_code_table3[ 152 ]; /* 152 */ +extern const opus_uint8 silk_shell_code_table_offsets[ SILK_MAX_PULSES + 1 ]; /* 17 */ + +extern const opus_uint8 silk_lsb_iCDF[ 2 ]; /* 2 */ + +extern const opus_uint8 silk_sign_iCDF[ 42 ]; /* 42 */ + +extern const opus_uint8 silk_uniform3_iCDF[ 3 ]; /* 3 */ +extern const opus_uint8 silk_uniform4_iCDF[ 4 ]; /* 4 */ +extern const opus_uint8 silk_uniform5_iCDF[ 5 ]; /* 5 */ +extern const opus_uint8 silk_uniform6_iCDF[ 6 ]; /* 6 */ +extern const opus_uint8 silk_uniform8_iCDF[ 8 ]; /* 8 */ + +extern const opus_uint8 silk_NLSF_EXT_iCDF[ 7 ]; /* 7 */ + +extern const opus_uint8 silk_LTP_per_index_iCDF[ 3 ]; /* 3 */ +extern const opus_uint8 * const silk_LTP_gain_iCDF_ptrs[ NB_LTP_CBKS ]; /* 3 */ +extern const opus_uint8 * const silk_LTP_gain_BITS_Q5_ptrs[ NB_LTP_CBKS ]; /* 3 */ +extern const opus_int8 * const silk_LTP_vq_ptrs_Q7[ NB_LTP_CBKS ]; /* 168 */ +extern const opus_uint8 * const silk_LTP_vq_gain_ptrs_Q7[NB_LTP_CBKS]; +extern const opus_int8 silk_LTP_vq_sizes[ NB_LTP_CBKS ]; /* 3 */ + +extern const opus_uint8 silk_LTPscale_iCDF[ 3 ]; /* 4 */ +extern const opus_int16 silk_LTPScales_table_Q14[ 3 ]; /* 6 */ + +extern const opus_uint8 silk_type_offset_VAD_iCDF[ 4 ]; /* 4 */ +extern const opus_uint8 silk_type_offset_no_VAD_iCDF[ 2 ]; /* 2 */ + +extern const opus_int16 silk_stereo_pred_quant_Q13[ STEREO_QUANT_TAB_SIZE ]; /* 32 */ +extern const opus_uint8 silk_stereo_pred_joint_iCDF[ 25 ]; /* 25 */ +extern const opus_uint8 silk_stereo_only_code_mid_iCDF[ 2 ]; /* 2 */ + +extern const opus_uint8 * const silk_LBRR_flags_iCDF_ptr[ 2 ]; /* 10 */ + +extern const opus_uint8 silk_NLSF_interpolation_factor_iCDF[ 5 ]; /* 5 */ + +extern const silk_NLSF_CB_struct silk_NLSF_CB_WB; /* 1040 */ +extern const silk_NLSF_CB_struct silk_NLSF_CB_NB_MB; /* 728 */ + +/* Quantization offsets */ +extern const opus_int16 silk_Quantization_Offsets_Q10[ 2 ][ 2 ]; /* 8 */ + +/* Interpolation points for filter coefficients used in the bandwidth transition smoother */ +extern const opus_int32 silk_Transition_LP_B_Q28[ TRANSITION_INT_NUM ][ TRANSITION_NB ]; /* 60 */ +extern const opus_int32 silk_Transition_LP_A_Q28[ TRANSITION_INT_NUM ][ TRANSITION_NA ]; /* 60 */ + +/* Rom table with cosine values */ +extern const opus_int16 silk_LSFCosTab_FIX_Q12[ LSF_COS_TAB_SZ_FIX + 1 ]; /* 258 */ + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/lib/opus/silk/tables_LTP.c b/lib/opus/silk/tables_LTP.c new file mode 100644 index 00000000..5e12c864 --- /dev/null +++ b/lib/opus/silk/tables_LTP.c @@ -0,0 +1,294 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "tables.h" + +const opus_uint8 silk_LTP_per_index_iCDF[3] = { + 179, 99, 0 +}; + +static const opus_uint8 silk_LTP_gain_iCDF_0[8] = { + 71, 56, 43, 30, 21, 12, 6, 0 +}; + +static const opus_uint8 silk_LTP_gain_iCDF_1[16] = { + 199, 165, 144, 124, 109, 96, 84, 71, + 61, 51, 42, 32, 23, 15, 8, 0 +}; + +static const opus_uint8 silk_LTP_gain_iCDF_2[32] = { + 241, 225, 211, 199, 187, 175, 164, 153, + 142, 132, 123, 114, 105, 96, 88, 80, + 72, 64, 57, 50, 44, 38, 33, 29, + 24, 20, 16, 12, 9, 5, 2, 0 +}; + +static const opus_uint8 silk_LTP_gain_BITS_Q5_0[8] = { + 15, 131, 138, 138, 155, 155, 173, 173 +}; + +static const opus_uint8 silk_LTP_gain_BITS_Q5_1[16] = { + 69, 93, 115, 118, 131, 138, 141, 138, + 150, 150, 155, 150, 155, 160, 166, 160 +}; + +static const opus_uint8 silk_LTP_gain_BITS_Q5_2[32] = { + 131, 128, 134, 141, 141, 141, 145, 145, + 145, 150, 155, 155, 155, 155, 160, 160, + 160, 160, 166, 166, 173, 173, 182, 192, + 182, 192, 192, 192, 205, 192, 205, 224 +}; + +const opus_uint8 * const silk_LTP_gain_iCDF_ptrs[NB_LTP_CBKS] = { + silk_LTP_gain_iCDF_0, + silk_LTP_gain_iCDF_1, + silk_LTP_gain_iCDF_2 +}; + +const opus_uint8 * const silk_LTP_gain_BITS_Q5_ptrs[NB_LTP_CBKS] = { + silk_LTP_gain_BITS_Q5_0, + silk_LTP_gain_BITS_Q5_1, + silk_LTP_gain_BITS_Q5_2 +}; + +static const opus_int8 silk_LTP_gain_vq_0[8][5] = +{ +{ + 4, 6, 24, 7, 5 +}, +{ + 0, 0, 2, 0, 0 +}, +{ + 12, 28, 41, 13, -4 +}, +{ + -9, 15, 42, 25, 14 +}, +{ + 1, -2, 62, 41, -9 +}, +{ + -10, 37, 65, -4, 3 +}, +{ + -6, 4, 66, 7, -8 +}, +{ + 16, 14, 38, -3, 33 +} +}; + +static const opus_int8 silk_LTP_gain_vq_1[16][5] = +{ +{ + 13, 22, 39, 23, 12 +}, +{ + -1, 36, 64, 27, -6 +}, +{ + -7, 10, 55, 43, 17 +}, +{ + 1, 1, 8, 1, 1 +}, +{ + 6, -11, 74, 53, -9 +}, +{ + -12, 55, 76, -12, 8 +}, +{ + -3, 3, 93, 27, -4 +}, +{ + 26, 39, 59, 3, -8 +}, +{ + 2, 0, 77, 11, 9 +}, +{ + -8, 22, 44, -6, 7 +}, +{ + 40, 9, 26, 3, 9 +}, +{ + -7, 20, 101, -7, 4 +}, +{ + 3, -8, 42, 26, 0 +}, +{ + -15, 33, 68, 2, 23 +}, +{ + -2, 55, 46, -2, 15 +}, +{ + 3, -1, 21, 16, 41 +} +}; + +static const opus_int8 silk_LTP_gain_vq_2[32][5] = +{ +{ + -6, 27, 61, 39, 5 +}, +{ + -11, 42, 88, 4, 1 +}, +{ + -2, 60, 65, 6, -4 +}, +{ + -1, -5, 73, 56, 1 +}, +{ + -9, 19, 94, 29, -9 +}, +{ + 0, 12, 99, 6, 4 +}, +{ + 8, -19, 102, 46, -13 +}, +{ + 3, 2, 13, 3, 2 +}, +{ + 9, -21, 84, 72, -18 +}, +{ + -11, 46, 104, -22, 8 +}, +{ + 18, 38, 48, 23, 0 +}, +{ + -16, 70, 83, -21, 11 +}, +{ + 5, -11, 117, 22, -8 +}, +{ + -6, 23, 117, -12, 3 +}, +{ + 3, -8, 95, 28, 4 +}, +{ + -10, 15, 77, 60, -15 +}, +{ + -1, 4, 124, 2, -4 +}, +{ + 3, 38, 84, 24, -25 +}, +{ + 2, 13, 42, 13, 31 +}, +{ + 21, -4, 56, 46, -1 +}, +{ + -1, 35, 79, -13, 19 +}, +{ + -7, 65, 88, -9, -14 +}, +{ + 20, 4, 81, 49, -29 +}, +{ + 20, 0, 75, 3, -17 +}, +{ + 5, -9, 44, 92, -8 +}, +{ + 1, -3, 22, 69, 31 +}, +{ + -6, 95, 41, -12, 5 +}, +{ + 39, 67, 16, -4, 1 +}, +{ + 0, -6, 120, 55, -36 +}, +{ + -13, 44, 122, 4, -24 +}, +{ + 81, 5, 11, 3, 7 +}, +{ + 2, 0, 9, 10, 88 +} +}; + +const opus_int8 * const silk_LTP_vq_ptrs_Q7[NB_LTP_CBKS] = { + (opus_int8 *)&silk_LTP_gain_vq_0[0][0], + (opus_int8 *)&silk_LTP_gain_vq_1[0][0], + (opus_int8 *)&silk_LTP_gain_vq_2[0][0] +}; + +/* Maximum frequency-dependent response of the pitch taps above, + computed as max(abs(freqz(taps))) */ +static const opus_uint8 silk_LTP_gain_vq_0_gain[8] = { + 46, 2, 90, 87, 93, 91, 82, 98 +}; + +static const opus_uint8 silk_LTP_gain_vq_1_gain[16] = { + 109, 120, 118, 12, 113, 115, 117, 119, + 99, 59, 87, 111, 63, 111, 112, 80 +}; + +static const opus_uint8 silk_LTP_gain_vq_2_gain[32] = { + 126, 124, 125, 124, 129, 121, 126, 23, + 132, 127, 127, 127, 126, 127, 122, 133, + 130, 134, 101, 118, 119, 145, 126, 86, + 124, 120, 123, 119, 170, 173, 107, 109 +}; + +const opus_uint8 * const silk_LTP_vq_gain_ptrs_Q7[NB_LTP_CBKS] = { + &silk_LTP_gain_vq_0_gain[0], + &silk_LTP_gain_vq_1_gain[0], + &silk_LTP_gain_vq_2_gain[0] +}; + +const opus_int8 silk_LTP_vq_sizes[NB_LTP_CBKS] = { + 8, 16, 32 +}; diff --git a/lib/opus/silk/tables_NLSF_CB_NB_MB.c b/lib/opus/silk/tables_NLSF_CB_NB_MB.c new file mode 100644 index 00000000..195d5b95 --- /dev/null +++ b/lib/opus/silk/tables_NLSF_CB_NB_MB.c @@ -0,0 +1,195 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "tables.h" + +static const opus_uint8 silk_NLSF_CB1_NB_MB_Q8[ 320 ] = { + 12, 35, 60, 83, 108, 132, 157, 180, + 206, 228, 15, 32, 55, 77, 101, 125, + 151, 175, 201, 225, 19, 42, 66, 89, + 114, 137, 162, 184, 209, 230, 12, 25, + 50, 72, 97, 120, 147, 172, 200, 223, + 26, 44, 69, 90, 114, 135, 159, 180, + 205, 225, 13, 22, 53, 80, 106, 130, + 156, 180, 205, 228, 15, 25, 44, 64, + 90, 115, 142, 168, 196, 222, 19, 24, + 62, 82, 100, 120, 145, 168, 190, 214, + 22, 31, 50, 79, 103, 120, 151, 170, + 203, 227, 21, 29, 45, 65, 106, 124, + 150, 171, 196, 224, 30, 49, 75, 97, + 121, 142, 165, 186, 209, 229, 19, 25, + 52, 70, 93, 116, 143, 166, 192, 219, + 26, 34, 62, 75, 97, 118, 145, 167, + 194, 217, 25, 33, 56, 70, 91, 113, + 143, 165, 196, 223, 21, 34, 51, 72, + 97, 117, 145, 171, 196, 222, 20, 29, + 50, 67, 90, 117, 144, 168, 197, 221, + 22, 31, 48, 66, 95, 117, 146, 168, + 196, 222, 24, 33, 51, 77, 116, 134, + 158, 180, 200, 224, 21, 28, 70, 87, + 106, 124, 149, 170, 194, 217, 26, 33, + 53, 64, 83, 117, 152, 173, 204, 225, + 27, 34, 65, 95, 108, 129, 155, 174, + 210, 225, 20, 26, 72, 99, 113, 131, + 154, 176, 200, 219, 34, 43, 61, 78, + 93, 114, 155, 177, 205, 229, 23, 29, + 54, 97, 124, 138, 163, 179, 209, 229, + 30, 38, 56, 89, 118, 129, 158, 178, + 200, 231, 21, 29, 49, 63, 85, 111, + 142, 163, 193, 222, 27, 48, 77, 103, + 133, 158, 179, 196, 215, 232, 29, 47, + 74, 99, 124, 151, 176, 198, 220, 237, + 33, 42, 61, 76, 93, 121, 155, 174, + 207, 225, 29, 53, 87, 112, 136, 154, + 170, 188, 208, 227, 24, 30, 52, 84, + 131, 150, 166, 186, 203, 229, 37, 48, + 64, 84, 104, 118, 156, 177, 201, 230 +}; + +static const opus_int16 silk_NLSF_CB1_Wght_Q9[ 320 ] = { + 2897, 2314, 2314, 2314, 2287, 2287, 2314, 2300, 2327, 2287, + 2888, 2580, 2394, 2367, 2314, 2274, 2274, 2274, 2274, 2194, + 2487, 2340, 2340, 2314, 2314, 2314, 2340, 2340, 2367, 2354, + 3216, 2766, 2340, 2340, 2314, 2274, 2221, 2207, 2261, 2194, + 2460, 2474, 2367, 2394, 2394, 2394, 2394, 2367, 2407, 2314, + 3479, 3056, 2127, 2207, 2274, 2274, 2274, 2287, 2314, 2261, + 3282, 3141, 2580, 2394, 2247, 2221, 2207, 2194, 2194, 2114, + 4096, 3845, 2221, 2620, 2620, 2407, 2314, 2394, 2367, 2074, + 3178, 3244, 2367, 2221, 2553, 2434, 2340, 2314, 2167, 2221, + 3338, 3488, 2726, 2194, 2261, 2460, 2354, 2367, 2207, 2101, + 2354, 2420, 2327, 2367, 2394, 2420, 2420, 2420, 2460, 2367, + 3779, 3629, 2434, 2527, 2367, 2274, 2274, 2300, 2207, 2048, + 3254, 3225, 2713, 2846, 2447, 2327, 2300, 2300, 2274, 2127, + 3263, 3300, 2753, 2806, 2447, 2261, 2261, 2247, 2127, 2101, + 2873, 2981, 2633, 2367, 2407, 2354, 2194, 2247, 2247, 2114, + 3225, 3197, 2633, 2580, 2274, 2181, 2247, 2221, 2221, 2141, + 3178, 3310, 2740, 2407, 2274, 2274, 2274, 2287, 2194, 2114, + 3141, 3272, 2460, 2061, 2287, 2500, 2367, 2487, 2434, 2181, + 3507, 3282, 2314, 2700, 2647, 2474, 2367, 2394, 2340, 2127, + 3423, 3535, 3038, 3056, 2300, 1950, 2221, 2274, 2274, 2274, + 3404, 3366, 2087, 2687, 2873, 2354, 2420, 2274, 2474, 2540, + 3760, 3488, 1950, 2660, 2897, 2527, 2394, 2367, 2460, 2261, + 3028, 3272, 2740, 2888, 2740, 2154, 2127, 2287, 2234, 2247, + 3695, 3657, 2025, 1969, 2660, 2700, 2580, 2500, 2327, 2367, + 3207, 3413, 2354, 2074, 2888, 2888, 2340, 2487, 2247, 2167, + 3338, 3366, 2846, 2780, 2327, 2154, 2274, 2287, 2114, 2061, + 2327, 2300, 2181, 2167, 2181, 2367, 2633, 2700, 2700, 2553, + 2407, 2434, 2221, 2261, 2221, 2221, 2340, 2420, 2607, 2700, + 3038, 3244, 2806, 2888, 2474, 2074, 2300, 2314, 2354, 2380, + 2221, 2154, 2127, 2287, 2500, 2793, 2793, 2620, 2580, 2367, + 3676, 3713, 2234, 1838, 2181, 2753, 2726, 2673, 2513, 2207, + 2793, 3160, 2726, 2553, 2846, 2513, 2181, 2394, 2221, 2181 +}; + +static const opus_uint8 silk_NLSF_CB1_iCDF_NB_MB[ 64 ] = { + 212, 178, 148, 129, 108, 96, 85, 82, + 79, 77, 61, 59, 57, 56, 51, 49, + 48, 45, 42, 41, 40, 38, 36, 34, + 31, 30, 21, 12, 10, 3, 1, 0, + 255, 245, 244, 236, 233, 225, 217, 203, + 190, 176, 175, 161, 149, 136, 125, 114, + 102, 91, 81, 71, 60, 52, 43, 35, + 28, 20, 19, 18, 12, 11, 5, 0 +}; + +static const opus_uint8 silk_NLSF_CB2_SELECT_NB_MB[ 160 ] = { + 16, 0, 0, 0, 0, 99, 66, 36, + 36, 34, 36, 34, 34, 34, 34, 83, + 69, 36, 52, 34, 116, 102, 70, 68, + 68, 176, 102, 68, 68, 34, 65, 85, + 68, 84, 36, 116, 141, 152, 139, 170, + 132, 187, 184, 216, 137, 132, 249, 168, + 185, 139, 104, 102, 100, 68, 68, 178, + 218, 185, 185, 170, 244, 216, 187, 187, + 170, 244, 187, 187, 219, 138, 103, 155, + 184, 185, 137, 116, 183, 155, 152, 136, + 132, 217, 184, 184, 170, 164, 217, 171, + 155, 139, 244, 169, 184, 185, 170, 164, + 216, 223, 218, 138, 214, 143, 188, 218, + 168, 244, 141, 136, 155, 170, 168, 138, + 220, 219, 139, 164, 219, 202, 216, 137, + 168, 186, 246, 185, 139, 116, 185, 219, + 185, 138, 100, 100, 134, 100, 102, 34, + 68, 68, 100, 68, 168, 203, 221, 218, + 168, 167, 154, 136, 104, 70, 164, 246, + 171, 137, 139, 137, 155, 218, 219, 139 +}; + +static const opus_uint8 silk_NLSF_CB2_iCDF_NB_MB[ 72 ] = { + 255, 254, 253, 238, 14, 3, 2, 1, + 0, 255, 254, 252, 218, 35, 3, 2, + 1, 0, 255, 254, 250, 208, 59, 4, + 2, 1, 0, 255, 254, 246, 194, 71, + 10, 2, 1, 0, 255, 252, 236, 183, + 82, 8, 2, 1, 0, 255, 252, 235, + 180, 90, 17, 2, 1, 0, 255, 248, + 224, 171, 97, 30, 4, 1, 0, 255, + 254, 236, 173, 95, 37, 7, 1, 0 +}; + +static const opus_uint8 silk_NLSF_CB2_BITS_NB_MB_Q5[ 72 ] = { + 255, 255, 255, 131, 6, 145, 255, 255, + 255, 255, 255, 236, 93, 15, 96, 255, + 255, 255, 255, 255, 194, 83, 25, 71, + 221, 255, 255, 255, 255, 162, 73, 34, + 66, 162, 255, 255, 255, 210, 126, 73, + 43, 57, 173, 255, 255, 255, 201, 125, + 71, 48, 58, 130, 255, 255, 255, 166, + 110, 73, 57, 62, 104, 210, 255, 255, + 251, 123, 65, 55, 68, 100, 171, 255 +}; + +static const opus_uint8 silk_NLSF_PRED_NB_MB_Q8[ 18 ] = { + 179, 138, 140, 148, 151, 149, 153, 151, + 163, 116, 67, 82, 59, 92, 72, 100, + 89, 92 +}; + +static const opus_int16 silk_NLSF_DELTA_MIN_NB_MB_Q15[ 11 ] = { + 250, 3, 6, 3, 3, 3, 4, 3, + 3, 3, 461 +}; + +const silk_NLSF_CB_struct silk_NLSF_CB_NB_MB = +{ + 32, + 10, + SILK_FIX_CONST( 0.18, 16 ), + SILK_FIX_CONST( 1.0 / 0.18, 6 ), + silk_NLSF_CB1_NB_MB_Q8, + silk_NLSF_CB1_Wght_Q9, + silk_NLSF_CB1_iCDF_NB_MB, + silk_NLSF_PRED_NB_MB_Q8, + silk_NLSF_CB2_SELECT_NB_MB, + silk_NLSF_CB2_iCDF_NB_MB, + silk_NLSF_CB2_BITS_NB_MB_Q5, + silk_NLSF_DELTA_MIN_NB_MB_Q15, +}; diff --git a/lib/opus/silk/tables_NLSF_CB_WB.c b/lib/opus/silk/tables_NLSF_CB_WB.c new file mode 100644 index 00000000..5cc9f57b --- /dev/null +++ b/lib/opus/silk/tables_NLSF_CB_WB.c @@ -0,0 +1,234 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "tables.h" + +static const opus_uint8 silk_NLSF_CB1_WB_Q8[ 512 ] = { + 7, 23, 38, 54, 69, 85, 100, 116, + 131, 147, 162, 178, 193, 208, 223, 239, + 13, 25, 41, 55, 69, 83, 98, 112, + 127, 142, 157, 171, 187, 203, 220, 236, + 15, 21, 34, 51, 61, 78, 92, 106, + 126, 136, 152, 167, 185, 205, 225, 240, + 10, 21, 36, 50, 63, 79, 95, 110, + 126, 141, 157, 173, 189, 205, 221, 237, + 17, 20, 37, 51, 59, 78, 89, 107, + 123, 134, 150, 164, 184, 205, 224, 240, + 10, 15, 32, 51, 67, 81, 96, 112, + 129, 142, 158, 173, 189, 204, 220, 236, + 8, 21, 37, 51, 65, 79, 98, 113, + 126, 138, 155, 168, 179, 192, 209, 218, + 12, 15, 34, 55, 63, 78, 87, 108, + 118, 131, 148, 167, 185, 203, 219, 236, + 16, 19, 32, 36, 56, 79, 91, 108, + 118, 136, 154, 171, 186, 204, 220, 237, + 11, 28, 43, 58, 74, 89, 105, 120, + 135, 150, 165, 180, 196, 211, 226, 241, + 6, 16, 33, 46, 60, 75, 92, 107, + 123, 137, 156, 169, 185, 199, 214, 225, + 11, 19, 30, 44, 57, 74, 89, 105, + 121, 135, 152, 169, 186, 202, 218, 234, + 12, 19, 29, 46, 57, 71, 88, 100, + 120, 132, 148, 165, 182, 199, 216, 233, + 17, 23, 35, 46, 56, 77, 92, 106, + 123, 134, 152, 167, 185, 204, 222, 237, + 14, 17, 45, 53, 63, 75, 89, 107, + 115, 132, 151, 171, 188, 206, 221, 240, + 9, 16, 29, 40, 56, 71, 88, 103, + 119, 137, 154, 171, 189, 205, 222, 237, + 16, 19, 36, 48, 57, 76, 87, 105, + 118, 132, 150, 167, 185, 202, 218, 236, + 12, 17, 29, 54, 71, 81, 94, 104, + 126, 136, 149, 164, 182, 201, 221, 237, + 15, 28, 47, 62, 79, 97, 115, 129, + 142, 155, 168, 180, 194, 208, 223, 238, + 8, 14, 30, 45, 62, 78, 94, 111, + 127, 143, 159, 175, 192, 207, 223, 239, + 17, 30, 49, 62, 79, 92, 107, 119, + 132, 145, 160, 174, 190, 204, 220, 235, + 14, 19, 36, 45, 61, 76, 91, 108, + 121, 138, 154, 172, 189, 205, 222, 238, + 12, 18, 31, 45, 60, 76, 91, 107, + 123, 138, 154, 171, 187, 204, 221, 236, + 13, 17, 31, 43, 53, 70, 83, 103, + 114, 131, 149, 167, 185, 203, 220, 237, + 17, 22, 35, 42, 58, 78, 93, 110, + 125, 139, 155, 170, 188, 206, 224, 240, + 8, 15, 34, 50, 67, 83, 99, 115, + 131, 146, 162, 178, 193, 209, 224, 239, + 13, 16, 41, 66, 73, 86, 95, 111, + 128, 137, 150, 163, 183, 206, 225, 241, + 17, 25, 37, 52, 63, 75, 92, 102, + 119, 132, 144, 160, 175, 191, 212, 231, + 19, 31, 49, 65, 83, 100, 117, 133, + 147, 161, 174, 187, 200, 213, 227, 242, + 18, 31, 52, 68, 88, 103, 117, 126, + 138, 149, 163, 177, 192, 207, 223, 239, + 16, 29, 47, 61, 76, 90, 106, 119, + 133, 147, 161, 176, 193, 209, 224, 240, + 15, 21, 35, 50, 61, 73, 86, 97, + 110, 119, 129, 141, 175, 198, 218, 237 +}; + +static const opus_int16 silk_NLSF_CB1_WB_Wght_Q9[ 512 ] = { + 3657, 2925, 2925, 2925, 2925, 2925, 2925, 2925, 2925, 2925, 2925, 2925, 2963, 2963, 2925, 2846, + 3216, 3085, 2972, 3056, 3056, 3010, 3010, 3010, 2963, 2963, 3010, 2972, 2888, 2846, 2846, 2726, + 3920, 4014, 2981, 3207, 3207, 2934, 3056, 2846, 3122, 3244, 2925, 2846, 2620, 2553, 2780, 2925, + 3516, 3197, 3010, 3103, 3019, 2888, 2925, 2925, 2925, 2925, 2888, 2888, 2888, 2888, 2888, 2753, + 5054, 5054, 2934, 3573, 3385, 3056, 3085, 2793, 3160, 3160, 2972, 2846, 2513, 2540, 2753, 2888, + 4428, 4149, 2700, 2753, 2972, 3010, 2925, 2846, 2981, 3019, 2925, 2925, 2925, 2925, 2888, 2726, + 3620, 3019, 2972, 3056, 3056, 2873, 2806, 3056, 3216, 3047, 2981, 3291, 3291, 2981, 3310, 2991, + 5227, 5014, 2540, 3338, 3526, 3385, 3197, 3094, 3376, 2981, 2700, 2647, 2687, 2793, 2846, 2673, + 5081, 5174, 4615, 4428, 2460, 2897, 3047, 3207, 3169, 2687, 2740, 2888, 2846, 2793, 2846, 2700, + 3122, 2888, 2963, 2925, 2925, 2925, 2925, 2963, 2963, 2963, 2963, 2925, 2925, 2963, 2963, 2963, + 4202, 3207, 2981, 3103, 3010, 2888, 2888, 2925, 2972, 2873, 2916, 3019, 2972, 3010, 3197, 2873, + 3760, 3760, 3244, 3103, 2981, 2888, 2925, 2888, 2972, 2934, 2793, 2793, 2846, 2888, 2888, 2660, + 3854, 4014, 3207, 3122, 3244, 2934, 3047, 2963, 2963, 3085, 2846, 2793, 2793, 2793, 2793, 2580, + 3845, 4080, 3357, 3516, 3094, 2740, 3010, 2934, 3122, 3085, 2846, 2846, 2647, 2647, 2846, 2806, + 5147, 4894, 3225, 3845, 3441, 3169, 2897, 3413, 3451, 2700, 2580, 2673, 2740, 2846, 2806, 2753, + 4109, 3789, 3291, 3160, 2925, 2888, 2888, 2925, 2793, 2740, 2793, 2740, 2793, 2846, 2888, 2806, + 5081, 5054, 3047, 3545, 3244, 3056, 3085, 2944, 3103, 2897, 2740, 2740, 2740, 2846, 2793, 2620, + 4309, 4309, 2860, 2527, 3207, 3376, 3376, 3075, 3075, 3376, 3056, 2846, 2647, 2580, 2726, 2753, + 3056, 2916, 2806, 2888, 2740, 2687, 2897, 3103, 3150, 3150, 3216, 3169, 3056, 3010, 2963, 2846, + 4375, 3882, 2925, 2888, 2846, 2888, 2846, 2846, 2888, 2888, 2888, 2846, 2888, 2925, 2888, 2846, + 2981, 2916, 2916, 2981, 2981, 3056, 3122, 3216, 3150, 3056, 3010, 2972, 2972, 2972, 2925, 2740, + 4229, 4149, 3310, 3347, 2925, 2963, 2888, 2981, 2981, 2846, 2793, 2740, 2846, 2846, 2846, 2793, + 4080, 4014, 3103, 3010, 2925, 2925, 2925, 2888, 2925, 2925, 2846, 2846, 2846, 2793, 2888, 2780, + 4615, 4575, 3169, 3441, 3207, 2981, 2897, 3038, 3122, 2740, 2687, 2687, 2687, 2740, 2793, 2700, + 4149, 4269, 3789, 3657, 2726, 2780, 2888, 2888, 3010, 2972, 2925, 2846, 2687, 2687, 2793, 2888, + 4215, 3554, 2753, 2846, 2846, 2888, 2888, 2888, 2925, 2925, 2888, 2925, 2925, 2925, 2963, 2888, + 5174, 4921, 2261, 3432, 3789, 3479, 3347, 2846, 3310, 3479, 3150, 2897, 2460, 2487, 2753, 2925, + 3451, 3685, 3122, 3197, 3357, 3047, 3207, 3207, 2981, 3216, 3085, 2925, 2925, 2687, 2540, 2434, + 2981, 3010, 2793, 2793, 2740, 2793, 2846, 2972, 3056, 3103, 3150, 3150, 3150, 3103, 3010, 3010, + 2944, 2873, 2687, 2726, 2780, 3010, 3432, 3545, 3357, 3244, 3056, 3010, 2963, 2925, 2888, 2846, + 3019, 2944, 2897, 3010, 3010, 2972, 3019, 3103, 3056, 3056, 3010, 2888, 2846, 2925, 2925, 2888, + 3920, 3967, 3010, 3197, 3357, 3216, 3291, 3291, 3479, 3704, 3441, 2726, 2181, 2460, 2580, 2607 +}; + +static const opus_uint8 silk_NLSF_CB1_iCDF_WB[ 64 ] = { + 225, 204, 201, 184, 183, 175, 158, 154, + 153, 135, 119, 115, 113, 110, 109, 99, + 98, 95, 79, 68, 52, 50, 48, 45, + 43, 32, 31, 27, 18, 10, 3, 0, + 255, 251, 235, 230, 212, 201, 196, 182, + 167, 166, 163, 151, 138, 124, 110, 104, + 90, 78, 76, 70, 69, 57, 45, 34, + 24, 21, 11, 6, 5, 4, 3, 0 +}; + +static const opus_uint8 silk_NLSF_CB2_SELECT_WB[ 256 ] = { + 0, 0, 0, 0, 0, 0, 0, 1, + 100, 102, 102, 68, 68, 36, 34, 96, + 164, 107, 158, 185, 180, 185, 139, 102, + 64, 66, 36, 34, 34, 0, 1, 32, + 208, 139, 141, 191, 152, 185, 155, 104, + 96, 171, 104, 166, 102, 102, 102, 132, + 1, 0, 0, 0, 0, 16, 16, 0, + 80, 109, 78, 107, 185, 139, 103, 101, + 208, 212, 141, 139, 173, 153, 123, 103, + 36, 0, 0, 0, 0, 0, 0, 1, + 48, 0, 0, 0, 0, 0, 0, 32, + 68, 135, 123, 119, 119, 103, 69, 98, + 68, 103, 120, 118, 118, 102, 71, 98, + 134, 136, 157, 184, 182, 153, 139, 134, + 208, 168, 248, 75, 189, 143, 121, 107, + 32, 49, 34, 34, 34, 0, 17, 2, + 210, 235, 139, 123, 185, 137, 105, 134, + 98, 135, 104, 182, 100, 183, 171, 134, + 100, 70, 68, 70, 66, 66, 34, 131, + 64, 166, 102, 68, 36, 2, 1, 0, + 134, 166, 102, 68, 34, 34, 66, 132, + 212, 246, 158, 139, 107, 107, 87, 102, + 100, 219, 125, 122, 137, 118, 103, 132, + 114, 135, 137, 105, 171, 106, 50, 34, + 164, 214, 141, 143, 185, 151, 121, 103, + 192, 34, 0, 0, 0, 0, 0, 1, + 208, 109, 74, 187, 134, 249, 159, 137, + 102, 110, 154, 118, 87, 101, 119, 101, + 0, 2, 0, 36, 36, 66, 68, 35, + 96, 164, 102, 100, 36, 0, 2, 33, + 167, 138, 174, 102, 100, 84, 2, 2, + 100, 107, 120, 119, 36, 197, 24, 0 +}; + +static const opus_uint8 silk_NLSF_CB2_iCDF_WB[ 72 ] = { + 255, 254, 253, 244, 12, 3, 2, 1, + 0, 255, 254, 252, 224, 38, 3, 2, + 1, 0, 255, 254, 251, 209, 57, 4, + 2, 1, 0, 255, 254, 244, 195, 69, + 4, 2, 1, 0, 255, 251, 232, 184, + 84, 7, 2, 1, 0, 255, 254, 240, + 186, 86, 14, 2, 1, 0, 255, 254, + 239, 178, 91, 30, 5, 1, 0, 255, + 248, 227, 177, 100, 19, 2, 1, 0 +}; + +static const opus_uint8 silk_NLSF_CB2_BITS_WB_Q5[ 72 ] = { + 255, 255, 255, 156, 4, 154, 255, 255, + 255, 255, 255, 227, 102, 15, 92, 255, + 255, 255, 255, 255, 213, 83, 24, 72, + 236, 255, 255, 255, 255, 150, 76, 33, + 63, 214, 255, 255, 255, 190, 121, 77, + 43, 55, 185, 255, 255, 255, 245, 137, + 71, 43, 59, 139, 255, 255, 255, 255, + 131, 66, 50, 66, 107, 194, 255, 255, + 166, 116, 76, 55, 53, 125, 255, 255 +}; + +static const opus_uint8 silk_NLSF_PRED_WB_Q8[ 30 ] = { + 175, 148, 160, 176, 178, 173, 174, 164, + 177, 174, 196, 182, 198, 192, 182, 68, + 62, 66, 60, 72, 117, 85, 90, 118, + 136, 151, 142, 160, 142, 155 +}; + +static const opus_int16 silk_NLSF_DELTA_MIN_WB_Q15[ 17 ] = { + 100, 3, 40, 3, 3, 3, 5, 14, + 14, 10, 11, 3, 8, 9, 7, 3, + 347 +}; + +const silk_NLSF_CB_struct silk_NLSF_CB_WB = +{ + 32, + 16, + SILK_FIX_CONST( 0.15, 16 ), + SILK_FIX_CONST( 1.0 / 0.15, 6 ), + silk_NLSF_CB1_WB_Q8, + silk_NLSF_CB1_WB_Wght_Q9, + silk_NLSF_CB1_iCDF_WB, + silk_NLSF_PRED_WB_Q8, + silk_NLSF_CB2_SELECT_WB, + silk_NLSF_CB2_iCDF_WB, + silk_NLSF_CB2_BITS_WB_Q5, + silk_NLSF_DELTA_MIN_WB_Q15, +}; + diff --git a/lib/opus/silk/tables_gain.c b/lib/opus/silk/tables_gain.c new file mode 100644 index 00000000..37e41d89 --- /dev/null +++ b/lib/opus/silk/tables_gain.c @@ -0,0 +1,63 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "tables.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +const opus_uint8 silk_gain_iCDF[ 3 ][ N_LEVELS_QGAIN / 8 ] = +{ +{ + 224, 112, 44, 15, 3, 2, 1, 0 +}, +{ + 254, 237, 192, 132, 70, 23, 4, 0 +}, +{ + 255, 252, 226, 155, 61, 11, 2, 0 +} +}; + +const opus_uint8 silk_delta_gain_iCDF[ MAX_DELTA_GAIN_QUANT - MIN_DELTA_GAIN_QUANT + 1 ] = { + 250, 245, 234, 203, 71, 50, 42, 38, + 35, 33, 31, 29, 28, 27, 26, 25, + 24, 23, 22, 21, 20, 19, 18, 17, + 16, 15, 14, 13, 12, 11, 10, 9, + 8, 7, 6, 5, 4, 3, 2, 1, + 0 +}; + +#ifdef __cplusplus +} +#endif diff --git a/lib/opus/silk/tables_other.c b/lib/opus/silk/tables_other.c new file mode 100644 index 00000000..e34d9077 --- /dev/null +++ b/lib/opus/silk/tables_other.c @@ -0,0 +1,124 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "structs.h" +#include "define.h" +#include "tables.h" + +#ifdef __cplusplus +extern "C" +{ +#endif + +/* Tables for stereo predictor coding */ +const opus_int16 silk_stereo_pred_quant_Q13[ STEREO_QUANT_TAB_SIZE ] = { + -13732, -10050, -8266, -7526, -6500, -5000, -2950, -820, + 820, 2950, 5000, 6500, 7526, 8266, 10050, 13732 +}; +const opus_uint8 silk_stereo_pred_joint_iCDF[ 25 ] = { + 249, 247, 246, 245, 244, + 234, 210, 202, 201, 200, + 197, 174, 82, 59, 56, + 55, 54, 46, 22, 12, + 11, 10, 9, 7, 0 +}; +const opus_uint8 silk_stereo_only_code_mid_iCDF[ 2 ] = { 64, 0 }; + +/* Tables for LBRR flags */ +static const opus_uint8 silk_LBRR_flags_2_iCDF[ 3 ] = { 203, 150, 0 }; +static const opus_uint8 silk_LBRR_flags_3_iCDF[ 7 ] = { 215, 195, 166, 125, 110, 82, 0 }; +const opus_uint8 * const silk_LBRR_flags_iCDF_ptr[ 2 ] = { + silk_LBRR_flags_2_iCDF, + silk_LBRR_flags_3_iCDF +}; + +/* Table for LSB coding */ +const opus_uint8 silk_lsb_iCDF[ 2 ] = { 120, 0 }; + +/* Tables for LTPScale */ +const opus_uint8 silk_LTPscale_iCDF[ 3 ] = { 128, 64, 0 }; + +/* Tables for signal type and offset coding */ +const opus_uint8 silk_type_offset_VAD_iCDF[ 4 ] = { + 232, 158, 10, 0 +}; +const opus_uint8 silk_type_offset_no_VAD_iCDF[ 2 ] = { + 230, 0 +}; + +/* Tables for NLSF interpolation factor */ +const opus_uint8 silk_NLSF_interpolation_factor_iCDF[ 5 ] = { 243, 221, 192, 181, 0 }; + +/* Quantization offsets */ +const opus_int16 silk_Quantization_Offsets_Q10[ 2 ][ 2 ] = { + { OFFSET_UVL_Q10, OFFSET_UVH_Q10 }, { OFFSET_VL_Q10, OFFSET_VH_Q10 } +}; + +/* Table for LTPScale */ +const opus_int16 silk_LTPScales_table_Q14[ 3 ] = { 15565, 12288, 8192 }; + +/* Uniform entropy tables */ +const opus_uint8 silk_uniform3_iCDF[ 3 ] = { 171, 85, 0 }; +const opus_uint8 silk_uniform4_iCDF[ 4 ] = { 192, 128, 64, 0 }; +const opus_uint8 silk_uniform5_iCDF[ 5 ] = { 205, 154, 102, 51, 0 }; +const opus_uint8 silk_uniform6_iCDF[ 6 ] = { 213, 171, 128, 85, 43, 0 }; +const opus_uint8 silk_uniform8_iCDF[ 8 ] = { 224, 192, 160, 128, 96, 64, 32, 0 }; + +const opus_uint8 silk_NLSF_EXT_iCDF[ 7 ] = { 100, 40, 16, 7, 3, 1, 0 }; + +/* Elliptic/Cauer filters designed with 0.1 dB passband ripple, + 80 dB minimum stopband attenuation, and + [0.95 : 0.15 : 0.35] normalized cut off frequencies. */ + +/* Interpolation points for filter coefficients used in the bandwidth transition smoother */ +const opus_int32 silk_Transition_LP_B_Q28[ TRANSITION_INT_NUM ][ TRANSITION_NB ] = +{ +{ 250767114, 501534038, 250767114 }, +{ 209867381, 419732057, 209867381 }, +{ 170987846, 341967853, 170987846 }, +{ 131531482, 263046905, 131531482 }, +{ 89306658, 178584282, 89306658 } +}; + +/* Interpolation points for filter coefficients used in the bandwidth transition smoother */ +const opus_int32 silk_Transition_LP_A_Q28[ TRANSITION_INT_NUM ][ TRANSITION_NA ] = +{ +{ 506393414, 239854379 }, +{ 411067935, 169683996 }, +{ 306733530, 116694253 }, +{ 185807084, 77959395 }, +{ 35497197, 57401098 } +}; + +#ifdef __cplusplus +} +#endif + diff --git a/lib/opus/silk/tables_pitch_lag.c b/lib/opus/silk/tables_pitch_lag.c new file mode 100644 index 00000000..e80cc59a --- /dev/null +++ b/lib/opus/silk/tables_pitch_lag.c @@ -0,0 +1,69 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "tables.h" + +const opus_uint8 silk_pitch_lag_iCDF[ 2 * ( PITCH_EST_MAX_LAG_MS - PITCH_EST_MIN_LAG_MS ) ] = { + 253, 250, 244, 233, 212, 182, 150, 131, + 120, 110, 98, 85, 72, 60, 49, 40, + 32, 25, 19, 15, 13, 11, 9, 8, + 7, 6, 5, 4, 3, 2, 1, 0 +}; + +const opus_uint8 silk_pitch_delta_iCDF[21] = { + 210, 208, 206, 203, 199, 193, 183, 168, + 142, 104, 74, 52, 37, 27, 20, 14, + 10, 6, 4, 2, 0 +}; + +const opus_uint8 silk_pitch_contour_iCDF[34] = { + 223, 201, 183, 167, 152, 138, 124, 111, + 98, 88, 79, 70, 62, 56, 50, 44, + 39, 35, 31, 27, 24, 21, 18, 16, + 14, 12, 10, 8, 6, 4, 3, 2, + 1, 0 +}; + +const opus_uint8 silk_pitch_contour_NB_iCDF[11] = { + 188, 176, 155, 138, 119, 97, 67, 43, + 26, 10, 0 +}; + +const opus_uint8 silk_pitch_contour_10_ms_iCDF[12] = { + 165, 119, 80, 61, 47, 35, 27, 20, + 14, 9, 4, 0 +}; + +const opus_uint8 silk_pitch_contour_10_ms_NB_iCDF[3] = { + 113, 63, 0 +}; + + diff --git a/lib/opus/silk/tables_pulses_per_block.c b/lib/opus/silk/tables_pulses_per_block.c new file mode 100644 index 00000000..c7c01c88 --- /dev/null +++ b/lib/opus/silk/tables_pulses_per_block.c @@ -0,0 +1,264 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "tables.h" + +const opus_uint8 silk_max_pulses_table[ 4 ] = { + 8, 10, 12, 16 +}; + +const opus_uint8 silk_pulses_per_block_iCDF[ 10 ][ 18 ] = { +{ + 125, 51, 26, 18, 15, 12, 11, 10, + 9, 8, 7, 6, 5, 4, 3, 2, + 1, 0 +}, +{ + 198, 105, 45, 22, 15, 12, 11, 10, + 9, 8, 7, 6, 5, 4, 3, 2, + 1, 0 +}, +{ + 213, 162, 116, 83, 59, 43, 32, 24, + 18, 15, 12, 9, 7, 6, 5, 3, + 2, 0 +}, +{ + 239, 187, 116, 59, 28, 16, 11, 10, + 9, 8, 7, 6, 5, 4, 3, 2, + 1, 0 +}, +{ + 250, 229, 188, 135, 86, 51, 30, 19, + 13, 10, 8, 6, 5, 4, 3, 2, + 1, 0 +}, +{ + 249, 235, 213, 185, 156, 128, 103, 83, + 66, 53, 42, 33, 26, 21, 17, 13, + 10, 0 +}, +{ + 254, 249, 235, 206, 164, 118, 77, 46, + 27, 16, 10, 7, 5, 4, 3, 2, + 1, 0 +}, +{ + 255, 253, 249, 239, 220, 191, 156, 119, + 85, 57, 37, 23, 15, 10, 6, 4, + 2, 0 +}, +{ + 255, 253, 251, 246, 237, 223, 203, 179, + 152, 124, 98, 75, 55, 40, 29, 21, + 15, 0 +}, +{ + 255, 254, 253, 247, 220, 162, 106, 67, + 42, 28, 18, 12, 9, 6, 4, 3, + 2, 0 +} +}; + +const opus_uint8 silk_pulses_per_block_BITS_Q5[ 9 ][ 18 ] = { +{ + 31, 57, 107, 160, 205, 205, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255 +}, +{ + 69, 47, 67, 111, 166, 205, 255, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255 +}, +{ + 82, 74, 79, 95, 109, 128, 145, 160, + 173, 205, 205, 205, 224, 255, 255, 224, + 255, 224 +}, +{ + 125, 74, 59, 69, 97, 141, 182, 255, + 255, 255, 255, 255, 255, 255, 255, 255, + 255, 255 +}, +{ + 173, 115, 85, 73, 76, 92, 115, 145, + 173, 205, 224, 224, 255, 255, 255, 255, + 255, 255 +}, +{ + 166, 134, 113, 102, 101, 102, 107, 118, + 125, 138, 145, 155, 166, 182, 192, 192, + 205, 150 +}, +{ + 224, 182, 134, 101, 83, 79, 85, 97, + 120, 145, 173, 205, 224, 255, 255, 255, + 255, 255 +}, +{ + 255, 224, 192, 150, 120, 101, 92, 89, + 93, 102, 118, 134, 160, 182, 192, 224, + 224, 224 +}, +{ + 255, 224, 224, 182, 155, 134, 118, 109, + 104, 102, 106, 111, 118, 131, 145, 160, + 173, 131 +} +}; + +const opus_uint8 silk_rate_levels_iCDF[ 2 ][ 9 ] = +{ +{ + 241, 190, 178, 132, 87, 74, 41, 14, + 0 +}, +{ + 223, 193, 157, 140, 106, 57, 39, 18, + 0 +} +}; + +const opus_uint8 silk_rate_levels_BITS_Q5[ 2 ][ 9 ] = +{ +{ + 131, 74, 141, 79, 80, 138, 95, 104, + 134 +}, +{ + 95, 99, 91, 125, 93, 76, 123, 115, + 123 +} +}; + +const opus_uint8 silk_shell_code_table0[ 152 ] = { + 128, 0, 214, 42, 0, 235, 128, 21, + 0, 244, 184, 72, 11, 0, 248, 214, + 128, 42, 7, 0, 248, 225, 170, 80, + 25, 5, 0, 251, 236, 198, 126, 54, + 18, 3, 0, 250, 238, 211, 159, 82, + 35, 15, 5, 0, 250, 231, 203, 168, + 128, 88, 53, 25, 6, 0, 252, 238, + 216, 185, 148, 108, 71, 40, 18, 4, + 0, 253, 243, 225, 199, 166, 128, 90, + 57, 31, 13, 3, 0, 254, 246, 233, + 212, 183, 147, 109, 73, 44, 23, 10, + 2, 0, 255, 250, 240, 223, 198, 166, + 128, 90, 58, 33, 16, 6, 1, 0, + 255, 251, 244, 231, 210, 181, 146, 110, + 75, 46, 25, 12, 5, 1, 0, 255, + 253, 248, 238, 221, 196, 164, 128, 92, + 60, 35, 18, 8, 3, 1, 0, 255, + 253, 249, 242, 229, 208, 180, 146, 110, + 76, 48, 27, 14, 7, 3, 1, 0 +}; + +const opus_uint8 silk_shell_code_table1[ 152 ] = { + 129, 0, 207, 50, 0, 236, 129, 20, + 0, 245, 185, 72, 10, 0, 249, 213, + 129, 42, 6, 0, 250, 226, 169, 87, + 27, 4, 0, 251, 233, 194, 130, 62, + 20, 4, 0, 250, 236, 207, 160, 99, + 47, 17, 3, 0, 255, 240, 217, 182, + 131, 81, 41, 11, 1, 0, 255, 254, + 233, 201, 159, 107, 61, 20, 2, 1, + 0, 255, 249, 233, 206, 170, 128, 86, + 50, 23, 7, 1, 0, 255, 250, 238, + 217, 186, 148, 108, 70, 39, 18, 6, + 1, 0, 255, 252, 243, 226, 200, 166, + 128, 90, 56, 30, 13, 4, 1, 0, + 255, 252, 245, 231, 209, 180, 146, 110, + 76, 47, 25, 11, 4, 1, 0, 255, + 253, 248, 237, 219, 194, 163, 128, 93, + 62, 37, 19, 8, 3, 1, 0, 255, + 254, 250, 241, 226, 205, 177, 145, 111, + 79, 51, 30, 15, 6, 2, 1, 0 +}; + +const opus_uint8 silk_shell_code_table2[ 152 ] = { + 129, 0, 203, 54, 0, 234, 129, 23, + 0, 245, 184, 73, 10, 0, 250, 215, + 129, 41, 5, 0, 252, 232, 173, 86, + 24, 3, 0, 253, 240, 200, 129, 56, + 15, 2, 0, 253, 244, 217, 164, 94, + 38, 10, 1, 0, 253, 245, 226, 189, + 132, 71, 27, 7, 1, 0, 253, 246, + 231, 203, 159, 105, 56, 23, 6, 1, + 0, 255, 248, 235, 213, 179, 133, 85, + 47, 19, 5, 1, 0, 255, 254, 243, + 221, 194, 159, 117, 70, 37, 12, 2, + 1, 0, 255, 254, 248, 234, 208, 171, + 128, 85, 48, 22, 8, 2, 1, 0, + 255, 254, 250, 240, 220, 189, 149, 107, + 67, 36, 16, 6, 2, 1, 0, 255, + 254, 251, 243, 227, 201, 166, 128, 90, + 55, 29, 13, 5, 2, 1, 0, 255, + 254, 252, 246, 234, 213, 183, 147, 109, + 73, 43, 22, 10, 4, 2, 1, 0 +}; + +const opus_uint8 silk_shell_code_table3[ 152 ] = { + 130, 0, 200, 58, 0, 231, 130, 26, + 0, 244, 184, 76, 12, 0, 249, 214, + 130, 43, 6, 0, 252, 232, 173, 87, + 24, 3, 0, 253, 241, 203, 131, 56, + 14, 2, 0, 254, 246, 221, 167, 94, + 35, 8, 1, 0, 254, 249, 232, 193, + 130, 65, 23, 5, 1, 0, 255, 251, + 239, 211, 162, 99, 45, 15, 4, 1, + 0, 255, 251, 243, 223, 186, 131, 74, + 33, 11, 3, 1, 0, 255, 252, 245, + 230, 202, 158, 105, 57, 24, 8, 2, + 1, 0, 255, 253, 247, 235, 214, 179, + 132, 84, 44, 19, 7, 2, 1, 0, + 255, 254, 250, 240, 223, 196, 159, 112, + 69, 36, 15, 6, 2, 1, 0, 255, + 254, 253, 245, 231, 209, 176, 136, 93, + 55, 27, 11, 3, 2, 1, 0, 255, + 254, 253, 252, 239, 221, 194, 158, 117, + 76, 42, 18, 4, 3, 2, 1, 0 +}; + +const opus_uint8 silk_shell_code_table_offsets[ 17 ] = { + 0, 0, 2, 5, 9, 14, 20, 27, + 35, 44, 54, 65, 77, 90, 104, 119, + 135 +}; + +const opus_uint8 silk_sign_iCDF[ 42 ] = { + 254, 49, 67, 77, 82, 93, 99, + 198, 11, 18, 24, 31, 36, 45, + 255, 46, 66, 78, 87, 94, 104, + 208, 14, 21, 32, 42, 51, 66, + 255, 94, 104, 109, 112, 115, 118, + 248, 53, 69, 80, 88, 95, 102 +}; diff --git a/lib/opus/silk/tests/meson.build b/lib/opus/silk/tests/meson.build new file mode 100644 index 00000000..b7c70f75 --- /dev/null +++ b/lib/opus/silk/tests/meson.build @@ -0,0 +1,8 @@ +exe = executable('test_unit_LPC_inv_pred_gain', + 'test_unit_LPC_inv_pred_gain.c', '../LPC_inv_pred_gain.c', + include_directories: opus_includes, + link_with: [celt_lib, celt_static_libs, silk_lib, silk_static_libs], + dependencies: libm, + install: false) + +test(test_name, exe) diff --git a/lib/opus/silk/tests/test_unit_LPC_inv_pred_gain.c b/lib/opus/silk/tests/test_unit_LPC_inv_pred_gain.c new file mode 100644 index 00000000..7ca902ad --- /dev/null +++ b/lib/opus/silk/tests/test_unit_LPC_inv_pred_gain.c @@ -0,0 +1,129 @@ +/*********************************************************************** +Copyright (c) 2017 Google Inc., Jean-Marc Valin +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include "celt/stack_alloc.h" +#include "cpu_support.h" +#include "SigProc_FIX.h" + +/* Computes the impulse response of the filter so we + can catch filters that are definitely unstable. Some + unstable filters may be classified as stable, but not + the other way around. */ +int check_stability(opus_int16 *A_Q12, int order) { + int i; + int j; + int sum_a, sum_abs_a; + double y[SILK_MAX_ORDER_LPC] = {0}; + sum_a = sum_abs_a = 0; + for( j = 0; j < order; j++ ) { + sum_a += A_Q12[ j ]; + sum_abs_a += silk_abs( A_Q12[ j ] ); + } + /* Check DC stability. */ + if( sum_a >= 4096 ) { + return 0; + } + /* If the sum of absolute values is less than 1, the filter + has to be stable. */ + if( sum_abs_a < 4096 ) { + return 1; + } + y[0] = 1; + for( i = 0; i < 10000; i++ ) { + double sum = 0; + for( j = 0; j < order; j++ ) { + sum += y[ j ]*A_Q12[ j ]; + } + for( j = order - 1; j > 0; j-- ) { + y[ j ] = y[ j - 1 ]; + } + y[ 0 ] = sum*(1./4096); + /* If impulse response reaches +/- 10000, the filter + is definitely unstable. */ + if( !(y[ 0 ] < 10000 && y[ 0 ] > -10000) ) { + return 0; + } + /* Test every 8 sample for low amplitude. */ + if( ( i & 0x7 ) == 0 ) { + double amp = 0; + for( j = 0; j < order; j++ ) { + amp += fabs(y[j]); + } + if( amp < 0.00001 ) { + return 1; + } + } + } + return 1; +} + +int main(void) { + const int arch = opus_select_arch(); + /* Set to 10000 so all branches in C function are triggered */ + const int loop_num = 10000; + int count = 0; + ALLOC_STACK; + + /* FIXME: Make the seed random (with option to set it explicitly) + so we get wider coverage. */ + srand(0); + + printf("Testing silk_LPC_inverse_pred_gain() optimization ...\n"); + for( count = 0; count < loop_num; count++ ) { + unsigned int i; + opus_int order; + unsigned int shift; + opus_int16 A_Q12[ SILK_MAX_ORDER_LPC ]; + opus_int32 gain; + + for( order = 2; order <= SILK_MAX_ORDER_LPC; order += 2 ) { /* order must be even. */ + for( shift = 0; shift < 16; shift++ ) { /* Different dynamic range. */ + for( i = 0; i < SILK_MAX_ORDER_LPC; i++ ) { + A_Q12[i] = ((opus_int16)rand()) >> shift; + } + gain = silk_LPC_inverse_pred_gain(A_Q12, order, arch); + /* Look for filters that silk_LPC_inverse_pred_gain() thinks are + stable but definitely aren't. */ + if( gain != 0 && !check_stability(A_Q12, order) ) { + fprintf(stderr, "**Loop %4d failed!**\n", count); + return 1; + } + } + } + if( !(count % 500) ) { + printf("Loop %4d passed\n", count); + } + } + printf("silk_LPC_inverse_pred_gain() optimization passed\n"); + return 0; +} diff --git a/lib/opus/silk/tuning_parameters.h b/lib/opus/silk/tuning_parameters.h new file mode 100644 index 00000000..d70275fd --- /dev/null +++ b/lib/opus/silk/tuning_parameters.h @@ -0,0 +1,155 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_TUNING_PARAMETERS_H +#define SILK_TUNING_PARAMETERS_H + +#ifdef __cplusplus +extern "C" +{ +#endif + +/* Decay time for bitreservoir */ +#define BITRESERVOIR_DECAY_TIME_MS 500 + +/*******************/ +/* Pitch estimator */ +/*******************/ + +/* Level of noise floor for whitening filter LPC analysis in pitch analysis */ +#define FIND_PITCH_WHITE_NOISE_FRACTION 1e-3f + +/* Bandwidth expansion for whitening filter in pitch analysis */ +#define FIND_PITCH_BANDWIDTH_EXPANSION 0.99f + +/*********************/ +/* Linear prediction */ +/*********************/ + +/* LPC analysis regularization */ +#define FIND_LPC_COND_FAC 1e-5f + +/* Max cumulative LTP gain */ +#define MAX_SUM_LOG_GAIN_DB 250.0f + +/* LTP analysis defines */ +#define LTP_CORR_INV_MAX 0.03f + +/***********************/ +/* High pass filtering */ +/***********************/ + +/* Smoothing parameters for low end of pitch frequency range estimation */ +#define VARIABLE_HP_SMTH_COEF1 0.1f +#define VARIABLE_HP_SMTH_COEF2 0.015f +#define VARIABLE_HP_MAX_DELTA_FREQ 0.4f + +/* Min and max cut-off frequency values (-3 dB points) */ +#define VARIABLE_HP_MIN_CUTOFF_HZ 60 +#define VARIABLE_HP_MAX_CUTOFF_HZ 100 + +/***********/ +/* Various */ +/***********/ + +/* VAD threshold */ +#define SPEECH_ACTIVITY_DTX_THRES 0.05f + +/* Speech Activity LBRR enable threshold */ +#define LBRR_SPEECH_ACTIVITY_THRES 0.3f + +/*************************/ +/* Perceptual parameters */ +/*************************/ + +/* reduction in coding SNR during low speech activity */ +#define BG_SNR_DECR_dB 2.0f + +/* factor for reducing quantization noise during voiced speech */ +#define HARM_SNR_INCR_dB 2.0f + +/* factor for reducing quantization noise for unvoiced sparse signals */ +#define SPARSE_SNR_INCR_dB 2.0f + +/* threshold for sparseness measure above which to use lower quantization offset during unvoiced */ +#define ENERGY_VARIATION_THRESHOLD_QNT_OFFSET 0.6f + +/* warping control */ +#define WARPING_MULTIPLIER 0.015f + +/* fraction added to first autocorrelation value */ +#define SHAPE_WHITE_NOISE_FRACTION 3e-5f + +/* noise shaping filter chirp factor */ +#define BANDWIDTH_EXPANSION 0.94f + +/* harmonic noise shaping */ +#define HARMONIC_SHAPING 0.3f + +/* extra harmonic noise shaping for high bitrates or noisy input */ +#define HIGH_RATE_OR_LOW_QUALITY_HARMONIC_SHAPING 0.2f + +/* parameter for shaping noise towards higher frequencies */ +#define HP_NOISE_COEF 0.25f + +/* parameter for shaping noise even more towards higher frequencies during voiced speech */ +#define HARM_HP_NOISE_COEF 0.35f + +/* parameter for applying a high-pass tilt to the input signal */ +#define INPUT_TILT 0.05f + +/* parameter for extra high-pass tilt to the input signal at high rates */ +#define HIGH_RATE_INPUT_TILT 0.1f + +/* parameter for reducing noise at the very low frequencies */ +#define LOW_FREQ_SHAPING 4.0f + +/* less reduction of noise at the very low frequencies for signals with low SNR at low frequencies */ +#define LOW_QUALITY_LOW_FREQ_SHAPING_DECR 0.5f + +/* subframe smoothing coefficient for HarmBoost, HarmShapeGain, Tilt (lower -> more smoothing) */ +#define SUBFR_SMTH_COEF 0.4f + +/* parameters defining the R/D tradeoff in the residual quantizer */ +#define LAMBDA_OFFSET 1.2f +#define LAMBDA_SPEECH_ACT -0.2f +#define LAMBDA_DELAYED_DECISIONS -0.05f +#define LAMBDA_INPUT_QUALITY -0.1f +#define LAMBDA_CODING_QUALITY -0.2f +#define LAMBDA_QUANT_OFFSET 0.8f + +/* Compensation in bitrate calculations for 10 ms modes */ +#define REDUCE_BITRATE_10_MS_BPS 2200 + +/* Maximum time before allowing a bandwidth transition */ +#define MAX_BANDWIDTH_SWITCH_DELAY_MS 5000 + +#ifdef __cplusplus +} +#endif + +#endif /* SILK_TUNING_PARAMETERS_H */ diff --git a/lib/opus/silk/typedef.h b/lib/opus/silk/typedef.h new file mode 100644 index 00000000..793d2c0c --- /dev/null +++ b/lib/opus/silk/typedef.h @@ -0,0 +1,81 @@ +/*********************************************************************** +Copyright (c) 2006-2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef SILK_TYPEDEF_H +#define SILK_TYPEDEF_H + +#include "opus_types.h" +#include "opus_defines.h" + +#ifndef FIXED_POINT +# include +# define silk_float float +# define silk_float_MAX FLT_MAX +#endif + +#define silk_int64_MAX ((opus_int64)0x7FFFFFFFFFFFFFFFLL) /* 2^63 - 1 */ +#define silk_int64_MIN ((opus_int64)0x8000000000000000LL) /* -2^63 */ +#define silk_int32_MAX 0x7FFFFFFF /* 2^31 - 1 = 2147483647 */ +#define silk_int32_MIN ((opus_int32)0x80000000) /* -2^31 = -2147483648 */ +#define silk_int16_MAX 0x7FFF /* 2^15 - 1 = 32767 */ +#define silk_int16_MIN ((opus_int16)0x8000) /* -2^15 = -32768 */ +#define silk_int8_MAX 0x7F /* 2^7 - 1 = 127 */ +#define silk_int8_MIN ((opus_int8)0x80) /* -2^7 = -128 */ +#define silk_uint8_MAX 0xFF /* 2^8 - 1 = 255 */ + +#define silk_TRUE 1 +#define silk_FALSE 0 + +/* assertions */ +#if (defined _WIN32 && !defined _WINCE && !defined(__GNUC__) && !defined(NO_ASSERTS)) +# ifndef silk_assert +# include /* ASSERTE() */ +# define silk_assert(COND) _ASSERTE(COND) +# endif +#else +# ifdef ENABLE_ASSERTIONS +# include +# include +#define silk_fatal(str) _silk_fatal(str, __FILE__, __LINE__); +#ifdef __GNUC__ +__attribute__((noreturn)) +#endif +static OPUS_INLINE void _silk_fatal(const char *str, const char *file, int line) +{ + fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str); +#if defined(_MSC_VER) + _set_abort_behavior( 0, _WRITE_ABORT_MSG); +#endif + abort(); +} +# define silk_assert(COND) {if (!(COND)) {silk_fatal("assertion failed: " #COND);}} +# else +# define silk_assert(COND) +# endif +#endif + +#endif /* SILK_TYPEDEF_H */ diff --git a/lib/opus/silk/x86/NSQ_del_dec_sse4_1.c b/lib/opus/silk/x86/NSQ_del_dec_sse4_1.c new file mode 100644 index 00000000..a58a76cd --- /dev/null +++ b/lib/opus/silk/x86/NSQ_del_dec_sse4_1.c @@ -0,0 +1,915 @@ +/* Copyright (c) 2014-2020, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang FrancisQuiers + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include "main.h" +#include "celt/x86/x86cpu.h" + +#include "stack_alloc.h" + +typedef struct { + opus_int32 sLPC_Q14[ MAX_SUB_FRAME_LENGTH + NSQ_LPC_BUF_LENGTH ]; + opus_int32 RandState[ DECISION_DELAY ]; + opus_int32 Q_Q10[ DECISION_DELAY ]; + opus_int32 Xq_Q14[ DECISION_DELAY ]; + opus_int32 Pred_Q15[ DECISION_DELAY ]; + opus_int32 Shape_Q14[ DECISION_DELAY ]; + opus_int32 sAR2_Q14[ MAX_SHAPE_LPC_ORDER ]; + opus_int32 LF_AR_Q14; + opus_int32 Diff_Q14; + opus_int32 Seed; + opus_int32 SeedInit; + opus_int32 RD_Q10; +} NSQ_del_dec_struct; + +typedef struct { + opus_int32 Q_Q10; + opus_int32 RD_Q10; + opus_int32 xq_Q14; + opus_int32 LF_AR_Q14; + opus_int32 Diff_Q14; + opus_int32 sLTP_shp_Q14; + opus_int32 LPC_exc_Q14; +} NSQ_sample_struct; + +typedef NSQ_sample_struct NSQ_sample_pair[ 2 ]; + +static OPUS_INLINE void silk_nsq_del_dec_scale_states_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + const opus_int16 x16[], /* I Input */ + opus_int32 x_sc_Q10[], /* O Input scaled with 1/Gain in Q10 */ + const opus_int16 sLTP[], /* I Re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I Subframe number */ + opus_int nStatesDelayedDecision, /* I Number of del dec states */ + const opus_int LTP_scale_Q14, /* I LTP state scaling */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type, /* I Signal type */ + const opus_int decisionDelay /* I Decision delay */ +); + +/******************************************/ +/* Noise shape quantizer for one subframe */ +/******************************************/ +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec_sse4_1( + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP filter state */ + opus_int32 delayedGain_Q10[], /* I/O Gain delay buffer */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int subfr, /* I Subframe number */ + opus_int shapingLPCOrder, /* I Shaping LPC filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + opus_int warping_Q16, /* I */ + opus_int nStatesDelayedDecision, /* I Number of states in decision tree */ + opus_int *smpl_buf_idx, /* I/O Index to newest samples in buffers */ + opus_int decisionDelay /* I */ +); + +void silk_NSQ_del_dec_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) +{ + opus_int i, k, lag, start_idx, LSF_interpolation_flag, Winner_ind, subfr; + opus_int last_smple_idx, smpl_buf_idx, decisionDelay; + const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; + opus_int16 *pxq; + VARDECL( opus_int32, sLTP_Q15 ); + VARDECL( opus_int16, sLTP ); + opus_int32 HarmShapeFIRPacked_Q14; + opus_int offset_Q10; + opus_int32 RDmin_Q10, Gain_Q10; + VARDECL( opus_int32, x_sc_Q10 ); + VARDECL( opus_int32, delayedGain_Q10 ); + VARDECL( NSQ_del_dec_struct, psDelDec ); + NSQ_del_dec_struct *psDD; +#ifdef OPUS_CHECK_ASM + silk_nsq_state NSQ_c; + SideInfoIndices psIndices_c; + opus_int8 pulses_c[ MAX_FRAME_LENGTH ]; + const opus_int8 *const pulses_a = pulses; +#endif + SAVE_STACK; + +#ifdef OPUS_CHECK_ASM + ( void )pulses_a; + silk_memcpy( &NSQ_c, NSQ, sizeof( NSQ_c ) ); + silk_memcpy( &psIndices_c, psIndices, sizeof( psIndices_c ) ); + silk_assert( psEncC->nb_subfr * psEncC->subfr_length <= MAX_FRAME_LENGTH ); + silk_memcpy( pulses_c, pulses, psEncC->nb_subfr * psEncC->subfr_length * sizeof( pulses[0] ) ); + silk_NSQ_del_dec_c( + psEncC, + &NSQ_c, + &psIndices_c, + x16, + pulses_c, + PredCoef_Q12, + LTPCoef_Q14, + AR_Q13, + HarmShapeGain_Q14, + Tilt_Q14, + LF_shp_Q14, + Gains_Q16, + pitchL, + Lambda_Q10, + LTP_scale_Q14 + ); +#endif + + /* Set unvoiced lag to the previous one, overwrite later for voiced */ + lag = NSQ->lagPrev; + + silk_assert( NSQ->prev_gain_Q16 != 0 ); + + /* Initialize delayed decision states */ + ALLOC( psDelDec, psEncC->nStatesDelayedDecision, NSQ_del_dec_struct ); + silk_memset( psDelDec, 0, psEncC->nStatesDelayedDecision * sizeof( NSQ_del_dec_struct ) ); + for( k = 0; k < psEncC->nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + psDD->Seed = ( k + psIndices->Seed ) & 3; + psDD->SeedInit = psDD->Seed; + psDD->RD_Q10 = 0; + psDD->LF_AR_Q14 = NSQ->sLF_AR_shp_Q14; + psDD->Diff_Q14 = NSQ->sDiff_shp_Q14; + psDD->Shape_Q14[ 0 ] = NSQ->sLTP_shp_Q14[ psEncC->ltp_mem_length - 1 ]; + silk_memcpy( psDD->sLPC_Q14, NSQ->sLPC_Q14, NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + silk_memcpy( psDD->sAR2_Q14, NSQ->sAR2_Q14, sizeof( NSQ->sAR2_Q14 ) ); + } + + offset_Q10 = silk_Quantization_Offsets_Q10[ psIndices->signalType >> 1 ][ psIndices->quantOffsetType ]; + smpl_buf_idx = 0; /* index of oldest samples */ + + decisionDelay = silk_min_int( DECISION_DELAY, psEncC->subfr_length ); + + /* For voiced frames limit the decision delay to lower than the pitch lag */ + if( psIndices->signalType == TYPE_VOICED ) { + for( k = 0; k < psEncC->nb_subfr; k++ ) { + decisionDelay = silk_min_int( decisionDelay, pitchL[ k ] - LTP_ORDER / 2 - 1 ); + } + } else { + if( lag > 0 ) { + decisionDelay = silk_min_int( decisionDelay, lag - LTP_ORDER / 2 - 1 ); + } + } + + if( psIndices->NLSFInterpCoef_Q2 == 4 ) { + LSF_interpolation_flag = 0; + } else { + LSF_interpolation_flag = 1; + } + + ALLOC( sLTP_Q15, psEncC->ltp_mem_length + psEncC->frame_length, opus_int32 ); + ALLOC( sLTP, psEncC->ltp_mem_length + psEncC->frame_length, opus_int16 ); + ALLOC( x_sc_Q10, psEncC->subfr_length, opus_int32 ); + ALLOC( delayedGain_Q10, DECISION_DELAY, opus_int32 ); + /* Set up pointers to start of sub frame */ + pxq = &NSQ->xq[ psEncC->ltp_mem_length ]; + NSQ->sLTP_shp_buf_idx = psEncC->ltp_mem_length; + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + subfr = 0; + for( k = 0; k < psEncC->nb_subfr; k++ ) { + A_Q12 = &PredCoef_Q12[ ( ( k >> 1 ) | ( 1 - LSF_interpolation_flag ) ) * MAX_LPC_ORDER ]; + B_Q14 = <PCoef_Q14[ k * LTP_ORDER ]; + AR_shp_Q13 = &AR_Q13[ k * MAX_SHAPE_LPC_ORDER ]; + + /* Noise shape parameters */ + silk_assert( HarmShapeGain_Q14[ k ] >= 0 ); + HarmShapeFIRPacked_Q14 = silk_RSHIFT( HarmShapeGain_Q14[ k ], 2 ); + HarmShapeFIRPacked_Q14 |= silk_LSHIFT( (opus_int32)silk_RSHIFT( HarmShapeGain_Q14[ k ], 1 ), 16 ); + + NSQ->rewhite_flag = 0; + if( psIndices->signalType == TYPE_VOICED ) { + /* Voiced */ + lag = pitchL[ k ]; + + /* Re-whitening */ + if( ( k & ( 3 - silk_LSHIFT( LSF_interpolation_flag, 1 ) ) ) == 0 ) { + if( k == 2 ) { + /* RESET DELAYED DECISIONS */ + /* Find winner */ + RDmin_Q10 = psDelDec[ 0 ].RD_Q10; + Winner_ind = 0; + for( i = 1; i < psEncC->nStatesDelayedDecision; i++ ) { + if( psDelDec[ i ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psDelDec[ i ].RD_Q10; + Winner_ind = i; + } + } + for( i = 0; i < psEncC->nStatesDelayedDecision; i++ ) { + if( i != Winner_ind ) { + psDelDec[ i ].RD_Q10 += ( silk_int32_MAX >> 4 ); + silk_assert( psDelDec[ i ].RD_Q10 >= 0 ); + } + } + + /* Copy final part of signals from winner state to output and long-term filter states */ + psDD = &psDelDec[ Winner_ind ]; + last_smple_idx = smpl_buf_idx + decisionDelay; + for( i = 0; i < decisionDelay; i++ ) { + last_smple_idx = ( last_smple_idx - 1 ) % DECISION_DELAY; + if( last_smple_idx < 0 ) last_smple_idx += DECISION_DELAY; + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDD->Q_Q10[ last_smple_idx ], 10 ); + pxq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDD->Xq_Q14[ last_smple_idx ], Gains_Q16[ 1 ] ), 14 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay + i ] = psDD->Shape_Q14[ last_smple_idx ]; + } + + subfr = 0; + } + + /* Rewhiten with new A coefs */ + start_idx = psEncC->ltp_mem_length - lag - psEncC->predictLPCOrder - LTP_ORDER / 2; + celt_assert( start_idx > 0 ); + + silk_LPC_analysis_filter( &sLTP[ start_idx ], &NSQ->xq[ start_idx + k * psEncC->subfr_length ], + A_Q12, psEncC->ltp_mem_length - start_idx, psEncC->predictLPCOrder, psEncC->arch ); + + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + NSQ->rewhite_flag = 1; + } + } + + silk_nsq_del_dec_scale_states_sse4_1( psEncC, NSQ, psDelDec, x16, x_sc_Q10, sLTP, sLTP_Q15, k, + psEncC->nStatesDelayedDecision, LTP_scale_Q14, Gains_Q16, pitchL, psIndices->signalType, decisionDelay ); + + silk_noise_shape_quantizer_del_dec_sse4_1( NSQ, psDelDec, psIndices->signalType, x_sc_Q10, pulses, pxq, sLTP_Q15, + delayedGain_Q10, A_Q12, B_Q14, AR_shp_Q13, lag, HarmShapeFIRPacked_Q14, Tilt_Q14[ k ], LF_shp_Q14[ k ], + Gains_Q16[ k ], Lambda_Q10, offset_Q10, psEncC->subfr_length, subfr++, psEncC->shapingLPCOrder, + psEncC->predictLPCOrder, psEncC->warping_Q16, psEncC->nStatesDelayedDecision, &smpl_buf_idx, decisionDelay ); + + x16 += psEncC->subfr_length; + pulses += psEncC->subfr_length; + pxq += psEncC->subfr_length; + } + + /* Find winner */ + RDmin_Q10 = psDelDec[ 0 ].RD_Q10; + Winner_ind = 0; + for( k = 1; k < psEncC->nStatesDelayedDecision; k++ ) { + if( psDelDec[ k ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psDelDec[ k ].RD_Q10; + Winner_ind = k; + } + } + + /* Copy final part of signals from winner state to output and long-term filter states */ + psDD = &psDelDec[ Winner_ind ]; + psIndices->Seed = psDD->SeedInit; + last_smple_idx = smpl_buf_idx + decisionDelay; + Gain_Q10 = silk_RSHIFT32( Gains_Q16[ psEncC->nb_subfr - 1 ], 6 ); + for( i = 0; i < decisionDelay; i++ ) { + last_smple_idx = ( last_smple_idx - 1 ) % DECISION_DELAY; + if( last_smple_idx < 0 ) last_smple_idx += DECISION_DELAY; + + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDD->Q_Q10[ last_smple_idx ], 10 ); + pxq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDD->Xq_Q14[ last_smple_idx ], Gain_Q10 ), 8 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay + i ] = psDD->Shape_Q14[ last_smple_idx ]; + } + silk_memcpy( NSQ->sLPC_Q14, &psDD->sLPC_Q14[ psEncC->subfr_length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + silk_memcpy( NSQ->sAR2_Q14, psDD->sAR2_Q14, sizeof( psDD->sAR2_Q14 ) ); + + /* Update states */ + NSQ->sLF_AR_shp_Q14 = psDD->LF_AR_Q14; + NSQ->sDiff_shp_Q14 = psDD->Diff_Q14; + NSQ->lagPrev = pitchL[ psEncC->nb_subfr - 1 ]; + + /* Save quantized speech signal */ + silk_memmove( NSQ->xq, &NSQ->xq[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int16 ) ); + silk_memmove( NSQ->sLTP_shp_Q14, &NSQ->sLTP_shp_Q14[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int32 ) ); + +#ifdef OPUS_CHECK_ASM + silk_assert( !memcmp( &NSQ_c, NSQ, sizeof( NSQ_c ) ) ); + silk_assert( !memcmp( &psIndices_c, psIndices, sizeof( psIndices_c ) ) ); + silk_assert( !memcmp( pulses_c, pulses_a, psEncC->nb_subfr * psEncC->subfr_length * sizeof( pulses[0] ) ) ); +#endif + + RESTORE_STACK; +} + +/******************************************/ +/* Noise shape quantizer for one subframe */ +/******************************************/ +static OPUS_INLINE void silk_noise_shape_quantizer_del_dec_sse4_1( + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP filter state */ + opus_int32 delayedGain_Q10[], /* I/O Gain delay buffer */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int subfr, /* I Subframe number */ + opus_int shapingLPCOrder, /* I Shaping LPC filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + opus_int warping_Q16, /* I */ + opus_int nStatesDelayedDecision, /* I Number of states in decision tree */ + opus_int *smpl_buf_idx, /* I/O Index to newest samples in buffers */ + opus_int decisionDelay /* I */ +) +{ + opus_int i, j, k, Winner_ind, RDmin_ind, RDmax_ind, last_smple_idx; + opus_int32 Winner_rand_state; + opus_int32 LTP_pred_Q14, LPC_pred_Q14, n_AR_Q14, n_LTP_Q14; + opus_int32 n_LF_Q14, r_Q10, rr_Q10, rd1_Q10, rd2_Q10, RDmin_Q10, RDmax_Q10; + opus_int32 q1_Q0, q1_Q10, q2_Q10, exc_Q14, LPC_exc_Q14, xq_Q14, Gain_Q10; + opus_int32 tmp1, tmp2, sLF_AR_shp_Q14; + opus_int32 *pred_lag_ptr, *shp_lag_ptr, *psLPC_Q14; + int rdo_offset; + + VARDECL( NSQ_sample_pair, psSampleState ); + NSQ_del_dec_struct *psDD; + NSQ_sample_struct *psSS; + + __m128i a_Q12_0123, a_Q12_4567, a_Q12_89AB, a_Q12_CDEF; + __m128i b_Q12_0123, b_sr_Q12_0123; + SAVE_STACK; + + celt_assert( nStatesDelayedDecision > 0 ); + ALLOC( psSampleState, nStatesDelayedDecision, NSQ_sample_pair ); + + rdo_offset = (Lambda_Q10 >> 1) - 512; + + shp_lag_ptr = &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - lag + HARM_SHAPE_FIR_TAPS / 2 ]; + pred_lag_ptr = &sLTP_Q15[ NSQ->sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + Gain_Q10 = silk_RSHIFT( Gain_Q16, 6 ); + + a_Q12_0123 = OP_CVTEPI16_EPI32_M64( a_Q12 ); + a_Q12_4567 = OP_CVTEPI16_EPI32_M64( a_Q12 + 4 ); + + if( opus_likely( predictLPCOrder == 16 ) ) { + a_Q12_89AB = OP_CVTEPI16_EPI32_M64( a_Q12 + 8 ); + a_Q12_CDEF = OP_CVTEPI16_EPI32_M64( a_Q12 + 12 ); + } + + if( signalType == TYPE_VOICED ){ + b_Q12_0123 = OP_CVTEPI16_EPI32_M64( b_Q14 ); + b_sr_Q12_0123 = _mm_shuffle_epi32( b_Q12_0123, _MM_SHUFFLE( 0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + } + for( i = 0; i < length; i++ ) { + /* Perform common calculations used in all states */ + + /* Long-term prediction */ + if( signalType == TYPE_VOICED ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LTP_pred_Q14 = 2; + { + __m128i tmpa, tmpb, pred_lag_ptr_tmp; + pred_lag_ptr_tmp = _mm_loadu_si128( (__m128i *)(&pred_lag_ptr[ -3 ] ) ); + pred_lag_ptr_tmp = _mm_shuffle_epi32( pred_lag_ptr_tmp, 0x1B ); + tmpa = _mm_mul_epi32( pred_lag_ptr_tmp, b_Q12_0123 ); + tmpa = _mm_srli_si128( tmpa, 2 ); + + pred_lag_ptr_tmp = _mm_shuffle_epi32( pred_lag_ptr_tmp, _MM_SHUFFLE( 0, 3, 2, 1 ) );/* equal shift right 4 bytes */ + pred_lag_ptr_tmp = _mm_mul_epi32( pred_lag_ptr_tmp, b_sr_Q12_0123 ); + pred_lag_ptr_tmp = _mm_srli_si128( pred_lag_ptr_tmp, 2 ); + pred_lag_ptr_tmp = _mm_add_epi32( pred_lag_ptr_tmp, tmpa ); + + tmpb = _mm_shuffle_epi32( pred_lag_ptr_tmp, _MM_SHUFFLE( 0, 0, 3, 2 ) );/* equal shift right 8 bytes */ + pred_lag_ptr_tmp = _mm_add_epi32( pred_lag_ptr_tmp, tmpb ); + LTP_pred_Q14 += _mm_cvtsi128_si32( pred_lag_ptr_tmp ); + + LTP_pred_Q14 = silk_SMLAWB( LTP_pred_Q14, pred_lag_ptr[ -4 ], b_Q14[ 4 ] ); + LTP_pred_Q14 = silk_LSHIFT( LTP_pred_Q14, 1 ); /* Q13 -> Q14 */ + pred_lag_ptr++; + } + } else { + LTP_pred_Q14 = 0; + } + + /* Long-term shaping */ + if( lag > 0 ) { + /* Symmetric, packed FIR coefficients */ + n_LTP_Q14 = silk_SMULWB( silk_ADD_SAT32( shp_lag_ptr[ 0 ], shp_lag_ptr[ -2 ] ), HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SMLAWT( n_LTP_Q14, shp_lag_ptr[ -1 ], HarmShapeFIRPacked_Q14 ); + n_LTP_Q14 = silk_SUB_LSHIFT32( LTP_pred_Q14, n_LTP_Q14, 2 ); /* Q12 -> Q14 */ + shp_lag_ptr++; + } else { + n_LTP_Q14 = 0; + } + { + __m128i tmpa, tmpb, psLPC_Q14_tmp, a_Q12_tmp; + + for( k = 0; k < nStatesDelayedDecision; k++ ) { + /* Delayed decision state */ + psDD = &psDelDec[ k ]; + + /* Sample state */ + psSS = psSampleState[ k ]; + + /* Generate dither */ + psDD->Seed = silk_RAND( psDD->Seed ); + + /* Pointer used in short term prediction and shaping */ + psLPC_Q14 = &psDD->sLPC_Q14[ NSQ_LPC_BUF_LENGTH - 1 + i ]; + /* Short-term prediction */ + silk_assert( predictLPCOrder == 10 || predictLPCOrder == 16 ); + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LPC_pred_Q14 = silk_RSHIFT( predictLPCOrder, 1 ); + + tmpb = _mm_setzero_si128(); + + /* step 1 */ + psLPC_Q14_tmp = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[ -3 ] ) ); /* -3, -2 , -1, 0 */ + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, 0x1B ); /* 0, -1, -2, -3 */ + tmpa = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_0123 ); /* 0, -1, -2, -3 * 0123 -> 0*0, 2*-2 */ + + tmpa = _mm_srli_epi64( tmpa, 16 ); + tmpb = _mm_add_epi32( tmpb, tmpa ); + + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, _MM_SHUFFLE( 0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + a_Q12_tmp = _mm_shuffle_epi32( a_Q12_0123, _MM_SHUFFLE(0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + psLPC_Q14_tmp = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_tmp ); /* 1*-1, 3*-3 */ + psLPC_Q14_tmp = _mm_srli_epi64( psLPC_Q14_tmp, 16 ); + tmpb = _mm_add_epi32( tmpb, psLPC_Q14_tmp ); + + /* step 2 */ + psLPC_Q14_tmp = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[ -7 ] ) ); + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, 0x1B ); + tmpa = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_4567 ); + tmpa = _mm_srli_epi64( tmpa, 16 ); + tmpb = _mm_add_epi32( tmpb, tmpa ); + + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, _MM_SHUFFLE( 0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + a_Q12_tmp = _mm_shuffle_epi32( a_Q12_4567, _MM_SHUFFLE(0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + psLPC_Q14_tmp = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_tmp ); + psLPC_Q14_tmp = _mm_srli_epi64( psLPC_Q14_tmp, 16 ); + tmpb = _mm_add_epi32( tmpb, psLPC_Q14_tmp ); + + if ( opus_likely( predictLPCOrder == 16 ) ) + { + /* step 3 */ + psLPC_Q14_tmp = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[ -11 ] ) ); + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, 0x1B ); + tmpa = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_89AB ); + tmpa = _mm_srli_epi64( tmpa, 16 ); + tmpb = _mm_add_epi32( tmpb, tmpa ); + + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, _MM_SHUFFLE( 0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + a_Q12_tmp = _mm_shuffle_epi32( a_Q12_89AB, _MM_SHUFFLE(0, 3, 2, 1 ) );/* equal shift right 4 bytes */ + psLPC_Q14_tmp = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_tmp ); + psLPC_Q14_tmp = _mm_srli_epi64( psLPC_Q14_tmp, 16 ); + tmpb = _mm_add_epi32( tmpb, psLPC_Q14_tmp ); + + /* step 4 */ + psLPC_Q14_tmp = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[ -15 ] ) ); + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, 0x1B ); + tmpa = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_CDEF ); + tmpa = _mm_srli_epi64( tmpa, 16 ); + tmpb = _mm_add_epi32( tmpb, tmpa ); + + psLPC_Q14_tmp = _mm_shuffle_epi32( psLPC_Q14_tmp, _MM_SHUFFLE( 0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + a_Q12_tmp = _mm_shuffle_epi32( a_Q12_CDEF, _MM_SHUFFLE(0, 3, 2, 1 ) ); /* equal shift right 4 bytes */ + psLPC_Q14_tmp = _mm_mul_epi32( psLPC_Q14_tmp, a_Q12_tmp ); + psLPC_Q14_tmp = _mm_srli_epi64( psLPC_Q14_tmp, 16 ); + tmpb = _mm_add_epi32( tmpb, psLPC_Q14_tmp ); + + /* add at last */ + /* equal shift right 8 bytes*/ + tmpa = _mm_shuffle_epi32( tmpb, _MM_SHUFFLE( 0, 0, 3, 2 ) ); + tmpb = _mm_add_epi32( tmpb, tmpa ); + LPC_pred_Q14 += _mm_cvtsi128_si32( tmpb ); + } + else + { + /* add at last */ + tmpa = _mm_shuffle_epi32( tmpb, _MM_SHUFFLE( 0, 0, 3, 2 ) ); /* equal shift right 8 bytes*/ + tmpb = _mm_add_epi32( tmpb, tmpa ); + LPC_pred_Q14 += _mm_cvtsi128_si32( tmpb ); + + LPC_pred_Q14 = silk_SMLAWB( LPC_pred_Q14, psLPC_Q14[ -8 ], a_Q12[ 8 ] ); + LPC_pred_Q14 = silk_SMLAWB( LPC_pred_Q14, psLPC_Q14[ -9 ], a_Q12[ 9 ] ); + } + + LPC_pred_Q14 = silk_LSHIFT( LPC_pred_Q14, 4 ); /* Q10 -> Q14 */ + + /* Noise shape feedback */ + celt_assert( ( shapingLPCOrder & 1 ) == 0 ); /* check that order is even */ + /* Output of lowpass section */ + tmp2 = silk_SMLAWB( psDD->Diff_Q14, psDD->sAR2_Q14[ 0 ], warping_Q16 ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( psDD->sAR2_Q14[ 0 ], psDD->sAR2_Q14[ 1 ] - tmp2, warping_Q16 ); + psDD->sAR2_Q14[ 0 ] = tmp2; + n_AR_Q14 = silk_RSHIFT( shapingLPCOrder, 1 ); + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp2, AR_shp_Q13[ 0 ] ); + /* Loop over allpass sections */ + for( j = 2; j < shapingLPCOrder; j += 2 ) { + /* Output of allpass section */ + tmp2 = silk_SMLAWB( psDD->sAR2_Q14[ j - 1 ], psDD->sAR2_Q14[ j + 0 ] - tmp1, warping_Q16 ); + psDD->sAR2_Q14[ j - 1 ] = tmp1; + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp1, AR_shp_Q13[ j - 1 ] ); + /* Output of allpass section */ + tmp1 = silk_SMLAWB( psDD->sAR2_Q14[ j + 0 ], psDD->sAR2_Q14[ j + 1 ] - tmp2, warping_Q16 ); + psDD->sAR2_Q14[ j + 0 ] = tmp2; + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp2, AR_shp_Q13[ j ] ); + } + psDD->sAR2_Q14[ shapingLPCOrder - 1 ] = tmp1; + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, tmp1, AR_shp_Q13[ shapingLPCOrder - 1 ] ); + + n_AR_Q14 = silk_LSHIFT( n_AR_Q14, 1 ); /* Q11 -> Q12 */ + n_AR_Q14 = silk_SMLAWB( n_AR_Q14, psDD->LF_AR_Q14, Tilt_Q14 ); /* Q12 */ + n_AR_Q14 = silk_LSHIFT( n_AR_Q14, 2 ); /* Q12 -> Q14 */ + + n_LF_Q14 = silk_SMULWB( psDD->Shape_Q14[ *smpl_buf_idx ], LF_shp_Q14 ); /* Q12 */ + n_LF_Q14 = silk_SMLAWT( n_LF_Q14, psDD->LF_AR_Q14, LF_shp_Q14 ); /* Q12 */ + n_LF_Q14 = silk_LSHIFT( n_LF_Q14, 2 ); /* Q12 -> Q14 */ + + /* Input minus prediction plus noise feedback */ + /* r = x[ i ] - LTP_pred - LPC_pred + n_AR + n_Tilt + n_LF + n_LTP */ + tmp1 = silk_ADD_SAT32( n_AR_Q14, n_LF_Q14 ); /* Q14 */ + tmp2 = silk_ADD32( n_LTP_Q14, LPC_pred_Q14 ); /* Q13 */ + tmp1 = silk_SUB_SAT32( tmp2, tmp1 ); /* Q13 */ + tmp1 = silk_RSHIFT_ROUND( tmp1, 4 ); /* Q10 */ + + r_Q10 = silk_SUB32( x_Q10[ i ], tmp1 ); /* residual error Q10 */ + + /* Flip sign depending on dither */ + if ( psDD->Seed < 0 ) { + r_Q10 = -r_Q10; + } + r_Q10 = silk_LIMIT_32( r_Q10, -(31 << 10), 30 << 10 ); + + /* Find two quantization level candidates and measure their rate-distortion */ + q1_Q10 = silk_SUB32( r_Q10, offset_Q10 ); + q1_Q0 = silk_RSHIFT( q1_Q10, 10 ); + if (Lambda_Q10 > 2048) { + /* For aggressive RDO, the bias becomes more than one pulse. */ + if (q1_Q10 > rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 - rdo_offset, 10 ); + } else if (q1_Q10 < -rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 + rdo_offset, 10 ); + } else if (q1_Q10 < 0) { + q1_Q0 = -1; + } else { + q1_Q0 = 0; + } + } + if( q1_Q0 > 0 ) { + q1_Q10 = silk_SUB32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q10 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == 0 ) { + q1_Q10 = offset_Q10; + q2_Q10 = silk_ADD32( q1_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q10 = silk_SMULBB( q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else if( q1_Q0 == -1 ) { + q2_Q10 = offset_Q10; + q1_Q10 = silk_SUB32( q2_Q10, 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q10 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( q2_Q10, Lambda_Q10 ); + } else { /* q1_Q0 < -1 */ + q1_Q10 = silk_ADD32( silk_LSHIFT( q1_Q0, 10 ), QUANT_LEVEL_ADJUST_Q10 ); + q1_Q10 = silk_ADD32( q1_Q10, offset_Q10 ); + q2_Q10 = silk_ADD32( q1_Q10, 1024 ); + rd1_Q10 = silk_SMULBB( -q1_Q10, Lambda_Q10 ); + rd2_Q10 = silk_SMULBB( -q2_Q10, Lambda_Q10 ); + } + rr_Q10 = silk_SUB32( r_Q10, q1_Q10 ); + rd1_Q10 = silk_RSHIFT( silk_SMLABB( rd1_Q10, rr_Q10, rr_Q10 ), 10 ); + rr_Q10 = silk_SUB32( r_Q10, q2_Q10 ); + rd2_Q10 = silk_RSHIFT( silk_SMLABB( rd2_Q10, rr_Q10, rr_Q10 ), 10 ); + + if( rd1_Q10 < rd2_Q10 ) { + psSS[ 0 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd1_Q10 ); + psSS[ 1 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd2_Q10 ); + psSS[ 0 ].Q_Q10 = q1_Q10; + psSS[ 1 ].Q_Q10 = q2_Q10; + } else { + psSS[ 0 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd2_Q10 ); + psSS[ 1 ].RD_Q10 = silk_ADD32( psDD->RD_Q10, rd1_Q10 ); + psSS[ 0 ].Q_Q10 = q2_Q10; + psSS[ 1 ].Q_Q10 = q1_Q10; + } + + /* Update states for best quantization */ + + /* Quantized excitation */ + exc_Q14 = silk_LSHIFT32( psSS[ 0 ].Q_Q10, 4 ); + if ( psDD->Seed < 0 ) { + exc_Q14 = -exc_Q14; + } + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD32( exc_Q14, LTP_pred_Q14 ); + xq_Q14 = silk_ADD32( LPC_exc_Q14, LPC_pred_Q14 ); + + /* Update states */ + psSS[ 0 ].Diff_Q14 = silk_SUB_LSHIFT32( xq_Q14, x_Q10[ i ], 4 ); + sLF_AR_shp_Q14 = silk_SUB32( psSS[ 0 ].Diff_Q14, n_AR_Q14 ); + psSS[ 0 ].sLTP_shp_Q14 = silk_SUB_SAT32( sLF_AR_shp_Q14, n_LF_Q14 ); + psSS[ 0 ].LF_AR_Q14 = sLF_AR_shp_Q14; + psSS[ 0 ].LPC_exc_Q14 = LPC_exc_Q14; + psSS[ 0 ].xq_Q14 = xq_Q14; + + /* Update states for second best quantization */ + + /* Quantized excitation */ + exc_Q14 = silk_LSHIFT32( psSS[ 1 ].Q_Q10, 4 ); + if ( psDD->Seed < 0 ) { + exc_Q14 = -exc_Q14; + } + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD32( exc_Q14, LTP_pred_Q14 ); + xq_Q14 = silk_ADD32( LPC_exc_Q14, LPC_pred_Q14 ); + + /* Update states */ + psSS[ 1 ].Diff_Q14 = silk_SUB_LSHIFT32( xq_Q14, x_Q10[ i ], 4 ); + sLF_AR_shp_Q14 = silk_SUB32( psSS[ 1 ].Diff_Q14, n_AR_Q14 ); + psSS[ 1 ].sLTP_shp_Q14 = silk_SUB_SAT32( sLF_AR_shp_Q14, n_LF_Q14 ); + psSS[ 1 ].LF_AR_Q14 = sLF_AR_shp_Q14; + psSS[ 1 ].LPC_exc_Q14 = LPC_exc_Q14; + psSS[ 1 ].xq_Q14 = xq_Q14; + } + } + *smpl_buf_idx = ( *smpl_buf_idx - 1 ) % DECISION_DELAY; + if( *smpl_buf_idx < 0 ) *smpl_buf_idx += DECISION_DELAY; + last_smple_idx = ( *smpl_buf_idx + decisionDelay ) % DECISION_DELAY; + + /* Find winner */ + RDmin_Q10 = psSampleState[ 0 ][ 0 ].RD_Q10; + Winner_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + if( psSampleState[ k ][ 0 ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ k ][ 0 ].RD_Q10; + Winner_ind = k; + } + } + + /* Increase RD values of expired states */ + Winner_rand_state = psDelDec[ Winner_ind ].RandState[ last_smple_idx ]; + for( k = 0; k < nStatesDelayedDecision; k++ ) { + if( psDelDec[ k ].RandState[ last_smple_idx ] != Winner_rand_state ) { + psSampleState[ k ][ 0 ].RD_Q10 = silk_ADD32( psSampleState[ k ][ 0 ].RD_Q10, silk_int32_MAX >> 4 ); + psSampleState[ k ][ 1 ].RD_Q10 = silk_ADD32( psSampleState[ k ][ 1 ].RD_Q10, silk_int32_MAX >> 4 ); + silk_assert( psSampleState[ k ][ 0 ].RD_Q10 >= 0 ); + } + } + + /* Find worst in first set and best in second set */ + RDmax_Q10 = psSampleState[ 0 ][ 0 ].RD_Q10; + RDmin_Q10 = psSampleState[ 0 ][ 1 ].RD_Q10; + RDmax_ind = 0; + RDmin_ind = 0; + for( k = 1; k < nStatesDelayedDecision; k++ ) { + /* find worst in first set */ + if( psSampleState[ k ][ 0 ].RD_Q10 > RDmax_Q10 ) { + RDmax_Q10 = psSampleState[ k ][ 0 ].RD_Q10; + RDmax_ind = k; + } + /* find best in second set */ + if( psSampleState[ k ][ 1 ].RD_Q10 < RDmin_Q10 ) { + RDmin_Q10 = psSampleState[ k ][ 1 ].RD_Q10; + RDmin_ind = k; + } + } + + /* Replace a state if best from second set outperforms worst in first set */ + if( RDmin_Q10 < RDmax_Q10 ) { + silk_memcpy( ( (opus_int32 *)&psDelDec[ RDmax_ind ] ) + i, + ( (opus_int32 *)&psDelDec[ RDmin_ind ] ) + i, sizeof( NSQ_del_dec_struct ) - i * sizeof( opus_int32) ); + silk_memcpy( &psSampleState[ RDmax_ind ][ 0 ], &psSampleState[ RDmin_ind ][ 1 ], sizeof( NSQ_sample_struct ) ); + } + + /* Write samples from winner to output and long-term filter states */ + psDD = &psDelDec[ Winner_ind ]; + if( subfr > 0 || i >= decisionDelay ) { + pulses[ i - decisionDelay ] = (opus_int8)silk_RSHIFT_ROUND( psDD->Q_Q10[ last_smple_idx ], 10 ); + xq[ i - decisionDelay ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( + silk_SMULWW( psDD->Xq_Q14[ last_smple_idx ], delayedGain_Q10[ last_smple_idx ] ), 8 ) ); + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - decisionDelay ] = psDD->Shape_Q14[ last_smple_idx ]; + sLTP_Q15[ NSQ->sLTP_buf_idx - decisionDelay ] = psDD->Pred_Q15[ last_smple_idx ]; + } + NSQ->sLTP_shp_buf_idx++; + NSQ->sLTP_buf_idx++; + + /* Update states */ + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + psSS = &psSampleState[ k ][ 0 ]; + psDD->LF_AR_Q14 = psSS->LF_AR_Q14; + psDD->Diff_Q14 = psSS->Diff_Q14; + psDD->sLPC_Q14[ NSQ_LPC_BUF_LENGTH + i ] = psSS->xq_Q14; + psDD->Xq_Q14[ *smpl_buf_idx ] = psSS->xq_Q14; + psDD->Q_Q10[ *smpl_buf_idx ] = psSS->Q_Q10; + psDD->Pred_Q15[ *smpl_buf_idx ] = silk_LSHIFT32( psSS->LPC_exc_Q14, 1 ); + psDD->Shape_Q14[ *smpl_buf_idx ] = psSS->sLTP_shp_Q14; + psDD->Seed = silk_ADD32_ovflw( psDD->Seed, silk_RSHIFT_ROUND( psSS->Q_Q10, 10 ) ); + psDD->RandState[ *smpl_buf_idx ] = psDD->Seed; + psDD->RD_Q10 = psSS->RD_Q10; + } + delayedGain_Q10[ *smpl_buf_idx ] = Gain_Q10; + } + /* Update LPC states */ + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + silk_memcpy( psDD->sLPC_Q14, &psDD->sLPC_Q14[ length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); + } + RESTORE_STACK; +} + +static OPUS_INLINE void silk_nsq_del_dec_scale_states_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + NSQ_del_dec_struct psDelDec[], /* I/O Delayed decision states */ + const opus_int16 x16[], /* I Input */ + opus_int32 x_sc_Q10[], /* O Input scaled with 1/Gain in Q10 */ + const opus_int16 sLTP[], /* I Re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I Subframe number */ + opus_int nStatesDelayedDecision, /* I Number of del dec states */ + const opus_int LTP_scale_Q14, /* I LTP state scaling */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type, /* I Signal type */ + const opus_int decisionDelay /* I Decision delay */ +) +{ + opus_int i, k, lag; + opus_int32 gain_adj_Q16, inv_gain_Q31, inv_gain_Q26; + NSQ_del_dec_struct *psDD; + __m128i xmm_inv_gain_Q26, xmm_x16_x2x0, xmm_x16_x3x1; + + lag = pitchL[ subfr ]; + inv_gain_Q31 = silk_INVERSE32_varQ( silk_max( Gains_Q16[ subfr ], 1 ), 47 ); + silk_assert( inv_gain_Q31 != 0 ); + + /* Scale input */ + inv_gain_Q26 = silk_RSHIFT_ROUND( inv_gain_Q31, 5 ); + + /* prepare inv_gain_Q26 in packed 4 32-bits */ + xmm_inv_gain_Q26 = _mm_set1_epi32(inv_gain_Q26); + + for( i = 0; i < psEncC->subfr_length - 3; i += 4 ) { + xmm_x16_x2x0 = OP_CVTEPI16_EPI32_M64( &(x16[ i ] ) ); + + /* equal shift right 4 bytes*/ + xmm_x16_x3x1 = _mm_shuffle_epi32( xmm_x16_x2x0, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + + xmm_x16_x2x0 = _mm_mul_epi32( xmm_x16_x2x0, xmm_inv_gain_Q26 ); + xmm_x16_x3x1 = _mm_mul_epi32( xmm_x16_x3x1, xmm_inv_gain_Q26 ); + + xmm_x16_x2x0 = _mm_srli_epi64( xmm_x16_x2x0, 16 ); + xmm_x16_x3x1 = _mm_slli_epi64( xmm_x16_x3x1, 16 ); + + xmm_x16_x2x0 = _mm_blend_epi16( xmm_x16_x2x0, xmm_x16_x3x1, 0xCC ); + + _mm_storeu_si128( (__m128i *)(&(x_sc_Q10[ i ] ) ), xmm_x16_x2x0 ); + } + + for( ; i < psEncC->subfr_length; i++ ) { + x_sc_Q10[ i ] = silk_SMULWW( x16[ i ], inv_gain_Q26 ); + } + + /* After rewhitening the LTP state is un-scaled, so scale with inv_gain_Q16 */ + if( NSQ->rewhite_flag ) { + if( subfr == 0 ) { + /* Do LTP downscaling */ + inv_gain_Q31 = silk_LSHIFT( silk_SMULWB( inv_gain_Q31, LTP_scale_Q14 ), 2 ); + } + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx; i++ ) { + silk_assert( i < MAX_FRAME_LENGTH ); + sLTP_Q15[ i ] = silk_SMULWB( inv_gain_Q31, sLTP[ i ] ); + } + } + + /* Adjust for changing gain */ + if( Gains_Q16[ subfr ] != NSQ->prev_gain_Q16 ) { + gain_adj_Q16 = silk_DIV32_varQ( NSQ->prev_gain_Q16, Gains_Q16[ subfr ], 16 ); + + /* Scale long-term shaping state */ + { + __m128i xmm_gain_adj_Q16, xmm_sLTP_shp_Q14_x2x0, xmm_sLTP_shp_Q14_x3x1; + + /* prepare gain_adj_Q16 in packed 4 32-bits */ + xmm_gain_adj_Q16 = _mm_set1_epi32( gain_adj_Q16 ); + + for( i = NSQ->sLTP_shp_buf_idx - psEncC->ltp_mem_length; i < NSQ->sLTP_shp_buf_idx - 3; i += 4 ) + { + xmm_sLTP_shp_Q14_x2x0 = _mm_loadu_si128( (__m128i *)(&(NSQ->sLTP_shp_Q14[ i ] ) ) ); + /* equal shift right 4 bytes*/ + xmm_sLTP_shp_Q14_x3x1 = _mm_shuffle_epi32( xmm_sLTP_shp_Q14_x2x0, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + + xmm_sLTP_shp_Q14_x2x0 = _mm_mul_epi32( xmm_sLTP_shp_Q14_x2x0, xmm_gain_adj_Q16 ); + xmm_sLTP_shp_Q14_x3x1 = _mm_mul_epi32( xmm_sLTP_shp_Q14_x3x1, xmm_gain_adj_Q16 ); + + xmm_sLTP_shp_Q14_x2x0 = _mm_srli_epi64( xmm_sLTP_shp_Q14_x2x0, 16 ); + xmm_sLTP_shp_Q14_x3x1 = _mm_slli_epi64( xmm_sLTP_shp_Q14_x3x1, 16 ); + + xmm_sLTP_shp_Q14_x2x0 = _mm_blend_epi16( xmm_sLTP_shp_Q14_x2x0, xmm_sLTP_shp_Q14_x3x1, 0xCC ); + + _mm_storeu_si128( (__m128i *)(&(NSQ->sLTP_shp_Q14[ i ] ) ), xmm_sLTP_shp_Q14_x2x0 ); + } + + for( ; i < NSQ->sLTP_shp_buf_idx; i++ ) { + NSQ->sLTP_shp_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLTP_shp_Q14[ i ] ); + } + + /* Scale long-term prediction state */ + if( signal_type == TYPE_VOICED && NSQ->rewhite_flag == 0 ) { + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx - decisionDelay; i++ ) { + sLTP_Q15[ i ] = silk_SMULWW( gain_adj_Q16, sLTP_Q15[ i ] ); + } + } + + for( k = 0; k < nStatesDelayedDecision; k++ ) { + psDD = &psDelDec[ k ]; + + /* Scale scalar states */ + psDD->LF_AR_Q14 = silk_SMULWW( gain_adj_Q16, psDD->LF_AR_Q14 ); + psDD->Diff_Q14 = silk_SMULWW( gain_adj_Q16, psDD->Diff_Q14 ); + + /* Scale short-term prediction and shaping states */ + for( i = 0; i < NSQ_LPC_BUF_LENGTH; i++ ) { + psDD->sLPC_Q14[ i ] = silk_SMULWW( gain_adj_Q16, psDD->sLPC_Q14[ i ] ); + } + for( i = 0; i < MAX_SHAPE_LPC_ORDER; i++ ) { + psDD->sAR2_Q14[ i ] = silk_SMULWW( gain_adj_Q16, psDD->sAR2_Q14[ i ] ); + } + for( i = 0; i < DECISION_DELAY; i++ ) { + psDD->Pred_Q15[ i ] = silk_SMULWW( gain_adj_Q16, psDD->Pred_Q15[ i ] ); + psDD->Shape_Q14[ i ] = silk_SMULWW( gain_adj_Q16, psDD->Shape_Q14[ i ] ); + } + } + } + + /* Save inverse gain */ + NSQ->prev_gain_Q16 = Gains_Q16[ subfr ]; + } +} diff --git a/lib/opus/silk/x86/NSQ_sse4_1.c b/lib/opus/silk/x86/NSQ_sse4_1.c new file mode 100644 index 00000000..d5ae1d3b --- /dev/null +++ b/lib/opus/silk/x86/NSQ_sse4_1.c @@ -0,0 +1,772 @@ +/* Copyright (c) 2014-2020, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang FrancisQuiers + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include "main.h" +#include "celt/x86/x86cpu.h" +#include "stack_alloc.h" + +static OPUS_INLINE void silk_nsq_scale_states_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + const opus_int16 x16[], /* I input */ + opus_int32 x_sc_Q10[], /* O input scaled with 1/Gain */ + const opus_int16 sLTP[], /* I re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I subframe number */ + const opus_int LTP_scale_Q14, /* I */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type /* I Signal type */ +); + +static OPUS_INLINE void silk_noise_shape_quantizer_10_16_sse4_1( + silk_nsq_state *NSQ, /* I/O NSQ state */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_sc_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP state */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping AR coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int32 table[][4] /* I */ +); + +void silk_NSQ_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) +{ + opus_int k, lag, start_idx, LSF_interpolation_flag; + const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13; + opus_int16 *pxq; + VARDECL( opus_int32, sLTP_Q15 ); + VARDECL( opus_int16, sLTP ); + opus_int32 HarmShapeFIRPacked_Q14; + opus_int offset_Q10; + VARDECL( opus_int32, x_sc_Q10 ); + + opus_int32 table[ 64 ][ 4 ]; + opus_int32 tmp1; + opus_int32 q1_Q10, q2_Q10, rd1_Q20, rd2_Q20; + +#ifdef OPUS_CHECK_ASM + silk_nsq_state NSQ_c; + SideInfoIndices psIndices_c; + opus_int8 pulses_c[ MAX_FRAME_LENGTH ]; + const opus_int8 *const pulses_a = pulses; +#endif + + SAVE_STACK; + +#ifdef OPUS_CHECK_ASM + ( void )pulses_a; + silk_memcpy( &NSQ_c, NSQ, sizeof( NSQ_c ) ); + silk_memcpy( &psIndices_c, psIndices, sizeof( psIndices_c ) ); + silk_assert( psEncC->nb_subfr * psEncC->subfr_length <= MAX_FRAME_LENGTH ); + silk_memcpy( pulses_c, pulses, psEncC->nb_subfr * psEncC->subfr_length * sizeof( pulses[0] ) ); + + silk_NSQ_c( + psEncC, + &NSQ_c, + &psIndices_c, + x16, + pulses_c, + PredCoef_Q12, + LTPCoef_Q14, + AR_Q13, + HarmShapeGain_Q14, + Tilt_Q14, + LF_shp_Q14, + Gains_Q16, + pitchL, + Lambda_Q10, + LTP_scale_Q14 + ); +#endif + + NSQ->rand_seed = psIndices->Seed; + + /* Set unvoiced lag to the previous one, overwrite later for voiced */ + lag = NSQ->lagPrev; + + silk_assert( NSQ->prev_gain_Q16 != 0 ); + + offset_Q10 = silk_Quantization_Offsets_Q10[ psIndices->signalType >> 1 ][ psIndices->quantOffsetType ]; + + /* 0 */ + q1_Q10 = offset_Q10; + q2_Q10 = offset_Q10 + ( 1024 - QUANT_LEVEL_ADJUST_Q10 ); + rd1_Q20 = q1_Q10 * Lambda_Q10; + rd2_Q20 = q2_Q10 * Lambda_Q10; + + table[ 32 ][ 0 ] = q1_Q10; + table[ 32 ][ 1 ] = q2_Q10; + table[ 32 ][ 2 ] = 2 * (q1_Q10 - q2_Q10); + table[ 32 ][ 3 ] = (rd1_Q20 - rd2_Q20) + (q1_Q10 * q1_Q10 - q2_Q10 * q2_Q10); + + /* -1 */ + q1_Q10 = offset_Q10 - ( 1024 - QUANT_LEVEL_ADJUST_Q10 ); + q2_Q10 = offset_Q10; + rd1_Q20 = - q1_Q10 * Lambda_Q10; + rd2_Q20 = q2_Q10 * Lambda_Q10; + + table[ 31 ][ 0 ] = q1_Q10; + table[ 31 ][ 1 ] = q2_Q10; + table[ 31 ][ 2 ] = 2 * (q1_Q10 - q2_Q10); + table[ 31 ][ 3 ] = (rd1_Q20 - rd2_Q20) + (q1_Q10 * q1_Q10 - q2_Q10 * q2_Q10); + + /* > 0 */ + for (k = 1; k <= 31; k++) + { + tmp1 = offset_Q10 + silk_LSHIFT( k, 10 ); + + q1_Q10 = tmp1 - QUANT_LEVEL_ADJUST_Q10; + q2_Q10 = tmp1 - QUANT_LEVEL_ADJUST_Q10 + 1024; + rd1_Q20 = q1_Q10 * Lambda_Q10; + rd2_Q20 = q2_Q10 * Lambda_Q10; + + table[ 32 + k ][ 0 ] = q1_Q10; + table[ 32 + k ][ 1 ] = q2_Q10; + table[ 32 + k ][ 2 ] = 2 * (q1_Q10 - q2_Q10); + table[ 32 + k ][ 3 ] = (rd1_Q20 - rd2_Q20) + (q1_Q10 * q1_Q10 - q2_Q10 * q2_Q10); + } + + /* < -1 */ + for (k = -32; k <= -2; k++) + { + tmp1 = offset_Q10 + silk_LSHIFT( k, 10 ); + + q1_Q10 = tmp1 + QUANT_LEVEL_ADJUST_Q10; + q2_Q10 = tmp1 + QUANT_LEVEL_ADJUST_Q10 + 1024; + rd1_Q20 = - q1_Q10 * Lambda_Q10; + rd2_Q20 = - q2_Q10 * Lambda_Q10; + + table[ 32 + k ][ 0 ] = q1_Q10; + table[ 32 + k ][ 1 ] = q2_Q10; + table[ 32 + k ][ 2 ] = 2 * (q1_Q10 - q2_Q10); + table[ 32 + k ][ 3 ] = (rd1_Q20 - rd2_Q20) + (q1_Q10 * q1_Q10 - q2_Q10 * q2_Q10); + } + + if( psIndices->NLSFInterpCoef_Q2 == 4 ) { + LSF_interpolation_flag = 0; + } else { + LSF_interpolation_flag = 1; + } + + ALLOC( sLTP_Q15, psEncC->ltp_mem_length + psEncC->frame_length, opus_int32 ); + ALLOC( sLTP, psEncC->ltp_mem_length + psEncC->frame_length, opus_int16 ); + ALLOC( x_sc_Q10, psEncC->subfr_length, opus_int32 ); + /* Set up pointers to start of sub frame */ + NSQ->sLTP_shp_buf_idx = psEncC->ltp_mem_length; + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + pxq = &NSQ->xq[ psEncC->ltp_mem_length ]; + for( k = 0; k < psEncC->nb_subfr; k++ ) { + A_Q12 = &PredCoef_Q12[ (( k >> 1 ) | ( 1 - LSF_interpolation_flag )) * MAX_LPC_ORDER ]; + B_Q14 = <PCoef_Q14[ k * LTP_ORDER ]; + AR_shp_Q13 = &AR_Q13[ k * MAX_SHAPE_LPC_ORDER ]; + + /* Noise shape parameters */ + silk_assert( HarmShapeGain_Q14[ k ] >= 0 ); + HarmShapeFIRPacked_Q14 = silk_RSHIFT( HarmShapeGain_Q14[ k ], 2 ); + HarmShapeFIRPacked_Q14 |= silk_LSHIFT( (opus_int32)silk_RSHIFT( HarmShapeGain_Q14[ k ], 1 ), 16 ); + + NSQ->rewhite_flag = 0; + if( psIndices->signalType == TYPE_VOICED ) { + /* Voiced */ + lag = pitchL[ k ]; + + /* Re-whitening */ + if( ( k & ( 3 - silk_LSHIFT( LSF_interpolation_flag, 1 ) ) ) == 0 ) { + /* Rewhiten with new A coefs */ + start_idx = psEncC->ltp_mem_length - lag - psEncC->predictLPCOrder - LTP_ORDER / 2; + celt_assert( start_idx > 0 ); + + silk_LPC_analysis_filter( &sLTP[ start_idx ], &NSQ->xq[ start_idx + k * psEncC->subfr_length ], + A_Q12, psEncC->ltp_mem_length - start_idx, psEncC->predictLPCOrder, psEncC->arch ); + + NSQ->rewhite_flag = 1; + NSQ->sLTP_buf_idx = psEncC->ltp_mem_length; + } + } + + silk_nsq_scale_states_sse4_1( psEncC, NSQ, x16, x_sc_Q10, sLTP, sLTP_Q15, k, LTP_scale_Q14, Gains_Q16, pitchL, psIndices->signalType ); + + if ( opus_likely( ( 10 == psEncC->shapingLPCOrder ) && ( 16 == psEncC->predictLPCOrder) ) ) + { + silk_noise_shape_quantizer_10_16_sse4_1( NSQ, psIndices->signalType, x_sc_Q10, pulses, pxq, sLTP_Q15, A_Q12, B_Q14, + AR_shp_Q13, lag, HarmShapeFIRPacked_Q14, Tilt_Q14[ k ], LF_shp_Q14[ k ], Gains_Q16[ k ], Lambda_Q10, + offset_Q10, psEncC->subfr_length, &(table[32]) ); + } + else + { + silk_noise_shape_quantizer( NSQ, psIndices->signalType, x_sc_Q10, pulses, pxq, sLTP_Q15, A_Q12, B_Q14, + AR_shp_Q13, lag, HarmShapeFIRPacked_Q14, Tilt_Q14[ k ], LF_shp_Q14[ k ], Gains_Q16[ k ], Lambda_Q10, + offset_Q10, psEncC->subfr_length, psEncC->shapingLPCOrder, psEncC->predictLPCOrder, psEncC->arch ); + } + + x16 += psEncC->subfr_length; + pulses += psEncC->subfr_length; + pxq += psEncC->subfr_length; + } + + /* Update lagPrev for next frame */ + NSQ->lagPrev = pitchL[ psEncC->nb_subfr - 1 ]; + + /* Save quantized speech and noise shaping signals */ + silk_memmove( NSQ->xq, &NSQ->xq[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int16 ) ); + silk_memmove( NSQ->sLTP_shp_Q14, &NSQ->sLTP_shp_Q14[ psEncC->frame_length ], psEncC->ltp_mem_length * sizeof( opus_int32 ) ); + +#ifdef OPUS_CHECK_ASM + silk_assert( !memcmp( &NSQ_c, NSQ, sizeof( NSQ_c ) ) ); + silk_assert( !memcmp( &psIndices_c, psIndices, sizeof( psIndices_c ) ) ); + silk_assert( !memcmp( pulses_c, pulses_a, psEncC->nb_subfr * psEncC->subfr_length * sizeof( pulses[0] ) ) ); +#endif + + RESTORE_STACK; +} + +/************************************/ +/* silk_noise_shape_quantizer_10_16 */ +/************************************/ +static OPUS_INLINE void silk_noise_shape_quantizer_10_16_sse4_1( + silk_nsq_state *NSQ, /* I/O NSQ state */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_sc_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP state */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping AR coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int32 table[][4] /* I */ +) +{ + opus_int i; + opus_int32 LTP_pred_Q13, LPC_pred_Q10, n_AR_Q12, n_LTP_Q13; + opus_int32 n_LF_Q12, r_Q10, q1_Q0, q1_Q10, q2_Q10; + opus_int32 exc_Q14, LPC_exc_Q14, xq_Q14, Gain_Q10, sDiff_shp_Q14; + opus_int32 tmp1, tmp2, sLF_AR_shp_Q14; + opus_int32 *psLPC_Q14, *shp_lag_ptr, *pred_lag_ptr; + + __m128i xmm_tempa, xmm_tempb; + + __m128i xmm_one; + + __m128i psLPC_Q14_hi_01234567, psLPC_Q14_hi_89ABCDEF; + __m128i psLPC_Q14_lo_01234567, psLPC_Q14_lo_89ABCDEF; + __m128i a_Q12_01234567, a_Q12_89ABCDEF; + + __m128i sAR2_Q14_hi_76543210, sAR2_Q14_lo_76543210; + __m128i AR_shp_Q13_76543210; + + int rdo_offset = (Lambda_Q10 >> 1) - 512; + + shp_lag_ptr = &NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - lag + HARM_SHAPE_FIR_TAPS / 2 ]; + pred_lag_ptr = &sLTP_Q15[ NSQ->sLTP_buf_idx - lag + LTP_ORDER / 2 ]; + Gain_Q10 = silk_RSHIFT( Gain_Q16, 6 ); + + /* Set up short term AR state */ + psLPC_Q14 = &NSQ->sLPC_Q14[ NSQ_LPC_BUF_LENGTH - 1 ]; + + sLF_AR_shp_Q14 = NSQ->sLF_AR_shp_Q14; + xq_Q14 = psLPC_Q14[ 0 ]; + sDiff_shp_Q14 = NSQ->sDiff_shp_Q14; + LTP_pred_Q13 = 0; + + /* load a_Q12 */ + xmm_one = _mm_set_epi8( 1, 0, 3, 2, 5, 4, 7, 6, 9, 8, 11, 10, 13, 12, 15, 14 ); + + /* load a_Q12[0] - a_Q12[7] */ + a_Q12_01234567 = _mm_loadu_si128( (__m128i *)(&a_Q12[ 0 ] ) ); + /* load a_Q12[ 8 ] - a_Q12[ 15 ] */ + a_Q12_89ABCDEF = _mm_loadu_si128( (__m128i *)(&a_Q12[ 8 ] ) ); + + a_Q12_01234567 = _mm_shuffle_epi8( a_Q12_01234567, xmm_one ); + a_Q12_89ABCDEF = _mm_shuffle_epi8( a_Q12_89ABCDEF, xmm_one ); + + /* load AR_shp_Q13 */ + AR_shp_Q13_76543210 = _mm_loadu_si128( (__m128i *)(&AR_shp_Q13[0] ) ); + + /* load psLPC_Q14 */ + xmm_one = _mm_set_epi8(15, 14, 11, 10, 7, 6, 3, 2, 13, 12, 9, 8, 5, 4, 1, 0 ); + + xmm_tempa = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[-16]) ); + xmm_tempb = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[-12]) ); + + xmm_tempa = _mm_shuffle_epi8( xmm_tempa, xmm_one ); + xmm_tempb = _mm_shuffle_epi8( xmm_tempb, xmm_one ); + + psLPC_Q14_hi_89ABCDEF = _mm_unpackhi_epi64( xmm_tempa, xmm_tempb ); + psLPC_Q14_lo_89ABCDEF = _mm_unpacklo_epi64( xmm_tempa, xmm_tempb ); + + xmm_tempa = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[ -8 ]) ); + xmm_tempb = _mm_loadu_si128( (__m128i *)(&psLPC_Q14[ -4 ]) ); + + xmm_tempa = _mm_shuffle_epi8( xmm_tempa, xmm_one ); + xmm_tempb = _mm_shuffle_epi8( xmm_tempb, xmm_one ); + + psLPC_Q14_hi_01234567 = _mm_unpackhi_epi64( xmm_tempa, xmm_tempb ); + psLPC_Q14_lo_01234567 = _mm_unpacklo_epi64( xmm_tempa, xmm_tempb ); + + /* load sAR2_Q14 */ + xmm_tempa = _mm_loadu_si128( (__m128i *)(&(NSQ->sAR2_Q14[ 0 ]) ) ); + xmm_tempb = _mm_loadu_si128( (__m128i *)(&(NSQ->sAR2_Q14[ 4 ]) ) ); + + xmm_tempa = _mm_shuffle_epi8( xmm_tempa, xmm_one ); + xmm_tempb = _mm_shuffle_epi8( xmm_tempb, xmm_one ); + + sAR2_Q14_hi_76543210 = _mm_unpackhi_epi64( xmm_tempa, xmm_tempb ); + sAR2_Q14_lo_76543210 = _mm_unpacklo_epi64( xmm_tempa, xmm_tempb ); + + /* prepare 1 in 8 * 16bit */ + xmm_one = _mm_set1_epi16(1); + + for( i = 0; i < length; i++ ) + { + /* Short-term prediction */ + __m128i xmm_hi_07, xmm_hi_8F, xmm_lo_07, xmm_lo_8F; + + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LPC_pred_Q10 = 8; /* silk_RSHIFT( predictLPCOrder, 1 ); */ + + /* shift psLPC_Q14 */ + psLPC_Q14_hi_89ABCDEF = _mm_alignr_epi8( psLPC_Q14_hi_01234567, psLPC_Q14_hi_89ABCDEF, 2 ); + psLPC_Q14_lo_89ABCDEF = _mm_alignr_epi8( psLPC_Q14_lo_01234567, psLPC_Q14_lo_89ABCDEF, 2 ); + + psLPC_Q14_hi_01234567 = _mm_srli_si128( psLPC_Q14_hi_01234567, 2 ); + psLPC_Q14_lo_01234567 = _mm_srli_si128( psLPC_Q14_lo_01234567, 2 ); + + psLPC_Q14_hi_01234567 = _mm_insert_epi16( psLPC_Q14_hi_01234567, (xq_Q14 >> 16), 7 ); + psLPC_Q14_lo_01234567 = _mm_insert_epi16( psLPC_Q14_lo_01234567, (xq_Q14), 7 ); + + /* high part, use pmaddwd, results in 4 32-bit */ + xmm_hi_07 = _mm_madd_epi16( psLPC_Q14_hi_01234567, a_Q12_01234567 ); + xmm_hi_8F = _mm_madd_epi16( psLPC_Q14_hi_89ABCDEF, a_Q12_89ABCDEF ); + + /* low part, use pmulhw, results in 8 16-bit, note we need simulate unsigned * signed, _mm_srai_epi16(psLPC_Q14_lo_01234567, 15) */ + xmm_tempa = _mm_cmpgt_epi16( _mm_setzero_si128(), psLPC_Q14_lo_01234567 ); + xmm_tempb = _mm_cmpgt_epi16( _mm_setzero_si128(), psLPC_Q14_lo_89ABCDEF ); + + xmm_tempa = _mm_and_si128( xmm_tempa, a_Q12_01234567 ); + xmm_tempb = _mm_and_si128( xmm_tempb, a_Q12_89ABCDEF ); + + xmm_lo_07 = _mm_mulhi_epi16( psLPC_Q14_lo_01234567, a_Q12_01234567 ); + xmm_lo_8F = _mm_mulhi_epi16( psLPC_Q14_lo_89ABCDEF, a_Q12_89ABCDEF ); + + xmm_lo_07 = _mm_add_epi16( xmm_lo_07, xmm_tempa ); + xmm_lo_8F = _mm_add_epi16( xmm_lo_8F, xmm_tempb ); + + xmm_lo_07 = _mm_madd_epi16( xmm_lo_07, xmm_one ); + xmm_lo_8F = _mm_madd_epi16( xmm_lo_8F, xmm_one ); + + /* accumulate */ + xmm_hi_07 = _mm_add_epi32( xmm_hi_07, xmm_hi_8F ); + xmm_lo_07 = _mm_add_epi32( xmm_lo_07, xmm_lo_8F ); + + xmm_hi_07 = _mm_add_epi32( xmm_hi_07, xmm_lo_07 ); + + xmm_hi_07 = _mm_add_epi32( xmm_hi_07, _mm_unpackhi_epi64(xmm_hi_07, xmm_hi_07 ) ); + xmm_hi_07 = _mm_add_epi32( xmm_hi_07, _mm_shufflelo_epi16(xmm_hi_07, 0x0E ) ); + + LPC_pred_Q10 += _mm_cvtsi128_si32( xmm_hi_07 ); + + /* Long-term prediction */ + if ( opus_likely( signalType == TYPE_VOICED ) ) { + /* Unrolled loop */ + /* Avoids introducing a bias because silk_SMLAWB() always rounds to -inf */ + LTP_pred_Q13 = 2; + { + __m128i b_Q14_3210, b_Q14_0123, pred_lag_ptr_0123; + + b_Q14_3210 = OP_CVTEPI16_EPI32_M64( b_Q14 ); + b_Q14_0123 = _mm_shuffle_epi32( b_Q14_3210, 0x1B ); + + /* loaded: [0] [-1] [-2] [-3] */ + pred_lag_ptr_0123 = _mm_loadu_si128( (__m128i *)(&pred_lag_ptr[ -3 ] ) ); + /* shuffle to [-3] [-2] [-1] [0] and to new xmm */ + xmm_tempa = _mm_shuffle_epi32( pred_lag_ptr_0123, 0x1B ); + /*64-bit multiply, a[2] * b[-2], a[0] * b[0] */ + xmm_tempa = _mm_mul_epi32( xmm_tempa, b_Q14_3210 ); + /* right shift 2 bytes (16 bits), zero extended */ + xmm_tempa = _mm_srli_si128( xmm_tempa, 2 ); + + /* a[1] * b[-1], a[3] * b[-3] */ + pred_lag_ptr_0123 = _mm_mul_epi32( pred_lag_ptr_0123, b_Q14_0123 ); + pred_lag_ptr_0123 = _mm_srli_si128( pred_lag_ptr_0123, 2 ); + + pred_lag_ptr_0123 = _mm_add_epi32( pred_lag_ptr_0123, xmm_tempa ); + /* equal shift right 8 bytes*/ + xmm_tempa = _mm_shuffle_epi32( pred_lag_ptr_0123, _MM_SHUFFLE( 0, 0, 3, 2 ) ); + xmm_tempa = _mm_add_epi32( xmm_tempa, pred_lag_ptr_0123 ); + + LTP_pred_Q13 += _mm_cvtsi128_si32( xmm_tempa ); + + LTP_pred_Q13 = silk_SMLAWB( LTP_pred_Q13, pred_lag_ptr[ -4 ], b_Q14[ 4 ] ); + pred_lag_ptr++; + } + } + + /* Noise shape feedback */ + NSQ->sAR2_Q14[ 9 ] = NSQ->sAR2_Q14[ 8 ]; + NSQ->sAR2_Q14[ 8 ] = _mm_cvtsi128_si32( _mm_srli_si128(_mm_unpackhi_epi16( sAR2_Q14_lo_76543210, sAR2_Q14_hi_76543210 ), 12 ) ); + + sAR2_Q14_hi_76543210 = _mm_slli_si128( sAR2_Q14_hi_76543210, 2 ); + sAR2_Q14_lo_76543210 = _mm_slli_si128( sAR2_Q14_lo_76543210, 2 ); + + sAR2_Q14_hi_76543210 = _mm_insert_epi16( sAR2_Q14_hi_76543210, (sDiff_shp_Q14 >> 16), 0 ); + sAR2_Q14_lo_76543210 = _mm_insert_epi16( sAR2_Q14_lo_76543210, (sDiff_shp_Q14), 0 ); + + /* high part, use pmaddwd, results in 4 32-bit */ + xmm_hi_07 = _mm_madd_epi16( sAR2_Q14_hi_76543210, AR_shp_Q13_76543210 ); + + /* low part, use pmulhw, results in 8 16-bit, note we need simulate unsigned * signed,_mm_srai_epi16(sAR2_Q14_lo_76543210, 15) */ + xmm_tempa = _mm_cmpgt_epi16( _mm_setzero_si128(), sAR2_Q14_lo_76543210 ); + xmm_tempa = _mm_and_si128( xmm_tempa, AR_shp_Q13_76543210 ); + + xmm_lo_07 = _mm_mulhi_epi16( sAR2_Q14_lo_76543210, AR_shp_Q13_76543210 ); + xmm_lo_07 = _mm_add_epi16( xmm_lo_07, xmm_tempa ); + + xmm_lo_07 = _mm_madd_epi16( xmm_lo_07, xmm_one ); + + /* accumulate */ + xmm_hi_07 = _mm_add_epi32( xmm_hi_07, xmm_lo_07 ); + + xmm_hi_07 = _mm_add_epi32( xmm_hi_07, _mm_unpackhi_epi64(xmm_hi_07, xmm_hi_07 ) ); + xmm_hi_07 = _mm_add_epi32( xmm_hi_07, _mm_shufflelo_epi16(xmm_hi_07, 0x0E ) ); + + n_AR_Q12 = 5 + _mm_cvtsi128_si32( xmm_hi_07 ); + + n_AR_Q12 = silk_SMLAWB( n_AR_Q12, NSQ->sAR2_Q14[ 8 ], AR_shp_Q13[ 8 ] ); + n_AR_Q12 = silk_SMLAWB( n_AR_Q12, NSQ->sAR2_Q14[ 9 ], AR_shp_Q13[ 9 ] ); + + n_AR_Q12 = silk_LSHIFT32( n_AR_Q12, 1 ); /* Q11 -> Q12 */ + n_AR_Q12 = silk_SMLAWB( n_AR_Q12, sLF_AR_shp_Q14, Tilt_Q14 ); + + n_LF_Q12 = silk_SMULWB( NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx - 1 ], LF_shp_Q14 ); + n_LF_Q12 = silk_SMLAWT( n_LF_Q12, sLF_AR_shp_Q14, LF_shp_Q14 ); + + celt_assert( lag > 0 || signalType != TYPE_VOICED ); + + /* Combine prediction and noise shaping signals */ + tmp1 = silk_SUB32( silk_LSHIFT32( LPC_pred_Q10, 2 ), n_AR_Q12 ); /* Q12 */ + tmp1 = silk_SUB32( tmp1, n_LF_Q12 ); /* Q12 */ + if( lag > 0 ) { + /* Symmetric, packed FIR coefficients */ + n_LTP_Q13 = silk_SMULWB( silk_ADD_SAT32( shp_lag_ptr[ 0 ], shp_lag_ptr[ -2 ] ), HarmShapeFIRPacked_Q14 ); + n_LTP_Q13 = silk_SMLAWT( n_LTP_Q13, shp_lag_ptr[ -1 ], HarmShapeFIRPacked_Q14 ); + n_LTP_Q13 = silk_LSHIFT( n_LTP_Q13, 1 ); + shp_lag_ptr++; + + tmp2 = silk_SUB32( LTP_pred_Q13, n_LTP_Q13 ); /* Q13 */ + tmp1 = silk_ADD_LSHIFT32( tmp2, tmp1, 1 ); /* Q13 */ + tmp1 = silk_RSHIFT_ROUND( tmp1, 3 ); /* Q10 */ + } else { + tmp1 = silk_RSHIFT_ROUND( tmp1, 2 ); /* Q10 */ + } + + r_Q10 = silk_SUB32( x_sc_Q10[ i ], tmp1 ); /* residual error Q10 */ + + /* Generate dither */ + NSQ->rand_seed = silk_RAND( NSQ->rand_seed ); + + /* Flip sign depending on dither */ + tmp2 = -r_Q10; + if ( NSQ->rand_seed < 0 ) r_Q10 = tmp2; + + r_Q10 = silk_LIMIT_32( r_Q10, -(31 << 10), 30 << 10 ); + + /* Find two quantization level candidates and measure their rate-distortion */ + q1_Q10 = silk_SUB32( r_Q10, offset_Q10 ); + q1_Q0 = silk_RSHIFT( q1_Q10, 10 ); + if (Lambda_Q10 > 2048) { + /* For aggressive RDO, the bias becomes more than one pulse. */ + if (q1_Q10 > rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 - rdo_offset, 10 ); + } else if (q1_Q10 < -rdo_offset) { + q1_Q0 = silk_RSHIFT( q1_Q10 + rdo_offset, 10 ); + } else if (q1_Q10 < 0) { + q1_Q0 = -1; + } else { + q1_Q0 = 0; + } + } + + q1_Q10 = table[q1_Q0][0]; + q2_Q10 = table[q1_Q0][1]; + + if (r_Q10 * table[q1_Q0][2] - table[q1_Q0][3] < 0) + { + q1_Q10 = q2_Q10; + } + + pulses[ i ] = (opus_int8)silk_RSHIFT_ROUND( q1_Q10, 10 ); + + /* Excitation */ + exc_Q14 = silk_LSHIFT( q1_Q10, 4 ); + + tmp2 = -exc_Q14; + if ( NSQ->rand_seed < 0 ) exc_Q14 = tmp2; + + /* Add predictions */ + LPC_exc_Q14 = silk_ADD_LSHIFT32( exc_Q14, LTP_pred_Q13, 1 ); + xq_Q14 = silk_ADD_LSHIFT32( LPC_exc_Q14, LPC_pred_Q10, 4 ); + + /* Update states */ + psLPC_Q14++; + *psLPC_Q14 = xq_Q14; + NSQ->sDiff_shp_Q14 = silk_SUB_LSHIFT32( xq_Q14, x_sc_Q10[ i ], 4 ); + sLF_AR_shp_Q14 = silk_SUB_LSHIFT32( NSQ->sDiff_shp_Q14, n_AR_Q12, 2 ); + + NSQ->sLTP_shp_Q14[ NSQ->sLTP_shp_buf_idx ] = silk_SUB_LSHIFT32( sLF_AR_shp_Q14, n_LF_Q12, 2 ); + sLTP_Q15[ NSQ->sLTP_buf_idx ] = silk_LSHIFT( LPC_exc_Q14, 1 ); + NSQ->sLTP_shp_buf_idx++; + NSQ->sLTP_buf_idx++; + + /* Make dither dependent on quantized signal */ + NSQ->rand_seed = silk_ADD32_ovflw( NSQ->rand_seed, pulses[ i ] ); + } + + NSQ->sLF_AR_shp_Q14 = sLF_AR_shp_Q14; + + /* Scale XQ back to normal level before saving */ + psLPC_Q14 = &NSQ->sLPC_Q14[ NSQ_LPC_BUF_LENGTH ]; + + /* write back sAR2_Q14 */ + xmm_tempa = _mm_unpackhi_epi16( sAR2_Q14_lo_76543210, sAR2_Q14_hi_76543210 ); + xmm_tempb = _mm_unpacklo_epi16( sAR2_Q14_lo_76543210, sAR2_Q14_hi_76543210 ); + _mm_storeu_si128( (__m128i *)(&NSQ->sAR2_Q14[ 4 ]), xmm_tempa ); + _mm_storeu_si128( (__m128i *)(&NSQ->sAR2_Q14[ 0 ]), xmm_tempb ); + + /* xq[ i ] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( psLPC_Q14[ i ], Gain_Q10 ), 8 ) ); */ + { + __m128i xmm_Gain_Q10; + __m128i xmm_xq_Q14_3210, xmm_xq_Q14_x3x1, xmm_xq_Q14_7654, xmm_xq_Q14_x7x5; + + /* prepare (1 << 7) in packed 4 32-bits */ + xmm_tempa = _mm_set1_epi32( (1 << 7) ); + + /* prepare Gain_Q10 in packed 4 32-bits */ + xmm_Gain_Q10 = _mm_set1_epi32( Gain_Q10 ); + + /* process xq */ + for (i = 0; i < length - 7; i += 8) + { + xmm_xq_Q14_3210 = _mm_loadu_si128( (__m128i *)(&(psLPC_Q14[ i + 0 ] ) ) ); + xmm_xq_Q14_7654 = _mm_loadu_si128( (__m128i *)(&(psLPC_Q14[ i + 4 ] ) ) ); + + /* equal shift right 4 bytes*/ + xmm_xq_Q14_x3x1 = _mm_shuffle_epi32( xmm_xq_Q14_3210, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + /* equal shift right 4 bytes*/ + xmm_xq_Q14_x7x5 = _mm_shuffle_epi32( xmm_xq_Q14_7654, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + + xmm_xq_Q14_3210 = _mm_mul_epi32( xmm_xq_Q14_3210, xmm_Gain_Q10 ); + xmm_xq_Q14_x3x1 = _mm_mul_epi32( xmm_xq_Q14_x3x1, xmm_Gain_Q10 ); + xmm_xq_Q14_7654 = _mm_mul_epi32( xmm_xq_Q14_7654, xmm_Gain_Q10 ); + xmm_xq_Q14_x7x5 = _mm_mul_epi32( xmm_xq_Q14_x7x5, xmm_Gain_Q10 ); + + xmm_xq_Q14_3210 = _mm_srli_epi64( xmm_xq_Q14_3210, 16 ); + xmm_xq_Q14_x3x1 = _mm_slli_epi64( xmm_xq_Q14_x3x1, 16 ); + xmm_xq_Q14_7654 = _mm_srli_epi64( xmm_xq_Q14_7654, 16 ); + xmm_xq_Q14_x7x5 = _mm_slli_epi64( xmm_xq_Q14_x7x5, 16 ); + + xmm_xq_Q14_3210 = _mm_blend_epi16( xmm_xq_Q14_3210, xmm_xq_Q14_x3x1, 0xCC ); + xmm_xq_Q14_7654 = _mm_blend_epi16( xmm_xq_Q14_7654, xmm_xq_Q14_x7x5, 0xCC ); + + /* silk_RSHIFT_ROUND(xq, 8) */ + xmm_xq_Q14_3210 = _mm_add_epi32( xmm_xq_Q14_3210, xmm_tempa ); + xmm_xq_Q14_7654 = _mm_add_epi32( xmm_xq_Q14_7654, xmm_tempa ); + + xmm_xq_Q14_3210 = _mm_srai_epi32( xmm_xq_Q14_3210, 8 ); + xmm_xq_Q14_7654 = _mm_srai_epi32( xmm_xq_Q14_7654, 8 ); + + /* silk_SAT16 */ + xmm_xq_Q14_3210 = _mm_packs_epi32( xmm_xq_Q14_3210, xmm_xq_Q14_7654 ); + + /* save to xq */ + _mm_storeu_si128( (__m128i *)(&xq[ i ] ), xmm_xq_Q14_3210 ); + } + } + for ( ; i < length; i++) + { + xq[i] = (opus_int16)silk_SAT16( silk_RSHIFT_ROUND( silk_SMULWW( psLPC_Q14[ i ], Gain_Q10 ), 8 ) ); + } + + /* Update LPC synth buffer */ + silk_memcpy( NSQ->sLPC_Q14, &NSQ->sLPC_Q14[ length ], NSQ_LPC_BUF_LENGTH * sizeof( opus_int32 ) ); +} + +static OPUS_INLINE void silk_nsq_scale_states_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + const opus_int16 x16[], /* I input */ + opus_int32 x_sc_Q10[], /* O input scaled with 1/Gain */ + const opus_int16 sLTP[], /* I re-whitened LTP state in Q0 */ + opus_int32 sLTP_Q15[], /* O LTP state matching scaled input */ + opus_int subfr, /* I subframe number */ + const opus_int LTP_scale_Q14, /* I */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lag */ + const opus_int signal_type /* I Signal type */ +) +{ + opus_int i, lag; + opus_int32 gain_adj_Q16, inv_gain_Q31, inv_gain_Q26; + __m128i xmm_inv_gain_Q26, xmm_x16_x2x0, xmm_x16_x3x1; + + lag = pitchL[ subfr ]; + inv_gain_Q31 = silk_INVERSE32_varQ( silk_max( Gains_Q16[ subfr ], 1 ), 47 ); + silk_assert( inv_gain_Q31 != 0 ); + + /* Scale input */ + inv_gain_Q26 = silk_RSHIFT_ROUND( inv_gain_Q31, 5 ); + + /* prepare inv_gain_Q26 in packed 4 32-bits */ + xmm_inv_gain_Q26 = _mm_set1_epi32(inv_gain_Q26); + + for( i = 0; i < psEncC->subfr_length - 3; i += 4 ) { + xmm_x16_x2x0 = OP_CVTEPI16_EPI32_M64( &(x16[ i ] ) ); + + /* equal shift right 4 bytes*/ + xmm_x16_x3x1 = _mm_shuffle_epi32( xmm_x16_x2x0, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + + xmm_x16_x2x0 = _mm_mul_epi32( xmm_x16_x2x0, xmm_inv_gain_Q26 ); + xmm_x16_x3x1 = _mm_mul_epi32( xmm_x16_x3x1, xmm_inv_gain_Q26 ); + + xmm_x16_x2x0 = _mm_srli_epi64( xmm_x16_x2x0, 16 ); + xmm_x16_x3x1 = _mm_slli_epi64( xmm_x16_x3x1, 16 ); + + xmm_x16_x2x0 = _mm_blend_epi16( xmm_x16_x2x0, xmm_x16_x3x1, 0xCC ); + + _mm_storeu_si128( (__m128i *)(&(x_sc_Q10[ i ] ) ), xmm_x16_x2x0 ); + } + + for( ; i < psEncC->subfr_length; i++ ) { + x_sc_Q10[ i ] = silk_SMULWW( x16[ i ], inv_gain_Q26 ); + } + + /* After rewhitening the LTP state is un-scaled, so scale with inv_gain_Q16 */ + if( NSQ->rewhite_flag ) { + if( subfr == 0 ) { + /* Do LTP downscaling */ + inv_gain_Q31 = silk_LSHIFT( silk_SMULWB( inv_gain_Q31, LTP_scale_Q14 ), 2 ); + } + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx; i++ ) { + silk_assert( i < MAX_FRAME_LENGTH ); + sLTP_Q15[ i ] = silk_SMULWB( inv_gain_Q31, sLTP[ i ] ); + } + } + + /* Adjust for changing gain */ + if( Gains_Q16[ subfr ] != NSQ->prev_gain_Q16 ) { + __m128i xmm_gain_adj_Q16, xmm_sLTP_shp_Q14_x2x0, xmm_sLTP_shp_Q14_x3x1; + gain_adj_Q16 = silk_DIV32_varQ( NSQ->prev_gain_Q16, Gains_Q16[ subfr ], 16 ); + + /* Scale long-term shaping state */ + + /* prepare gain_adj_Q16 in packed 4 32-bits */ + xmm_gain_adj_Q16 = _mm_set1_epi32(gain_adj_Q16); + + for( i = NSQ->sLTP_shp_buf_idx - psEncC->ltp_mem_length; i < NSQ->sLTP_shp_buf_idx - 3; i += 4 ) + { + xmm_sLTP_shp_Q14_x2x0 = _mm_loadu_si128( (__m128i *)(&(NSQ->sLTP_shp_Q14[ i ] ) ) ); + /* equal shift right 4 bytes*/ + xmm_sLTP_shp_Q14_x3x1 = _mm_shuffle_epi32( xmm_sLTP_shp_Q14_x2x0, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + + xmm_sLTP_shp_Q14_x2x0 = _mm_mul_epi32( xmm_sLTP_shp_Q14_x2x0, xmm_gain_adj_Q16 ); + xmm_sLTP_shp_Q14_x3x1 = _mm_mul_epi32( xmm_sLTP_shp_Q14_x3x1, xmm_gain_adj_Q16 ); + + xmm_sLTP_shp_Q14_x2x0 = _mm_srli_epi64( xmm_sLTP_shp_Q14_x2x0, 16 ); + xmm_sLTP_shp_Q14_x3x1 = _mm_slli_epi64( xmm_sLTP_shp_Q14_x3x1, 16 ); + + xmm_sLTP_shp_Q14_x2x0 = _mm_blend_epi16( xmm_sLTP_shp_Q14_x2x0, xmm_sLTP_shp_Q14_x3x1, 0xCC ); + + _mm_storeu_si128( (__m128i *)(&(NSQ->sLTP_shp_Q14[ i ] ) ), xmm_sLTP_shp_Q14_x2x0 ); + } + + for( ; i < NSQ->sLTP_shp_buf_idx; i++ ) { + NSQ->sLTP_shp_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLTP_shp_Q14[ i ] ); + } + + /* Scale long-term prediction state */ + if( signal_type == TYPE_VOICED && NSQ->rewhite_flag == 0 ) { + for( i = NSQ->sLTP_buf_idx - lag - LTP_ORDER / 2; i < NSQ->sLTP_buf_idx; i++ ) { + sLTP_Q15[ i ] = silk_SMULWW( gain_adj_Q16, sLTP_Q15[ i ] ); + } + } + + NSQ->sLF_AR_shp_Q14 = silk_SMULWW( gain_adj_Q16, NSQ->sLF_AR_shp_Q14 ); + NSQ->sDiff_shp_Q14 = silk_SMULWW( gain_adj_Q16, NSQ->sDiff_shp_Q14 ); + + /* Scale short-term prediction and shaping states */ + for( i = 0; i < NSQ_LPC_BUF_LENGTH; i++ ) { + NSQ->sLPC_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sLPC_Q14[ i ] ); + } + for( i = 0; i < MAX_SHAPE_LPC_ORDER; i++ ) { + NSQ->sAR2_Q14[ i ] = silk_SMULWW( gain_adj_Q16, NSQ->sAR2_Q14[ i ] ); + } + + /* Save inverse gain */ + NSQ->prev_gain_Q16 = Gains_Q16[ subfr ]; + } +} diff --git a/lib/opus/silk/x86/SigProc_FIX_sse.h b/lib/opus/silk/x86/SigProc_FIX_sse.h new file mode 100644 index 00000000..89a5ec88 --- /dev/null +++ b/lib/opus/silk/x86/SigProc_FIX_sse.h @@ -0,0 +1,99 @@ +/* Copyright (c) 2014, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef SIGPROC_FIX_SSE_H +# define SIGPROC_FIX_SSE_H + +# ifdef HAVE_CONFIG_H +# include "config.h" +# endif + +# if defined(OPUS_X86_MAY_HAVE_SSE4_1) +void silk_burg_modified_sse4_1( + opus_int32 *res_nrg, /* O Residual energy */ + opus_int *res_nrg_Q, /* O Residual energy Q value */ + opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ + const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ + const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ + const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I Number of subframes stacked in x */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */ +); + +# if defined(OPUS_X86_PRESUME_SSE4_1) + +# define OVERRIDE_silk_burg_modified +# define silk_burg_modified(res_nrg, res_nrg_Q, A_Q16, x, minInvGain_Q30, subfr_length, nb_subfr, D, arch) \ + ((void)(arch), silk_burg_modified_sse4_1(res_nrg, res_nrg_Q, A_Q16, x, minInvGain_Q30, subfr_length, nb_subfr, D, arch)) + +# elif defined(OPUS_HAVE_RTCD) + +extern void (*const SILK_BURG_MODIFIED_IMPL[OPUS_ARCHMASK + 1])( + opus_int32 *res_nrg, /* O Residual energy */ + opus_int *res_nrg_Q, /* O Residual energy Q value */ + opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ + const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ + const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ + const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I Number of subframes stacked in x */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */); + +# define OVERRIDE_silk_burg_modified +# define silk_burg_modified(res_nrg, res_nrg_Q, A_Q16, x, minInvGain_Q30, subfr_length, nb_subfr, D, arch) \ + ((*SILK_BURG_MODIFIED_IMPL[(arch) & OPUS_ARCHMASK])(res_nrg, res_nrg_Q, A_Q16, x, minInvGain_Q30, subfr_length, nb_subfr, D, arch)) + +# endif + +opus_int64 silk_inner_prod16_sse4_1( + const opus_int16 *inVec1, + const opus_int16 *inVec2, + const opus_int len +); + + +# if defined(OPUS_X86_PRESUME_SSE4_1) + +# define OVERRIDE_silk_inner_prod16 +# define silk_inner_prod16(inVec1, inVec2, len, arch) \ + ((void)(arch),silk_inner_prod16_sse4_1(inVec1, inVec2, len)) + +# elif defined(OPUS_HAVE_RTCD) + +extern opus_int64 (*const SILK_INNER_PROD16_IMPL[OPUS_ARCHMASK + 1])( + const opus_int16 *inVec1, + const opus_int16 *inVec2, + const opus_int len); + +# define OVERRIDE_silk_inner_prod16 +# define silk_inner_prod16(inVec1, inVec2, len, arch) \ + ((*SILK_INNER_PROD16_IMPL[(arch) & OPUS_ARCHMASK])(inVec1, inVec2, len)) + +# endif +# endif +#endif diff --git a/lib/opus/silk/x86/VAD_sse4_1.c b/lib/opus/silk/x86/VAD_sse4_1.c new file mode 100644 index 00000000..e7eaf971 --- /dev/null +++ b/lib/opus/silk/x86/VAD_sse4_1.c @@ -0,0 +1,289 @@ +/* Copyright (c) 2014-2020, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang FrancisQuiers + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include + +#include "main.h" +#include "stack_alloc.h" + +/* Weighting factors for tilt measure */ +static const opus_int32 tiltWeights[ VAD_N_BANDS ] = { 30000, 6000, -12000, -12000 }; + +/***************************************/ +/* Get the speech activity level in Q8 */ +/***************************************/ +opus_int silk_VAD_GetSA_Q8_sse4_1( /* O Return value, 0 if success */ + silk_encoder_state *psEncC, /* I/O Encoder state */ + const opus_int16 pIn[] /* I PCM input */ +) +{ + opus_int SA_Q15, pSNR_dB_Q7, input_tilt; + opus_int decimated_framelength1, decimated_framelength2; + opus_int decimated_framelength; + opus_int dec_subframe_length, dec_subframe_offset, SNR_Q7, i, b, s; + opus_int32 sumSquared, smooth_coef_Q16; + opus_int16 HPstateTmp; + VARDECL( opus_int16, X ); + opus_int32 Xnrg[ VAD_N_BANDS ]; + opus_int32 NrgToNoiseRatio_Q8[ VAD_N_BANDS ]; + opus_int32 speech_nrg, x_tmp; + opus_int X_offset[ VAD_N_BANDS ]; + opus_int ret = 0; + silk_VAD_state *psSilk_VAD = &psEncC->sVAD; + + SAVE_STACK; + +#ifdef OPUS_CHECK_ASM + silk_encoder_state psEncC_c; + opus_int ret_c; + + silk_memcpy( &psEncC_c, psEncC, sizeof( psEncC_c ) ); + ret_c = silk_VAD_GetSA_Q8_c( &psEncC_c, pIn ); +#endif + + /* Safety checks */ + silk_assert( VAD_N_BANDS == 4 ); + celt_assert( MAX_FRAME_LENGTH >= psEncC->frame_length ); + celt_assert( psEncC->frame_length <= 512 ); + celt_assert( psEncC->frame_length == 8 * silk_RSHIFT( psEncC->frame_length, 3 ) ); + + /***********************/ + /* Filter and Decimate */ + /***********************/ + decimated_framelength1 = silk_RSHIFT( psEncC->frame_length, 1 ); + decimated_framelength2 = silk_RSHIFT( psEncC->frame_length, 2 ); + decimated_framelength = silk_RSHIFT( psEncC->frame_length, 3 ); + /* Decimate into 4 bands: + 0 L 3L L 3L 5L + - -- - -- -- + 8 8 2 4 4 + + [0-1 kHz| temp. |1-2 kHz| 2-4 kHz | 4-8 kHz | + + They're arranged to allow the minimal ( frame_length / 4 ) extra + scratch space during the downsampling process */ + X_offset[ 0 ] = 0; + X_offset[ 1 ] = decimated_framelength + decimated_framelength2; + X_offset[ 2 ] = X_offset[ 1 ] + decimated_framelength; + X_offset[ 3 ] = X_offset[ 2 ] + decimated_framelength2; + ALLOC( X, X_offset[ 3 ] + decimated_framelength1, opus_int16 ); + + /* 0-8 kHz to 0-4 kHz and 4-8 kHz */ + silk_ana_filt_bank_1( pIn, &psSilk_VAD->AnaState[ 0 ], + X, &X[ X_offset[ 3 ] ], psEncC->frame_length ); + + /* 0-4 kHz to 0-2 kHz and 2-4 kHz */ + silk_ana_filt_bank_1( X, &psSilk_VAD->AnaState1[ 0 ], + X, &X[ X_offset[ 2 ] ], decimated_framelength1 ); + + /* 0-2 kHz to 0-1 kHz and 1-2 kHz */ + silk_ana_filt_bank_1( X, &psSilk_VAD->AnaState2[ 0 ], + X, &X[ X_offset[ 1 ] ], decimated_framelength2 ); + + /*********************************************/ + /* HP filter on lowest band (differentiator) */ + /*********************************************/ + X[ decimated_framelength - 1 ] = silk_RSHIFT( X[ decimated_framelength - 1 ], 1 ); + HPstateTmp = X[ decimated_framelength - 1 ]; + for( i = decimated_framelength - 1; i > 0; i-- ) { + X[ i - 1 ] = silk_RSHIFT( X[ i - 1 ], 1 ); + X[ i ] -= X[ i - 1 ]; + } + X[ 0 ] -= psSilk_VAD->HPstate; + psSilk_VAD->HPstate = HPstateTmp; + + /*************************************/ + /* Calculate the energy in each band */ + /*************************************/ + for( b = 0; b < VAD_N_BANDS; b++ ) { + /* Find the decimated framelength in the non-uniformly divided bands */ + decimated_framelength = silk_RSHIFT( psEncC->frame_length, silk_min_int( VAD_N_BANDS - b, VAD_N_BANDS - 1 ) ); + + /* Split length into subframe lengths */ + dec_subframe_length = silk_RSHIFT( decimated_framelength, VAD_INTERNAL_SUBFRAMES_LOG2 ); + dec_subframe_offset = 0; + + /* Compute energy per sub-frame */ + /* initialize with summed energy of last subframe */ + Xnrg[ b ] = psSilk_VAD->XnrgSubfr[ b ]; + for( s = 0; s < VAD_INTERNAL_SUBFRAMES; s++ ) { + __m128i xmm_X, xmm_acc; + sumSquared = 0; + + xmm_acc = _mm_setzero_si128(); + + for( i = 0; i < dec_subframe_length - 7; i += 8 ) + { + xmm_X = _mm_loadu_si128( (__m128i *)&(X[ X_offset[ b ] + i + dec_subframe_offset ] ) ); + xmm_X = _mm_srai_epi16( xmm_X, 3 ); + xmm_X = _mm_madd_epi16( xmm_X, xmm_X ); + xmm_acc = _mm_add_epi32( xmm_acc, xmm_X ); + } + + xmm_acc = _mm_add_epi32( xmm_acc, _mm_unpackhi_epi64( xmm_acc, xmm_acc ) ); + xmm_acc = _mm_add_epi32( xmm_acc, _mm_shufflelo_epi16( xmm_acc, 0x0E ) ); + + sumSquared += _mm_cvtsi128_si32( xmm_acc ); + + for( ; i < dec_subframe_length; i++ ) { + /* The energy will be less than dec_subframe_length * ( silk_int16_MIN / 8 ) ^ 2. */ + /* Therefore we can accumulate with no risk of overflow (unless dec_subframe_length > 128) */ + x_tmp = silk_RSHIFT( + X[ X_offset[ b ] + i + dec_subframe_offset ], 3 ); + sumSquared = silk_SMLABB( sumSquared, x_tmp, x_tmp ); + + /* Safety check */ + silk_assert( sumSquared >= 0 ); + } + + /* Add/saturate summed energy of current subframe */ + if( s < VAD_INTERNAL_SUBFRAMES - 1 ) { + Xnrg[ b ] = silk_ADD_POS_SAT32( Xnrg[ b ], sumSquared ); + } else { + /* Look-ahead subframe */ + Xnrg[ b ] = silk_ADD_POS_SAT32( Xnrg[ b ], silk_RSHIFT( sumSquared, 1 ) ); + } + + dec_subframe_offset += dec_subframe_length; + } + psSilk_VAD->XnrgSubfr[ b ] = sumSquared; + } + + /********************/ + /* Noise estimation */ + /********************/ + silk_VAD_GetNoiseLevels( &Xnrg[ 0 ], psSilk_VAD ); + + /***********************************************/ + /* Signal-plus-noise to noise ratio estimation */ + /***********************************************/ + sumSquared = 0; + input_tilt = 0; + for( b = 0; b < VAD_N_BANDS; b++ ) { + speech_nrg = Xnrg[ b ] - psSilk_VAD->NL[ b ]; + if( speech_nrg > 0 ) { + /* Divide, with sufficient resolution */ + if( ( Xnrg[ b ] & 0xFF800000 ) == 0 ) { + NrgToNoiseRatio_Q8[ b ] = silk_DIV32( silk_LSHIFT( Xnrg[ b ], 8 ), psSilk_VAD->NL[ b ] + 1 ); + } else { + NrgToNoiseRatio_Q8[ b ] = silk_DIV32( Xnrg[ b ], silk_RSHIFT( psSilk_VAD->NL[ b ], 8 ) + 1 ); + } + + /* Convert to log domain */ + SNR_Q7 = silk_lin2log( NrgToNoiseRatio_Q8[ b ] ) - 8 * 128; + + /* Sum-of-squares */ + sumSquared = silk_SMLABB( sumSquared, SNR_Q7, SNR_Q7 ); /* Q14 */ + + /* Tilt measure */ + if( speech_nrg < ( (opus_int32)1 << 20 ) ) { + /* Scale down SNR value for small subband speech energies */ + SNR_Q7 = silk_SMULWB( silk_LSHIFT( silk_SQRT_APPROX( speech_nrg ), 6 ), SNR_Q7 ); + } + input_tilt = silk_SMLAWB( input_tilt, tiltWeights[ b ], SNR_Q7 ); + } else { + NrgToNoiseRatio_Q8[ b ] = 256; + } + } + + /* Mean-of-squares */ + sumSquared = silk_DIV32_16( sumSquared, VAD_N_BANDS ); /* Q14 */ + + /* Root-mean-square approximation, scale to dBs, and write to output pointer */ + pSNR_dB_Q7 = (opus_int16)( 3 * silk_SQRT_APPROX( sumSquared ) ); /* Q7 */ + + /*********************************/ + /* Speech Probability Estimation */ + /*********************************/ + SA_Q15 = silk_sigm_Q15( silk_SMULWB( VAD_SNR_FACTOR_Q16, pSNR_dB_Q7 ) - VAD_NEGATIVE_OFFSET_Q5 ); + + /**************************/ + /* Frequency Tilt Measure */ + /**************************/ + psEncC->input_tilt_Q15 = silk_LSHIFT( silk_sigm_Q15( input_tilt ) - 16384, 1 ); + + /**************************************************/ + /* Scale the sigmoid output based on power levels */ + /**************************************************/ + speech_nrg = 0; + for( b = 0; b < VAD_N_BANDS; b++ ) { + /* Accumulate signal-without-noise energies, higher frequency bands have more weight */ + speech_nrg += ( b + 1 ) * silk_RSHIFT( Xnrg[ b ] - psSilk_VAD->NL[ b ], 4 ); + } + + if( psEncC->frame_length == 20 * psEncC->fs_kHz ) { + speech_nrg = silk_RSHIFT32( speech_nrg, 1 ); + } + /* Power scaling */ + if( speech_nrg <= 0 ) { + SA_Q15 = silk_RSHIFT( SA_Q15, 1 ); + } else if( speech_nrg < 16384 ) { + speech_nrg = silk_LSHIFT32( speech_nrg, 16 ); + + /* square-root */ + speech_nrg = silk_SQRT_APPROX( speech_nrg ); + SA_Q15 = silk_SMULWB( 32768 + speech_nrg, SA_Q15 ); + } + + /* Copy the resulting speech activity in Q8 */ + psEncC->speech_activity_Q8 = silk_min_int( silk_RSHIFT( SA_Q15, 7 ), silk_uint8_MAX ); + + /***********************************/ + /* Energy Level and SNR estimation */ + /***********************************/ + /* Smoothing coefficient */ + smooth_coef_Q16 = silk_SMULWB( VAD_SNR_SMOOTH_COEF_Q18, silk_SMULWB( (opus_int32)SA_Q15, SA_Q15 ) ); + + if( psEncC->frame_length == 10 * psEncC->fs_kHz ) { + smooth_coef_Q16 >>= 1; + } + + for( b = 0; b < VAD_N_BANDS; b++ ) { + /* compute smoothed energy-to-noise ratio per band */ + psSilk_VAD->NrgRatioSmth_Q8[ b ] = silk_SMLAWB( psSilk_VAD->NrgRatioSmth_Q8[ b ], + NrgToNoiseRatio_Q8[ b ] - psSilk_VAD->NrgRatioSmth_Q8[ b ], smooth_coef_Q16 ); + + /* signal to noise ratio in dB per band */ + SNR_Q7 = 3 * ( silk_lin2log( psSilk_VAD->NrgRatioSmth_Q8[b] ) - 8 * 128 ); + /* quality = sigmoid( 0.25 * ( SNR_dB - 16 ) ); */ + psEncC->input_quality_bands_Q15[ b ] = silk_sigm_Q15( silk_RSHIFT( SNR_Q7 - 16 * 128, 4 ) ); + } + +#ifdef OPUS_CHECK_ASM + silk_assert( ret == ret_c ); + silk_assert( !memcmp( &psEncC_c, psEncC, sizeof( psEncC_c ) ) ); +#endif + + RESTORE_STACK; + return( ret ); +} diff --git a/lib/opus/silk/x86/VQ_WMat_EC_sse4_1.c b/lib/opus/silk/x86/VQ_WMat_EC_sse4_1.c new file mode 100644 index 00000000..2c7d18d0 --- /dev/null +++ b/lib/opus/silk/x86/VQ_WMat_EC_sse4_1.c @@ -0,0 +1,173 @@ +/* Copyright (c) 2014-2020, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang FrancisQuiers + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include "main.h" +#include "celt/x86/x86cpu.h" + +/* Entropy constrained matrix-weighted VQ, hard-coded to 5-element vectors, for a single input data vector */ +void silk_VQ_WMat_EC_sse4_1( + opus_int8 *ind, /* O index of best codebook vector */ + opus_int32 *res_nrg_Q15, /* O best residual energy */ + opus_int32 *rate_dist_Q8, /* O best total bitrate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ + const opus_int32 *XX_Q17, /* I correlation matrix */ + const opus_int32 *xX_Q17, /* I correlation vector */ + const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ + const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ + const opus_int subfr_len, /* I number of samples per subframe */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ + const opus_int L /* I number of vectors in codebook */ +) +{ + opus_int k, gain_tmp_Q7; + const opus_int8 *cb_row_Q7; + opus_int32 neg_xX_Q24[ 5 ]; + opus_int32 sum1_Q15, sum2_Q24; + opus_int32 bits_res_Q8, bits_tot_Q8; + __m128i v_XX_31_Q17, v_XX_42_Q17, v_cb_row_31_Q7, v_cb_row_42_Q7, v_acc1_Q24, v_acc2_Q24; + + /* Negate and convert to new Q domain */ + neg_xX_Q24[ 0 ] = -silk_LSHIFT32( xX_Q17[ 0 ], 7 ); + neg_xX_Q24[ 1 ] = -silk_LSHIFT32( xX_Q17[ 1 ], 7 ); + neg_xX_Q24[ 2 ] = -silk_LSHIFT32( xX_Q17[ 2 ], 7 ); + neg_xX_Q24[ 3 ] = -silk_LSHIFT32( xX_Q17[ 3 ], 7 ); + neg_xX_Q24[ 4 ] = -silk_LSHIFT32( xX_Q17[ 4 ], 7 ); + + v_XX_31_Q17 = _mm_loadu_si128( (__m128i *)(&XX_Q17[ 1 ] ) ); + v_XX_42_Q17 = _mm_shuffle_epi32( v_XX_31_Q17, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + + /* Loop over codebook */ + *rate_dist_Q8 = silk_int32_MAX; + *res_nrg_Q15 = silk_int32_MAX; + cb_row_Q7 = cb_Q7; + /* If things go really bad, at least *ind is set to something safe. */ + *ind = 0; + for( k = 0; k < L; k++ ) { + opus_int32 penalty; + gain_tmp_Q7 = cb_gain_Q7[k]; + /* Weighted rate */ + /* Quantization error: 1 - 2 * xX * cb + cb' * XX * cb */ + sum1_Q15 = SILK_FIX_CONST( 1.001, 15 ); + + /* Penalty for too large gain */ + penalty = silk_LSHIFT32( silk_max( silk_SUB32( gain_tmp_Q7, max_gain_Q7 ), 0 ), 11 ); + + /* first row of XX_Q17 */ + v_cb_row_31_Q7 = OP_CVTEPI8_EPI32_M32( &cb_row_Q7[ 1 ] ); + v_cb_row_42_Q7 = _mm_shuffle_epi32( v_cb_row_31_Q7, _MM_SHUFFLE( 0, 3, 2, 1 ) ); + v_cb_row_31_Q7 = _mm_mul_epi32( v_XX_31_Q17, v_cb_row_31_Q7 ); + v_cb_row_42_Q7 = _mm_mul_epi32( v_XX_42_Q17, v_cb_row_42_Q7 ); + v_acc1_Q24 = _mm_add_epi64( v_cb_row_31_Q7, v_cb_row_42_Q7); + v_acc2_Q24 = _mm_shuffle_epi32( v_acc1_Q24, _MM_SHUFFLE( 1, 0, 3, 2 ) ); + v_acc1_Q24 = _mm_add_epi64( v_acc1_Q24, v_acc2_Q24); + sum2_Q24 = _mm_cvtsi128_si32( v_acc1_Q24 ); + sum2_Q24 = silk_ADD32( neg_xX_Q24[ 0 ], sum2_Q24 ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 0 ], cb_row_Q7[ 0 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 0 ] ); + + /* second row of XX_Q17 */ + sum2_Q24 = silk_MLA( neg_xX_Q24[ 1 ], XX_Q17[ 7 ], cb_row_Q7[ 2 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 8 ], cb_row_Q7[ 3 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 9 ], cb_row_Q7[ 4 ] ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 6 ], cb_row_Q7[ 1 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 1 ] ); + + /* third row of XX_Q17 */ + sum2_Q24 = silk_MLA( neg_xX_Q24[ 2 ], XX_Q17[ 13 ], cb_row_Q7[ 3 ] ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 14 ], cb_row_Q7[ 4 ] ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 12 ], cb_row_Q7[ 2 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 2 ] ); + + /* fourth row of XX_Q17 */ + sum2_Q24 = silk_MLA( neg_xX_Q24[ 3 ], XX_Q17[ 19 ], cb_row_Q7[ 4 ] ); + sum2_Q24 = silk_LSHIFT32( sum2_Q24, 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 18 ], cb_row_Q7[ 3 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 3 ] ); + + /* last row of XX_Q17 */ + sum2_Q24 = silk_LSHIFT32( neg_xX_Q24[ 4 ], 1 ); + sum2_Q24 = silk_MLA( sum2_Q24, XX_Q17[ 24 ], cb_row_Q7[ 4 ] ); + sum1_Q15 = silk_SMLAWB( sum1_Q15, sum2_Q24, cb_row_Q7[ 4 ] ); + + /* find best */ + if( sum1_Q15 >= 0 ) { + /* Translate residual energy to bits using high-rate assumption (6 dB ==> 1 bit/sample) */ + bits_res_Q8 = silk_SMULBB( subfr_len, silk_lin2log( sum1_Q15 + penalty) - (15 << 7) ); + /* In the following line we reduce the codelength component by half ("-1"); seems to slightly improve quality */ + bits_tot_Q8 = silk_ADD_LSHIFT32( bits_res_Q8, cl_Q5[ k ], 3-1 ); + if( bits_tot_Q8 <= *rate_dist_Q8 ) { + *rate_dist_Q8 = bits_tot_Q8; + *res_nrg_Q15 = sum1_Q15 + penalty; + *ind = (opus_int8)k; + *gain_Q7 = gain_tmp_Q7; + } + } + + /* Go to next cbk vector */ + cb_row_Q7 += LTP_ORDER; + } + +#ifdef OPUS_CHECK_ASM + { + opus_int8 ind_c = 0; + opus_int32 res_nrg_Q15_c = 0; + opus_int32 rate_dist_Q8_c = 0; + opus_int gain_Q7_c = 0; + + silk_VQ_WMat_EC_c( + &ind_c, + &res_nrg_Q15_c, + &rate_dist_Q8_c, + &gain_Q7_c, + XX_Q17, + xX_Q17, + cb_Q7, + cb_gain_Q7, + cl_Q5, + subfr_len, + max_gain_Q7, + L + ); + + silk_assert( *ind == ind_c ); + silk_assert( *res_nrg_Q15 == res_nrg_Q15_c ); + silk_assert( *rate_dist_Q8 == rate_dist_Q8_c ); + silk_assert( *gain_Q7 == gain_Q7_c ); + } +#endif +} diff --git a/lib/opus/silk/x86/main_sse.h b/lib/opus/silk/x86/main_sse.h new file mode 100644 index 00000000..a01d7f6c --- /dev/null +++ b/lib/opus/silk/x86/main_sse.h @@ -0,0 +1,247 @@ +/* Copyright (c) 2014, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef MAIN_SSE_H +# define MAIN_SSE_H + +# ifdef HAVE_CONFIG_H +# include "config.h" +# endif + +# if defined(OPUS_X86_MAY_HAVE_SSE4_1) + +void silk_VQ_WMat_EC_sse4_1( + opus_int8 *ind, /* O index of best codebook vector */ + opus_int32 *res_nrg_Q15, /* O best residual energy */ + opus_int32 *rate_dist_Q8, /* O best total bitrate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ + const opus_int32 *XX_Q17, /* I correlation matrix */ + const opus_int32 *xX_Q17, /* I correlation vector */ + const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ + const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ + const opus_int subfr_len, /* I number of samples per subframe */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ + const opus_int L /* I number of vectors in codebook */ +); + +# if defined OPUS_X86_PRESUME_SSE4_1 + +# define OVERRIDE_silk_VQ_WMat_EC +# define silk_VQ_WMat_EC(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, \ + subfr_len, max_gain_Q7, L, arch) \ + ((void)(arch),silk_VQ_WMat_EC_sse4_1(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, \ + subfr_len, max_gain_Q7, L)) + +# elif defined(OPUS_HAVE_RTCD) + +extern void (*const SILK_VQ_WMAT_EC_IMPL[OPUS_ARCHMASK + 1])( + opus_int8 *ind, /* O index of best codebook vector */ + opus_int32 *res_nrg_Q15, /* O best residual energy */ + opus_int32 *rate_dist_Q8, /* O best total bitrate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ + const opus_int32 *XX_Q17, /* I correlation matrix */ + const opus_int32 *xX_Q17, /* I correlation vector */ + const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ + const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ + const opus_int subfr_len, /* I number of samples per subframe */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ + const opus_int L /* I number of vectors in codebook */ +); + +# define OVERRIDE_silk_VQ_WMat_EC +# define silk_VQ_WMat_EC(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, \ + subfr_len, max_gain_Q7, L, arch) \ + ((*SILK_VQ_WMAT_EC_IMPL[(arch) & OPUS_ARCHMASK])(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, \ + subfr_len, max_gain_Q7, L)) + +# endif + +void silk_NSQ_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +); + +# if defined OPUS_X86_PRESUME_SSE4_1 + +# define OVERRIDE_silk_NSQ +# define silk_NSQ(psEncC, NSQ, psIndices, x_Q3, pulses, PredCoef_Q12, LTPCoef_Q14, AR2_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14, arch) \ + ((void)(arch),silk_NSQ_sse4_1(psEncC, NSQ, psIndices, x_Q3, pulses, PredCoef_Q12, LTPCoef_Q14, AR2_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14)) + +# elif defined(OPUS_HAVE_RTCD) + +extern void (*const SILK_NSQ_IMPL[OPUS_ARCHMASK + 1])( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +); + +# define OVERRIDE_silk_NSQ +# define silk_NSQ(psEncC, NSQ, psIndices, x_Q3, pulses, PredCoef_Q12, LTPCoef_Q14, AR2_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14, arch) \ + ((*SILK_NSQ_IMPL[(arch) & OPUS_ARCHMASK])(psEncC, NSQ, psIndices, x_Q3, pulses, PredCoef_Q12, LTPCoef_Q14, AR2_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14)) + +# endif + +void silk_NSQ_del_dec_sse4_1( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +); + +# if defined OPUS_X86_PRESUME_SSE4_1 + +# define OVERRIDE_silk_NSQ_del_dec +# define silk_NSQ_del_dec(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14, arch) \ + ((void)(arch),silk_NSQ_del_dec_sse4_1(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14)) + +# elif defined(OPUS_HAVE_RTCD) + +extern void (*const SILK_NSQ_DEL_DEC_IMPL[OPUS_ARCHMASK + 1])( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +); + +# define OVERRIDE_silk_NSQ_del_dec +# define silk_NSQ_del_dec(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14, arch) \ + ((*SILK_NSQ_DEL_DEC_IMPL[(arch) & OPUS_ARCHMASK])(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \ + HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14)) + +# endif + +void silk_noise_shape_quantizer( + silk_nsq_state *NSQ, /* I/O NSQ state */ + opus_int signalType, /* I Signal type */ + const opus_int32 x_sc_Q10[], /* I */ + opus_int8 pulses[], /* O */ + opus_int16 xq[], /* O */ + opus_int32 sLTP_Q15[], /* I/O LTP state */ + const opus_int16 a_Q12[], /* I Short term prediction coefs */ + const opus_int16 b_Q14[], /* I Long term prediction coefs */ + const opus_int16 AR_shp_Q13[], /* I Noise shaping AR coefs */ + opus_int lag, /* I Pitch lag */ + opus_int32 HarmShapeFIRPacked_Q14, /* I */ + opus_int Tilt_Q14, /* I Spectral tilt */ + opus_int32 LF_shp_Q14, /* I */ + opus_int32 Gain_Q16, /* I */ + opus_int Lambda_Q10, /* I */ + opus_int offset_Q10, /* I */ + opus_int length, /* I Input length */ + opus_int shapingLPCOrder, /* I Noise shaping AR filter order */ + opus_int predictLPCOrder, /* I Prediction filter order */ + int arch /* I Architecture */ +); + +/**************************/ +/* Noise level estimation */ +/**************************/ +void silk_VAD_GetNoiseLevels( + const opus_int32 pX[ VAD_N_BANDS ], /* I subband energies */ + silk_VAD_state *psSilk_VAD /* I/O Pointer to Silk VAD state */ +); + +opus_int silk_VAD_GetSA_Q8_sse4_1( + silk_encoder_state *psEnC, + const opus_int16 pIn[] +); + +# if defined(OPUS_X86_PRESUME_SSE4_1) + +# define OVERRIDE_silk_VAD_GetSA_Q8 +# define silk_VAD_GetSA_Q8(psEnC, pIn, arch) ((void)(arch),silk_VAD_GetSA_Q8_sse4_1(psEnC, pIn)) + +# elif defined(OPUS_HAVE_RTCD) + +extern opus_int (*const SILK_VAD_GETSA_Q8_IMPL[OPUS_ARCHMASK + 1])( + silk_encoder_state *psEnC, + const opus_int16 pIn[]); + +# define OVERRIDE_silk_VAD_GetSA_Q8 +# define silk_VAD_GetSA_Q8(psEnC, pIn, arch) \ + ((*SILK_VAD_GETSA_Q8_IMPL[(arch) & OPUS_ARCHMASK])(psEnC, pIn)) + +# endif + +# endif +#endif diff --git a/lib/opus/silk/x86/x86_silk_map.c b/lib/opus/silk/x86/x86_silk_map.c new file mode 100644 index 00000000..70f60078 --- /dev/null +++ b/lib/opus/silk/x86/x86_silk_map.c @@ -0,0 +1,159 @@ +/* Copyright (c) 2014, Cisco Systems, INC + Written by XiangMingZhu WeiZhou MinPeng YanWang + + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#if defined(HAVE_CONFIG_H) +#include "config.h" +#endif + +#include "celt/x86/x86cpu.h" +#include "structs.h" +#include "SigProc_FIX.h" +#include "pitch.h" +#include "main.h" + +#if defined(OPUS_HAVE_RTCD) && !defined(OPUS_X86_PRESUME_SSE4_1) + +#if defined(FIXED_POINT) + +#include "fixed/main_FIX.h" + +opus_int64 (*const SILK_INNER_PROD16_IMPL[ OPUS_ARCHMASK + 1 ] )( + const opus_int16 *inVec1, + const opus_int16 *inVec2, + const opus_int len +) = { + silk_inner_prod16_c, /* non-sse */ + silk_inner_prod16_c, + silk_inner_prod16_c, + MAY_HAVE_SSE4_1( silk_inner_prod16 ), /* sse4.1 */ + MAY_HAVE_SSE4_1( silk_inner_prod16 ) /* avx */ +}; + +#endif + +opus_int (*const SILK_VAD_GETSA_Q8_IMPL[ OPUS_ARCHMASK + 1 ] )( + silk_encoder_state *psEncC, + const opus_int16 pIn[] +) = { + silk_VAD_GetSA_Q8_c, /* non-sse */ + silk_VAD_GetSA_Q8_c, + silk_VAD_GetSA_Q8_c, + MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 ), /* sse4.1 */ + MAY_HAVE_SSE4_1( silk_VAD_GetSA_Q8 ) /* avx */ +}; + +void (*const SILK_NSQ_IMPL[ OPUS_ARCHMASK + 1 ] )( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) = { + silk_NSQ_c, /* non-sse */ + silk_NSQ_c, + silk_NSQ_c, + MAY_HAVE_SSE4_1( silk_NSQ ), /* sse4.1 */ + MAY_HAVE_SSE4_1( silk_NSQ ) /* avx */ +}; + +void (*const SILK_VQ_WMAT_EC_IMPL[ OPUS_ARCHMASK + 1 ] )( + opus_int8 *ind, /* O index of best codebook vector */ + opus_int32 *res_nrg_Q15, /* O best residual energy */ + opus_int32 *rate_dist_Q8, /* O best total bitrate */ + opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ + const opus_int32 *XX_Q17, /* I correlation matrix */ + const opus_int32 *xX_Q17, /* I correlation vector */ + const opus_int8 *cb_Q7, /* I codebook */ + const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ + const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ + const opus_int subfr_len, /* I number of samples per subframe */ + const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ + const opus_int L /* I number of vectors in codebook */ +) = { + silk_VQ_WMat_EC_c, /* non-sse */ + silk_VQ_WMat_EC_c, + silk_VQ_WMat_EC_c, + MAY_HAVE_SSE4_1( silk_VQ_WMat_EC ), /* sse4.1 */ + MAY_HAVE_SSE4_1( silk_VQ_WMat_EC ) /* avx */ +}; + +void (*const SILK_NSQ_DEL_DEC_IMPL[ OPUS_ARCHMASK + 1 ] )( + const silk_encoder_state *psEncC, /* I Encoder State */ + silk_nsq_state *NSQ, /* I/O NSQ state */ + SideInfoIndices *psIndices, /* I/O Quantization Indices */ + const opus_int16 x16[], /* I Input */ + opus_int8 pulses[], /* O Quantized pulse signal */ + const opus_int16 PredCoef_Q12[ 2 * MAX_LPC_ORDER ], /* I Short term prediction coefs */ + const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ + const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ + const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ + const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ + const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ + const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ + const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ + const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ + const opus_int LTP_scale_Q14 /* I LTP state scaling */ +) = { + silk_NSQ_del_dec_c, /* non-sse */ + silk_NSQ_del_dec_c, + silk_NSQ_del_dec_c, + MAY_HAVE_SSE4_1( silk_NSQ_del_dec ), /* sse4.1 */ + MAY_HAVE_SSE4_1( silk_NSQ_del_dec ) /* avx */ +}; + +#if defined(FIXED_POINT) + +void (*const SILK_BURG_MODIFIED_IMPL[ OPUS_ARCHMASK + 1 ] )( + opus_int32 *res_nrg, /* O Residual energy */ + opus_int *res_nrg_Q, /* O Residual energy Q value */ + opus_int32 A_Q16[], /* O Prediction coefficients (length order) */ + const opus_int16 x[], /* I Input signal, length: nb_subfr * ( D + subfr_length ) */ + const opus_int32 minInvGain_Q30, /* I Inverse of max prediction gain */ + const opus_int subfr_length, /* I Input signal subframe length (incl. D preceding samples) */ + const opus_int nb_subfr, /* I Number of subframes stacked in x */ + const opus_int D, /* I Order */ + int arch /* I Run-time architecture */ +) = { + silk_burg_modified_c, /* non-sse */ + silk_burg_modified_c, + silk_burg_modified_c, + MAY_HAVE_SSE4_1( silk_burg_modified ), /* sse4.1 */ + MAY_HAVE_SSE4_1( silk_burg_modified ) /* avx */ +}; + +#endif +#endif diff --git a/lib/opus/silk_headers.mk b/lib/opus/silk_headers.mk new file mode 100644 index 00000000..2588067c --- /dev/null +++ b/lib/opus/silk_headers.mk @@ -0,0 +1,44 @@ +SILK_HEAD = \ +silk/debug.h \ +silk/control.h \ +silk/errors.h \ +silk/API.h \ +silk/typedef.h \ +silk/define.h \ +silk/main.h \ +silk/x86/main_sse.h \ +silk/PLC.h \ +silk/structs.h \ +silk/tables.h \ +silk/tuning_parameters.h \ +silk/Inlines.h \ +silk/MacroCount.h \ +silk/MacroDebug.h \ +silk/macros.h \ +silk/NSQ.h \ +silk/pitch_est_defines.h \ +silk/resampler_private.h \ +silk/resampler_rom.h \ +silk/resampler_structs.h \ +silk/SigProc_FIX.h \ +silk/x86/SigProc_FIX_sse.h \ +silk/arm/biquad_alt_arm.h \ +silk/arm/LPC_inv_pred_gain_arm.h \ +silk/arm/macros_armv4.h \ +silk/arm/macros_armv5e.h \ +silk/arm/macros_arm64.h \ +silk/arm/SigProc_FIX_armv4.h \ +silk/arm/SigProc_FIX_armv5e.h \ +silk/arm/NSQ_del_dec_arm.h \ +silk/arm/NSQ_neon.h \ +silk/fixed/main_FIX.h \ +silk/fixed/structs_FIX.h \ +silk/fixed/arm/warped_autocorrelation_FIX_arm.h \ +silk/fixed/mips/noise_shape_analysis_FIX_mipsr1.h \ +silk/fixed/mips/warped_autocorrelation_FIX_mipsr1.h \ +silk/float/main_FLP.h \ +silk/float/structs_FLP.h \ +silk/float/SigProc_FLP.h \ +silk/mips/macros_mipsr1.h \ +silk/mips/NSQ_del_dec_mipsr1.h \ +silk/mips/sigproc_fix_mipsr1.h diff --git a/lib/opus/silk_sources.mk b/lib/opus/silk_sources.mk new file mode 100644 index 00000000..3df24816 --- /dev/null +++ b/lib/opus/silk_sources.mk @@ -0,0 +1,158 @@ +SILK_SOURCES = \ +silk/CNG.c \ +silk/code_signs.c \ +silk/init_decoder.c \ +silk/decode_core.c \ +silk/decode_frame.c \ +silk/decode_parameters.c \ +silk/decode_indices.c \ +silk/decode_pulses.c \ +silk/decoder_set_fs.c \ +silk/dec_API.c \ +silk/enc_API.c \ +silk/encode_indices.c \ +silk/encode_pulses.c \ +silk/gain_quant.c \ +silk/interpolate.c \ +silk/LP_variable_cutoff.c \ +silk/NLSF_decode.c \ +silk/NSQ.c \ +silk/NSQ_del_dec.c \ +silk/PLC.c \ +silk/shell_coder.c \ +silk/tables_gain.c \ +silk/tables_LTP.c \ +silk/tables_NLSF_CB_NB_MB.c \ +silk/tables_NLSF_CB_WB.c \ +silk/tables_other.c \ +silk/tables_pitch_lag.c \ +silk/tables_pulses_per_block.c \ +silk/VAD.c \ +silk/control_audio_bandwidth.c \ +silk/quant_LTP_gains.c \ +silk/VQ_WMat_EC.c \ +silk/HP_variable_cutoff.c \ +silk/NLSF_encode.c \ +silk/NLSF_VQ.c \ +silk/NLSF_unpack.c \ +silk/NLSF_del_dec_quant.c \ +silk/process_NLSFs.c \ +silk/stereo_LR_to_MS.c \ +silk/stereo_MS_to_LR.c \ +silk/check_control_input.c \ +silk/control_SNR.c \ +silk/init_encoder.c \ +silk/control_codec.c \ +silk/A2NLSF.c \ +silk/ana_filt_bank_1.c \ +silk/biquad_alt.c \ +silk/bwexpander_32.c \ +silk/bwexpander.c \ +silk/debug.c \ +silk/decode_pitch.c \ +silk/inner_prod_aligned.c \ +silk/lin2log.c \ +silk/log2lin.c \ +silk/LPC_analysis_filter.c \ +silk/LPC_inv_pred_gain.c \ +silk/table_LSF_cos.c \ +silk/NLSF2A.c \ +silk/NLSF_stabilize.c \ +silk/NLSF_VQ_weights_laroia.c \ +silk/pitch_est_tables.c \ +silk/resampler.c \ +silk/resampler_down2_3.c \ +silk/resampler_down2.c \ +silk/resampler_private_AR2.c \ +silk/resampler_private_down_FIR.c \ +silk/resampler_private_IIR_FIR.c \ +silk/resampler_private_up2_HQ.c \ +silk/resampler_rom.c \ +silk/sigm_Q15.c \ +silk/sort.c \ +silk/sum_sqr_shift.c \ +silk/stereo_decode_pred.c \ +silk/stereo_encode_pred.c \ +silk/stereo_find_predictor.c \ +silk/stereo_quant_pred.c \ +silk/LPC_fit.c + +SILK_SOURCES_X86_RTCD = \ +silk/x86/x86_silk_map.c + +SILK_SOURCES_SSE4_1 = \ +silk/x86/NSQ_sse4_1.c \ +silk/x86/NSQ_del_dec_sse4_1.c \ +silk/x86/VAD_sse4_1.c \ +silk/x86/VQ_WMat_EC_sse4_1.c + +SILK_SOURCES_ARM_RTCD = \ +silk/arm/arm_silk_map.c + +SILK_SOURCES_ARM_NEON_INTR = \ +silk/arm/biquad_alt_neon_intr.c \ +silk/arm/LPC_inv_pred_gain_neon_intr.c \ +silk/arm/NSQ_del_dec_neon_intr.c \ +silk/arm/NSQ_neon.c + +SILK_SOURCES_FIXED = \ +silk/fixed/LTP_analysis_filter_FIX.c \ +silk/fixed/LTP_scale_ctrl_FIX.c \ +silk/fixed/corrMatrix_FIX.c \ +silk/fixed/encode_frame_FIX.c \ +silk/fixed/find_LPC_FIX.c \ +silk/fixed/find_LTP_FIX.c \ +silk/fixed/find_pitch_lags_FIX.c \ +silk/fixed/find_pred_coefs_FIX.c \ +silk/fixed/noise_shape_analysis_FIX.c \ +silk/fixed/process_gains_FIX.c \ +silk/fixed/regularize_correlations_FIX.c \ +silk/fixed/residual_energy16_FIX.c \ +silk/fixed/residual_energy_FIX.c \ +silk/fixed/warped_autocorrelation_FIX.c \ +silk/fixed/apply_sine_window_FIX.c \ +silk/fixed/autocorr_FIX.c \ +silk/fixed/burg_modified_FIX.c \ +silk/fixed/k2a_FIX.c \ +silk/fixed/k2a_Q16_FIX.c \ +silk/fixed/pitch_analysis_core_FIX.c \ +silk/fixed/vector_ops_FIX.c \ +silk/fixed/schur64_FIX.c \ +silk/fixed/schur_FIX.c + +SILK_SOURCES_FIXED_SSE4_1 = \ +silk/fixed/x86/vector_ops_FIX_sse4_1.c \ +silk/fixed/x86/burg_modified_FIX_sse4_1.c + +SILK_SOURCES_FIXED_ARM_NEON_INTR = \ +silk/fixed/arm/warped_autocorrelation_FIX_neon_intr.c + +SILK_SOURCES_FLOAT = \ +silk/float/apply_sine_window_FLP.c \ +silk/float/corrMatrix_FLP.c \ +silk/float/encode_frame_FLP.c \ +silk/float/find_LPC_FLP.c \ +silk/float/find_LTP_FLP.c \ +silk/float/find_pitch_lags_FLP.c \ +silk/float/find_pred_coefs_FLP.c \ +silk/float/LPC_analysis_filter_FLP.c \ +silk/float/LTP_analysis_filter_FLP.c \ +silk/float/LTP_scale_ctrl_FLP.c \ +silk/float/noise_shape_analysis_FLP.c \ +silk/float/process_gains_FLP.c \ +silk/float/regularize_correlations_FLP.c \ +silk/float/residual_energy_FLP.c \ +silk/float/warped_autocorrelation_FLP.c \ +silk/float/wrappers_FLP.c \ +silk/float/autocorrelation_FLP.c \ +silk/float/burg_modified_FLP.c \ +silk/float/bwexpander_FLP.c \ +silk/float/energy_FLP.c \ +silk/float/inner_product_FLP.c \ +silk/float/k2a_FLP.c \ +silk/float/LPC_inv_pred_gain_FLP.c \ +silk/float/pitch_analysis_core_FLP.c \ +silk/float/scale_copy_vector_FLP.c \ +silk/float/scale_vector_FLP.c \ +silk/float/schur_FLP.c \ +silk/float/sort_FLP.c diff --git a/lib/opus/src/analysis.c b/lib/opus/src/analysis.c new file mode 100644 index 00000000..058328f0 --- /dev/null +++ b/lib/opus/src/analysis.c @@ -0,0 +1,983 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#define ANALYSIS_C + +#ifdef MLP_TRAINING +#include +#endif + +#include "mathops.h" +#include "kiss_fft.h" +#include "celt.h" +#include "modes.h" +#include "arch.h" +#include "quant_bands.h" +#include "analysis.h" +#include "mlp.h" +#include "stack_alloc.h" +#include "float_cast.h" + +#ifndef M_PI +#define M_PI 3.141592653 +#endif + +#ifndef DISABLE_FLOAT_API + +#define TRANSITION_PENALTY 10 + +static const float dct_table[128] = { + 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, + 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, + 0.351851f, 0.338330f, 0.311806f, 0.273300f, 0.224292f, 0.166664f, 0.102631f, 0.034654f, + -0.034654f,-0.102631f,-0.166664f,-0.224292f,-0.273300f,-0.311806f,-0.338330f,-0.351851f, + 0.346760f, 0.293969f, 0.196424f, 0.068975f,-0.068975f,-0.196424f,-0.293969f,-0.346760f, + -0.346760f,-0.293969f,-0.196424f,-0.068975f, 0.068975f, 0.196424f, 0.293969f, 0.346760f, + 0.338330f, 0.224292f, 0.034654f,-0.166664f,-0.311806f,-0.351851f,-0.273300f,-0.102631f, + 0.102631f, 0.273300f, 0.351851f, 0.311806f, 0.166664f,-0.034654f,-0.224292f,-0.338330f, + 0.326641f, 0.135299f,-0.135299f,-0.326641f,-0.326641f,-0.135299f, 0.135299f, 0.326641f, + 0.326641f, 0.135299f,-0.135299f,-0.326641f,-0.326641f,-0.135299f, 0.135299f, 0.326641f, + 0.311806f, 0.034654f,-0.273300f,-0.338330f,-0.102631f, 0.224292f, 0.351851f, 0.166664f, + -0.166664f,-0.351851f,-0.224292f, 0.102631f, 0.338330f, 0.273300f,-0.034654f,-0.311806f, + 0.293969f,-0.068975f,-0.346760f,-0.196424f, 0.196424f, 0.346760f, 0.068975f,-0.293969f, + -0.293969f, 0.068975f, 0.346760f, 0.196424f,-0.196424f,-0.346760f,-0.068975f, 0.293969f, + 0.273300f,-0.166664f,-0.338330f, 0.034654f, 0.351851f, 0.102631f,-0.311806f,-0.224292f, + 0.224292f, 0.311806f,-0.102631f,-0.351851f,-0.034654f, 0.338330f, 0.166664f,-0.273300f, +}; + +static const float analysis_window[240] = { + 0.000043f, 0.000171f, 0.000385f, 0.000685f, 0.001071f, 0.001541f, 0.002098f, 0.002739f, + 0.003466f, 0.004278f, 0.005174f, 0.006156f, 0.007222f, 0.008373f, 0.009607f, 0.010926f, + 0.012329f, 0.013815f, 0.015385f, 0.017037f, 0.018772f, 0.020590f, 0.022490f, 0.024472f, + 0.026535f, 0.028679f, 0.030904f, 0.033210f, 0.035595f, 0.038060f, 0.040604f, 0.043227f, + 0.045928f, 0.048707f, 0.051564f, 0.054497f, 0.057506f, 0.060591f, 0.063752f, 0.066987f, + 0.070297f, 0.073680f, 0.077136f, 0.080665f, 0.084265f, 0.087937f, 0.091679f, 0.095492f, + 0.099373f, 0.103323f, 0.107342f, 0.111427f, 0.115579f, 0.119797f, 0.124080f, 0.128428f, + 0.132839f, 0.137313f, 0.141849f, 0.146447f, 0.151105f, 0.155823f, 0.160600f, 0.165435f, + 0.170327f, 0.175276f, 0.180280f, 0.185340f, 0.190453f, 0.195619f, 0.200838f, 0.206107f, + 0.211427f, 0.216797f, 0.222215f, 0.227680f, 0.233193f, 0.238751f, 0.244353f, 0.250000f, + 0.255689f, 0.261421f, 0.267193f, 0.273005f, 0.278856f, 0.284744f, 0.290670f, 0.296632f, + 0.302628f, 0.308658f, 0.314721f, 0.320816f, 0.326941f, 0.333097f, 0.339280f, 0.345492f, + 0.351729f, 0.357992f, 0.364280f, 0.370590f, 0.376923f, 0.383277f, 0.389651f, 0.396044f, + 0.402455f, 0.408882f, 0.415325f, 0.421783f, 0.428254f, 0.434737f, 0.441231f, 0.447736f, + 0.454249f, 0.460770f, 0.467298f, 0.473832f, 0.480370f, 0.486912f, 0.493455f, 0.500000f, + 0.506545f, 0.513088f, 0.519630f, 0.526168f, 0.532702f, 0.539230f, 0.545751f, 0.552264f, + 0.558769f, 0.565263f, 0.571746f, 0.578217f, 0.584675f, 0.591118f, 0.597545f, 0.603956f, + 0.610349f, 0.616723f, 0.623077f, 0.629410f, 0.635720f, 0.642008f, 0.648271f, 0.654508f, + 0.660720f, 0.666903f, 0.673059f, 0.679184f, 0.685279f, 0.691342f, 0.697372f, 0.703368f, + 0.709330f, 0.715256f, 0.721144f, 0.726995f, 0.732807f, 0.738579f, 0.744311f, 0.750000f, + 0.755647f, 0.761249f, 0.766807f, 0.772320f, 0.777785f, 0.783203f, 0.788573f, 0.793893f, + 0.799162f, 0.804381f, 0.809547f, 0.814660f, 0.819720f, 0.824724f, 0.829673f, 0.834565f, + 0.839400f, 0.844177f, 0.848895f, 0.853553f, 0.858151f, 0.862687f, 0.867161f, 0.871572f, + 0.875920f, 0.880203f, 0.884421f, 0.888573f, 0.892658f, 0.896677f, 0.900627f, 0.904508f, + 0.908321f, 0.912063f, 0.915735f, 0.919335f, 0.922864f, 0.926320f, 0.929703f, 0.933013f, + 0.936248f, 0.939409f, 0.942494f, 0.945503f, 0.948436f, 0.951293f, 0.954072f, 0.956773f, + 0.959396f, 0.961940f, 0.964405f, 0.966790f, 0.969096f, 0.971321f, 0.973465f, 0.975528f, + 0.977510f, 0.979410f, 0.981228f, 0.982963f, 0.984615f, 0.986185f, 0.987671f, 0.989074f, + 0.990393f, 0.991627f, 0.992778f, 0.993844f, 0.994826f, 0.995722f, 0.996534f, 0.997261f, + 0.997902f, 0.998459f, 0.998929f, 0.999315f, 0.999615f, 0.999829f, 0.999957f, 1.000000f, +}; + +static const int tbands[NB_TBANDS+1] = { + 4, 8, 12, 16, 20, 24, 28, 32, 40, 48, 56, 64, 80, 96, 112, 136, 160, 192, 240 +}; + +#define NB_TONAL_SKIP_BANDS 9 + +static opus_val32 silk_resampler_down2_hp( + opus_val32 *S, /* I/O State vector [ 2 ] */ + opus_val32 *out, /* O Output signal [ floor(len/2) ] */ + const opus_val32 *in, /* I Input signal [ len ] */ + int inLen /* I Number of input samples */ +) +{ + int k, len2 = inLen/2; + opus_val32 in32, out32, out32_hp, Y, X; + opus_val64 hp_ener = 0; + /* Internal variables and state are in Q10 format */ + for( k = 0; k < len2; k++ ) { + /* Convert to Q10 */ + in32 = in[ 2 * k ]; + + /* All-pass section for even input sample */ + Y = SUB32( in32, S[ 0 ] ); + X = MULT16_32_Q15(QCONST16(0.6074371f, 15), Y); + out32 = ADD32( S[ 0 ], X ); + S[ 0 ] = ADD32( in32, X ); + out32_hp = out32; + /* Convert to Q10 */ + in32 = in[ 2 * k + 1 ]; + + /* All-pass section for odd input sample, and add to output of previous section */ + Y = SUB32( in32, S[ 1 ] ); + X = MULT16_32_Q15(QCONST16(0.15063f, 15), Y); + out32 = ADD32( out32, S[ 1 ] ); + out32 = ADD32( out32, X ); + S[ 1 ] = ADD32( in32, X ); + + Y = SUB32( -in32, S[ 2 ] ); + X = MULT16_32_Q15(QCONST16(0.15063f, 15), Y); + out32_hp = ADD32( out32_hp, S[ 2 ] ); + out32_hp = ADD32( out32_hp, X ); + S[ 2 ] = ADD32( -in32, X ); + + hp_ener += out32_hp*(opus_val64)out32_hp; + /* Add, convert back to int16 and store to output */ + out[ k ] = HALF32(out32); + } +#ifdef FIXED_POINT + /* len2 can be up to 480, so we shift by 8 more to make it fit. */ + hp_ener = hp_ener >> (2*SIG_SHIFT + 8); +#endif + return (opus_val32)hp_ener; +} + +static opus_val32 downmix_and_resample(downmix_func downmix, const void *_x, opus_val32 *y, opus_val32 S[3], int subframe, int offset, int c1, int c2, int C, int Fs) +{ + VARDECL(opus_val32, tmp); + opus_val32 scale; + int j; + opus_val32 ret = 0; + SAVE_STACK; + + if (subframe==0) return 0; + if (Fs == 48000) + { + subframe *= 2; + offset *= 2; + } else if (Fs == 16000) { + subframe = subframe*2/3; + offset = offset*2/3; + } + ALLOC(tmp, subframe, opus_val32); + + downmix(_x, tmp, subframe, offset, c1, c2, C); +#ifdef FIXED_POINT + scale = (1<-1) + scale /= 2; + for (j=0;jarch = opus_select_arch(); + tonal->Fs = Fs; + /* Clear remaining fields. */ + tonality_analysis_reset(tonal); +} + +void tonality_analysis_reset(TonalityAnalysisState *tonal) +{ + /* Clear non-reusable fields. */ + char *start = (char*)&tonal->TONALITY_ANALYSIS_RESET_START; + OPUS_CLEAR(start, sizeof(TonalityAnalysisState) - (start - (char*)tonal)); +} + +void tonality_get_info(TonalityAnalysisState *tonal, AnalysisInfo *info_out, int len) +{ + int pos; + int curr_lookahead; + float tonality_max; + float tonality_avg; + int tonality_count; + int i; + int pos0; + float prob_avg; + float prob_count; + float prob_min, prob_max; + float vad_prob; + int mpos, vpos; + int bandwidth_span; + + pos = tonal->read_pos; + curr_lookahead = tonal->write_pos-tonal->read_pos; + if (curr_lookahead<0) + curr_lookahead += DETECT_SIZE; + + tonal->read_subframe += len/(tonal->Fs/400); + while (tonal->read_subframe>=8) + { + tonal->read_subframe -= 8; + tonal->read_pos++; + } + if (tonal->read_pos>=DETECT_SIZE) + tonal->read_pos-=DETECT_SIZE; + + /* On long frames, look at the second analysis window rather than the first. */ + if (len > tonal->Fs/50 && pos != tonal->write_pos) + { + pos++; + if (pos==DETECT_SIZE) + pos=0; + } + if (pos == tonal->write_pos) + pos--; + if (pos<0) + pos = DETECT_SIZE-1; + pos0 = pos; + OPUS_COPY(info_out, &tonal->info[pos], 1); + if (!info_out->valid) + return; + tonality_max = tonality_avg = info_out->tonality; + tonality_count = 1; + /* Look at the neighbouring frames and pick largest bandwidth found (to be safe). */ + bandwidth_span = 6; + /* If possible, look ahead for a tone to compensate for the delay in the tone detector. */ + for (i=0;i<3;i++) + { + pos++; + if (pos==DETECT_SIZE) + pos = 0; + if (pos == tonal->write_pos) + break; + tonality_max = MAX32(tonality_max, tonal->info[pos].tonality); + tonality_avg += tonal->info[pos].tonality; + tonality_count++; + info_out->bandwidth = IMAX(info_out->bandwidth, tonal->info[pos].bandwidth); + bandwidth_span--; + } + pos = pos0; + /* Look back in time to see if any has a wider bandwidth than the current frame. */ + for (i=0;iwrite_pos) + break; + info_out->bandwidth = IMAX(info_out->bandwidth, tonal->info[pos].bandwidth); + } + info_out->tonality = MAX32(tonality_avg/tonality_count, tonality_max-.2f); + + mpos = vpos = pos0; + /* If we have enough look-ahead, compensate for the ~5-frame delay in the music prob and + ~1 frame delay in the VAD prob. */ + if (curr_lookahead > 15) + { + mpos += 5; + if (mpos>=DETECT_SIZE) + mpos -= DETECT_SIZE; + vpos += 1; + if (vpos>=DETECT_SIZE) + vpos -= DETECT_SIZE; + } + + /* The following calculations attempt to minimize a "badness function" + for the transition. When switching from speech to music, the badness + of switching at frame k is + b_k = S*v_k + \sum_{i=0}^{k-1} v_i*(p_i - T) + where + v_i is the activity probability (VAD) at frame i, + p_i is the music probability at frame i + T is the probability threshold for switching + S is the penalty for switching during active audio rather than silence + the current frame has index i=0 + + Rather than apply badness to directly decide when to switch, what we compute + instead is the threshold for which the optimal switching point is now. When + considering whether to switch now (frame 0) or at frame k, we have: + S*v_0 = S*v_k + \sum_{i=0}^{k-1} v_i*(p_i - T) + which gives us: + T = ( \sum_{i=0}^{k-1} v_i*p_i + S*(v_k-v_0) ) / ( \sum_{i=0}^{k-1} v_i ) + We take the min threshold across all positive values of k (up to the maximum + amount of lookahead we have) to give us the threshold for which the current + frame is the optimal switch point. + + The last step is that we need to consider whether we want to switch at all. + For that we use the average of the music probability over the entire window. + If the threshold is higher than that average we're not going to + switch, so we compute a min with the average as well. The result of all these + min operations is music_prob_min, which gives the threshold for switching to music + if we're currently encoding for speech. + + We do the exact opposite to compute music_prob_max which is used for switching + from music to speech. + */ + prob_min = 1.f; + prob_max = 0.f; + vad_prob = tonal->info[vpos].activity_probability; + prob_count = MAX16(.1f, vad_prob); + prob_avg = MAX16(.1f, vad_prob)*tonal->info[mpos].music_prob; + while (1) + { + float pos_vad; + mpos++; + if (mpos==DETECT_SIZE) + mpos = 0; + if (mpos == tonal->write_pos) + break; + vpos++; + if (vpos==DETECT_SIZE) + vpos = 0; + if (vpos == tonal->write_pos) + break; + pos_vad = tonal->info[vpos].activity_probability; + prob_min = MIN16((prob_avg - TRANSITION_PENALTY*(vad_prob - pos_vad))/prob_count, prob_min); + prob_max = MAX16((prob_avg + TRANSITION_PENALTY*(vad_prob - pos_vad))/prob_count, prob_max); + prob_count += MAX16(.1f, pos_vad); + prob_avg += MAX16(.1f, pos_vad)*tonal->info[mpos].music_prob; + } + info_out->music_prob = prob_avg/prob_count; + prob_min = MIN16(prob_avg/prob_count, prob_min); + prob_max = MAX16(prob_avg/prob_count, prob_max); + prob_min = MAX16(prob_min, 0.f); + prob_max = MIN16(prob_max, 1.f); + + /* If we don't have enough look-ahead, do our best to make a decent decision. */ + if (curr_lookahead < 10) + { + float pmin, pmax; + pmin = prob_min; + pmax = prob_max; + pos = pos0; + /* Look for min/max in the past. */ + for (i=0;icount-1, 15);i++) + { + pos--; + if (pos < 0) + pos = DETECT_SIZE-1; + pmin = MIN16(pmin, tonal->info[pos].music_prob); + pmax = MAX16(pmax, tonal->info[pos].music_prob); + } + /* Bias against switching on active audio. */ + pmin = MAX16(0.f, pmin - .1f*vad_prob); + pmax = MIN16(1.f, pmax + .1f*vad_prob); + prob_min += (1.f-.1f*curr_lookahead)*(pmin - prob_min); + prob_max += (1.f-.1f*curr_lookahead)*(pmax - prob_max); + } + info_out->music_prob_min = prob_min; + info_out->music_prob_max = prob_max; + + /* printf("%f %f %f %f %f\n", prob_min, prob_max, prob_avg/prob_count, vad_prob, info_out->music_prob); */ +} + +static const float std_feature_bias[9] = { + 5.684947f, 3.475288f, 1.770634f, 1.599784f, 3.773215f, + 2.163313f, 1.260756f, 1.116868f, 1.918795f +}; + +#define LEAKAGE_OFFSET 2.5f +#define LEAKAGE_SLOPE 2.f + +#ifdef FIXED_POINT +/* For fixed-point, the input is +/-2^15 shifted up by SIG_SHIFT, so we need to + compensate for that in the energy. */ +#define SCALE_COMPENS (1.f/((opus_int32)1<<(15+SIG_SHIFT))) +#define SCALE_ENER(e) ((SCALE_COMPENS*SCALE_COMPENS)*(e)) +#else +#define SCALE_ENER(e) (e) +#endif + +#ifdef FIXED_POINT +static int is_digital_silence32(const opus_val32* pcm, int frame_size, int channels, int lsb_depth) +{ + int silence = 0; + opus_val32 sample_max = 0; +#ifdef MLP_TRAINING + return 0; +#endif + sample_max = celt_maxabs32(pcm, frame_size*channels); + + silence = (sample_max == 0); + (void)lsb_depth; + return silence; +} +#else +#define is_digital_silence32(pcm, frame_size, channels, lsb_depth) is_digital_silence(pcm, frame_size, channels, lsb_depth) +#endif + +static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt_mode, const void *x, int len, int offset, int c1, int c2, int C, int lsb_depth, downmix_func downmix) +{ + int i, b; + const kiss_fft_state *kfft; + VARDECL(kiss_fft_cpx, in); + VARDECL(kiss_fft_cpx, out); + int N = 480, N2=240; + float * OPUS_RESTRICT A = tonal->angle; + float * OPUS_RESTRICT dA = tonal->d_angle; + float * OPUS_RESTRICT d2A = tonal->d2_angle; + VARDECL(float, tonality); + VARDECL(float, noisiness); + float band_tonality[NB_TBANDS]; + float logE[NB_TBANDS]; + float BFCC[8]; + float features[25]; + float frame_tonality; + float max_frame_tonality; + /*float tw_sum=0;*/ + float frame_noisiness; + const float pi4 = (float)(M_PI*M_PI*M_PI*M_PI); + float slope=0; + float frame_stationarity; + float relativeE; + float frame_probs[2]; + float alpha, alphaE, alphaE2; + float frame_loudness; + float bandwidth_mask; + int is_masked[NB_TBANDS+1]; + int bandwidth=0; + float maxE = 0; + float noise_floor; + int remaining; + AnalysisInfo *info; + float hp_ener; + float tonality2[240]; + float midE[8]; + float spec_variability=0; + float band_log2[NB_TBANDS+1]; + float leakage_from[NB_TBANDS+1]; + float leakage_to[NB_TBANDS+1]; + float layer_out[MAX_NEURONS]; + float below_max_pitch; + float above_max_pitch; + int is_silence; + SAVE_STACK; + + if (!tonal->initialized) + { + tonal->mem_fill = 240; + tonal->initialized = 1; + } + alpha = 1.f/IMIN(10, 1+tonal->count); + alphaE = 1.f/IMIN(25, 1+tonal->count); + /* Noise floor related decay for bandwidth detection: -2.2 dB/second */ + alphaE2 = 1.f/IMIN(100, 1+tonal->count); + if (tonal->count <= 1) alphaE2 = 1; + + if (tonal->Fs == 48000) + { + /* len and offset are now at 24 kHz. */ + len/= 2; + offset /= 2; + } else if (tonal->Fs == 16000) { + len = 3*len/2; + offset = 3*offset/2; + } + + kfft = celt_mode->mdct.kfft[0]; + tonal->hp_ener_accum += (float)downmix_and_resample(downmix, x, + &tonal->inmem[tonal->mem_fill], tonal->downmix_state, + IMIN(len, ANALYSIS_BUF_SIZE-tonal->mem_fill), offset, c1, c2, C, tonal->Fs); + if (tonal->mem_fill+len < ANALYSIS_BUF_SIZE) + { + tonal->mem_fill += len; + /* Don't have enough to update the analysis */ + RESTORE_STACK; + return; + } + hp_ener = tonal->hp_ener_accum; + info = &tonal->info[tonal->write_pos++]; + if (tonal->write_pos>=DETECT_SIZE) + tonal->write_pos-=DETECT_SIZE; + + is_silence = is_digital_silence32(tonal->inmem, ANALYSIS_BUF_SIZE, 1, lsb_depth); + + ALLOC(in, 480, kiss_fft_cpx); + ALLOC(out, 480, kiss_fft_cpx); + ALLOC(tonality, 240, float); + ALLOC(noisiness, 240, float); + for (i=0;iinmem[i]); + in[i].i = (kiss_fft_scalar)(w*tonal->inmem[N2+i]); + in[N-i-1].r = (kiss_fft_scalar)(w*tonal->inmem[N-i-1]); + in[N-i-1].i = (kiss_fft_scalar)(w*tonal->inmem[N+N2-i-1]); + } + OPUS_MOVE(tonal->inmem, tonal->inmem+ANALYSIS_BUF_SIZE-240, 240); + remaining = len - (ANALYSIS_BUF_SIZE-tonal->mem_fill); + tonal->hp_ener_accum = (float)downmix_and_resample(downmix, x, + &tonal->inmem[240], tonal->downmix_state, remaining, + offset+ANALYSIS_BUF_SIZE-tonal->mem_fill, c1, c2, C, tonal->Fs); + tonal->mem_fill = 240 + remaining; + if (is_silence) + { + /* On silence, copy the previous analysis. */ + int prev_pos = tonal->write_pos-2; + if (prev_pos < 0) + prev_pos += DETECT_SIZE; + OPUS_COPY(info, &tonal->info[prev_pos], 1); + RESTORE_STACK; + return; + } + opus_fft(kfft, in, out, tonal->arch); +#ifndef FIXED_POINT + /* If there's any NaN on the input, the entire output will be NaN, so we only need to check one value. */ + if (celt_isnan(out[0].r)) + { + info->valid = 0; + RESTORE_STACK; + return; + } +#endif + + for (i=1;iactivity = 0; + frame_noisiness = 0; + frame_stationarity = 0; + if (!tonal->count) + { + for (b=0;blowE[b] = 1e10; + tonal->highE[b] = -1e10; + } + } + relativeE = 0; + frame_loudness = 0; + /* The energy of the very first band is special because of DC. */ + { + float E = 0; + float X1r, X2r; + X1r = 2*(float)out[0].r; + X2r = 2*(float)out[0].i; + E = X1r*X1r + X2r*X2r; + for (i=1;i<4;i++) + { + float binE = out[i].r*(float)out[i].r + out[N-i].r*(float)out[N-i].r + + out[i].i*(float)out[i].i + out[N-i].i*(float)out[N-i].i; + E += binE; + } + E = SCALE_ENER(E); + band_log2[0] = .5f*1.442695f*(float)log(E+1e-10f); + } + for (b=0;bvalid = 0; + RESTORE_STACK; + return; + } +#endif + + tonal->E[tonal->E_count][b] = E; + frame_noisiness += nE/(1e-15f+E); + + frame_loudness += (float)sqrt(E+1e-10f); + logE[b] = (float)log(E+1e-10f); + band_log2[b+1] = .5f*1.442695f*(float)log(E+1e-10f); + tonal->logE[tonal->E_count][b] = logE[b]; + if (tonal->count==0) + tonal->highE[b] = tonal->lowE[b] = logE[b]; + if (tonal->highE[b] > tonal->lowE[b] + 7.5) + { + if (tonal->highE[b] - logE[b] > logE[b] - tonal->lowE[b]) + tonal->highE[b] -= .01f; + else + tonal->lowE[b] += .01f; + } + if (logE[b] > tonal->highE[b]) + { + tonal->highE[b] = logE[b]; + tonal->lowE[b] = MAX32(tonal->highE[b]-15, tonal->lowE[b]); + } else if (logE[b] < tonal->lowE[b]) + { + tonal->lowE[b] = logE[b]; + tonal->highE[b] = MIN32(tonal->lowE[b]+15, tonal->highE[b]); + } + relativeE += (logE[b]-tonal->lowE[b])/(1e-5f + (tonal->highE[b]-tonal->lowE[b])); + + L1=L2=0; + for (i=0;iE[i][b]); + L2 += tonal->E[i][b]; + } + + stationarity = MIN16(0.99f,L1/(float)sqrt(1e-15+NB_FRAMES*L2)); + stationarity *= stationarity; + stationarity *= stationarity; + frame_stationarity += stationarity; + /*band_tonality[b] = tE/(1e-15+E)*/; + band_tonality[b] = MAX16(tE/(1e-15f+E), stationarity*tonal->prev_band_tonality[b]); +#if 0 + if (b>=NB_TONAL_SKIP_BANDS) + { + frame_tonality += tweight[b]*band_tonality[b]; + tw_sum += tweight[b]; + } +#else + frame_tonality += band_tonality[b]; + if (b>=NB_TBANDS-NB_TONAL_SKIP_BANDS) + frame_tonality -= band_tonality[b-NB_TBANDS+NB_TONAL_SKIP_BANDS]; +#endif + max_frame_tonality = MAX16(max_frame_tonality, (1.f+.03f*(b-NB_TBANDS))*frame_tonality); + slope += band_tonality[b]*(b-8); + /*printf("%f %f ", band_tonality[b], stationarity);*/ + tonal->prev_band_tonality[b] = band_tonality[b]; + } + + leakage_from[0] = band_log2[0]; + leakage_to[0] = band_log2[0] - LEAKAGE_OFFSET; + for (b=1;b=0;b--) + { + float leak_slope = LEAKAGE_SLOPE*(tbands[b+1]-tbands[b])/4; + leakage_from[b] = MIN16(leakage_from[b+1]+leak_slope, leakage_from[b]); + leakage_to[b] = MAX16(leakage_to[b+1]-leak_slope, leakage_to[b]); + } + celt_assert(NB_TBANDS+1 <= LEAK_BANDS); + for (b=0;bleak_boost[b] = IMIN(255, (int)floor(.5 + 64.f*boost)); + } + for (;bleak_boost[b] = 0; + + for (i=0;ilogE[i][k] - tonal->logE[j][k]; + dist += tmp*tmp; + } + if (j!=i) + mindist = MIN32(mindist, dist); + } + spec_variability += mindist; + } + spec_variability = (float)sqrt(spec_variability/NB_FRAMES/NB_TBANDS); + bandwidth_mask = 0; + bandwidth = 0; + maxE = 0; + noise_floor = 5.7e-4f/(1<<(IMAX(0,lsb_depth-8))); + noise_floor *= noise_floor; + below_max_pitch=0; + above_max_pitch=0; + for (b=0;bmeanE[b] = MAX32((1-alphaE2)*tonal->meanE[b], E); + Em = MAX32(E, tonal->meanE[b]); + /* Consider the band "active" only if all these conditions are met: + 1) less than 90 dB below the peak band (maximal masking possible considering + both the ATH and the loudness-dependent slope of the spreading function) + 2) above the PCM quantization noise floor + We use b+1 because the first CELT band isn't included in tbands[] + */ + if (E*1e9f > maxE && (Em > 3*noise_floor*(band_end-band_start) || E > noise_floor*(band_end-band_start))) + bandwidth = b+1; + /* Check if the band is masked (see below). */ + is_masked[b] = E < (tonal->prev_bandwidth >= b+1 ? .01f : .05f)*bandwidth_mask; + /* Use a simple follower with 13 dB/Bark slope for spreading function. */ + bandwidth_mask = MAX32(.05f*bandwidth_mask, E); + } + /* Special case for the last two bands, for which we don't have spectrum but only + the energy above 12 kHz. The difficulty here is that the high-pass we use + leaks some LF energy, so we need to increase the threshold without accidentally cutting + off the band. */ + if (tonal->Fs == 48000) { + float noise_ratio; + float Em; + float E = hp_ener*(1.f/(60*60)); + noise_ratio = tonal->prev_bandwidth==20 ? 10.f : 30.f; + +#ifdef FIXED_POINT + /* silk_resampler_down2_hp() shifted right by an extra 8 bits. */ + E *= 256.f*(1.f/Q15ONE)*(1.f/Q15ONE); +#endif + above_max_pitch += E; + tonal->meanE[b] = MAX32((1-alphaE2)*tonal->meanE[b], E); + Em = MAX32(E, tonal->meanE[b]); + if (Em > 3*noise_ratio*noise_floor*160 || E > noise_ratio*noise_floor*160) + bandwidth = 20; + /* Check if the band is masked (see below). */ + is_masked[b] = E < (tonal->prev_bandwidth == 20 ? .01f : .05f)*bandwidth_mask; + } + if (above_max_pitch > below_max_pitch) + info->max_pitch_ratio = below_max_pitch/above_max_pitch; + else + info->max_pitch_ratio = 1; + /* In some cases, resampling aliasing can create a small amount of energy in the first band + being cut. So if the last band is masked, we don't include it. */ + if (bandwidth == 20 && is_masked[NB_TBANDS]) + bandwidth-=2; + else if (bandwidth > 0 && bandwidth <= NB_TBANDS && is_masked[bandwidth-1]) + bandwidth--; + if (tonal->count<=2) + bandwidth = 20; + frame_loudness = 20*(float)log10(frame_loudness); + tonal->Etracker = MAX32(tonal->Etracker-.003f, frame_loudness); + tonal->lowECount *= (1-alphaE); + if (frame_loudness < tonal->Etracker-30) + tonal->lowECount += alphaE; + + for (i=0;i<8;i++) + { + float sum=0; + for (b=0;b<16;b++) + sum += dct_table[i*16+b]*logE[b]; + BFCC[i] = sum; + } + for (i=0;i<8;i++) + { + float sum=0; + for (b=0;b<16;b++) + sum += dct_table[i*16+b]*.5f*(tonal->highE[b]+tonal->lowE[b]); + midE[i] = sum; + } + + frame_stationarity /= NB_TBANDS; + relativeE /= NB_TBANDS; + if (tonal->count<10) + relativeE = .5f; + frame_noisiness /= NB_TBANDS; +#if 1 + info->activity = frame_noisiness + (1-frame_noisiness)*relativeE; +#else + info->activity = .5*(1+frame_noisiness-frame_stationarity); +#endif + frame_tonality = (max_frame_tonality/(NB_TBANDS-NB_TONAL_SKIP_BANDS)); + frame_tonality = MAX16(frame_tonality, tonal->prev_tonality*.8f); + tonal->prev_tonality = frame_tonality; + + slope /= 8*8; + info->tonality_slope = slope; + + tonal->E_count = (tonal->E_count+1)%NB_FRAMES; + tonal->count = IMIN(tonal->count+1, ANALYSIS_COUNT_MAX); + info->tonality = frame_tonality; + + for (i=0;i<4;i++) + features[i] = -0.12299f*(BFCC[i]+tonal->mem[i+24]) + 0.49195f*(tonal->mem[i]+tonal->mem[i+16]) + 0.69693f*tonal->mem[i+8] - 1.4349f*tonal->cmean[i]; + + for (i=0;i<4;i++) + tonal->cmean[i] = (1-alpha)*tonal->cmean[i] + alpha*BFCC[i]; + + for (i=0;i<4;i++) + features[4+i] = 0.63246f*(BFCC[i]-tonal->mem[i+24]) + 0.31623f*(tonal->mem[i]-tonal->mem[i+16]); + for (i=0;i<3;i++) + features[8+i] = 0.53452f*(BFCC[i]+tonal->mem[i+24]) - 0.26726f*(tonal->mem[i]+tonal->mem[i+16]) -0.53452f*tonal->mem[i+8]; + + if (tonal->count > 5) + { + for (i=0;i<9;i++) + tonal->std[i] = (1-alpha)*tonal->std[i] + alpha*features[i]*features[i]; + } + for (i=0;i<4;i++) + features[i] = BFCC[i]-midE[i]; + + for (i=0;i<8;i++) + { + tonal->mem[i+24] = tonal->mem[i+16]; + tonal->mem[i+16] = tonal->mem[i+8]; + tonal->mem[i+8] = tonal->mem[i]; + tonal->mem[i] = BFCC[i]; + } + for (i=0;i<9;i++) + features[11+i] = (float)sqrt(tonal->std[i]) - std_feature_bias[i]; + features[18] = spec_variability - 0.78f; + features[20] = info->tonality - 0.154723f; + features[21] = info->activity - 0.724643f; + features[22] = frame_stationarity - 0.743717f; + features[23] = info->tonality_slope + 0.069216f; + features[24] = tonal->lowECount - 0.067930f; + + compute_dense(&layer0, layer_out, features); + compute_gru(&layer1, tonal->rnn_state, layer_out); + compute_dense(&layer2, frame_probs, tonal->rnn_state); + + /* Probability of speech or music vs noise */ + info->activity_probability = frame_probs[1]; + info->music_prob = frame_probs[0]; + + /*printf("%f %f %f\n", frame_probs[0], frame_probs[1], info->music_prob);*/ +#ifdef MLP_TRAINING + for (i=0;i<25;i++) + printf("%f ", features[i]); + printf("\n"); +#endif + + info->bandwidth = bandwidth; + tonal->prev_bandwidth = bandwidth; + /*printf("%d %d\n", info->bandwidth, info->opus_bandwidth);*/ + info->noisiness = frame_noisiness; + info->valid = 1; + RESTORE_STACK; +} + +void run_analysis(TonalityAnalysisState *analysis, const CELTMode *celt_mode, const void *analysis_pcm, + int analysis_frame_size, int frame_size, int c1, int c2, int C, opus_int32 Fs, + int lsb_depth, downmix_func downmix, AnalysisInfo *analysis_info) +{ + int offset; + int pcm_len; + + analysis_frame_size -= analysis_frame_size&1; + if (analysis_pcm != NULL) + { + /* Avoid overflow/wrap-around of the analysis buffer */ + analysis_frame_size = IMIN((DETECT_SIZE-5)*Fs/50, analysis_frame_size); + + pcm_len = analysis_frame_size - analysis->analysis_offset; + offset = analysis->analysis_offset; + while (pcm_len>0) { + tonality_analysis(analysis, celt_mode, analysis_pcm, IMIN(Fs/50, pcm_len), offset, c1, c2, C, lsb_depth, downmix); + offset += Fs/50; + pcm_len -= Fs/50; + } + analysis->analysis_offset = analysis_frame_size; + + analysis->analysis_offset -= frame_size; + } + + tonality_get_info(analysis, analysis_info, frame_size); +} + +#endif /* DISABLE_FLOAT_API */ diff --git a/lib/opus/src/analysis.h b/lib/opus/src/analysis.h new file mode 100644 index 00000000..0b66555f --- /dev/null +++ b/lib/opus/src/analysis.h @@ -0,0 +1,103 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifndef ANALYSIS_H +#define ANALYSIS_H + +#include "celt.h" +#include "opus_private.h" +#include "mlp.h" + +#define NB_FRAMES 8 +#define NB_TBANDS 18 +#define ANALYSIS_BUF_SIZE 720 /* 30 ms at 24 kHz */ + +/* At that point we can stop counting frames because it no longer matters. */ +#define ANALYSIS_COUNT_MAX 10000 + +#define DETECT_SIZE 100 + +/* Uncomment this to print the MLP features on stdout. */ +/*#define MLP_TRAINING*/ + +typedef struct { + int arch; + int application; + opus_int32 Fs; +#define TONALITY_ANALYSIS_RESET_START angle + float angle[240]; + float d_angle[240]; + float d2_angle[240]; + opus_val32 inmem[ANALYSIS_BUF_SIZE]; + int mem_fill; /* number of usable samples in the buffer */ + float prev_band_tonality[NB_TBANDS]; + float prev_tonality; + int prev_bandwidth; + float E[NB_FRAMES][NB_TBANDS]; + float logE[NB_FRAMES][NB_TBANDS]; + float lowE[NB_TBANDS]; + float highE[NB_TBANDS]; + float meanE[NB_TBANDS+1]; + float mem[32]; + float cmean[8]; + float std[9]; + float Etracker; + float lowECount; + int E_count; + int count; + int analysis_offset; + int write_pos; + int read_pos; + int read_subframe; + float hp_ener_accum; + int initialized; + float rnn_state[MAX_NEURONS]; + opus_val32 downmix_state[3]; + AnalysisInfo info[DETECT_SIZE]; +} TonalityAnalysisState; + +/** Initialize a TonalityAnalysisState struct. + * + * This performs some possibly slow initialization steps which should + * not be repeated every analysis step. No allocated memory is retained + * by the state struct, so no cleanup call is required. + */ +void tonality_analysis_init(TonalityAnalysisState *analysis, opus_int32 Fs); + +/** Reset a TonalityAnalysisState stuct. + * + * Call this when there's a discontinuity in the data. + */ +void tonality_analysis_reset(TonalityAnalysisState *analysis); + +void tonality_get_info(TonalityAnalysisState *tonal, AnalysisInfo *info_out, int len); + +void run_analysis(TonalityAnalysisState *analysis, const CELTMode *celt_mode, const void *analysis_pcm, + int analysis_frame_size, int frame_size, int c1, int c2, int C, opus_int32 Fs, + int lsb_depth, downmix_func downmix, AnalysisInfo *analysis_info); + +#endif diff --git a/lib/opus/src/mapping_matrix.c b/lib/opus/src/mapping_matrix.c new file mode 100644 index 00000000..31298af0 --- /dev/null +++ b/lib/opus/src/mapping_matrix.c @@ -0,0 +1,378 @@ +/* Copyright (c) 2017 Google Inc. + Written by Andrew Allen */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "arch.h" +#include "float_cast.h" +#include "opus_private.h" +#include "opus_defines.h" +#include "mapping_matrix.h" + +#define MATRIX_INDEX(nb_rows, row, col) (nb_rows * col + row) + +opus_int32 mapping_matrix_get_size(int rows, int cols) +{ + opus_int32 size; + + /* Mapping Matrix must only support up to 255 channels in or out. + * Additionally, the total cell count must be <= 65004 octets in order + * for the matrix to be stored in an OGG header. + */ + if (rows > 255 || cols > 255) + return 0; + size = rows * (opus_int32)cols * sizeof(opus_int16); + if (size > 65004) + return 0; + + return align(sizeof(MappingMatrix)) + align(size); +} + +opus_int16 *mapping_matrix_get_data(const MappingMatrix *matrix) +{ + /* void* cast avoids clang -Wcast-align warning */ + return (opus_int16*)(void*)((char*)matrix + align(sizeof(MappingMatrix))); +} + +void mapping_matrix_init(MappingMatrix * const matrix, + int rows, int cols, int gain, const opus_int16 *data, opus_int32 data_size) +{ + int i; + opus_int16 *ptr; + +#if !defined(ENABLE_ASSERTIONS) + (void)data_size; +#endif + celt_assert(align(data_size) == align(rows * cols * sizeof(opus_int16))); + + matrix->rows = rows; + matrix->cols = cols; + matrix->gain = gain; + ptr = mapping_matrix_get_data(matrix); + for (i = 0; i < rows * cols; i++) + { + ptr[i] = data[i]; + } +} + +#ifndef DISABLE_FLOAT_API +void mapping_matrix_multiply_channel_in_float( + const MappingMatrix *matrix, + const float *input, + int input_rows, + opus_val16 *output, + int output_row, + int output_rows, + int frame_size) +{ + /* Matrix data is ordered col-wise. */ + opus_int16* matrix_data; + int i, col; + + celt_assert(input_rows <= matrix->cols && output_rows <= matrix->rows); + + matrix_data = mapping_matrix_get_data(matrix); + + for (i = 0; i < frame_size; i++) + { + float tmp = 0; + for (col = 0; col < input_rows; col++) + { + tmp += + matrix_data[MATRIX_INDEX(matrix->rows, output_row, col)] * + input[MATRIX_INDEX(input_rows, col, i)]; + } +#if defined(FIXED_POINT) + output[output_rows * i] = FLOAT2INT16((1/32768.f)*tmp); +#else + output[output_rows * i] = (1/32768.f)*tmp; +#endif + } +} + +void mapping_matrix_multiply_channel_out_float( + const MappingMatrix *matrix, + const opus_val16 *input, + int input_row, + int input_rows, + float *output, + int output_rows, + int frame_size +) +{ + /* Matrix data is ordered col-wise. */ + opus_int16* matrix_data; + int i, row; + float input_sample; + + celt_assert(input_rows <= matrix->cols && output_rows <= matrix->rows); + + matrix_data = mapping_matrix_get_data(matrix); + + for (i = 0; i < frame_size; i++) + { +#if defined(FIXED_POINT) + input_sample = (1/32768.f)*input[input_rows * i]; +#else + input_sample = input[input_rows * i]; +#endif + for (row = 0; row < output_rows; row++) + { + float tmp = + (1/32768.f)*matrix_data[MATRIX_INDEX(matrix->rows, row, input_row)] * + input_sample; + output[MATRIX_INDEX(output_rows, row, i)] += tmp; + } + } +} +#endif /* DISABLE_FLOAT_API */ + +void mapping_matrix_multiply_channel_in_short( + const MappingMatrix *matrix, + const opus_int16 *input, + int input_rows, + opus_val16 *output, + int output_row, + int output_rows, + int frame_size) +{ + /* Matrix data is ordered col-wise. */ + opus_int16* matrix_data; + int i, col; + + celt_assert(input_rows <= matrix->cols && output_rows <= matrix->rows); + + matrix_data = mapping_matrix_get_data(matrix); + + for (i = 0; i < frame_size; i++) + { + opus_val32 tmp = 0; + for (col = 0; col < input_rows; col++) + { +#if defined(FIXED_POINT) + tmp += + ((opus_int32)matrix_data[MATRIX_INDEX(matrix->rows, output_row, col)] * + (opus_int32)input[MATRIX_INDEX(input_rows, col, i)]) >> 8; +#else + tmp += + matrix_data[MATRIX_INDEX(matrix->rows, output_row, col)] * + input[MATRIX_INDEX(input_rows, col, i)]; +#endif + } +#if defined(FIXED_POINT) + output[output_rows * i] = (opus_int16)((tmp + 64) >> 7); +#else + output[output_rows * i] = (1/(32768.f*32768.f))*tmp; +#endif + } +} + +void mapping_matrix_multiply_channel_out_short( + const MappingMatrix *matrix, + const opus_val16 *input, + int input_row, + int input_rows, + opus_int16 *output, + int output_rows, + int frame_size) +{ + /* Matrix data is ordered col-wise. */ + opus_int16* matrix_data; + int i, row; + opus_int32 input_sample; + + celt_assert(input_rows <= matrix->cols && output_rows <= matrix->rows); + + matrix_data = mapping_matrix_get_data(matrix); + + for (i = 0; i < frame_size; i++) + { +#if defined(FIXED_POINT) + input_sample = (opus_int32)input[input_rows * i]; +#else + input_sample = (opus_int32)FLOAT2INT16(input[input_rows * i]); +#endif + for (row = 0; row < output_rows; row++) + { + opus_int32 tmp = + (opus_int32)matrix_data[MATRIX_INDEX(matrix->rows, row, input_row)] * + input_sample; + output[MATRIX_INDEX(output_rows, row, i)] += (tmp + 16384) >> 15; + } + } +} + +const MappingMatrix mapping_matrix_foa_mixing = { 6, 6, 0 }; +const opus_int16 mapping_matrix_foa_mixing_data[36] = { + 16384, 0, -16384, 23170, 0, 0, 16384, 23170, + 16384, 0, 0, 0, 16384, 0, -16384, -23170, + 0, 0, 16384, -23170, 16384, 0, 0, 0, + 0, 0, 0, 0, 32767, 0, 0, 0, + 0, 0, 0, 32767 +}; + +const MappingMatrix mapping_matrix_soa_mixing = { 11, 11, 0 }; +const opus_int16 mapping_matrix_soa_mixing_data[121] = { + 10923, 7723, 13377, -13377, 11585, 9459, 7723, -16384, + -6689, 0, 0, 10923, 7723, 13377, 13377, -11585, + 9459, 7723, 16384, -6689, 0, 0, 10923, -15447, + 13377, 0, 0, -18919, 7723, 0, 13377, 0, + 0, 10923, 7723, -13377, -13377, 11585, -9459, 7723, + 16384, -6689, 0, 0, 10923, -7723, 0, 13377, + -16384, 0, -15447, 0, 9459, 0, 0, 10923, + -7723, 0, -13377, 16384, 0, -15447, 0, 9459, + 0, 0, 10923, 15447, 0, 0, 0, 0, + -15447, 0, -18919, 0, 0, 10923, 7723, -13377, + 13377, -11585, -9459, 7723, -16384, -6689, 0, 0, + 10923, -15447, -13377, 0, 0, 18919, 7723, 0, + 13377, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 32767, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 32767 +}; + +const MappingMatrix mapping_matrix_toa_mixing = { 18, 18, 0 }; +const opus_int16 mapping_matrix_toa_mixing_data[324] = { + 8208, 0, -881, 14369, 0, 0, -8192, -4163, + 13218, 0, 0, 0, 11095, -8836, -6218, 14833, + 0, 0, 8208, -10161, 881, 10161, -13218, -2944, + -8192, 2944, 0, -10488, -6218, 6248, -11095, -6248, + 0, -10488, 0, 0, 8208, 10161, 881, -10161, + -13218, 2944, -8192, -2944, 0, 10488, -6218, -6248, + -11095, 6248, 0, 10488, 0, 0, 8176, 5566, + -11552, 5566, 9681, -11205, 8192, -11205, 0, 4920, + -15158, 9756, -3334, 9756, 0, -4920, 0, 0, + 8176, 7871, 11552, 0, 0, 15846, 8192, 0, + -9681, -6958, 0, 13797, 3334, 0, -15158, 0, + 0, 0, 8176, 0, 11552, 7871, 0, 0, + 8192, 15846, 9681, 0, 0, 0, 3334, 13797, + 15158, 6958, 0, 0, 8176, 5566, -11552, -5566, + -9681, -11205, 8192, 11205, 0, 4920, 15158, 9756, + -3334, -9756, 0, 4920, 0, 0, 8208, 14369, + -881, 0, 0, -4163, -8192, 0, -13218, -14833, + 0, -8836, 11095, 0, 6218, 0, 0, 0, + 8208, 10161, 881, 10161, 13218, 2944, -8192, 2944, + 0, 10488, 6218, -6248, -11095, -6248, 0, -10488, + 0, 0, 8208, -14369, -881, 0, 0, 4163, + -8192, 0, -13218, 14833, 0, 8836, 11095, 0, + 6218, 0, 0, 0, 8208, 0, -881, -14369, + 0, 0, -8192, 4163, 13218, 0, 0, 0, + 11095, 8836, -6218, -14833, 0, 0, 8176, -5566, + -11552, 5566, -9681, 11205, 8192, -11205, 0, -4920, + 15158, -9756, -3334, 9756, 0, -4920, 0, 0, + 8176, 0, 11552, -7871, 0, 0, 8192, -15846, + 9681, 0, 0, 0, 3334, -13797, 15158, -6958, + 0, 0, 8176, -7871, 11552, 0, 0, -15846, + 8192, 0, -9681, 6958, 0, -13797, 3334, 0, + -15158, 0, 0, 0, 8176, -5566, -11552, -5566, + 9681, 11205, 8192, 11205, 0, -4920, -15158, -9756, + -3334, -9756, 0, 4920, 0, 0, 8208, -10161, + 881, -10161, 13218, -2944, -8192, -2944, 0, -10488, + 6218, 6248, -11095, 6248, 0, 10488, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 32767, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 32767 +}; + +const MappingMatrix mapping_matrix_foa_demixing = { 6, 6, 0 }; +const opus_int16 mapping_matrix_foa_demixing_data[36] = { + 16384, 16384, 16384, 16384, 0, 0, 0, 23170, + 0, -23170, 0, 0, -16384, 16384, -16384, 16384, + 0, 0, 23170, 0, -23170, 0, 0, 0, + 0, 0, 0, 0, 32767, 0, 0, 0, + 0, 0, 0, 32767 +}; + +const MappingMatrix mapping_matrix_soa_demixing = { 11, 11, 3050 }; +const opus_int16 mapping_matrix_soa_demixing_data[121] = { + 2771, 2771, 2771, 2771, 2771, 2771, 2771, 2771, + 2771, 0, 0, 10033, 10033, -20066, 10033, 14189, + 14189, -28378, 10033, -20066, 0, 0, 3393, 3393, + 3393, -3393, 0, 0, 0, -3393, -3393, 0, + 0, -17378, 17378, 0, -17378, -24576, 24576, 0, + 17378, 0, 0, 0, -14189, 14189, 0, -14189, + -28378, 28378, 0, 14189, 0, 0, 0, 2399, + 2399, -4799, -2399, 0, 0, 0, -2399, 4799, + 0, 0, 1959, 1959, 1959, 1959, -3918, -3918, + -3918, 1959, 1959, 0, 0, -4156, 4156, 0, + 4156, 0, 0, 0, -4156, 0, 0, 0, + 8192, 8192, -16384, 8192, 16384, 16384, -32768, 8192, + -16384, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 8312, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 8312 +}; + +const MappingMatrix mapping_matrix_toa_demixing = { 18, 18, 0 }; +const opus_int16 mapping_matrix_toa_demixing_data[324] = { + 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, + 8192, 8192, 8192, 8192, 8192, 8192, 8192, 8192, + 0, 0, 0, -9779, 9779, 6263, 8857, 0, + 6263, 13829, 9779, -13829, 0, -6263, 0, -8857, + -6263, -9779, 0, 0, -3413, 3413, 3413, -11359, + 11359, 11359, -11359, -3413, 3413, -3413, -3413, -11359, + 11359, 11359, -11359, 3413, 0, 0, 13829, 9779, + -9779, 6263, 0, 8857, -6263, 0, 9779, 0, + -13829, 6263, -8857, 0, -6263, -9779, 0, 0, + 0, -15617, -15617, 6406, 0, 0, -6406, 0, + 15617, 0, 0, -6406, 0, 0, 6406, 15617, + 0, 0, 0, -5003, 5003, -10664, 15081, 0, + -10664, -7075, 5003, 7075, 0, 10664, 0, -15081, + 10664, -5003, 0, 0, -8176, -8176, -8176, 8208, + 8208, 8208, 8208, -8176, -8176, -8176, -8176, 8208, + 8208, 8208, 8208, -8176, 0, 0, -7075, 5003, + -5003, -10664, 0, 15081, 10664, 0, 5003, 0, + 7075, -10664, -15081, 0, 10664, -5003, 0, 0, + 15617, 0, 0, 0, -6406, 6406, 0, -15617, + 0, -15617, 15617, 0, 6406, -6406, 0, 0, + 0, 0, 0, -11393, 11393, 2993, -4233, 0, + 2993, -16112, 11393, 16112, 0, -2993, 0, 4233, + -2993, -11393, 0, 0, 0, -9974, -9974, -13617, + 0, 0, 13617, 0, 9974, 0, 0, 13617, + 0, 0, -13617, 9974, 0, 0, 0, 5579, + -5579, 10185, 14403, 0, 10185, -7890, -5579, 7890, + 0, -10185, 0, -14403, -10185, 5579, 0, 0, + 11826, -11826, -11826, -901, 901, 901, -901, 11826, + -11826, 11826, 11826, -901, 901, 901, -901, -11826, + 0, 0, -7890, -5579, 5579, 10185, 0, 14403, + -10185, 0, -5579, 0, 7890, 10185, -14403, 0, + -10185, 5579, 0, 0, -9974, 0, 0, 0, + -13617, 13617, 0, 9974, 0, 9974, -9974, 0, + 13617, -13617, 0, 0, 0, 0, 16112, -11393, + 11393, -2993, 0, 4233, 2993, 0, -11393, 0, + -16112, -2993, -4233, 0, 2993, 11393, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 32767, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 32767 +}; + diff --git a/lib/opus/src/mapping_matrix.h b/lib/opus/src/mapping_matrix.h new file mode 100644 index 00000000..98bc82df --- /dev/null +++ b/lib/opus/src/mapping_matrix.h @@ -0,0 +1,133 @@ +/* Copyright (c) 2017 Google Inc. + Written by Andrew Allen */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/** + * @file mapping_matrix.h + * @brief Opus reference implementation mapping matrix API + */ + +#ifndef MAPPING_MATRIX_H +#define MAPPING_MATRIX_H + +#include "opus_types.h" +#include "opus_projection.h" + +#ifdef __cplusplus +extern "C" { +#endif + +typedef struct MappingMatrix +{ + int rows; /* number of channels outputted from matrix. */ + int cols; /* number of channels inputted to matrix. */ + int gain; /* in dB. S7.8-format. */ + /* Matrix cell data goes here using col-wise ordering. */ +} MappingMatrix; + +opus_int32 mapping_matrix_get_size(int rows, int cols); + +opus_int16 *mapping_matrix_get_data(const MappingMatrix *matrix); + +void mapping_matrix_init( + MappingMatrix * const matrix, + int rows, + int cols, + int gain, + const opus_int16 *data, + opus_int32 data_size +); + +#ifndef DISABLE_FLOAT_API +void mapping_matrix_multiply_channel_in_float( + const MappingMatrix *matrix, + const float *input, + int input_rows, + opus_val16 *output, + int output_row, + int output_rows, + int frame_size +); + +void mapping_matrix_multiply_channel_out_float( + const MappingMatrix *matrix, + const opus_val16 *input, + int input_row, + int input_rows, + float *output, + int output_rows, + int frame_size +); +#endif /* DISABLE_FLOAT_API */ + +void mapping_matrix_multiply_channel_in_short( + const MappingMatrix *matrix, + const opus_int16 *input, + int input_rows, + opus_val16 *output, + int output_row, + int output_rows, + int frame_size +); + +void mapping_matrix_multiply_channel_out_short( + const MappingMatrix *matrix, + const opus_val16 *input, + int input_row, + int input_rows, + opus_int16 *output, + int output_rows, + int frame_size +); + +/* Pre-computed mixing and demixing matrices for 1st to 3rd-order ambisonics. + * foa: first-order ambisonics + * soa: second-order ambisonics + * toa: third-order ambisonics + */ +extern const MappingMatrix mapping_matrix_foa_mixing; +extern const opus_int16 mapping_matrix_foa_mixing_data[36]; + +extern const MappingMatrix mapping_matrix_soa_mixing; +extern const opus_int16 mapping_matrix_soa_mixing_data[121]; + +extern const MappingMatrix mapping_matrix_toa_mixing; +extern const opus_int16 mapping_matrix_toa_mixing_data[324]; + +extern const MappingMatrix mapping_matrix_foa_demixing; +extern const opus_int16 mapping_matrix_foa_demixing_data[36]; + +extern const MappingMatrix mapping_matrix_soa_demixing; +extern const opus_int16 mapping_matrix_soa_demixing_data[121]; + +extern const MappingMatrix mapping_matrix_toa_demixing; +extern const opus_int16 mapping_matrix_toa_demixing_data[324]; + +#ifdef __cplusplus +} +#endif + +#endif /* MAPPING_MATRIX_H */ diff --git a/lib/opus/src/meson.build b/lib/opus/src/meson.build new file mode 100644 index 00000000..cc07ff06 --- /dev/null +++ b/lib/opus/src/meson.build @@ -0,0 +1,45 @@ +opus_sources = sources['OPUS_SOURCES'] + +opus_sources_float = sources['OPUS_SOURCES_FLOAT'] + +if not disable_float_api + opus_sources += opus_sources_float +endif + +opus_lib_c_args = [] +if host_machine.system() == 'windows' + opus_lib_c_args += ['-DDLL_EXPORT'] +endif + +opus_lib = library('opus', + opus_sources, + version: libversion, + darwin_versions: macosversion, + c_args: opus_lib_c_args, + include_directories: opus_includes, + link_with: [celt_lib, silk_lib], + dependencies: libm, + install: true) + +opus_dep = declare_dependency(link_with: opus_lib, + include_directories: opus_public_includes) + +# Extra uninstalled Opus programs +if not extra_programs.disabled() + foreach prog : ['opus_compare', 'opus_demo', 'repacketizer_demo'] + executable(prog, '@0@.c'.format(prog), + include_directories: opus_includes, + link_with: opus_lib, + dependencies: libm, + install: false) + endforeach + + if opt_custom_modes + executable('opus_custom_demo', '../celt/opus_custom_demo.c', + include_directories: opus_includes, + link_with: opus_lib, + dependencies: libm, + install: false) + endif + +endif diff --git a/lib/opus/src/mlp.c b/lib/opus/src/mlp.c new file mode 100644 index 00000000..964c6a98 --- /dev/null +++ b/lib/opus/src/mlp.c @@ -0,0 +1,144 @@ +/* Copyright (c) 2008-2011 Octasic Inc. + 2012-2017 Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR + CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "opus_types.h" +#include "opus_defines.h" +#include "arch.h" +#include "tansig_table.h" +#include "mlp.h" + +static OPUS_INLINE float tansig_approx(float x) +{ + int i; + float y, dy; + float sign=1; + /* Tests are reversed to catch NaNs */ + if (!(x<8)) + return 1; + if (!(x>-8)) + return -1; +#ifndef FIXED_POINT + /* Another check in case of -ffast-math */ + if (celt_isnan(x)) + return 0; +#endif + if (x<0) + { + x=-x; + sign=-1; + } + i = (int)floor(.5f+25*x); + x -= .04f*i; + y = tansig_table[i]; + dy = 1-y*y; + y = y + x*dy*(1 - y*x); + return sign*y; +} + +static OPUS_INLINE float sigmoid_approx(float x) +{ + return .5f + .5f*tansig_approx(.5f*x); +} + +static void gemm_accum(float *out, const opus_int8 *weights, int rows, int cols, int col_stride, const float *x) +{ + int i, j; + for (i=0;inb_inputs; + N = layer->nb_neurons; + stride = N; + for (i=0;ibias[i]; + gemm_accum(output, layer->input_weights, N, M, stride, input); + for (i=0;isigmoid) { + for (i=0;inb_inputs; + N = gru->nb_neurons; + stride = 3*N; + /* Compute update gate. */ + for (i=0;ibias[i]; + gemm_accum(z, gru->input_weights, N, M, stride, input); + gemm_accum(z, gru->recurrent_weights, N, N, stride, state); + for (i=0;ibias[N + i]; + gemm_accum(r, &gru->input_weights[N], N, M, stride, input); + gemm_accum(r, &gru->recurrent_weights[N], N, N, stride, state); + for (i=0;ibias[2*N + i]; + for (i=0;iinput_weights[2*N], N, M, stride, input); + gemm_accum(h, &gru->recurrent_weights[2*N], N, N, stride, tmp); + for (i=0;i=0) + break; + x[i*C] = x[i*C]+a*x[i*C]*x[i*C]; + } + + curr=0; + x0 = x[0]; + while(1) + { + int start, end; + float maxval; + int special=0; + int peak_pos; + for (i=curr;i1 || x[i*C]<-1) + break; + } + if (i==N) + { + a=0; + break; + } + peak_pos = i; + start=end=i; + maxval=ABS16(x[i*C]); + /* Look for first zero crossing before clipping */ + while (start>0 && x[i*C]*x[(start-1)*C]>=0) + start--; + /* Look for first zero crossing after clipping */ + while (end=0) + { + /* Look for other peaks until the next zero-crossing. */ + if (ABS16(x[end*C])>maxval) + { + maxval = ABS16(x[end*C]); + peak_pos = end; + } + end++; + } + /* Detect the special case where we clip before the first zero crossing */ + special = (start==0 && x[i*C]*x[0]>=0); + + /* Compute a such that maxval + a*maxval^2 = 1 */ + a=(maxval-1)/(maxval*maxval); + /* Slightly boost "a" by 2^-22. This is just enough to ensure -ffast-math + does not cause output values larger than +/-1, but small enough not + to matter even for 24-bit output. */ + a += a*2.4e-7f; + if (x[i*C]>0) + a = -a; + /* Apply soft clipping */ + for (i=start;i=2) + { + /* Add a linear ramp from the first sample to the signal peak. + This avoids a discontinuity at the beginning of the frame. */ + float delta; + float offset = x0-x[0]; + delta = offset / peak_pos; + for (i=curr;i>2; + return 2; + } +} + +static int parse_size(const unsigned char *data, opus_int32 len, opus_int16 *size) +{ + if (len<1) + { + *size = -1; + return -1; + } else if (data[0]<252) + { + *size = data[0]; + return 1; + } else if (len<2) + { + *size = -1; + return -1; + } else { + *size = 4*data[1] + data[0]; + return 2; + } +} + +int opus_packet_get_samples_per_frame(const unsigned char *data, + opus_int32 Fs) +{ + int audiosize; + if (data[0]&0x80) + { + audiosize = ((data[0]>>3)&0x3); + audiosize = (Fs<>3)&0x3); + if (audiosize == 3) + audiosize = Fs*60/1000; + else + audiosize = (Fs< len) + return OPUS_INVALID_PACKET; + data += bytes; + last_size = len-size[0]; + break; + /* Multiple CBR/VBR frames (from 0 to 120 ms) */ + default: /*case 3:*/ + if (len<1) + return OPUS_INVALID_PACKET; + /* Number of frames encoded in bits 0 to 5 */ + ch = *data++; + count = ch&0x3F; + if (count <= 0 || framesize*(opus_int32)count > 5760) + return OPUS_INVALID_PACKET; + len--; + /* Padding flag is bit 6 */ + if (ch&0x40) + { + int p; + do { + int tmp; + if (len<=0) + return OPUS_INVALID_PACKET; + p = *data++; + len--; + tmp = p==255 ? 254: p; + len -= tmp; + pad += tmp; + } while (p==255); + } + if (len<0) + return OPUS_INVALID_PACKET; + /* VBR flag is bit 7 */ + cbr = !(ch&0x80); + if (!cbr) + { + /* VBR case */ + last_size = len; + for (i=0;i len) + return OPUS_INVALID_PACKET; + data += bytes; + last_size -= bytes+size[i]; + } + if (last_size<0) + return OPUS_INVALID_PACKET; + } else if (!self_delimited) + { + /* CBR case */ + last_size = len/count; + if (last_size*count!=len) + return OPUS_INVALID_PACKET; + for (i=0;i len) + return OPUS_INVALID_PACKET; + data += bytes; + /* For CBR packets, apply the size to all the frames. */ + if (cbr) + { + if (size[count-1]*count > len) + return OPUS_INVALID_PACKET; + for (i=0;i last_size) + return OPUS_INVALID_PACKET; + } else + { + /* Because it's not encoded explicitly, it's possible the size of the + last packet (or all the packets, for the CBR case) is larger than + 1275. Reject them here.*/ + if (last_size > 1275) + return OPUS_INVALID_PACKET; + size[count-1] = (opus_int16)last_size; + } + + if (payload_offset) + *payload_offset = (int)(data-data0); + + for (i=0;i +#include +#include +#include + +#define OPUS_PI (3.14159265F) + +#define OPUS_COSF(_x) ((float)cos(_x)) +#define OPUS_SINF(_x) ((float)sin(_x)) + +static void *check_alloc(void *_ptr){ + if(_ptr==NULL){ + fprintf(stderr,"Out of memory.\n"); + exit(EXIT_FAILURE); + } + return _ptr; +} + +static void *opus_malloc(size_t _size){ + return check_alloc(malloc(_size)); +} + +static void *opus_realloc(void *_ptr,size_t _size){ + return check_alloc(realloc(_ptr,_size)); +} + +static size_t read_pcm16(float **_samples,FILE *_fin,int _nchannels){ + unsigned char buf[1024]; + float *samples; + size_t nsamples; + size_t csamples; + size_t xi; + size_t nread; + samples=NULL; + nsamples=csamples=0; + for(;;){ + nread=fread(buf,2*_nchannels,1024/(2*_nchannels),_fin); + if(nread<=0)break; + if(nsamples+nread>csamples){ + do csamples=csamples<<1|1; + while(nsamples+nread>csamples); + samples=(float *)opus_realloc(samples, + _nchannels*csamples*sizeof(*samples)); + } + for(xi=0;xi=_window_sz)ti-=_window_sz; + } + re*=_downsample; + im*=_downsample; + _ps[(xi*ps_sz+xj)*_nchannels+ci]=re*re+im*im+100000; + p[ci]+=_ps[(xi*ps_sz+xj)*_nchannels+ci]; + } + } + if(_out){ + _out[(xi*_nbands+bi)*_nchannels]=p[0]/(_bands[bi+1]-_bands[bi]); + if(_nchannels==2){ + _out[(xi*_nbands+bi)*_nchannels+1]=p[1]/(_bands[bi+1]-_bands[bi]); + } + } + } + } + free(window); +} + +#define NBANDS (21) +#define NFREQS (240) + +/*Bands on which we compute the pseudo-NMR (Bark-derived + CELT bands).*/ +static const int BANDS[NBANDS+1]={ + 0,2,4,6,8,10,12,14,16,20,24,28,32,40,48,56,68,80,96,120,156,200 +}; + +#define TEST_WIN_SIZE (480) +#define TEST_WIN_STEP (120) + +int main(int _argc,const char **_argv){ + FILE *fin1; + FILE *fin2; + float *x; + float *y; + float *xb; + float *X; + float *Y; + double err; + float Q; + size_t xlength; + size_t ylength; + size_t nframes; + size_t xi; + int ci; + int xj; + int bi; + int nchannels; + unsigned rate; + int downsample; + int ybands; + int yfreqs; + int max_compare; + if(_argc<3||_argc>6){ + fprintf(stderr,"Usage: %s [-s] [-r rate2] \n", + _argv[0]); + return EXIT_FAILURE; + } + nchannels=1; + if(strcmp(_argv[1],"-s")==0){ + nchannels=2; + _argv++; + } + rate=48000; + ybands=NBANDS; + yfreqs=NFREQS; + downsample=1; + if(strcmp(_argv[1],"-r")==0){ + rate=atoi(_argv[2]); + if(rate!=8000&&rate!=12000&&rate!=16000&&rate!=24000&&rate!=48000){ + fprintf(stderr, + "Sampling rate must be 8000, 12000, 16000, 24000, or 48000\n"); + return EXIT_FAILURE; + } + downsample=48000/rate; + switch(rate){ + case 8000:ybands=13;break; + case 12000:ybands=15;break; + case 16000:ybands=17;break; + case 24000:ybands=19;break; + } + yfreqs=NFREQS/downsample; + _argv+=2; + } + fin1=fopen(_argv[1],"rb"); + if(fin1==NULL){ + fprintf(stderr,"Error opening '%s'.\n",_argv[1]); + return EXIT_FAILURE; + } + fin2=fopen(_argv[2],"rb"); + if(fin2==NULL){ + fprintf(stderr,"Error opening '%s'.\n",_argv[2]); + fclose(fin1); + return EXIT_FAILURE; + } + /*Read in the data and allocate scratch space.*/ + xlength=read_pcm16(&x,fin1,2); + if(nchannels==1){ + for(xi=0;xi0;){ + for(ci=0;ci0){ + /*Temporal masking: -3 dB/2.5ms slope.*/ + for(bi=0;bi=79&&xj<=81)im*=0.1F; + if(xj==80)im*=0.1F; + Eb+=im; + } + } + Eb /= (BANDS[bi+1]-BANDS[bi])*nchannels; + Ef += Eb*Eb; + } + /*Using a fixed normalization value means we're willing to accept slightly + lower quality for lower sampling rates.*/ + Ef/=NBANDS; + Ef*=Ef; + err+=Ef*Ef; + } + free(xb); + free(X); + free(Y); + err=pow(err/nframes,1.0/16); + Q=100*(1-0.5*log(1+err)/log(1.13)); + if(Q<0){ + fprintf(stderr,"Test vector FAILS\n"); + fprintf(stderr,"Internal weighted error is %f\n",err); + return EXIT_FAILURE; + } + else{ + fprintf(stderr,"Test vector PASSES\n"); + fprintf(stderr, + "Opus quality metric: %.1f %% (internal weighted error is %f)\n",Q,err); + return EXIT_SUCCESS; + } +} diff --git a/lib/opus/src/opus_decoder.c b/lib/opus/src/opus_decoder.c new file mode 100644 index 00000000..6520e748 --- /dev/null +++ b/lib/opus/src/opus_decoder.c @@ -0,0 +1,1041 @@ +/* Copyright (c) 2010 Xiph.Org Foundation, Skype Limited + Written by Jean-Marc Valin and Koen Vos */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +# include "config.h" +#endif + +#ifndef OPUS_BUILD +# error "OPUS_BUILD _MUST_ be defined to build Opus. This probably means you need other defines as well, as in a config.h. See the included build files for details." +#endif + +#if defined(__GNUC__) && (__GNUC__ >= 2) && !defined(__OPTIMIZE__) && !defined(OPUS_WILL_BE_SLOW) +# pragma message "You appear to be compiling without optimization, if so opus will be very slow." +#endif + +#include +#include "celt.h" +#include "opus.h" +#include "entdec.h" +#include "modes.h" +#include "API.h" +#include "stack_alloc.h" +#include "float_cast.h" +#include "opus_private.h" +#include "os_support.h" +#include "structs.h" +#include "define.h" +#include "mathops.h" +#include "cpu_support.h" + +struct OpusDecoder { + int celt_dec_offset; + int silk_dec_offset; + int channels; + opus_int32 Fs; /** Sampling rate (at the API level) */ + silk_DecControlStruct DecControl; + int decode_gain; + int arch; + + /* Everything beyond this point gets cleared on a reset */ +#define OPUS_DECODER_RESET_START stream_channels + int stream_channels; + + int bandwidth; + int mode; + int prev_mode; + int frame_size; + int prev_redundancy; + int last_packet_duration; +#ifndef FIXED_POINT + opus_val16 softclip_mem[2]; +#endif + + opus_uint32 rangeFinal; +}; + +#if defined(ENABLE_HARDENING) || defined(ENABLE_ASSERTIONS) +static void validate_opus_decoder(OpusDecoder *st) +{ + celt_assert(st->channels == 1 || st->channels == 2); + celt_assert(st->Fs == 48000 || st->Fs == 24000 || st->Fs == 16000 || st->Fs == 12000 || st->Fs == 8000); + celt_assert(st->DecControl.API_sampleRate == st->Fs); + celt_assert(st->DecControl.internalSampleRate == 0 || st->DecControl.internalSampleRate == 16000 || st->DecControl.internalSampleRate == 12000 || st->DecControl.internalSampleRate == 8000); + celt_assert(st->DecControl.nChannelsAPI == st->channels); + celt_assert(st->DecControl.nChannelsInternal == 0 || st->DecControl.nChannelsInternal == 1 || st->DecControl.nChannelsInternal == 2); + celt_assert(st->DecControl.payloadSize_ms == 0 || st->DecControl.payloadSize_ms == 10 || st->DecControl.payloadSize_ms == 20 || st->DecControl.payloadSize_ms == 40 || st->DecControl.payloadSize_ms == 60); +#ifdef OPUS_ARCHMASK + celt_assert(st->arch >= 0); + celt_assert(st->arch <= OPUS_ARCHMASK); +#endif + celt_assert(st->stream_channels == 1 || st->stream_channels == 2); +} +#define VALIDATE_OPUS_DECODER(st) validate_opus_decoder(st) +#else +#define VALIDATE_OPUS_DECODER(st) +#endif + +int opus_decoder_get_size(int channels) +{ + int silkDecSizeBytes, celtDecSizeBytes; + int ret; + if (channels<1 || channels > 2) + return 0; + ret = silk_Get_Decoder_Size( &silkDecSizeBytes ); + if(ret) + return 0; + silkDecSizeBytes = align(silkDecSizeBytes); + celtDecSizeBytes = celt_decoder_get_size(channels); + return align(sizeof(OpusDecoder))+silkDecSizeBytes+celtDecSizeBytes; +} + +int opus_decoder_init(OpusDecoder *st, opus_int32 Fs, int channels) +{ + void *silk_dec; + CELTDecoder *celt_dec; + int ret, silkDecSizeBytes; + + if ((Fs!=48000&&Fs!=24000&&Fs!=16000&&Fs!=12000&&Fs!=8000) + || (channels!=1&&channels!=2)) + return OPUS_BAD_ARG; + + OPUS_CLEAR((char*)st, opus_decoder_get_size(channels)); + /* Initialize SILK decoder */ + ret = silk_Get_Decoder_Size(&silkDecSizeBytes); + if (ret) + return OPUS_INTERNAL_ERROR; + + silkDecSizeBytes = align(silkDecSizeBytes); + st->silk_dec_offset = align(sizeof(OpusDecoder)); + st->celt_dec_offset = st->silk_dec_offset+silkDecSizeBytes; + silk_dec = (char*)st+st->silk_dec_offset; + celt_dec = (CELTDecoder*)((char*)st+st->celt_dec_offset); + st->stream_channels = st->channels = channels; + + st->Fs = Fs; + st->DecControl.API_sampleRate = st->Fs; + st->DecControl.nChannelsAPI = st->channels; + + /* Reset decoder */ + ret = silk_InitDecoder( silk_dec ); + if(ret)return OPUS_INTERNAL_ERROR; + + /* Initialize CELT decoder */ + ret = celt_decoder_init(celt_dec, Fs, channels); + if(ret!=OPUS_OK)return OPUS_INTERNAL_ERROR; + + celt_decoder_ctl(celt_dec, CELT_SET_SIGNALLING(0)); + + st->prev_mode = 0; + st->frame_size = Fs/400; + st->arch = opus_select_arch(); + return OPUS_OK; +} + +OpusDecoder *opus_decoder_create(opus_int32 Fs, int channels, int *error) +{ + int ret; + OpusDecoder *st; + if ((Fs!=48000&&Fs!=24000&&Fs!=16000&&Fs!=12000&&Fs!=8000) + || (channels!=1&&channels!=2)) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + st = (OpusDecoder *)opus_alloc(opus_decoder_get_size(channels)); + if (st == NULL) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + ret = opus_decoder_init(st, Fs, channels); + if (error) + *error = ret; + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + return st; +} + +static void smooth_fade(const opus_val16 *in1, const opus_val16 *in2, + opus_val16 *out, int overlap, int channels, + const opus_val16 *window, opus_int32 Fs) +{ + int i, c; + int inc = 48000/Fs; + for (c=0;csilk_dec_offset; + celt_dec = (CELTDecoder*)((char*)st+st->celt_dec_offset); + F20 = st->Fs/50; + F10 = F20>>1; + F5 = F10>>1; + F2_5 = F5>>1; + if (frame_size < F2_5) + { + RESTORE_STACK; + return OPUS_BUFFER_TOO_SMALL; + } + /* Limit frame_size to avoid excessive stack allocations. */ + frame_size = IMIN(frame_size, st->Fs/25*3); + /* Payloads of 1 (2 including ToC) or 0 trigger the PLC/DTX */ + if (len<=1) + { + data = NULL; + /* In that case, don't conceal more than what the ToC says */ + frame_size = IMIN(frame_size, st->frame_size); + } + if (data != NULL) + { + audiosize = st->frame_size; + mode = st->mode; + bandwidth = st->bandwidth; + ec_dec_init(&dec,(unsigned char*)data,len); + } else { + audiosize = frame_size; + /* Run PLC using last used mode (CELT if we ended with CELT redundancy) */ + mode = st->prev_redundancy ? MODE_CELT_ONLY : st->prev_mode; + bandwidth = 0; + + if (mode == 0) + { + /* If we haven't got any packet yet, all we can do is return zeros */ + for (i=0;ichannels;i++) + pcm[i] = 0; + RESTORE_STACK; + return audiosize; + } + + /* Avoids trying to run the PLC on sizes other than 2.5 (CELT), 5 (CELT), + 10, or 20 (e.g. 12.5 or 30 ms). */ + if (audiosize > F20) + { + do { + int ret = opus_decode_frame(st, NULL, 0, pcm, IMIN(audiosize, F20), 0); + if (ret<0) + { + RESTORE_STACK; + return ret; + } + pcm += ret*st->channels; + audiosize -= ret; + } while (audiosize > 0); + RESTORE_STACK; + return frame_size; + } else if (audiosize < F20) + { + if (audiosize > F10) + audiosize = F10; + else if (mode != MODE_SILK_ONLY && audiosize > F5 && audiosize < F10) + audiosize = F5; + } + } + + /* In fixed-point, we can tell CELT to do the accumulation on top of the + SILK PCM buffer. This saves some stack space. */ +#ifdef FIXED_POINT + celt_accum = (mode != MODE_CELT_ONLY) && (frame_size >= F10); +#else + celt_accum = 0; +#endif + + pcm_transition_silk_size = ALLOC_NONE; + pcm_transition_celt_size = ALLOC_NONE; + if (data!=NULL && st->prev_mode > 0 && ( + (mode == MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY && !st->prev_redundancy) + || (mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) ) + ) + { + transition = 1; + /* Decide where to allocate the stack memory for pcm_transition */ + if (mode == MODE_CELT_ONLY) + pcm_transition_celt_size = F5*st->channels; + else + pcm_transition_silk_size = F5*st->channels; + } + ALLOC(pcm_transition_celt, pcm_transition_celt_size, opus_val16); + if (transition && mode == MODE_CELT_ONLY) + { + pcm_transition = pcm_transition_celt; + opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); + } + if (audiosize > frame_size) + { + /*fprintf(stderr, "PCM buffer too small: %d vs %d (mode = %d)\n", audiosize, frame_size, mode);*/ + RESTORE_STACK; + return OPUS_BAD_ARG; + } else { + frame_size = audiosize; + } + + /* Don't allocate any memory when in CELT-only mode */ + pcm_silk_size = (mode != MODE_CELT_ONLY && !celt_accum) ? IMAX(F10, frame_size)*st->channels : ALLOC_NONE; + ALLOC(pcm_silk, pcm_silk_size, opus_int16); + + /* SILK processing */ + if (mode != MODE_CELT_ONLY) + { + int lost_flag, decoded_samples; + opus_int16 *pcm_ptr; +#ifdef FIXED_POINT + if (celt_accum) + pcm_ptr = pcm; + else +#endif + pcm_ptr = pcm_silk; + + if (st->prev_mode==MODE_CELT_ONLY) + silk_InitDecoder( silk_dec ); + + /* The SILK PLC cannot produce frames of less than 10 ms */ + st->DecControl.payloadSize_ms = IMAX(10, 1000 * audiosize / st->Fs); + + if (data != NULL) + { + st->DecControl.nChannelsInternal = st->stream_channels; + if( mode == MODE_SILK_ONLY ) { + if( bandwidth == OPUS_BANDWIDTH_NARROWBAND ) { + st->DecControl.internalSampleRate = 8000; + } else if( bandwidth == OPUS_BANDWIDTH_MEDIUMBAND ) { + st->DecControl.internalSampleRate = 12000; + } else if( bandwidth == OPUS_BANDWIDTH_WIDEBAND ) { + st->DecControl.internalSampleRate = 16000; + } else { + st->DecControl.internalSampleRate = 16000; + celt_assert( 0 ); + } + } else { + /* Hybrid mode */ + st->DecControl.internalSampleRate = 16000; + } + } + + lost_flag = data == NULL ? 1 : 2 * decode_fec; + decoded_samples = 0; + do { + /* Call SILK decoder */ + int first_frame = decoded_samples == 0; + silk_ret = silk_Decode( silk_dec, &st->DecControl, + lost_flag, first_frame, &dec, pcm_ptr, &silk_frame_size, st->arch ); + if( silk_ret ) { + if (lost_flag) { + /* PLC failure should not be fatal */ + silk_frame_size = frame_size; + for (i=0;ichannels;i++) + pcm_ptr[i] = 0; + } else { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + } + pcm_ptr += silk_frame_size * st->channels; + decoded_samples += silk_frame_size; + } while( decoded_samples < frame_size ); + } + + start_band = 0; + if (!decode_fec && mode != MODE_CELT_ONLY && data != NULL + && ec_tell(&dec)+17+20*(mode == MODE_HYBRID) <= 8*len) + { + /* Check if we have a redundant 0-8 kHz band */ + if (mode == MODE_HYBRID) + redundancy = ec_dec_bit_logp(&dec, 12); + else + redundancy = 1; + if (redundancy) + { + celt_to_silk = ec_dec_bit_logp(&dec, 1); + /* redundancy_bytes will be at least two, in the non-hybrid + case due to the ec_tell() check above */ + redundancy_bytes = mode==MODE_HYBRID ? + (opus_int32)ec_dec_uint(&dec, 256)+2 : + len-((ec_tell(&dec)+7)>>3); + len -= redundancy_bytes; + /* This is a sanity check. It should never happen for a valid + packet, so the exact behaviour is not normative. */ + if (len*8 < ec_tell(&dec)) + { + len = 0; + redundancy_bytes = 0; + redundancy = 0; + } + /* Shrink decoder because of raw bits */ + dec.storage -= redundancy_bytes; + } + } + if (mode != MODE_CELT_ONLY) + start_band = 17; + + if (redundancy) + { + transition = 0; + pcm_transition_silk_size=ALLOC_NONE; + } + + ALLOC(pcm_transition_silk, pcm_transition_silk_size, opus_val16); + + if (transition && mode != MODE_CELT_ONLY) + { + pcm_transition = pcm_transition_silk; + opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); + } + + + if (bandwidth) + { + int endband=21; + + switch(bandwidth) + { + case OPUS_BANDWIDTH_NARROWBAND: + endband = 13; + break; + case OPUS_BANDWIDTH_MEDIUMBAND: + case OPUS_BANDWIDTH_WIDEBAND: + endband = 17; + break; + case OPUS_BANDWIDTH_SUPERWIDEBAND: + endband = 19; + break; + case OPUS_BANDWIDTH_FULLBAND: + endband = 21; + break; + default: + celt_assert(0); + break; + } + MUST_SUCCEED(celt_decoder_ctl(celt_dec, CELT_SET_END_BAND(endband))); + } + MUST_SUCCEED(celt_decoder_ctl(celt_dec, CELT_SET_CHANNELS(st->stream_channels))); + + /* Only allocation memory for redundancy if/when needed */ + redundant_audio_size = redundancy ? F5*st->channels : ALLOC_NONE; + ALLOC(redundant_audio, redundant_audio_size, opus_val16); + + /* 5 ms redundant frame for CELT->SILK*/ + if (redundancy && celt_to_silk) + { + /* If the previous frame did not use CELT (the first redundancy frame in + a transition from SILK may have been lost) then the CELT decoder is + stale at this point and the redundancy audio is not useful, however + the final range is still needed (for testing), so the redundancy is + always decoded but the decoded audio may not be used */ + MUST_SUCCEED(celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0))); + celt_decode_with_ec(celt_dec, data+len, redundancy_bytes, + redundant_audio, F5, NULL, 0); + MUST_SUCCEED(celt_decoder_ctl(celt_dec, OPUS_GET_FINAL_RANGE(&redundant_rng))); + } + + /* MUST be after PLC */ + MUST_SUCCEED(celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(start_band))); + + if (mode != MODE_SILK_ONLY) + { + int celt_frame_size = IMIN(F20, frame_size); + /* Make sure to discard any previous CELT state */ + if (mode != st->prev_mode && st->prev_mode > 0 && !st->prev_redundancy) + MUST_SUCCEED(celt_decoder_ctl(celt_dec, OPUS_RESET_STATE)); + /* Decode CELT */ + celt_ret = celt_decode_with_ec(celt_dec, decode_fec ? NULL : data, + len, pcm, celt_frame_size, &dec, celt_accum); + } else { + unsigned char silence[2] = {0xFF, 0xFF}; + if (!celt_accum) + { + for (i=0;ichannels;i++) + pcm[i] = 0; + } + /* For hybrid -> SILK transitions, we let the CELT MDCT + do a fade-out by decoding a silence frame */ + if (st->prev_mode == MODE_HYBRID && !(redundancy && celt_to_silk && st->prev_redundancy) ) + { + MUST_SUCCEED(celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0))); + celt_decode_with_ec(celt_dec, silence, 2, pcm, F2_5, NULL, celt_accum); + } + } + + if (mode != MODE_CELT_ONLY && !celt_accum) + { +#ifdef FIXED_POINT + for (i=0;ichannels;i++) + pcm[i] = SAT16(ADD32(pcm[i], pcm_silk[i])); +#else + for (i=0;ichannels;i++) + pcm[i] = pcm[i] + (opus_val16)((1.f/32768.f)*pcm_silk[i]); +#endif + } + + { + const CELTMode *celt_mode; + MUST_SUCCEED(celt_decoder_ctl(celt_dec, CELT_GET_MODE(&celt_mode))); + window = celt_mode->window; + } + + /* 5 ms redundant frame for SILK->CELT */ + if (redundancy && !celt_to_silk) + { + MUST_SUCCEED(celt_decoder_ctl(celt_dec, OPUS_RESET_STATE)); + MUST_SUCCEED(celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0))); + + celt_decode_with_ec(celt_dec, data+len, redundancy_bytes, redundant_audio, F5, NULL, 0); + MUST_SUCCEED(celt_decoder_ctl(celt_dec, OPUS_GET_FINAL_RANGE(&redundant_rng))); + smooth_fade(pcm+st->channels*(frame_size-F2_5), redundant_audio+st->channels*F2_5, + pcm+st->channels*(frame_size-F2_5), F2_5, st->channels, window, st->Fs); + } + /* 5ms redundant frame for CELT->SILK; ignore if the previous frame did not + use CELT (the first redundancy frame in a transition from SILK may have + been lost) */ + if (redundancy && celt_to_silk && (st->prev_mode != MODE_SILK_ONLY || st->prev_redundancy)) + { + for (c=0;cchannels;c++) + { + for (i=0;ichannels*i+c] = redundant_audio[st->channels*i+c]; + } + smooth_fade(redundant_audio+st->channels*F2_5, pcm+st->channels*F2_5, + pcm+st->channels*F2_5, F2_5, st->channels, window, st->Fs); + } + if (transition) + { + if (audiosize >= F5) + { + for (i=0;ichannels*F2_5;i++) + pcm[i] = pcm_transition[i]; + smooth_fade(pcm_transition+st->channels*F2_5, pcm+st->channels*F2_5, + pcm+st->channels*F2_5, F2_5, + st->channels, window, st->Fs); + } else { + /* Not enough time to do a clean transition, but we do it anyway + This will not preserve amplitude perfectly and may introduce + a bit of temporal aliasing, but it shouldn't be too bad and + that's pretty much the best we can do. In any case, generating this + transition it pretty silly in the first place */ + smooth_fade(pcm_transition, pcm, + pcm, F2_5, + st->channels, window, st->Fs); + } + } + + if(st->decode_gain) + { + opus_val32 gain; + gain = celt_exp2(MULT16_16_P15(QCONST16(6.48814081e-4f, 25), st->decode_gain)); + for (i=0;ichannels;i++) + { + opus_val32 x; + x = MULT16_32_P16(pcm[i],gain); + pcm[i] = SATURATE(x, 32767); + } + } + + if (len <= 1) + st->rangeFinal = 0; + else + st->rangeFinal = dec.rng ^ redundant_rng; + + st->prev_mode = mode; + st->prev_redundancy = redundancy && !celt_to_silk; + + if (celt_ret>=0) + { + if (OPUS_CHECK_ARRAY(pcm, audiosize*st->channels)) + OPUS_PRINT_INT(audiosize); + } + + RESTORE_STACK; + return celt_ret < 0 ? celt_ret : audiosize; + +} + +int opus_decode_native(OpusDecoder *st, const unsigned char *data, + opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec, + int self_delimited, opus_int32 *packet_offset, int soft_clip) +{ + int i, nb_samples; + int count, offset; + unsigned char toc; + int packet_frame_size, packet_bandwidth, packet_mode, packet_stream_channels; + /* 48 x 2.5 ms = 120 ms */ + opus_int16 size[48]; + VALIDATE_OPUS_DECODER(st); + if (decode_fec<0 || decode_fec>1) + return OPUS_BAD_ARG; + /* For FEC/PLC, frame_size has to be to have a multiple of 2.5 ms */ + if ((decode_fec || len==0 || data==NULL) && frame_size%(st->Fs/400)!=0) + return OPUS_BAD_ARG; + if (len==0 || data==NULL) + { + int pcm_count=0; + do { + int ret; + ret = opus_decode_frame(st, NULL, 0, pcm+pcm_count*st->channels, frame_size-pcm_count, 0); + if (ret<0) + return ret; + pcm_count += ret; + } while (pcm_count < frame_size); + celt_assert(pcm_count == frame_size); + if (OPUS_CHECK_ARRAY(pcm, pcm_count*st->channels)) + OPUS_PRINT_INT(pcm_count); + st->last_packet_duration = pcm_count; + return pcm_count; + } else if (len<0) + return OPUS_BAD_ARG; + + packet_mode = opus_packet_get_mode(data); + packet_bandwidth = opus_packet_get_bandwidth(data); + packet_frame_size = opus_packet_get_samples_per_frame(data, st->Fs); + packet_stream_channels = opus_packet_get_nb_channels(data); + + count = opus_packet_parse_impl(data, len, self_delimited, &toc, NULL, + size, &offset, packet_offset); + if (count<0) + return count; + + data += offset; + + if (decode_fec) + { + int duration_copy; + int ret; + /* If no FEC can be present, run the PLC (recursive call) */ + if (frame_size < packet_frame_size || packet_mode == MODE_CELT_ONLY || st->mode == MODE_CELT_ONLY) + return opus_decode_native(st, NULL, 0, pcm, frame_size, 0, 0, NULL, soft_clip); + /* Otherwise, run the PLC on everything except the size for which we might have FEC */ + duration_copy = st->last_packet_duration; + if (frame_size-packet_frame_size!=0) + { + ret = opus_decode_native(st, NULL, 0, pcm, frame_size-packet_frame_size, 0, 0, NULL, soft_clip); + if (ret<0) + { + st->last_packet_duration = duration_copy; + return ret; + } + celt_assert(ret==frame_size-packet_frame_size); + } + /* Complete with FEC */ + st->mode = packet_mode; + st->bandwidth = packet_bandwidth; + st->frame_size = packet_frame_size; + st->stream_channels = packet_stream_channels; + ret = opus_decode_frame(st, data, size[0], pcm+st->channels*(frame_size-packet_frame_size), + packet_frame_size, 1); + if (ret<0) + return ret; + else { + if (OPUS_CHECK_ARRAY(pcm, frame_size*st->channels)) + OPUS_PRINT_INT(frame_size); + st->last_packet_duration = frame_size; + return frame_size; + } + } + + if (count*packet_frame_size > frame_size) + return OPUS_BUFFER_TOO_SMALL; + + /* Update the state as the last step to avoid updating it on an invalid packet */ + st->mode = packet_mode; + st->bandwidth = packet_bandwidth; + st->frame_size = packet_frame_size; + st->stream_channels = packet_stream_channels; + + nb_samples=0; + for (i=0;ichannels, frame_size-nb_samples, 0); + if (ret<0) + return ret; + celt_assert(ret==packet_frame_size); + data += size[i]; + nb_samples += ret; + } + st->last_packet_duration = nb_samples; + if (OPUS_CHECK_ARRAY(pcm, nb_samples*st->channels)) + OPUS_PRINT_INT(nb_samples); +#ifndef FIXED_POINT + if (soft_clip) + opus_pcm_soft_clip(pcm, nb_samples, st->channels, st->softclip_mem); + else + st->softclip_mem[0]=st->softclip_mem[1]=0; +#endif + return nb_samples; +} + +#ifdef FIXED_POINT + +int opus_decode(OpusDecoder *st, const unsigned char *data, + opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) +{ + if(frame_size<=0) + return OPUS_BAD_ARG; + return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL, 0); +} + +#ifndef DISABLE_FLOAT_API +int opus_decode_float(OpusDecoder *st, const unsigned char *data, + opus_int32 len, float *pcm, int frame_size, int decode_fec) +{ + VARDECL(opus_int16, out); + int ret, i; + int nb_samples; + ALLOC_STACK; + + if(frame_size<=0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + if (data != NULL && len > 0 && !decode_fec) + { + nb_samples = opus_decoder_get_nb_samples(st, data, len); + if (nb_samples>0) + frame_size = IMIN(frame_size, nb_samples); + else + return OPUS_INVALID_PACKET; + } + celt_assert(st->channels == 1 || st->channels == 2); + ALLOC(out, frame_size*st->channels, opus_int16); + + ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL, 0); + if (ret > 0) + { + for (i=0;ichannels;i++) + pcm[i] = (1.f/32768.f)*(out[i]); + } + RESTORE_STACK; + return ret; +} +#endif + + +#else +int opus_decode(OpusDecoder *st, const unsigned char *data, + opus_int32 len, opus_int16 *pcm, int frame_size, int decode_fec) +{ + VARDECL(float, out); + int ret, i; + int nb_samples; + ALLOC_STACK; + + if(frame_size<=0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + + if (data != NULL && len > 0 && !decode_fec) + { + nb_samples = opus_decoder_get_nb_samples(st, data, len); + if (nb_samples>0) + frame_size = IMIN(frame_size, nb_samples); + else + return OPUS_INVALID_PACKET; + } + celt_assert(st->channels == 1 || st->channels == 2); + ALLOC(out, frame_size*st->channels, float); + + ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL, 1); + if (ret > 0) + { + for (i=0;ichannels;i++) + pcm[i] = FLOAT2INT16(out[i]); + } + RESTORE_STACK; + return ret; +} + +int opus_decode_float(OpusDecoder *st, const unsigned char *data, + opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) +{ + if(frame_size<=0) + return OPUS_BAD_ARG; + return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL, 0); +} + +#endif + +int opus_decoder_ctl(OpusDecoder *st, int request, ...) +{ + int ret = OPUS_OK; + va_list ap; + void *silk_dec; + CELTDecoder *celt_dec; + + silk_dec = (char*)st+st->silk_dec_offset; + celt_dec = (CELTDecoder*)((char*)st+st->celt_dec_offset); + + + va_start(ap, request); + + switch (request) + { + case OPUS_GET_BANDWIDTH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->bandwidth; + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + opus_uint32 *value = va_arg(ap, opus_uint32*); + if (!value) + { + goto bad_arg; + } + *value = st->rangeFinal; + } + break; + case OPUS_RESET_STATE: + { + OPUS_CLEAR((char*)&st->OPUS_DECODER_RESET_START, + sizeof(OpusDecoder)- + ((char*)&st->OPUS_DECODER_RESET_START - (char*)st)); + + celt_decoder_ctl(celt_dec, OPUS_RESET_STATE); + silk_InitDecoder( silk_dec ); + st->stream_channels = st->channels; + st->frame_size = st->Fs/400; + } + break; + case OPUS_GET_SAMPLE_RATE_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->Fs; + } + break; + case OPUS_GET_PITCH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + if (st->prev_mode == MODE_CELT_ONLY) + ret = celt_decoder_ctl(celt_dec, OPUS_GET_PITCH(value)); + else + *value = st->DecControl.prevPitchLag; + } + break; + case OPUS_GET_GAIN_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->decode_gain; + } + break; + case OPUS_SET_GAIN_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<-32768 || value>32767) + { + goto bad_arg; + } + st->decode_gain = value; + } + break; + case OPUS_GET_LAST_PACKET_DURATION_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->last_packet_duration; + } + break; + case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>1) + { + goto bad_arg; + } + ret = celt_decoder_ctl(celt_dec, OPUS_SET_PHASE_INVERSION_DISABLED(value)); + } + break; + case OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + ret = celt_decoder_ctl(celt_dec, OPUS_GET_PHASE_INVERSION_DISABLED(value)); + } + break; + default: + /*fprintf(stderr, "unknown opus_decoder_ctl() request: %d", request);*/ + ret = OPUS_UNIMPLEMENTED; + break; + } + + va_end(ap); + return ret; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +} + +void opus_decoder_destroy(OpusDecoder *st) +{ + opus_free(st); +} + + +int opus_packet_get_bandwidth(const unsigned char *data) +{ + int bandwidth; + if (data[0]&0x80) + { + bandwidth = OPUS_BANDWIDTH_MEDIUMBAND + ((data[0]>>5)&0x3); + if (bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) + bandwidth = OPUS_BANDWIDTH_NARROWBAND; + } else if ((data[0]&0x60) == 0x60) + { + bandwidth = (data[0]&0x10) ? OPUS_BANDWIDTH_FULLBAND : + OPUS_BANDWIDTH_SUPERWIDEBAND; + } else { + bandwidth = OPUS_BANDWIDTH_NARROWBAND + ((data[0]>>5)&0x3); + } + return bandwidth; +} + +int opus_packet_get_nb_channels(const unsigned char *data) +{ + return (data[0]&0x4) ? 2 : 1; +} + +int opus_packet_get_nb_frames(const unsigned char packet[], opus_int32 len) +{ + int count; + if (len<1) + return OPUS_BAD_ARG; + count = packet[0]&0x3; + if (count==0) + return 1; + else if (count!=3) + return 2; + else if (len<2) + return OPUS_INVALID_PACKET; + else + return packet[1]&0x3F; +} + +int opus_packet_get_nb_samples(const unsigned char packet[], opus_int32 len, + opus_int32 Fs) +{ + int samples; + int count = opus_packet_get_nb_frames(packet, len); + + if (count<0) + return count; + + samples = count*opus_packet_get_samples_per_frame(packet, Fs); + /* Can't have more than 120 ms */ + if (samples*25 > Fs*3) + return OPUS_INVALID_PACKET; + else + return samples; +} + +int opus_decoder_get_nb_samples(const OpusDecoder *dec, + const unsigned char packet[], opus_int32 len) +{ + return opus_packet_get_nb_samples(packet, len, dec->Fs); +} diff --git a/lib/opus/src/opus_demo.c b/lib/opus/src/opus_demo.c new file mode 100644 index 00000000..4cc26a6c --- /dev/null +++ b/lib/opus/src/opus_demo.c @@ -0,0 +1,892 @@ +/* Copyright (c) 2007-2008 CSIRO + Copyright (c) 2007-2009 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include +#include "opus.h" +#include "debug.h" +#include "opus_types.h" +#include "opus_private.h" +#include "opus_multistream.h" + +#define MAX_PACKET 1500 + +void print_usage( char* argv[] ) +{ + fprintf(stderr, "Usage: %s [-e] " + " [options] \n", argv[0]); + fprintf(stderr, " %s -d " + "[options] \n\n", argv[0]); + fprintf(stderr, "application: voip | audio | restricted-lowdelay\n" ); + fprintf(stderr, "options:\n" ); + fprintf(stderr, "-e : only runs the encoder (output the bit-stream)\n" ); + fprintf(stderr, "-d : only runs the decoder (reads the bit-stream as input)\n" ); + fprintf(stderr, "-cbr : enable constant bitrate; default: variable bitrate\n" ); + fprintf(stderr, "-cvbr : enable constrained variable bitrate; default: unconstrained\n" ); + fprintf(stderr, "-delayed-decision : use look-ahead for speech/music detection (experts only); default: disabled\n" ); + fprintf(stderr, "-bandwidth : audio bandwidth (from narrowband to fullband); default: sampling rate\n" ); + fprintf(stderr, "-framesize <2.5|5|10|20|40|60|80|100|120> : frame size in ms; default: 20 \n" ); + fprintf(stderr, "-max_payload : maximum payload size in bytes, default: 1024\n" ); + fprintf(stderr, "-complexity : complexity, 0 (lowest) ... 10 (highest); default: 10\n" ); + fprintf(stderr, "-inbandfec : enable SILK inband FEC\n" ); + fprintf(stderr, "-forcemono : force mono encoding, even for stereo input\n" ); + fprintf(stderr, "-dtx : enable SILK DTX\n" ); + fprintf(stderr, "-loss : simulate packet loss, in percent (0-100); default: 0\n" ); +} + +static void int_to_char(opus_uint32 i, unsigned char ch[4]) +{ + ch[0] = i>>24; + ch[1] = (i>>16)&0xFF; + ch[2] = (i>>8)&0xFF; + ch[3] = i&0xFF; +} + +static opus_uint32 char_to_int(unsigned char ch[4]) +{ + return ((opus_uint32)ch[0]<<24) | ((opus_uint32)ch[1]<<16) + | ((opus_uint32)ch[2]<< 8) | (opus_uint32)ch[3]; +} + +#define check_encoder_option(decode_only, opt) do {if (decode_only) {fprintf(stderr, "option %s is only for encoding\n", opt); goto failure;}} while(0) + +static const int silk8_test[][4] = { + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960*3, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960*2, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 480, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960*3, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960*2, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_NARROWBAND, 480, 2} +}; + +static const int silk12_test[][4] = { + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 960*3, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 960*2, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 960, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 480, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 960*3, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 960*2, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 960, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_MEDIUMBAND, 480, 2} +}; + +static const int silk16_test[][4] = { + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960*3, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960*2, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 480, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960*3, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960*2, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_WIDEBAND, 480, 2} +}; + +static const int hybrid24_test[][4] = { + {MODE_SILK_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 960, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 480, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 960, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 480, 2} +}; + +static const int hybrid48_test[][4] = { + {MODE_SILK_ONLY, OPUS_BANDWIDTH_FULLBAND, 960, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_FULLBAND, 480, 1}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_FULLBAND, 960, 2}, + {MODE_SILK_ONLY, OPUS_BANDWIDTH_FULLBAND, 480, 2} +}; + +static const int celt_test[][4] = { + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 960, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 960, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960, 1}, + + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 480, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 480, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 480, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 480, 1}, + + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 240, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 240, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 240, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 240, 1}, + + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 120, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 120, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 120, 1}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 120, 1}, + + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 960, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 960, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 960, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 960, 2}, + + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 480, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 480, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 480, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 480, 2}, + + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 240, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 240, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 240, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 240, 2}, + + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 120, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_SUPERWIDEBAND, 120, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_WIDEBAND, 120, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_NARROWBAND, 120, 2}, + +}; + +static const int celt_hq_test[][4] = { + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 960, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 480, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 240, 2}, + {MODE_CELT_ONLY, OPUS_BANDWIDTH_FULLBAND, 120, 2}, +}; + +#if 0 /* This is a hack that replaces the normal encoder/decoder with the multistream version */ +#define OpusEncoder OpusMSEncoder +#define OpusDecoder OpusMSDecoder +#define opus_encode opus_multistream_encode +#define opus_decode opus_multistream_decode +#define opus_encoder_ctl opus_multistream_encoder_ctl +#define opus_decoder_ctl opus_multistream_decoder_ctl +#define opus_encoder_create ms_opus_encoder_create +#define opus_decoder_create ms_opus_decoder_create +#define opus_encoder_destroy opus_multistream_encoder_destroy +#define opus_decoder_destroy opus_multistream_decoder_destroy + +static OpusEncoder *ms_opus_encoder_create(opus_int32 Fs, int channels, int application, int *error) +{ + int streams, coupled_streams; + unsigned char mapping[256]; + return (OpusEncoder *)opus_multistream_surround_encoder_create(Fs, channels, 1, &streams, &coupled_streams, mapping, application, error); +} +static OpusDecoder *ms_opus_decoder_create(opus_int32 Fs, int channels, int *error) +{ + int streams; + int coupled_streams; + unsigned char mapping[256]={0,1}; + streams = 1; + coupled_streams = channels==2; + return (OpusDecoder *)opus_multistream_decoder_create(Fs, channels, streams, coupled_streams, mapping, error); +} +#endif + +int main(int argc, char *argv[]) +{ + int err; + char *inFile, *outFile; + FILE *fin=NULL; + FILE *fout=NULL; + OpusEncoder *enc=NULL; + OpusDecoder *dec=NULL; + int args; + int len[2]; + int frame_size, channels; + opus_int32 bitrate_bps=0; + unsigned char *data[2] = {NULL, NULL}; + unsigned char *fbytes=NULL; + opus_int32 sampling_rate; + int use_vbr; + int max_payload_bytes; + int complexity; + int use_inbandfec; + int use_dtx; + int forcechannels; + int cvbr = 0; + int packet_loss_perc; + opus_int32 count=0, count_act=0; + int k; + opus_int32 skip=0; + int stop=0; + short *in=NULL; + short *out=NULL; + int application=OPUS_APPLICATION_AUDIO; + double bits=0.0, bits_max=0.0, bits_act=0.0, bits2=0.0, nrg; + double tot_samples=0; + opus_uint64 tot_in, tot_out; + int bandwidth=OPUS_AUTO; + const char *bandwidth_string; + int lost = 0, lost_prev = 1; + int toggle = 0; + opus_uint32 enc_final_range[2]; + opus_uint32 dec_final_range; + int encode_only=0, decode_only=0; + int max_frame_size = 48000*2; + size_t num_read; + int curr_read=0; + int sweep_bps = 0; + int random_framesize=0, newsize=0, delayed_celt=0; + int sweep_max=0, sweep_min=0; + int random_fec=0; + const int (*mode_list)[4]=NULL; + int nb_modes_in_list=0; + int curr_mode=0; + int curr_mode_count=0; + int mode_switch_time = 48000; + int nb_encoded=0; + int remaining=0; + int variable_duration=OPUS_FRAMESIZE_ARG; + int delayed_decision=0; + int ret = EXIT_FAILURE; + + if (argc < 5 ) + { + print_usage( argv ); + goto failure; + } + + tot_in=tot_out=0; + fprintf(stderr, "%s\n", opus_get_version_string()); + + args = 1; + if (strcmp(argv[args], "-e")==0) + { + encode_only = 1; + args++; + } else if (strcmp(argv[args], "-d")==0) + { + decode_only = 1; + args++; + } + if (!decode_only && argc < 7 ) + { + print_usage( argv ); + goto failure; + } + + if (!decode_only) + { + if (strcmp(argv[args], "voip")==0) + application = OPUS_APPLICATION_VOIP; + else if (strcmp(argv[args], "restricted-lowdelay")==0) + application = OPUS_APPLICATION_RESTRICTED_LOWDELAY; + else if (strcmp(argv[args], "audio")!=0) { + fprintf(stderr, "unknown application: %s\n", argv[args]); + print_usage(argv); + goto failure; + } + args++; + } + sampling_rate = (opus_int32)atol(argv[args]); + args++; + + if (sampling_rate != 8000 && sampling_rate != 12000 + && sampling_rate != 16000 && sampling_rate != 24000 + && sampling_rate != 48000) + { + fprintf(stderr, "Supported sampling rates are 8000, 12000, " + "16000, 24000 and 48000.\n"); + goto failure; + } + frame_size = sampling_rate/50; + + channels = atoi(argv[args]); + args++; + + if (channels < 1 || channels > 2) + { + fprintf(stderr, "Opus_demo supports only 1 or 2 channels.\n"); + goto failure; + } + + if (!decode_only) + { + bitrate_bps = (opus_int32)atol(argv[args]); + args++; + } + + /* defaults: */ + use_vbr = 1; + max_payload_bytes = MAX_PACKET; + complexity = 10; + use_inbandfec = 0; + forcechannels = OPUS_AUTO; + use_dtx = 0; + packet_loss_perc = 0; + + while( args < argc - 2 ) { + /* process command line options */ + if( strcmp( argv[ args ], "-cbr" ) == 0 ) { + check_encoder_option(decode_only, "-cbr"); + use_vbr = 0; + args++; + } else if( strcmp( argv[ args ], "-bandwidth" ) == 0 ) { + check_encoder_option(decode_only, "-bandwidth"); + if (strcmp(argv[ args + 1 ], "NB")==0) + bandwidth = OPUS_BANDWIDTH_NARROWBAND; + else if (strcmp(argv[ args + 1 ], "MB")==0) + bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + else if (strcmp(argv[ args + 1 ], "WB")==0) + bandwidth = OPUS_BANDWIDTH_WIDEBAND; + else if (strcmp(argv[ args + 1 ], "SWB")==0) + bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; + else if (strcmp(argv[ args + 1 ], "FB")==0) + bandwidth = OPUS_BANDWIDTH_FULLBAND; + else { + fprintf(stderr, "Unknown bandwidth %s. " + "Supported are NB, MB, WB, SWB, FB.\n", + argv[ args + 1 ]); + goto failure; + } + args += 2; + } else if( strcmp( argv[ args ], "-framesize" ) == 0 ) { + check_encoder_option(decode_only, "-framesize"); + if (strcmp(argv[ args + 1 ], "2.5")==0) + frame_size = sampling_rate/400; + else if (strcmp(argv[ args + 1 ], "5")==0) + frame_size = sampling_rate/200; + else if (strcmp(argv[ args + 1 ], "10")==0) + frame_size = sampling_rate/100; + else if (strcmp(argv[ args + 1 ], "20")==0) + frame_size = sampling_rate/50; + else if (strcmp(argv[ args + 1 ], "40")==0) + frame_size = sampling_rate/25; + else if (strcmp(argv[ args + 1 ], "60")==0) + frame_size = 3*sampling_rate/50; + else if (strcmp(argv[ args + 1 ], "80")==0) + frame_size = 4*sampling_rate/50; + else if (strcmp(argv[ args + 1 ], "100")==0) + frame_size = 5*sampling_rate/50; + else if (strcmp(argv[ args + 1 ], "120")==0) + frame_size = 6*sampling_rate/50; + else { + fprintf(stderr, "Unsupported frame size: %s ms. " + "Supported are 2.5, 5, 10, 20, 40, 60, 80, 100, 120.\n", + argv[ args + 1 ]); + goto failure; + } + args += 2; + } else if( strcmp( argv[ args ], "-max_payload" ) == 0 ) { + check_encoder_option(decode_only, "-max_payload"); + max_payload_bytes = atoi( argv[ args + 1 ] ); + args += 2; + } else if( strcmp( argv[ args ], "-complexity" ) == 0 ) { + check_encoder_option(decode_only, "-complexity"); + complexity = atoi( argv[ args + 1 ] ); + args += 2; + } else if( strcmp( argv[ args ], "-inbandfec" ) == 0 ) { + use_inbandfec = 1; + args++; + } else if( strcmp( argv[ args ], "-forcemono" ) == 0 ) { + check_encoder_option(decode_only, "-forcemono"); + forcechannels = 1; + args++; + } else if( strcmp( argv[ args ], "-cvbr" ) == 0 ) { + check_encoder_option(decode_only, "-cvbr"); + cvbr = 1; + args++; + } else if( strcmp( argv[ args ], "-delayed-decision" ) == 0 ) { + check_encoder_option(decode_only, "-delayed-decision"); + delayed_decision = 1; + args++; + } else if( strcmp( argv[ args ], "-dtx") == 0 ) { + check_encoder_option(decode_only, "-dtx"); + use_dtx = 1; + args++; + } else if( strcmp( argv[ args ], "-loss" ) == 0 ) { + packet_loss_perc = atoi( argv[ args + 1 ] ); + args += 2; + } else if( strcmp( argv[ args ], "-sweep" ) == 0 ) { + check_encoder_option(decode_only, "-sweep"); + sweep_bps = atoi( argv[ args + 1 ] ); + args += 2; + } else if( strcmp( argv[ args ], "-random_framesize" ) == 0 ) { + check_encoder_option(decode_only, "-random_framesize"); + random_framesize = 1; + args++; + } else if( strcmp( argv[ args ], "-sweep_max" ) == 0 ) { + check_encoder_option(decode_only, "-sweep_max"); + sweep_max = atoi( argv[ args + 1 ] ); + args += 2; + } else if( strcmp( argv[ args ], "-random_fec" ) == 0 ) { + check_encoder_option(decode_only, "-random_fec"); + random_fec = 1; + args++; + } else if( strcmp( argv[ args ], "-silk8k_test" ) == 0 ) { + check_encoder_option(decode_only, "-silk8k_test"); + mode_list = silk8_test; + nb_modes_in_list = 8; + args++; + } else if( strcmp( argv[ args ], "-silk12k_test" ) == 0 ) { + check_encoder_option(decode_only, "-silk12k_test"); + mode_list = silk12_test; + nb_modes_in_list = 8; + args++; + } else if( strcmp( argv[ args ], "-silk16k_test" ) == 0 ) { + check_encoder_option(decode_only, "-silk16k_test"); + mode_list = silk16_test; + nb_modes_in_list = 8; + args++; + } else if( strcmp( argv[ args ], "-hybrid24k_test" ) == 0 ) { + check_encoder_option(decode_only, "-hybrid24k_test"); + mode_list = hybrid24_test; + nb_modes_in_list = 4; + args++; + } else if( strcmp( argv[ args ], "-hybrid48k_test" ) == 0 ) { + check_encoder_option(decode_only, "-hybrid48k_test"); + mode_list = hybrid48_test; + nb_modes_in_list = 4; + args++; + } else if( strcmp( argv[ args ], "-celt_test" ) == 0 ) { + check_encoder_option(decode_only, "-celt_test"); + mode_list = celt_test; + nb_modes_in_list = 32; + args++; + } else if( strcmp( argv[ args ], "-celt_hq_test" ) == 0 ) { + check_encoder_option(decode_only, "-celt_hq_test"); + mode_list = celt_hq_test; + nb_modes_in_list = 4; + args++; + } else { + printf( "Error: unrecognized setting: %s\n\n", argv[ args ] ); + print_usage( argv ); + goto failure; + } + } + + if (sweep_max) + sweep_min = bitrate_bps; + + if (max_payload_bytes < 0 || max_payload_bytes > MAX_PACKET) + { + fprintf (stderr, "max_payload_bytes must be between 0 and %d\n", + MAX_PACKET); + goto failure; + } + + inFile = argv[argc-2]; + fin = fopen(inFile, "rb"); + if (!fin) + { + fprintf (stderr, "Could not open input file %s\n", argv[argc-2]); + goto failure; + } + if (mode_list) + { + int size; + fseek(fin, 0, SEEK_END); + size = ftell(fin); + fprintf(stderr, "File size is %d bytes\n", size); + fseek(fin, 0, SEEK_SET); + mode_switch_time = size/sizeof(short)/channels/nb_modes_in_list; + fprintf(stderr, "Switching mode every %d samples\n", mode_switch_time); + } + + outFile = argv[argc-1]; + fout = fopen(outFile, "wb+"); + if (!fout) + { + fprintf (stderr, "Could not open output file %s\n", argv[argc-1]); + goto failure; + } + + if (!decode_only) + { + enc = opus_encoder_create(sampling_rate, channels, application, &err); + if (err != OPUS_OK) + { + fprintf(stderr, "Cannot create encoder: %s\n", opus_strerror(err)); + goto failure; + } + opus_encoder_ctl(enc, OPUS_SET_BITRATE(bitrate_bps)); + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(bandwidth)); + opus_encoder_ctl(enc, OPUS_SET_VBR(use_vbr)); + opus_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(cvbr)); + opus_encoder_ctl(enc, OPUS_SET_COMPLEXITY(complexity)); + opus_encoder_ctl(enc, OPUS_SET_INBAND_FEC(use_inbandfec)); + opus_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(forcechannels)); + opus_encoder_ctl(enc, OPUS_SET_DTX(use_dtx)); + opus_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(packet_loss_perc)); + + opus_encoder_ctl(enc, OPUS_GET_LOOKAHEAD(&skip)); + opus_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(16)); + opus_encoder_ctl(enc, OPUS_SET_EXPERT_FRAME_DURATION(variable_duration)); + } + if (!encode_only) + { + dec = opus_decoder_create(sampling_rate, channels, &err); + if (err != OPUS_OK) + { + fprintf(stderr, "Cannot create decoder: %s\n", opus_strerror(err)); + goto failure; + } + } + + + switch(bandwidth) + { + case OPUS_BANDWIDTH_NARROWBAND: + bandwidth_string = "narrowband"; + break; + case OPUS_BANDWIDTH_MEDIUMBAND: + bandwidth_string = "mediumband"; + break; + case OPUS_BANDWIDTH_WIDEBAND: + bandwidth_string = "wideband"; + break; + case OPUS_BANDWIDTH_SUPERWIDEBAND: + bandwidth_string = "superwideband"; + break; + case OPUS_BANDWIDTH_FULLBAND: + bandwidth_string = "fullband"; + break; + case OPUS_AUTO: + bandwidth_string = "auto bandwidth"; + break; + default: + bandwidth_string = "unknown"; + break; + } + + if (decode_only) + fprintf(stderr, "Decoding with %ld Hz output (%d channels)\n", + (long)sampling_rate, channels); + else + fprintf(stderr, "Encoding %ld Hz input at %.3f kb/s " + "in %s with %d-sample frames.\n", + (long)sampling_rate, bitrate_bps*0.001, + bandwidth_string, frame_size); + + in = (short*)malloc(max_frame_size*channels*sizeof(short)); + out = (short*)malloc(max_frame_size*channels*sizeof(short)); + /* We need to allocate for 16-bit PCM data, but we store it as unsigned char. */ + fbytes = (unsigned char*)malloc(max_frame_size*channels*sizeof(short)); + data[0] = (unsigned char*)calloc(max_payload_bytes,sizeof(unsigned char)); + if ( use_inbandfec ) { + data[1] = (unsigned char*)calloc(max_payload_bytes,sizeof(unsigned char)); + } + if(delayed_decision) + { + if (frame_size==sampling_rate/400) + variable_duration = OPUS_FRAMESIZE_2_5_MS; + else if (frame_size==sampling_rate/200) + variable_duration = OPUS_FRAMESIZE_5_MS; + else if (frame_size==sampling_rate/100) + variable_duration = OPUS_FRAMESIZE_10_MS; + else if (frame_size==sampling_rate/50) + variable_duration = OPUS_FRAMESIZE_20_MS; + else if (frame_size==sampling_rate/25) + variable_duration = OPUS_FRAMESIZE_40_MS; + else if (frame_size==3*sampling_rate/50) + variable_duration = OPUS_FRAMESIZE_60_MS; + else if (frame_size==4*sampling_rate/50) + variable_duration = OPUS_FRAMESIZE_80_MS; + else if (frame_size==5*sampling_rate/50) + variable_duration = OPUS_FRAMESIZE_100_MS; + else + variable_duration = OPUS_FRAMESIZE_120_MS; + opus_encoder_ctl(enc, OPUS_SET_EXPERT_FRAME_DURATION(variable_duration)); + frame_size = 2*48000; + } + while (!stop) + { + if (delayed_celt) + { + frame_size = newsize; + delayed_celt = 0; + } else if (random_framesize && rand()%20==0) + { + newsize = rand()%6; + switch(newsize) + { + case 0: newsize=sampling_rate/400; break; + case 1: newsize=sampling_rate/200; break; + case 2: newsize=sampling_rate/100; break; + case 3: newsize=sampling_rate/50; break; + case 4: newsize=sampling_rate/25; break; + case 5: newsize=3*sampling_rate/50; break; + } + while (newsize < sampling_rate/25 && bitrate_bps-abs(sweep_bps) <= 3*12*sampling_rate/newsize) + newsize*=2; + if (newsize < sampling_rate/100 && frame_size >= sampling_rate/100) + { + opus_encoder_ctl(enc, OPUS_SET_FORCE_MODE(MODE_CELT_ONLY)); + delayed_celt=1; + } else { + frame_size = newsize; + } + } + if (random_fec && rand()%30==0) + { + opus_encoder_ctl(enc, OPUS_SET_INBAND_FEC(rand()%4==0)); + } + if (decode_only) + { + unsigned char ch[4]; + num_read = fread(ch, 1, 4, fin); + if (num_read!=4) + break; + len[toggle] = char_to_int(ch); + if (len[toggle]>max_payload_bytes || len[toggle]<0) + { + fprintf(stderr, "Invalid payload length: %d\n",len[toggle]); + break; + } + num_read = fread(ch, 1, 4, fin); + if (num_read!=4) + break; + enc_final_range[toggle] = char_to_int(ch); + num_read = fread(data[toggle], 1, len[toggle], fin); + if (num_read!=(size_t)len[toggle]) + { + fprintf(stderr, "Ran out of input, " + "expecting %d bytes got %d\n", + len[toggle],(int)num_read); + break; + } + } else { + int i; + if (mode_list!=NULL) + { + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(mode_list[curr_mode][1])); + opus_encoder_ctl(enc, OPUS_SET_FORCE_MODE(mode_list[curr_mode][0])); + opus_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(mode_list[curr_mode][3])); + frame_size = mode_list[curr_mode][2]; + } + num_read = fread(fbytes, sizeof(short)*channels, frame_size-remaining, fin); + curr_read = (int)num_read; + tot_in += curr_read; + for(i=0;i sweep_max) + sweep_bps = -sweep_bps; + else if (bitrate_bps < sweep_min) + sweep_bps = -sweep_bps; + } + /* safety */ + if (bitrate_bps<1000) + bitrate_bps = 1000; + opus_encoder_ctl(enc, OPUS_SET_BITRATE(bitrate_bps)); + } + opus_encoder_ctl(enc, OPUS_GET_FINAL_RANGE(&enc_final_range[toggle])); + if (len[toggle] < 0) + { + fprintf (stderr, "opus_encode() returned %d\n", len[toggle]); + goto failure; + } + curr_mode_count += frame_size; + if (curr_mode_count > mode_switch_time && curr_mode < nb_modes_in_list-1) + { + curr_mode++; + curr_mode_count = 0; + } + } + +#if 0 /* This is for testing the padding code, do not enable by default */ + if (len[toggle]<1275) + { + int new_len = len[toggle]+rand()%(max_payload_bytes-len[toggle]); + if ((err = opus_packet_pad(data[toggle], len[toggle], new_len)) != OPUS_OK) + { + fprintf(stderr, "padding failed: %s\n", opus_strerror(err)); + goto failure; + } + len[toggle] = new_len; + } +#endif + if (encode_only) + { + unsigned char int_field[4]; + int_to_char(len[toggle], int_field); + if (fwrite(int_field, 1, 4, fout) != 4) { + fprintf(stderr, "Error writing.\n"); + goto failure; + } + int_to_char(enc_final_range[toggle], int_field); + if (fwrite(int_field, 1, 4, fout) != 4) { + fprintf(stderr, "Error writing.\n"); + goto failure; + } + if (fwrite(data[toggle], 1, len[toggle], fout) != (unsigned)len[toggle]) { + fprintf(stderr, "Error writing.\n"); + goto failure; + } + tot_samples += nb_encoded; + } else { + opus_int32 output_samples; + lost = len[toggle]==0 || (packet_loss_perc>0 && rand()%100 < packet_loss_perc); + if (lost) + opus_decoder_ctl(dec, OPUS_GET_LAST_PACKET_DURATION(&output_samples)); + else + output_samples = max_frame_size; + if( count >= use_inbandfec ) { + /* delay by one packet when using in-band FEC */ + if( use_inbandfec ) { + if( lost_prev ) { + /* attempt to decode with in-band FEC from next packet */ + opus_decoder_ctl(dec, OPUS_GET_LAST_PACKET_DURATION(&output_samples)); + output_samples = opus_decode(dec, lost ? NULL : data[toggle], len[toggle], out, output_samples, 1); + } else { + /* regular decode */ + output_samples = max_frame_size; + output_samples = opus_decode(dec, data[1-toggle], len[1-toggle], out, output_samples, 0); + } + } else { + output_samples = opus_decode(dec, lost ? NULL : data[toggle], len[toggle], out, output_samples, 0); + } + if (output_samples>0) + { + if (!decode_only && tot_out + output_samples > tot_in) + { + stop=1; + output_samples = (opus_int32)(tot_in - tot_out); + } + if (output_samples>skip) { + int i; + for(i=0;i<(output_samples-skip)*channels;i++) + { + short s; + s=out[i+(skip*channels)]; + fbytes[2*i]=s&0xFF; + fbytes[2*i+1]=(s>>8)&0xFF; + } + if (fwrite(fbytes, sizeof(short)*channels, output_samples-skip, fout) != (unsigned)(output_samples-skip)){ + fprintf(stderr, "Error writing.\n"); + goto failure; + } + tot_out += output_samples-skip; + } + if (output_samples= use_inbandfec ) { + /* count bits */ + bits += len[toggle]*8; + bits_max = ( len[toggle]*8 > bits_max ) ? len[toggle]*8 : bits_max; + bits2 += len[toggle]*len[toggle]*64; + if (!decode_only) + { + nrg = 0.0; + for ( k = 0; k < frame_size * channels; k++ ) { + nrg += in[ k ] * (double)in[ k ]; + } + nrg /= frame_size * channels; + if( nrg > 1e5 ) { + bits_act += len[toggle]*8; + count_act++; + } + } + } + count++; + toggle = (toggle + use_inbandfec) & 1; + } + + if(decode_only && count > 0) + frame_size = (int)(tot_samples / count); + count -= use_inbandfec; + if (tot_samples >= 1 && count > 0 && frame_size) + { + /* Print out bitrate statistics */ + double var; + fprintf (stderr, "average bitrate: %7.3f kb/s\n", + 1e-3*bits*sampling_rate/tot_samples); + fprintf (stderr, "maximum bitrate: %7.3f kb/s\n", + 1e-3*bits_max*sampling_rate/frame_size); + if (!decode_only) + fprintf (stderr, "active bitrate: %7.3f kb/s\n", + 1e-3*bits_act*sampling_rate/(1e-15+frame_size*(double)count_act)); + var = bits2/count - bits*bits/(count*(double)count); + if (var < 0) + var = 0; + fprintf (stderr, "bitrate standard deviation: %7.3f kb/s\n", + 1e-3*sqrt(var)*sampling_rate/frame_size); + } else { + fprintf(stderr, "bitrate statistics are undefined\n"); + } + silk_TimerSave("opus_timing.txt"); + ret = EXIT_SUCCESS; +failure: + opus_encoder_destroy(enc); + opus_decoder_destroy(dec); + free(data[0]); + free(data[1]); + if (fin) + fclose(fin); + if (fout) + fclose(fout); + free(in); + free(out); + free(fbytes); + return ret; +} diff --git a/lib/opus/src/opus_encoder.c b/lib/opus/src/opus_encoder.c new file mode 100644 index 00000000..8c8db5a5 --- /dev/null +++ b/lib/opus/src/opus_encoder.c @@ -0,0 +1,2780 @@ +/* Copyright (c) 2010-2011 Xiph.Org Foundation, Skype Limited + Written by Jean-Marc Valin and Koen Vos */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include "celt.h" +#include "entenc.h" +#include "modes.h" +#include "API.h" +#include "stack_alloc.h" +#include "float_cast.h" +#include "opus.h" +#include "arch.h" +#include "pitch.h" +#include "opus_private.h" +#include "os_support.h" +#include "cpu_support.h" +#include "analysis.h" +#include "mathops.h" +#include "tuning_parameters.h" +#ifdef FIXED_POINT +#include "fixed/structs_FIX.h" +#else +#include "float/structs_FLP.h" +#endif + +#define MAX_ENCODER_BUFFER 480 + +#ifndef DISABLE_FLOAT_API +#define PSEUDO_SNR_THRESHOLD 316.23f /* 10^(25/10) */ +#endif + +typedef struct { + opus_val32 XX, XY, YY; + opus_val16 smoothed_width; + opus_val16 max_follower; +} StereoWidthState; + +struct OpusEncoder { + int celt_enc_offset; + int silk_enc_offset; + silk_EncControlStruct silk_mode; + int application; + int channels; + int delay_compensation; + int force_channels; + int signal_type; + int user_bandwidth; + int max_bandwidth; + int user_forced_mode; + int voice_ratio; + opus_int32 Fs; + int use_vbr; + int vbr_constraint; + int variable_duration; + opus_int32 bitrate_bps; + opus_int32 user_bitrate_bps; + int lsb_depth; + int encoder_buffer; + int lfe; + int arch; + int use_dtx; /* general DTX for both SILK and CELT */ + int fec_config; +#ifndef DISABLE_FLOAT_API + TonalityAnalysisState analysis; +#endif + +#define OPUS_ENCODER_RESET_START stream_channels + int stream_channels; + opus_int16 hybrid_stereo_width_Q14; + opus_int32 variable_HP_smth2_Q15; + opus_val16 prev_HB_gain; + opus_val32 hp_mem[4]; + int mode; + int prev_mode; + int prev_channels; + int prev_framesize; + int bandwidth; + /* Bandwidth determined automatically from the rate (before any other adjustment) */ + int auto_bandwidth; + int silk_bw_switch; + /* Sampling rate (at the API level) */ + int first; + opus_val16 * energy_masking; + StereoWidthState width_mem; + opus_val16 delay_buffer[MAX_ENCODER_BUFFER*2]; +#ifndef DISABLE_FLOAT_API + int detected_bandwidth; + int nb_no_activity_ms_Q1; + opus_val32 peak_signal_energy; +#endif + int nonfinal_frame; /* current frame is not the final in a packet */ + opus_uint32 rangeFinal; +}; + +/* Transition tables for the voice and music. First column is the + middle (memoriless) threshold. The second column is the hysteresis + (difference with the middle) */ +static const opus_int32 mono_voice_bandwidth_thresholds[8] = { + 9000, 700, /* NB<->MB */ + 9000, 700, /* MB<->WB */ + 13500, 1000, /* WB<->SWB */ + 14000, 2000, /* SWB<->FB */ +}; +static const opus_int32 mono_music_bandwidth_thresholds[8] = { + 9000, 700, /* NB<->MB */ + 9000, 700, /* MB<->WB */ + 11000, 1000, /* WB<->SWB */ + 12000, 2000, /* SWB<->FB */ +}; +static const opus_int32 stereo_voice_bandwidth_thresholds[8] = { + 9000, 700, /* NB<->MB */ + 9000, 700, /* MB<->WB */ + 13500, 1000, /* WB<->SWB */ + 14000, 2000, /* SWB<->FB */ +}; +static const opus_int32 stereo_music_bandwidth_thresholds[8] = { + 9000, 700, /* NB<->MB */ + 9000, 700, /* MB<->WB */ + 11000, 1000, /* WB<->SWB */ + 12000, 2000, /* SWB<->FB */ +}; +/* Threshold bit-rates for switching between mono and stereo */ +static const opus_int32 stereo_voice_threshold = 19000; +static const opus_int32 stereo_music_threshold = 17000; + +/* Threshold bit-rate for switching between SILK/hybrid and CELT-only */ +static const opus_int32 mode_thresholds[2][2] = { + /* voice */ /* music */ + { 64000, 10000}, /* mono */ + { 44000, 10000}, /* stereo */ +}; + +static const opus_int32 fec_thresholds[] = { + 12000, 1000, /* NB */ + 14000, 1000, /* MB */ + 16000, 1000, /* WB */ + 20000, 1000, /* SWB */ + 22000, 1000, /* FB */ +}; + +int opus_encoder_get_size(int channels) +{ + int silkEncSizeBytes, celtEncSizeBytes; + int ret; + if (channels<1 || channels > 2) + return 0; + ret = silk_Get_Encoder_Size( &silkEncSizeBytes ); + if (ret) + return 0; + silkEncSizeBytes = align(silkEncSizeBytes); + celtEncSizeBytes = celt_encoder_get_size(channels); + return align(sizeof(OpusEncoder))+silkEncSizeBytes+celtEncSizeBytes; +} + +int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int application) +{ + void *silk_enc; + CELTEncoder *celt_enc; + int err; + int ret, silkEncSizeBytes; + + if((Fs!=48000&&Fs!=24000&&Fs!=16000&&Fs!=12000&&Fs!=8000)||(channels!=1&&channels!=2)|| + (application != OPUS_APPLICATION_VOIP && application != OPUS_APPLICATION_AUDIO + && application != OPUS_APPLICATION_RESTRICTED_LOWDELAY)) + return OPUS_BAD_ARG; + + OPUS_CLEAR((char*)st, opus_encoder_get_size(channels)); + /* Create SILK encoder */ + ret = silk_Get_Encoder_Size( &silkEncSizeBytes ); + if (ret) + return OPUS_BAD_ARG; + silkEncSizeBytes = align(silkEncSizeBytes); + st->silk_enc_offset = align(sizeof(OpusEncoder)); + st->celt_enc_offset = st->silk_enc_offset+silkEncSizeBytes; + silk_enc = (char*)st+st->silk_enc_offset; + celt_enc = (CELTEncoder*)((char*)st+st->celt_enc_offset); + + st->stream_channels = st->channels = channels; + + st->Fs = Fs; + + st->arch = opus_select_arch(); + + ret = silk_InitEncoder( silk_enc, st->arch, &st->silk_mode ); + if(ret)return OPUS_INTERNAL_ERROR; + + /* default SILK parameters */ + st->silk_mode.nChannelsAPI = channels; + st->silk_mode.nChannelsInternal = channels; + st->silk_mode.API_sampleRate = st->Fs; + st->silk_mode.maxInternalSampleRate = 16000; + st->silk_mode.minInternalSampleRate = 8000; + st->silk_mode.desiredInternalSampleRate = 16000; + st->silk_mode.payloadSize_ms = 20; + st->silk_mode.bitRate = 25000; + st->silk_mode.packetLossPercentage = 0; + st->silk_mode.complexity = 9; + st->silk_mode.useInBandFEC = 0; + st->silk_mode.useDTX = 0; + st->silk_mode.useCBR = 0; + st->silk_mode.reducedDependency = 0; + + /* Create CELT encoder */ + /* Initialize CELT encoder */ + err = celt_encoder_init(celt_enc, Fs, channels, st->arch); + if(err!=OPUS_OK)return OPUS_INTERNAL_ERROR; + + celt_encoder_ctl(celt_enc, CELT_SET_SIGNALLING(0)); + celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(st->silk_mode.complexity)); + + st->use_vbr = 1; + /* Makes constrained VBR the default (safer for real-time use) */ + st->vbr_constraint = 1; + st->user_bitrate_bps = OPUS_AUTO; + st->bitrate_bps = 3000+Fs*channels; + st->application = application; + st->signal_type = OPUS_AUTO; + st->user_bandwidth = OPUS_AUTO; + st->max_bandwidth = OPUS_BANDWIDTH_FULLBAND; + st->force_channels = OPUS_AUTO; + st->user_forced_mode = OPUS_AUTO; + st->voice_ratio = -1; + st->encoder_buffer = st->Fs/100; + st->lsb_depth = 24; + st->variable_duration = OPUS_FRAMESIZE_ARG; + + /* Delay compensation of 4 ms (2.5 ms for SILK's extra look-ahead + + 1.5 ms for SILK resamplers and stereo prediction) */ + st->delay_compensation = st->Fs/250; + + st->hybrid_stereo_width_Q14 = 1 << 14; + st->prev_HB_gain = Q15ONE; + st->variable_HP_smth2_Q15 = silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 ); + st->first = 1; + st->mode = MODE_HYBRID; + st->bandwidth = OPUS_BANDWIDTH_FULLBAND; + +#ifndef DISABLE_FLOAT_API + tonality_analysis_init(&st->analysis, st->Fs); + st->analysis.application = st->application; +#endif + + return OPUS_OK; +} + +static unsigned char gen_toc(int mode, int framerate, int bandwidth, int channels) +{ + int period; + unsigned char toc; + period = 0; + while (framerate < 400) + { + framerate <<= 1; + period++; + } + if (mode == MODE_SILK_ONLY) + { + toc = (bandwidth-OPUS_BANDWIDTH_NARROWBAND)<<5; + toc |= (period-2)<<3; + } else if (mode == MODE_CELT_ONLY) + { + int tmp = bandwidth-OPUS_BANDWIDTH_MEDIUMBAND; + if (tmp < 0) + tmp = 0; + toc = 0x80; + toc |= tmp << 5; + toc |= period<<3; + } else /* Hybrid */ + { + toc = 0x60; + toc |= (bandwidth-OPUS_BANDWIDTH_SUPERWIDEBAND)<<4; + toc |= (period-2)<<3; + } + toc |= (channels==2)<<2; + return toc; +} + +#ifndef FIXED_POINT +static void silk_biquad_float( + const opus_val16 *in, /* I: Input signal */ + const opus_int32 *B_Q28, /* I: MA coefficients [3] */ + const opus_int32 *A_Q28, /* I: AR coefficients [2] */ + opus_val32 *S, /* I/O: State vector [2] */ + opus_val16 *out, /* O: Output signal */ + const opus_int32 len, /* I: Signal length (must be even) */ + int stride +) +{ + /* DIRECT FORM II TRANSPOSED (uses 2 element state vector) */ + opus_int k; + opus_val32 vout; + opus_val32 inval; + opus_val32 A[2], B[3]; + + A[0] = (opus_val32)(A_Q28[0] * (1.f/((opus_int32)1<<28))); + A[1] = (opus_val32)(A_Q28[1] * (1.f/((opus_int32)1<<28))); + B[0] = (opus_val32)(B_Q28[0] * (1.f/((opus_int32)1<<28))); + B[1] = (opus_val32)(B_Q28[1] * (1.f/((opus_int32)1<<28))); + B[2] = (opus_val32)(B_Q28[2] * (1.f/((opus_int32)1<<28))); + + /* Negate A_Q28 values and split in two parts */ + + for( k = 0; k < len; k++ ) { + /* S[ 0 ], S[ 1 ]: Q12 */ + inval = in[ k*stride ]; + vout = S[ 0 ] + B[0]*inval; + + S[ 0 ] = S[1] - vout*A[0] + B[1]*inval; + + S[ 1 ] = - vout*A[1] + B[2]*inval + VERY_SMALL; + + /* Scale back to Q0 and saturate */ + out[ k*stride ] = vout; + } +} +#endif + +static void hp_cutoff(const opus_val16 *in, opus_int32 cutoff_Hz, opus_val16 *out, opus_val32 *hp_mem, int len, int channels, opus_int32 Fs, int arch) +{ + opus_int32 B_Q28[ 3 ], A_Q28[ 2 ]; + opus_int32 Fc_Q19, r_Q28, r_Q22; + (void)arch; + + silk_assert( cutoff_Hz <= silk_int32_MAX / SILK_FIX_CONST( 1.5 * 3.14159 / 1000, 19 ) ); + Fc_Q19 = silk_DIV32_16( silk_SMULBB( SILK_FIX_CONST( 1.5 * 3.14159 / 1000, 19 ), cutoff_Hz ), Fs/1000 ); + silk_assert( Fc_Q19 > 0 && Fc_Q19 < 32768 ); + + r_Q28 = SILK_FIX_CONST( 1.0, 28 ) - silk_MUL( SILK_FIX_CONST( 0.92, 9 ), Fc_Q19 ); + + /* b = r * [ 1; -2; 1 ]; */ + /* a = [ 1; -2 * r * ( 1 - 0.5 * Fc^2 ); r^2 ]; */ + B_Q28[ 0 ] = r_Q28; + B_Q28[ 1 ] = silk_LSHIFT( -r_Q28, 1 ); + B_Q28[ 2 ] = r_Q28; + + /* -r * ( 2 - Fc * Fc ); */ + r_Q22 = silk_RSHIFT( r_Q28, 6 ); + A_Q28[ 0 ] = silk_SMULWW( r_Q22, silk_SMULWW( Fc_Q19, Fc_Q19 ) - SILK_FIX_CONST( 2.0, 22 ) ); + A_Q28[ 1 ] = silk_SMULWW( r_Q22, r_Q22 ); + +#ifdef FIXED_POINT + if( channels == 1 ) { + silk_biquad_alt_stride1( in, B_Q28, A_Q28, hp_mem, out, len ); + } else { + silk_biquad_alt_stride2( in, B_Q28, A_Q28, hp_mem, out, len, arch ); + } +#else + silk_biquad_float( in, B_Q28, A_Q28, hp_mem, out, len, channels ); + if( channels == 2 ) { + silk_biquad_float( in+1, B_Q28, A_Q28, hp_mem+2, out+1, len, channels ); + } +#endif +} + +#ifdef FIXED_POINT +static void dc_reject(const opus_val16 *in, opus_int32 cutoff_Hz, opus_val16 *out, opus_val32 *hp_mem, int len, int channels, opus_int32 Fs) +{ + int c, i; + int shift; + + /* Approximates -round(log2(6.3*cutoff_Hz/Fs)) */ + shift=celt_ilog2(Fs/(cutoff_Hz*4)); + for (c=0;cFs/400; + if (st->user_bitrate_bps==OPUS_AUTO) + return 60*st->Fs/frame_size + st->Fs*st->channels; + else if (st->user_bitrate_bps==OPUS_BITRATE_MAX) + return max_data_bytes*8*st->Fs/frame_size; + else + return st->user_bitrate_bps; +} + +#ifndef DISABLE_FLOAT_API +#ifdef FIXED_POINT +#define PCM2VAL(x) FLOAT2INT16(x) +#else +#define PCM2VAL(x) SCALEIN(x) +#endif + +void downmix_float(const void *_x, opus_val32 *y, int subframe, int offset, int c1, int c2, int C) +{ + const float *x; + int j; + + x = (const float *)_x; + for (j=0;j-1) + { + for (j=0;j-1) + { + for (j=0;j= OPUS_FRAMESIZE_2_5_MS && variable_duration <= OPUS_FRAMESIZE_120_MS) + { + if (variable_duration <= OPUS_FRAMESIZE_40_MS) + new_size = (Fs/400)<<(variable_duration-OPUS_FRAMESIZE_2_5_MS); + else + new_size = (variable_duration-OPUS_FRAMESIZE_2_5_MS-2)*Fs/50; + } + else + return -1; + if (new_size>frame_size) + return -1; + if (400*new_size!=Fs && 200*new_size!=Fs && 100*new_size!=Fs && + 50*new_size!=Fs && 25*new_size!=Fs && 50*new_size!=3*Fs && + 50*new_size!=4*Fs && 50*new_size!=5*Fs && 50*new_size!=6*Fs) + return -1; + return new_size; +} + +opus_val16 compute_stereo_width(const opus_val16 *pcm, int frame_size, opus_int32 Fs, StereoWidthState *mem) +{ + opus_val32 xx, xy, yy; + opus_val16 sqrt_xx, sqrt_yy; + opus_val16 qrrt_xx, qrrt_yy; + int frame_rate; + int i; + opus_val16 short_alpha; + + frame_rate = Fs/frame_size; + short_alpha = Q15ONE - MULT16_16(25, Q15ONE)/IMAX(50,frame_rate); + xx=xy=yy=0; + /* Unroll by 4. The frame size is always a multiple of 4 *except* for + 2.5 ms frames at 12 kHz. Since this setting is very rare (and very + stupid), we just discard the last two samples. */ + for (i=0;iXX += MULT16_32_Q15(short_alpha, xx-mem->XX); + mem->XY += MULT16_32_Q15(short_alpha, xy-mem->XY); + mem->YY += MULT16_32_Q15(short_alpha, yy-mem->YY); + mem->XX = MAX32(0, mem->XX); + mem->XY = MAX32(0, mem->XY); + mem->YY = MAX32(0, mem->YY); + if (MAX32(mem->XX, mem->YY)>QCONST16(8e-4f, 18)) + { + opus_val16 corr; + opus_val16 ldiff; + opus_val16 width; + sqrt_xx = celt_sqrt(mem->XX); + sqrt_yy = celt_sqrt(mem->YY); + qrrt_xx = celt_sqrt(sqrt_xx); + qrrt_yy = celt_sqrt(sqrt_yy); + /* Inter-channel correlation */ + mem->XY = MIN32(mem->XY, sqrt_xx*sqrt_yy); + corr = SHR32(frac_div32(mem->XY,EPSILON+MULT16_16(sqrt_xx,sqrt_yy)),16); + /* Approximate loudness difference */ + ldiff = MULT16_16(Q15ONE, ABS16(qrrt_xx-qrrt_yy))/(EPSILON+qrrt_xx+qrrt_yy); + width = MULT16_16_Q15(celt_sqrt(QCONST32(1.f,30)-MULT16_16(corr,corr)), ldiff); + /* Smoothing over one second */ + mem->smoothed_width += (width-mem->smoothed_width)/frame_rate; + /* Peak follower */ + mem->max_follower = MAX16(mem->max_follower-QCONST16(.02f,15)/frame_rate, mem->smoothed_width); + } + /*printf("%f %f %f %f %f ", corr/(float)Q15ONE, ldiff/(float)Q15ONE, width/(float)Q15ONE, mem->smoothed_width/(float)Q15ONE, mem->max_follower/(float)Q15ONE);*/ + return EXTRACT16(MIN32(Q15ONE, MULT16_16(20, mem->max_follower))); +} + +static int decide_fec(int useInBandFEC, int PacketLoss_perc, int last_fec, int mode, int *bandwidth, opus_int32 rate) +{ + int orig_bandwidth; + if (!useInBandFEC || PacketLoss_perc == 0 || mode == MODE_CELT_ONLY) + return 0; + orig_bandwidth = *bandwidth; + for (;;) + { + opus_int32 hysteresis; + opus_int32 LBRR_rate_thres_bps; + /* Compute threshold for using FEC at the current bandwidth setting */ + LBRR_rate_thres_bps = fec_thresholds[2*(*bandwidth - OPUS_BANDWIDTH_NARROWBAND)]; + hysteresis = fec_thresholds[2*(*bandwidth - OPUS_BANDWIDTH_NARROWBAND) + 1]; + if (last_fec == 1) LBRR_rate_thres_bps -= hysteresis; + if (last_fec == 0) LBRR_rate_thres_bps += hysteresis; + LBRR_rate_thres_bps = silk_SMULWB( silk_MUL( LBRR_rate_thres_bps, + 125 - silk_min( PacketLoss_perc, 25 ) ), SILK_FIX_CONST( 0.01, 16 ) ); + /* If loss <= 5%, we look at whether we have enough rate to enable FEC. + If loss > 5%, we decrease the bandwidth until we can enable FEC. */ + if (rate > LBRR_rate_thres_bps) + return 1; + else if (PacketLoss_perc <= 5) + return 0; + else if (*bandwidth > OPUS_BANDWIDTH_NARROWBAND) + (*bandwidth)--; + else + break; + } + /* Couldn't find any bandwidth to enable FEC, keep original bandwidth. */ + *bandwidth = orig_bandwidth; + return 0; +} + +static int compute_silk_rate_for_hybrid(int rate, int bandwidth, int frame20ms, int vbr, int fec, int channels) { + int entry; + int i; + int N; + int silk_rate; + static int rate_table[][5] = { + /* |total| |-------- SILK------------| + |-- No FEC -| |--- FEC ---| + 10ms 20ms 10ms 20ms */ + { 0, 0, 0, 0, 0}, + {12000, 10000, 10000, 11000, 11000}, + {16000, 13500, 13500, 15000, 15000}, + {20000, 16000, 16000, 18000, 18000}, + {24000, 18000, 18000, 21000, 21000}, + {32000, 22000, 22000, 28000, 28000}, + {64000, 38000, 38000, 50000, 50000} + }; + /* Do the allocation per-channel. */ + rate /= channels; + entry = 1 + frame20ms + 2*fec; + N = sizeof(rate_table)/sizeof(rate_table[0]); + for (i=1;i rate) break; + } + if (i == N) + { + silk_rate = rate_table[i-1][entry]; + /* For now, just give 50% of the extra bits to SILK. */ + silk_rate += (rate-rate_table[i-1][0])/2; + } else { + opus_int32 lo, hi, x0, x1; + lo = rate_table[i-1][entry]; + hi = rate_table[i][entry]; + x0 = rate_table[i-1][0]; + x1 = rate_table[i][0]; + silk_rate = (lo*(x1-rate) + hi*(rate-x0))/(x1-x0); + } + if (!vbr) + { + /* Tiny boost to SILK for CBR. We should probably tune this better. */ + silk_rate += 100; + } + if (bandwidth==OPUS_BANDWIDTH_SUPERWIDEBAND) + silk_rate += 300; + silk_rate *= channels; + /* Small adjustment for stereo (calibrated for 32 kb/s, haven't tried other bitrates). */ + if (channels == 2 && rate >= 12000) + silk_rate -= 1000; + return silk_rate; +} + +/* Returns the equivalent bitrate corresponding to 20 ms frames, + complexity 10 VBR operation. */ +static opus_int32 compute_equiv_rate(opus_int32 bitrate, int channels, + int frame_rate, int vbr, int mode, int complexity, int loss) +{ + opus_int32 equiv; + equiv = bitrate; + /* Take into account overhead from smaller frames. */ + if (frame_rate > 50) + equiv -= (40*channels+20)*(frame_rate - 50); + /* CBR is about a 8% penalty for both SILK and CELT. */ + if (!vbr) + equiv -= equiv/12; + /* Complexity makes about 10% difference (from 0 to 10) in general. */ + equiv = equiv * (90+complexity)/100; + if (mode == MODE_SILK_ONLY || mode == MODE_HYBRID) + { + /* SILK complexity 0-1 uses the non-delayed-decision NSQ, which + costs about 20%. */ + if (complexity<2) + equiv = equiv*4/5; + equiv -= equiv*loss/(6*loss + 10); + } else if (mode == MODE_CELT_ONLY) { + /* CELT complexity 0-4 doesn't have the pitch filter, which costs + about 10%. */ + if (complexity<5) + equiv = equiv*9/10; + } else { + /* Mode not known yet */ + /* Half the SILK loss*/ + equiv -= equiv*loss/(12*loss + 20); + } + return equiv; +} + +#ifndef DISABLE_FLOAT_API + +int is_digital_silence(const opus_val16* pcm, int frame_size, int channels, int lsb_depth) +{ + int silence = 0; + opus_val32 sample_max = 0; +#ifdef MLP_TRAINING + return 0; +#endif + sample_max = celt_maxabs16(pcm, frame_size*channels); + +#ifdef FIXED_POINT + silence = (sample_max == 0); + (void)lsb_depth; +#else + silence = (sample_max <= (opus_val16) 1 / (1 << lsb_depth)); +#endif + + return silence; +} + +#ifdef FIXED_POINT +static opus_val32 compute_frame_energy(const opus_val16 *pcm, int frame_size, int channels, int arch) +{ + int i; + opus_val32 sample_max; + int max_shift; + int shift; + opus_val32 energy = 0; + int len = frame_size*channels; + (void)arch; + /* Max amplitude in the signal */ + sample_max = celt_maxabs16(pcm, len); + + /* Compute the right shift required in the MAC to avoid an overflow */ + max_shift = celt_ilog2(len); + shift = IMAX(0, (celt_ilog2(sample_max) << 1) + max_shift - 28); + + /* Compute the energy */ + for (i=0; i NB_SPEECH_FRAMES_BEFORE_DTX*20*2) + { + if (*nb_no_activity_ms_Q1 <= (NB_SPEECH_FRAMES_BEFORE_DTX + MAX_CONSECUTIVE_DTX)*20*2) + /* Valid frame for DTX! */ + return 1; + else + (*nb_no_activity_ms_Q1) = NB_SPEECH_FRAMES_BEFORE_DTX*20*2; + } + } else + (*nb_no_activity_ms_Q1) = 0; + + return 0; +} + +#endif + +static opus_int32 encode_multiframe_packet(OpusEncoder *st, + const opus_val16 *pcm, + int nb_frames, + int frame_size, + unsigned char *data, + opus_int32 out_data_bytes, + int to_celt, + int lsb_depth, + int float_api) +{ + int i; + int ret = 0; + VARDECL(unsigned char, tmp_data); + int bak_mode, bak_bandwidth, bak_channels, bak_to_mono; + VARDECL(OpusRepacketizer, rp); + int max_header_bytes; + opus_int32 bytes_per_frame; + opus_int32 cbr_bytes; + opus_int32 repacketize_len; + int tmp_len; + ALLOC_STACK; + + /* Worst cases: + * 2 frames: Code 2 with different compressed sizes + * >2 frames: Code 3 VBR */ + max_header_bytes = nb_frames == 2 ? 3 : (2+(nb_frames-1)*2); + + if (st->use_vbr || st->user_bitrate_bps==OPUS_BITRATE_MAX) + repacketize_len = out_data_bytes; + else { + cbr_bytes = 3*st->bitrate_bps/(3*8*st->Fs/(frame_size*nb_frames)); + repacketize_len = IMIN(cbr_bytes, out_data_bytes); + } + bytes_per_frame = IMIN(1276, 1+(repacketize_len-max_header_bytes)/nb_frames); + + ALLOC(tmp_data, nb_frames*bytes_per_frame, unsigned char); + ALLOC(rp, 1, OpusRepacketizer); + opus_repacketizer_init(rp); + + bak_mode = st->user_forced_mode; + bak_bandwidth = st->user_bandwidth; + bak_channels = st->force_channels; + + st->user_forced_mode = st->mode; + st->user_bandwidth = st->bandwidth; + st->force_channels = st->stream_channels; + + bak_to_mono = st->silk_mode.toMono; + if (bak_to_mono) + st->force_channels = 1; + else + st->prev_channels = st->stream_channels; + + for (i=0;isilk_mode.toMono = 0; + st->nonfinal_frame = i<(nb_frames-1); + + /* When switching from SILK/Hybrid to CELT, only ask for a switch at the last frame */ + if (to_celt && i==nb_frames-1) + st->user_forced_mode = MODE_CELT_ONLY; + + tmp_len = opus_encode_native(st, pcm+i*(st->channels*frame_size), frame_size, + tmp_data+i*bytes_per_frame, bytes_per_frame, lsb_depth, NULL, 0, 0, 0, 0, + NULL, float_api); + + if (tmp_len<0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + + ret = opus_repacketizer_cat(rp, tmp_data+i*bytes_per_frame, tmp_len); + + if (ret<0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + } + + ret = opus_repacketizer_out_range_impl(rp, 0, nb_frames, data, repacketize_len, 0, !st->use_vbr); + + if (ret<0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + + /* Discard configs that were forced locally for the purpose of repacketization */ + st->user_forced_mode = bak_mode; + st->user_bandwidth = bak_bandwidth; + st->force_channels = bak_channels; + st->silk_mode.toMono = bak_to_mono; + + RESTORE_STACK; + return ret; +} + +static int compute_redundancy_bytes(opus_int32 max_data_bytes, opus_int32 bitrate_bps, int frame_rate, int channels) +{ + int redundancy_bytes_cap; + int redundancy_bytes; + opus_int32 redundancy_rate; + int base_bits; + opus_int32 available_bits; + base_bits = (40*channels+20); + + /* Equivalent rate for 5 ms frames. */ + redundancy_rate = bitrate_bps + base_bits*(200 - frame_rate); + /* For VBR, further increase the bitrate if we can afford it. It's pretty short + and we'll avoid artefacts. */ + redundancy_rate = 3*redundancy_rate/2; + redundancy_bytes = redundancy_rate/1600; + + /* Compute the max rate we can use given CBR or VBR with cap. */ + available_bits = max_data_bytes*8 - 2*base_bits; + redundancy_bytes_cap = (available_bits*240/(240+48000/frame_rate) + base_bits)/8; + redundancy_bytes = IMIN(redundancy_bytes, redundancy_bytes_cap); + /* It we can't get enough bits for redundancy to be worth it, rely on the decoder PLC. */ + if (redundancy_bytes > 4 + 8*channels) + redundancy_bytes = IMIN(257, redundancy_bytes); + else + redundancy_bytes = 0; + return redundancy_bytes; +} + +opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_size, + unsigned char *data, opus_int32 out_data_bytes, int lsb_depth, + const void *analysis_pcm, opus_int32 analysis_size, int c1, int c2, + int analysis_channels, downmix_func downmix, int float_api) +{ + void *silk_enc; + CELTEncoder *celt_enc; + int i; + int ret=0; + opus_int32 nBytes; + ec_enc enc; + int bytes_target; + int prefill=0; + int start_band = 0; + int redundancy = 0; + int redundancy_bytes = 0; /* Number of bytes to use for redundancy frame */ + int celt_to_silk = 0; + VARDECL(opus_val16, pcm_buf); + int nb_compr_bytes; + int to_celt = 0; + opus_uint32 redundant_rng = 0; + int cutoff_Hz, hp_freq_smth1; + int voice_est; /* Probability of voice in Q7 */ + opus_int32 equiv_rate; + int delay_compensation; + int frame_rate; + opus_int32 max_rate; /* Max bitrate we're allowed to use */ + int curr_bandwidth; + opus_val16 HB_gain; + opus_int32 max_data_bytes; /* Max number of bytes we're allowed to use */ + int total_buffer; + opus_val16 stereo_width; + const CELTMode *celt_mode; +#ifndef DISABLE_FLOAT_API + AnalysisInfo analysis_info; + int analysis_read_pos_bak=-1; + int analysis_read_subframe_bak=-1; + int is_silence = 0; +#endif + opus_int activity = VAD_NO_DECISION; + + VARDECL(opus_val16, tmp_prefill); + + ALLOC_STACK; + + max_data_bytes = IMIN(1276, out_data_bytes); + + st->rangeFinal = 0; + if (frame_size <= 0 || max_data_bytes <= 0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + + /* Cannot encode 100 ms in 1 byte */ + if (max_data_bytes==1 && st->Fs==(frame_size*10)) + { + RESTORE_STACK; + return OPUS_BUFFER_TOO_SMALL; + } + + silk_enc = (char*)st+st->silk_enc_offset; + celt_enc = (CELTEncoder*)((char*)st+st->celt_enc_offset); + if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) + delay_compensation = 0; + else + delay_compensation = st->delay_compensation; + + lsb_depth = IMIN(lsb_depth, st->lsb_depth); + + celt_encoder_ctl(celt_enc, CELT_GET_MODE(&celt_mode)); +#ifndef DISABLE_FLOAT_API + analysis_info.valid = 0; +#ifdef FIXED_POINT + if (st->silk_mode.complexity >= 10 && st->Fs>=16000) +#else + if (st->silk_mode.complexity >= 7 && st->Fs>=16000) +#endif + { + is_silence = is_digital_silence(pcm, frame_size, st->channels, lsb_depth); + analysis_read_pos_bak = st->analysis.read_pos; + analysis_read_subframe_bak = st->analysis.read_subframe; + run_analysis(&st->analysis, celt_mode, analysis_pcm, analysis_size, frame_size, + c1, c2, analysis_channels, st->Fs, + lsb_depth, downmix, &analysis_info); + + /* Track the peak signal energy */ + if (!is_silence && analysis_info.activity_probability > DTX_ACTIVITY_THRESHOLD) + st->peak_signal_energy = MAX32(MULT16_32_Q15(QCONST16(0.999f, 15), st->peak_signal_energy), + compute_frame_energy(pcm, frame_size, st->channels, st->arch)); + } else if (st->analysis.initialized) { + tonality_analysis_reset(&st->analysis); + } +#else + (void)analysis_pcm; + (void)analysis_size; + (void)c1; + (void)c2; + (void)analysis_channels; + (void)downmix; +#endif + +#ifndef DISABLE_FLOAT_API + /* Reset voice_ratio if this frame is not silent or if analysis is disabled. + * Otherwise, preserve voice_ratio from the last non-silent frame */ + if (!is_silence) + st->voice_ratio = -1; + + if (is_silence) + { + activity = !is_silence; + } else if (analysis_info.valid) + { + activity = analysis_info.activity_probability >= DTX_ACTIVITY_THRESHOLD; + if (!activity) + { + /* Mark as active if this noise frame is sufficiently loud */ + opus_val32 noise_energy = compute_frame_energy(pcm, frame_size, st->channels, st->arch); + activity = st->peak_signal_energy < (PSEUDO_SNR_THRESHOLD * noise_energy); + } + } + + st->detected_bandwidth = 0; + if (analysis_info.valid) + { + int analysis_bandwidth; + if (st->signal_type == OPUS_AUTO) + { + float prob; + if (st->prev_mode == 0) + prob = analysis_info.music_prob; + else if (st->prev_mode == MODE_CELT_ONLY) + prob = analysis_info.music_prob_max; + else + prob = analysis_info.music_prob_min; + st->voice_ratio = (int)floor(.5+100*(1-prob)); + } + + analysis_bandwidth = analysis_info.bandwidth; + if (analysis_bandwidth<=12) + st->detected_bandwidth = OPUS_BANDWIDTH_NARROWBAND; + else if (analysis_bandwidth<=14) + st->detected_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + else if (analysis_bandwidth<=16) + st->detected_bandwidth = OPUS_BANDWIDTH_WIDEBAND; + else if (analysis_bandwidth<=18) + st->detected_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; + else + st->detected_bandwidth = OPUS_BANDWIDTH_FULLBAND; + } +#else + st->voice_ratio = -1; +#endif + + if (st->channels==2 && st->force_channels!=1) + stereo_width = compute_stereo_width(pcm, frame_size, st->Fs, &st->width_mem); + else + stereo_width = 0; + total_buffer = delay_compensation; + st->bitrate_bps = user_bitrate_to_bitrate(st, frame_size, max_data_bytes); + + frame_rate = st->Fs/frame_size; + if (!st->use_vbr) + { + int cbrBytes; + /* Multiply by 12 to make sure the division is exact. */ + int frame_rate12 = 12*st->Fs/frame_size; + /* We need to make sure that "int" values always fit in 16 bits. */ + cbrBytes = IMIN( (12*st->bitrate_bps/8 + frame_rate12/2)/frame_rate12, max_data_bytes); + st->bitrate_bps = cbrBytes*(opus_int32)frame_rate12*8/12; + /* Make sure we provide at least one byte to avoid failing. */ + max_data_bytes = IMAX(1, cbrBytes); + } + if (max_data_bytes<3 || st->bitrate_bps < 3*frame_rate*8 + || (frame_rate<50 && (max_data_bytes*frame_rate<300 || st->bitrate_bps < 2400))) + { + /*If the space is too low to do something useful, emit 'PLC' frames.*/ + int tocmode = st->mode; + int bw = st->bandwidth == 0 ? OPUS_BANDWIDTH_NARROWBAND : st->bandwidth; + int packet_code = 0; + int num_multiframes = 0; + + if (tocmode==0) + tocmode = MODE_SILK_ONLY; + if (frame_rate>100) + tocmode = MODE_CELT_ONLY; + /* 40 ms -> 2 x 20 ms if in CELT_ONLY or HYBRID mode */ + if (frame_rate==25 && tocmode!=MODE_SILK_ONLY) + { + frame_rate = 50; + packet_code = 1; + } + + /* >= 60 ms frames */ + if (frame_rate<=16) + { + /* 1 x 60 ms, 2 x 40 ms, 2 x 60 ms */ + if (out_data_bytes==1 || (tocmode==MODE_SILK_ONLY && frame_rate!=10)) + { + tocmode = MODE_SILK_ONLY; + + packet_code = frame_rate <= 12; + frame_rate = frame_rate == 12 ? 25 : 16; + } + else + { + num_multiframes = 50/frame_rate; + frame_rate = 50; + packet_code = 3; + } + } + + if(tocmode==MODE_SILK_ONLY&&bw>OPUS_BANDWIDTH_WIDEBAND) + bw=OPUS_BANDWIDTH_WIDEBAND; + else if (tocmode==MODE_CELT_ONLY&&bw==OPUS_BANDWIDTH_MEDIUMBAND) + bw=OPUS_BANDWIDTH_NARROWBAND; + else if (tocmode==MODE_HYBRID&&bw<=OPUS_BANDWIDTH_SUPERWIDEBAND) + bw=OPUS_BANDWIDTH_SUPERWIDEBAND; + + data[0] = gen_toc(tocmode, frame_rate, bw, st->stream_channels); + data[0] |= packet_code; + + ret = packet_code <= 1 ? 1 : 2; + + max_data_bytes = IMAX(max_data_bytes, ret); + + if (packet_code==3) + data[1] = num_multiframes; + + if (!st->use_vbr) + { + ret = opus_packet_pad(data, ret, max_data_bytes); + if (ret == OPUS_OK) + ret = max_data_bytes; + else + ret = OPUS_INTERNAL_ERROR; + } + RESTORE_STACK; + return ret; + } + max_rate = frame_rate*max_data_bytes*8; + + /* Equivalent 20-ms rate for mode/channel/bandwidth decisions */ + equiv_rate = compute_equiv_rate(st->bitrate_bps, st->channels, st->Fs/frame_size, + st->use_vbr, 0, st->silk_mode.complexity, st->silk_mode.packetLossPercentage); + + if (st->signal_type == OPUS_SIGNAL_VOICE) + voice_est = 127; + else if (st->signal_type == OPUS_SIGNAL_MUSIC) + voice_est = 0; + else if (st->voice_ratio >= 0) + { + voice_est = st->voice_ratio*327>>8; + /* For AUDIO, never be more than 90% confident of having speech */ + if (st->application == OPUS_APPLICATION_AUDIO) + voice_est = IMIN(voice_est, 115); + } else if (st->application == OPUS_APPLICATION_VOIP) + voice_est = 115; + else + voice_est = 48; + + if (st->force_channels!=OPUS_AUTO && st->channels == 2) + { + st->stream_channels = st->force_channels; + } else { +#ifdef FUZZING + (void)stereo_music_threshold; + (void)stereo_voice_threshold; + /* Random mono/stereo decision */ + if (st->channels == 2 && (rand()&0x1F)==0) + st->stream_channels = 3-st->stream_channels; +#else + /* Rate-dependent mono-stereo decision */ + if (st->channels == 2) + { + opus_int32 stereo_threshold; + stereo_threshold = stereo_music_threshold + ((voice_est*voice_est*(stereo_voice_threshold-stereo_music_threshold))>>14); + if (st->stream_channels == 2) + stereo_threshold -= 1000; + else + stereo_threshold += 1000; + st->stream_channels = (equiv_rate > stereo_threshold) ? 2 : 1; + } else { + st->stream_channels = st->channels; + } +#endif + } + /* Update equivalent rate for channels decision. */ + equiv_rate = compute_equiv_rate(st->bitrate_bps, st->stream_channels, st->Fs/frame_size, + st->use_vbr, 0, st->silk_mode.complexity, st->silk_mode.packetLossPercentage); + + /* Allow SILK DTX if DTX is enabled but the generalized DTX cannot be used, + e.g. because of the complexity setting or sample rate. */ +#ifndef DISABLE_FLOAT_API + st->silk_mode.useDTX = st->use_dtx && !(analysis_info.valid || is_silence); +#else + st->silk_mode.useDTX = st->use_dtx; +#endif + + /* Mode selection depending on application and signal type */ + if (st->application == OPUS_APPLICATION_RESTRICTED_LOWDELAY) + { + st->mode = MODE_CELT_ONLY; + } else if (st->user_forced_mode == OPUS_AUTO) + { +#ifdef FUZZING + (void)stereo_width; + (void)mode_thresholds; + /* Random mode switching */ + if ((rand()&0xF)==0) + { + if ((rand()&0x1)==0) + st->mode = MODE_CELT_ONLY; + else + st->mode = MODE_SILK_ONLY; + } else { + if (st->prev_mode==MODE_CELT_ONLY) + st->mode = MODE_CELT_ONLY; + else + st->mode = MODE_SILK_ONLY; + } +#else + opus_int32 mode_voice, mode_music; + opus_int32 threshold; + + /* Interpolate based on stereo width */ + mode_voice = (opus_int32)(MULT16_32_Q15(Q15ONE-stereo_width,mode_thresholds[0][0]) + + MULT16_32_Q15(stereo_width,mode_thresholds[1][0])); + mode_music = (opus_int32)(MULT16_32_Q15(Q15ONE-stereo_width,mode_thresholds[1][1]) + + MULT16_32_Q15(stereo_width,mode_thresholds[1][1])); + /* Interpolate based on speech/music probability */ + threshold = mode_music + ((voice_est*voice_est*(mode_voice-mode_music))>>14); + /* Bias towards SILK for VoIP because of some useful features */ + if (st->application == OPUS_APPLICATION_VOIP) + threshold += 8000; + + /*printf("%f %d\n", stereo_width/(float)Q15ONE, threshold);*/ + /* Hysteresis */ + if (st->prev_mode == MODE_CELT_ONLY) + threshold -= 4000; + else if (st->prev_mode>0) + threshold += 4000; + + st->mode = (equiv_rate >= threshold) ? MODE_CELT_ONLY: MODE_SILK_ONLY; + + /* When FEC is enabled and there's enough packet loss, use SILK. + Unless the FEC is set to 2, in which case we don't switch to SILK if we're confident we have music. */ + if (st->silk_mode.useInBandFEC && st->silk_mode.packetLossPercentage > (128-voice_est)>>4 && (st->fec_config != 2 || voice_est > 25)) + st->mode = MODE_SILK_ONLY; + /* When encoding voice and DTX is enabled but the generalized DTX cannot be used, + use SILK in order to make use of its DTX. */ + if (st->silk_mode.useDTX && voice_est > 100) + st->mode = MODE_SILK_ONLY; +#endif + + /* If max_data_bytes represents less than 6 kb/s, switch to CELT-only mode */ + if (max_data_bytes < (frame_rate > 50 ? 9000 : 6000)*frame_size / (st->Fs * 8)) + st->mode = MODE_CELT_ONLY; + } else { + st->mode = st->user_forced_mode; + } + + /* Override the chosen mode to make sure we meet the requested frame size */ + if (st->mode != MODE_CELT_ONLY && frame_size < st->Fs/100) + st->mode = MODE_CELT_ONLY; + if (st->lfe) + st->mode = MODE_CELT_ONLY; + + if (st->prev_mode > 0 && + ((st->mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) || + (st->mode == MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY))) + { + redundancy = 1; + celt_to_silk = (st->mode != MODE_CELT_ONLY); + if (!celt_to_silk) + { + /* Switch to SILK/hybrid if frame size is 10 ms or more*/ + if (frame_size >= st->Fs/100) + { + st->mode = st->prev_mode; + to_celt = 1; + } else { + redundancy=0; + } + } + } + + /* When encoding multiframes, we can ask for a switch to CELT only in the last frame. This switch + * is processed above as the requested mode shouldn't interrupt stereo->mono transition. */ + if (st->stream_channels == 1 && st->prev_channels ==2 && st->silk_mode.toMono==0 + && st->mode != MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY) + { + /* Delay stereo->mono transition by two frames so that SILK can do a smooth downmix */ + st->silk_mode.toMono = 1; + st->stream_channels = 2; + } else { + st->silk_mode.toMono = 0; + } + + /* Update equivalent rate with mode decision. */ + equiv_rate = compute_equiv_rate(st->bitrate_bps, st->stream_channels, st->Fs/frame_size, + st->use_vbr, st->mode, st->silk_mode.complexity, st->silk_mode.packetLossPercentage); + + if (st->mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) + { + silk_EncControlStruct dummy; + silk_InitEncoder( silk_enc, st->arch, &dummy); + prefill=1; + } + + /* Automatic (rate-dependent) bandwidth selection */ + if (st->mode == MODE_CELT_ONLY || st->first || st->silk_mode.allowBandwidthSwitch) + { + const opus_int32 *voice_bandwidth_thresholds, *music_bandwidth_thresholds; + opus_int32 bandwidth_thresholds[8]; + int bandwidth = OPUS_BANDWIDTH_FULLBAND; + + if (st->channels==2 && st->force_channels!=1) + { + voice_bandwidth_thresholds = stereo_voice_bandwidth_thresholds; + music_bandwidth_thresholds = stereo_music_bandwidth_thresholds; + } else { + voice_bandwidth_thresholds = mono_voice_bandwidth_thresholds; + music_bandwidth_thresholds = mono_music_bandwidth_thresholds; + } + /* Interpolate bandwidth thresholds depending on voice estimation */ + for (i=0;i<8;i++) + { + bandwidth_thresholds[i] = music_bandwidth_thresholds[i] + + ((voice_est*voice_est*(voice_bandwidth_thresholds[i]-music_bandwidth_thresholds[i]))>>14); + } + do { + int threshold, hysteresis; + threshold = bandwidth_thresholds[2*(bandwidth-OPUS_BANDWIDTH_MEDIUMBAND)]; + hysteresis = bandwidth_thresholds[2*(bandwidth-OPUS_BANDWIDTH_MEDIUMBAND)+1]; + if (!st->first) + { + if (st->auto_bandwidth >= bandwidth) + threshold -= hysteresis; + else + threshold += hysteresis; + } + if (equiv_rate >= threshold) + break; + } while (--bandwidth>OPUS_BANDWIDTH_NARROWBAND); + /* We don't use mediumband anymore, except when explicitly requested or during + mode transitions. */ + if (bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) + bandwidth = OPUS_BANDWIDTH_WIDEBAND; + st->bandwidth = st->auto_bandwidth = bandwidth; + /* Prevents any transition to SWB/FB until the SILK layer has fully + switched to WB mode and turned the variable LP filter off */ + if (!st->first && st->mode != MODE_CELT_ONLY && !st->silk_mode.inWBmodeWithoutVariableLP && st->bandwidth > OPUS_BANDWIDTH_WIDEBAND) + st->bandwidth = OPUS_BANDWIDTH_WIDEBAND; + } + + if (st->bandwidth>st->max_bandwidth) + st->bandwidth = st->max_bandwidth; + + if (st->user_bandwidth != OPUS_AUTO) + st->bandwidth = st->user_bandwidth; + + /* This prevents us from using hybrid at unsafe CBR/max rates */ + if (st->mode != MODE_CELT_ONLY && max_rate < 15000) + { + st->bandwidth = IMIN(st->bandwidth, OPUS_BANDWIDTH_WIDEBAND); + } + + /* Prevents Opus from wasting bits on frequencies that are above + the Nyquist rate of the input signal */ + if (st->Fs <= 24000 && st->bandwidth > OPUS_BANDWIDTH_SUPERWIDEBAND) + st->bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; + if (st->Fs <= 16000 && st->bandwidth > OPUS_BANDWIDTH_WIDEBAND) + st->bandwidth = OPUS_BANDWIDTH_WIDEBAND; + if (st->Fs <= 12000 && st->bandwidth > OPUS_BANDWIDTH_MEDIUMBAND) + st->bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + if (st->Fs <= 8000 && st->bandwidth > OPUS_BANDWIDTH_NARROWBAND) + st->bandwidth = OPUS_BANDWIDTH_NARROWBAND; +#ifndef DISABLE_FLOAT_API + /* Use detected bandwidth to reduce the encoded bandwidth. */ + if (st->detected_bandwidth && st->user_bandwidth == OPUS_AUTO) + { + int min_detected_bandwidth; + /* Makes bandwidth detection more conservative just in case the detector + gets it wrong when we could have coded a high bandwidth transparently. + When operating in SILK/hybrid mode, we don't go below wideband to avoid + more complicated switches that require redundancy. */ + if (equiv_rate <= 18000*st->stream_channels && st->mode == MODE_CELT_ONLY) + min_detected_bandwidth = OPUS_BANDWIDTH_NARROWBAND; + else if (equiv_rate <= 24000*st->stream_channels && st->mode == MODE_CELT_ONLY) + min_detected_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + else if (equiv_rate <= 30000*st->stream_channels) + min_detected_bandwidth = OPUS_BANDWIDTH_WIDEBAND; + else if (equiv_rate <= 44000*st->stream_channels) + min_detected_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; + else + min_detected_bandwidth = OPUS_BANDWIDTH_FULLBAND; + + st->detected_bandwidth = IMAX(st->detected_bandwidth, min_detected_bandwidth); + st->bandwidth = IMIN(st->bandwidth, st->detected_bandwidth); + } +#endif + st->silk_mode.LBRR_coded = decide_fec(st->silk_mode.useInBandFEC, st->silk_mode.packetLossPercentage, + st->silk_mode.LBRR_coded, st->mode, &st->bandwidth, equiv_rate); + celt_encoder_ctl(celt_enc, OPUS_SET_LSB_DEPTH(lsb_depth)); + + /* CELT mode doesn't support mediumband, use wideband instead */ + if (st->mode == MODE_CELT_ONLY && st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) + st->bandwidth = OPUS_BANDWIDTH_WIDEBAND; + if (st->lfe) + st->bandwidth = OPUS_BANDWIDTH_NARROWBAND; + + curr_bandwidth = st->bandwidth; + + /* Chooses the appropriate mode for speech + *NEVER* switch to/from CELT-only mode here as this will invalidate some assumptions */ + if (st->mode == MODE_SILK_ONLY && curr_bandwidth > OPUS_BANDWIDTH_WIDEBAND) + st->mode = MODE_HYBRID; + if (st->mode == MODE_HYBRID && curr_bandwidth <= OPUS_BANDWIDTH_WIDEBAND) + st->mode = MODE_SILK_ONLY; + + /* Can't support higher than >60 ms frames, and >20 ms when in Hybrid or CELT-only modes */ + if ((frame_size > st->Fs/50 && (st->mode != MODE_SILK_ONLY)) || frame_size > 3*st->Fs/50) + { + int enc_frame_size; + int nb_frames; + + if (st->mode == MODE_SILK_ONLY) + { + if (frame_size == 2*st->Fs/25) /* 80 ms -> 2x 40 ms */ + enc_frame_size = st->Fs/25; + else if (frame_size == 3*st->Fs/25) /* 120 ms -> 2x 60 ms */ + enc_frame_size = 3*st->Fs/50; + else /* 100 ms -> 5x 20 ms */ + enc_frame_size = st->Fs/50; + } + else + enc_frame_size = st->Fs/50; + + nb_frames = frame_size/enc_frame_size; + +#ifndef DISABLE_FLOAT_API + if (analysis_read_pos_bak!= -1) + { + st->analysis.read_pos = analysis_read_pos_bak; + st->analysis.read_subframe = analysis_read_subframe_bak; + } +#endif + + ret = encode_multiframe_packet(st, pcm, nb_frames, enc_frame_size, data, + out_data_bytes, to_celt, lsb_depth, float_api); + + RESTORE_STACK; + return ret; + } + + /* For the first frame at a new SILK bandwidth */ + if (st->silk_bw_switch) + { + redundancy = 1; + celt_to_silk = 1; + st->silk_bw_switch = 0; + /* Do a prefill without reseting the sampling rate control. */ + prefill=2; + } + + /* If we decided to go with CELT, make sure redundancy is off, no matter what + we decided earlier. */ + if (st->mode == MODE_CELT_ONLY) + redundancy = 0; + + if (redundancy) + { + redundancy_bytes = compute_redundancy_bytes(max_data_bytes, st->bitrate_bps, frame_rate, st->stream_channels); + if (redundancy_bytes == 0) + redundancy = 0; + } + + /* printf("%d %d %d %d\n", st->bitrate_bps, st->stream_channels, st->mode, curr_bandwidth); */ + bytes_target = IMIN(max_data_bytes-redundancy_bytes, st->bitrate_bps * frame_size / (st->Fs * 8)) - 1; + + data += 1; + + ec_enc_init(&enc, data, max_data_bytes-1); + + ALLOC(pcm_buf, (total_buffer+frame_size)*st->channels, opus_val16); + OPUS_COPY(pcm_buf, &st->delay_buffer[(st->encoder_buffer-total_buffer)*st->channels], total_buffer*st->channels); + + if (st->mode == MODE_CELT_ONLY) + hp_freq_smth1 = silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 ); + else + hp_freq_smth1 = ((silk_encoder*)silk_enc)->state_Fxx[0].sCmn.variable_HP_smth1_Q15; + + st->variable_HP_smth2_Q15 = silk_SMLAWB( st->variable_HP_smth2_Q15, + hp_freq_smth1 - st->variable_HP_smth2_Q15, SILK_FIX_CONST( VARIABLE_HP_SMTH_COEF2, 16 ) ); + + /* convert from log scale to Hertz */ + cutoff_Hz = silk_log2lin( silk_RSHIFT( st->variable_HP_smth2_Q15, 8 ) ); + + if (st->application == OPUS_APPLICATION_VOIP) + { + hp_cutoff(pcm, cutoff_Hz, &pcm_buf[total_buffer*st->channels], st->hp_mem, frame_size, st->channels, st->Fs, st->arch); + } else { + dc_reject(pcm, 3, &pcm_buf[total_buffer*st->channels], st->hp_mem, frame_size, st->channels, st->Fs); + } +#ifndef FIXED_POINT + if (float_api) + { + opus_val32 sum; + sum = celt_inner_prod(&pcm_buf[total_buffer*st->channels], &pcm_buf[total_buffer*st->channels], frame_size*st->channels, st->arch); + /* This should filter out both NaNs and ridiculous signals that could + cause NaNs further down. */ + if (!(sum < 1e9f) || celt_isnan(sum)) + { + OPUS_CLEAR(&pcm_buf[total_buffer*st->channels], frame_size*st->channels); + st->hp_mem[0] = st->hp_mem[1] = st->hp_mem[2] = st->hp_mem[3] = 0; + } + } +#endif + + + /* SILK processing */ + HB_gain = Q15ONE; + if (st->mode != MODE_CELT_ONLY) + { + opus_int32 total_bitRate, celt_rate; +#ifdef FIXED_POINT + const opus_int16 *pcm_silk; +#else + VARDECL(opus_int16, pcm_silk); + ALLOC(pcm_silk, st->channels*frame_size, opus_int16); +#endif + + /* Distribute bits between SILK and CELT */ + total_bitRate = 8 * bytes_target * frame_rate; + if( st->mode == MODE_HYBRID ) { + /* Base rate for SILK */ + st->silk_mode.bitRate = compute_silk_rate_for_hybrid(total_bitRate, + curr_bandwidth, st->Fs == 50 * frame_size, st->use_vbr, st->silk_mode.LBRR_coded, + st->stream_channels); + if (!st->energy_masking) + { + /* Increasingly attenuate high band when it gets allocated fewer bits */ + celt_rate = total_bitRate - st->silk_mode.bitRate; + HB_gain = Q15ONE - SHR32(celt_exp2(-celt_rate * QCONST16(1.f/1024, 10)), 1); + } + } else { + /* SILK gets all bits */ + st->silk_mode.bitRate = total_bitRate; + } + + /* Surround masking for SILK */ + if (st->energy_masking && st->use_vbr && !st->lfe) + { + opus_val32 mask_sum=0; + opus_val16 masking_depth; + opus_int32 rate_offset; + int c; + int end = 17; + opus_int16 srate = 16000; + if (st->bandwidth == OPUS_BANDWIDTH_NARROWBAND) + { + end = 13; + srate = 8000; + } else if (st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) + { + end = 15; + srate = 12000; + } + for (c=0;cchannels;c++) + { + for(i=0;ienergy_masking[21*c+i], + QCONST16(.5f, DB_SHIFT)), -QCONST16(2.0f, DB_SHIFT)); + if (mask > 0) + mask = HALF16(mask); + mask_sum += mask; + } + } + /* Conservative rate reduction, we cut the masking in half */ + masking_depth = mask_sum / end*st->channels; + masking_depth += QCONST16(.2f, DB_SHIFT); + rate_offset = (opus_int32)PSHR32(MULT16_16(srate, masking_depth), DB_SHIFT); + rate_offset = MAX32(rate_offset, -2*st->silk_mode.bitRate/3); + /* Split the rate change between the SILK and CELT part for hybrid. */ + if (st->bandwidth==OPUS_BANDWIDTH_SUPERWIDEBAND || st->bandwidth==OPUS_BANDWIDTH_FULLBAND) + st->silk_mode.bitRate += 3*rate_offset/5; + else + st->silk_mode.bitRate += rate_offset; + } + + st->silk_mode.payloadSize_ms = 1000 * frame_size / st->Fs; + st->silk_mode.nChannelsAPI = st->channels; + st->silk_mode.nChannelsInternal = st->stream_channels; + if (curr_bandwidth == OPUS_BANDWIDTH_NARROWBAND) { + st->silk_mode.desiredInternalSampleRate = 8000; + } else if (curr_bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) { + st->silk_mode.desiredInternalSampleRate = 12000; + } else { + celt_assert( st->mode == MODE_HYBRID || curr_bandwidth == OPUS_BANDWIDTH_WIDEBAND ); + st->silk_mode.desiredInternalSampleRate = 16000; + } + if( st->mode == MODE_HYBRID ) { + /* Don't allow bandwidth reduction at lowest bitrates in hybrid mode */ + st->silk_mode.minInternalSampleRate = 16000; + } else { + st->silk_mode.minInternalSampleRate = 8000; + } + + st->silk_mode.maxInternalSampleRate = 16000; + if (st->mode == MODE_SILK_ONLY) + { + opus_int32 effective_max_rate = max_rate; + if (frame_rate > 50) + effective_max_rate = effective_max_rate*2/3; + if (effective_max_rate < 8000) + { + st->silk_mode.maxInternalSampleRate = 12000; + st->silk_mode.desiredInternalSampleRate = IMIN(12000, st->silk_mode.desiredInternalSampleRate); + } + if (effective_max_rate < 7000) + { + st->silk_mode.maxInternalSampleRate = 8000; + st->silk_mode.desiredInternalSampleRate = IMIN(8000, st->silk_mode.desiredInternalSampleRate); + } + } + + st->silk_mode.useCBR = !st->use_vbr; + + /* Call SILK encoder for the low band */ + + /* Max bits for SILK, counting ToC, redundancy bytes, and optionally redundancy. */ + st->silk_mode.maxBits = (max_data_bytes-1)*8; + if (redundancy && redundancy_bytes >= 2) + { + /* Counting 1 bit for redundancy position and 20 bits for flag+size (only for hybrid). */ + st->silk_mode.maxBits -= redundancy_bytes*8 + 1; + if (st->mode == MODE_HYBRID) + st->silk_mode.maxBits -= 20; + } + if (st->silk_mode.useCBR) + { + if (st->mode == MODE_HYBRID) + { + st->silk_mode.maxBits = IMIN(st->silk_mode.maxBits, st->silk_mode.bitRate * frame_size / st->Fs); + } + } else { + /* Constrained VBR. */ + if (st->mode == MODE_HYBRID) + { + /* Compute SILK bitrate corresponding to the max total bits available */ + opus_int32 maxBitRate = compute_silk_rate_for_hybrid(st->silk_mode.maxBits*st->Fs / frame_size, + curr_bandwidth, st->Fs == 50 * frame_size, st->use_vbr, st->silk_mode.LBRR_coded, + st->stream_channels); + st->silk_mode.maxBits = maxBitRate * frame_size / st->Fs; + } + } + + if (prefill) + { + opus_int32 zero=0; + int prefill_offset; + /* Use a smooth onset for the SILK prefill to avoid the encoder trying to encode + a discontinuity. The exact location is what we need to avoid leaving any "gap" + in the audio when mixing with the redundant CELT frame. Here we can afford to + overwrite st->delay_buffer because the only thing that uses it before it gets + rewritten is tmp_prefill[] and even then only the part after the ramp really + gets used (rather than sent to the encoder and discarded) */ + prefill_offset = st->channels*(st->encoder_buffer-st->delay_compensation-st->Fs/400); + gain_fade(st->delay_buffer+prefill_offset, st->delay_buffer+prefill_offset, + 0, Q15ONE, celt_mode->overlap, st->Fs/400, st->channels, celt_mode->window, st->Fs); + OPUS_CLEAR(st->delay_buffer, prefill_offset); +#ifdef FIXED_POINT + pcm_silk = st->delay_buffer; +#else + for (i=0;iencoder_buffer*st->channels;i++) + pcm_silk[i] = FLOAT2INT16(st->delay_buffer[i]); +#endif + silk_Encode( silk_enc, &st->silk_mode, pcm_silk, st->encoder_buffer, NULL, &zero, prefill, activity ); + /* Prevent a second switch in the real encode call. */ + st->silk_mode.opusCanSwitch = 0; + } + +#ifdef FIXED_POINT + pcm_silk = pcm_buf+total_buffer*st->channels; +#else + for (i=0;ichannels;i++) + pcm_silk[i] = FLOAT2INT16(pcm_buf[total_buffer*st->channels + i]); +#endif + ret = silk_Encode( silk_enc, &st->silk_mode, pcm_silk, frame_size, &enc, &nBytes, 0, activity ); + if( ret ) { + /*fprintf (stderr, "SILK encode error: %d\n", ret);*/ + /* Handle error */ + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + + /* Extract SILK internal bandwidth for signaling in first byte */ + if( st->mode == MODE_SILK_ONLY ) { + if( st->silk_mode.internalSampleRate == 8000 ) { + curr_bandwidth = OPUS_BANDWIDTH_NARROWBAND; + } else if( st->silk_mode.internalSampleRate == 12000 ) { + curr_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; + } else if( st->silk_mode.internalSampleRate == 16000 ) { + curr_bandwidth = OPUS_BANDWIDTH_WIDEBAND; + } + } else { + celt_assert( st->silk_mode.internalSampleRate == 16000 ); + } + + st->silk_mode.opusCanSwitch = st->silk_mode.switchReady && !st->nonfinal_frame; + + if (nBytes==0) + { + st->rangeFinal = 0; + data[-1] = gen_toc(st->mode, st->Fs/frame_size, curr_bandwidth, st->stream_channels); + RESTORE_STACK; + return 1; + } + + /* FIXME: How do we allocate the redundancy for CBR? */ + if (st->silk_mode.opusCanSwitch) + { + redundancy_bytes = compute_redundancy_bytes(max_data_bytes, st->bitrate_bps, frame_rate, st->stream_channels); + redundancy = (redundancy_bytes != 0); + celt_to_silk = 0; + st->silk_bw_switch = 1; + } + } + + /* CELT processing */ + { + int endband=21; + + switch(curr_bandwidth) + { + case OPUS_BANDWIDTH_NARROWBAND: + endband = 13; + break; + case OPUS_BANDWIDTH_MEDIUMBAND: + case OPUS_BANDWIDTH_WIDEBAND: + endband = 17; + break; + case OPUS_BANDWIDTH_SUPERWIDEBAND: + endband = 19; + break; + case OPUS_BANDWIDTH_FULLBAND: + endband = 21; + break; + } + celt_encoder_ctl(celt_enc, CELT_SET_END_BAND(endband)); + celt_encoder_ctl(celt_enc, CELT_SET_CHANNELS(st->stream_channels)); + } + celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(OPUS_BITRATE_MAX)); + if (st->mode != MODE_SILK_ONLY) + { + opus_val32 celt_pred=2; + celt_encoder_ctl(celt_enc, OPUS_SET_VBR(0)); + /* We may still decide to disable prediction later */ + if (st->silk_mode.reducedDependency) + celt_pred = 0; + celt_encoder_ctl(celt_enc, CELT_SET_PREDICTION(celt_pred)); + + if (st->mode == MODE_HYBRID) + { + if( st->use_vbr ) { + celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(st->bitrate_bps-st->silk_mode.bitRate)); + celt_encoder_ctl(celt_enc, OPUS_SET_VBR_CONSTRAINT(0)); + } + } else { + if (st->use_vbr) + { + celt_encoder_ctl(celt_enc, OPUS_SET_VBR(1)); + celt_encoder_ctl(celt_enc, OPUS_SET_VBR_CONSTRAINT(st->vbr_constraint)); + celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(st->bitrate_bps)); + } + } + } + + ALLOC(tmp_prefill, st->channels*st->Fs/400, opus_val16); + if (st->mode != MODE_SILK_ONLY && st->mode != st->prev_mode && st->prev_mode > 0) + { + OPUS_COPY(tmp_prefill, &st->delay_buffer[(st->encoder_buffer-total_buffer-st->Fs/400)*st->channels], st->channels*st->Fs/400); + } + + if (st->channels*(st->encoder_buffer-(frame_size+total_buffer)) > 0) + { + OPUS_MOVE(st->delay_buffer, &st->delay_buffer[st->channels*frame_size], st->channels*(st->encoder_buffer-frame_size-total_buffer)); + OPUS_COPY(&st->delay_buffer[st->channels*(st->encoder_buffer-frame_size-total_buffer)], + &pcm_buf[0], + (frame_size+total_buffer)*st->channels); + } else { + OPUS_COPY(st->delay_buffer, &pcm_buf[(frame_size+total_buffer-st->encoder_buffer)*st->channels], st->encoder_buffer*st->channels); + } + /* gain_fade() and stereo_fade() need to be after the buffer copying + because we don't want any of this to affect the SILK part */ + if( st->prev_HB_gain < Q15ONE || HB_gain < Q15ONE ) { + gain_fade(pcm_buf, pcm_buf, + st->prev_HB_gain, HB_gain, celt_mode->overlap, frame_size, st->channels, celt_mode->window, st->Fs); + } + st->prev_HB_gain = HB_gain; + if (st->mode != MODE_HYBRID || st->stream_channels==1) + { + if (equiv_rate > 32000) + st->silk_mode.stereoWidth_Q14 = 16384; + else if (equiv_rate < 16000) + st->silk_mode.stereoWidth_Q14 = 0; + else + st->silk_mode.stereoWidth_Q14 = 16384 - 2048*(opus_int32)(32000-equiv_rate)/(equiv_rate-14000); + } + if( !st->energy_masking && st->channels == 2 ) { + /* Apply stereo width reduction (at low bitrates) */ + if( st->hybrid_stereo_width_Q14 < (1 << 14) || st->silk_mode.stereoWidth_Q14 < (1 << 14) ) { + opus_val16 g1, g2; + g1 = st->hybrid_stereo_width_Q14; + g2 = (opus_val16)(st->silk_mode.stereoWidth_Q14); +#ifdef FIXED_POINT + g1 = g1==16384 ? Q15ONE : SHL16(g1,1); + g2 = g2==16384 ? Q15ONE : SHL16(g2,1); +#else + g1 *= (1.f/16384); + g2 *= (1.f/16384); +#endif + stereo_fade(pcm_buf, pcm_buf, g1, g2, celt_mode->overlap, + frame_size, st->channels, celt_mode->window, st->Fs); + st->hybrid_stereo_width_Q14 = st->silk_mode.stereoWidth_Q14; + } + } + + if ( st->mode != MODE_CELT_ONLY && ec_tell(&enc)+17+20*(st->mode == MODE_HYBRID) <= 8*(max_data_bytes-1)) + { + /* For SILK mode, the redundancy is inferred from the length */ + if (st->mode == MODE_HYBRID) + ec_enc_bit_logp(&enc, redundancy, 12); + if (redundancy) + { + int max_redundancy; + ec_enc_bit_logp(&enc, celt_to_silk, 1); + if (st->mode == MODE_HYBRID) + { + /* Reserve the 8 bits needed for the redundancy length, + and at least a few bits for CELT if possible */ + max_redundancy = (max_data_bytes-1)-((ec_tell(&enc)+8+3+7)>>3); + } + else + max_redundancy = (max_data_bytes-1)-((ec_tell(&enc)+7)>>3); + /* Target the same bit-rate for redundancy as for the rest, + up to a max of 257 bytes */ + redundancy_bytes = IMIN(max_redundancy, redundancy_bytes); + redundancy_bytes = IMIN(257, IMAX(2, redundancy_bytes)); + if (st->mode == MODE_HYBRID) + ec_enc_uint(&enc, redundancy_bytes-2, 256); + } + } else { + redundancy = 0; + } + + if (!redundancy) + { + st->silk_bw_switch = 0; + redundancy_bytes = 0; + } + if (st->mode != MODE_CELT_ONLY)start_band=17; + + if (st->mode == MODE_SILK_ONLY) + { + ret = (ec_tell(&enc)+7)>>3; + ec_enc_done(&enc); + nb_compr_bytes = ret; + } else { + nb_compr_bytes = (max_data_bytes-1)-redundancy_bytes; + ec_enc_shrink(&enc, nb_compr_bytes); + } + +#ifndef DISABLE_FLOAT_API + if (redundancy || st->mode != MODE_SILK_ONLY) + celt_encoder_ctl(celt_enc, CELT_SET_ANALYSIS(&analysis_info)); +#endif + if (st->mode == MODE_HYBRID) { + SILKInfo info; + info.signalType = st->silk_mode.signalType; + info.offset = st->silk_mode.offset; + celt_encoder_ctl(celt_enc, CELT_SET_SILK_INFO(&info)); + } + + /* 5 ms redundant frame for CELT->SILK */ + if (redundancy && celt_to_silk) + { + int err; + celt_encoder_ctl(celt_enc, CELT_SET_START_BAND(0)); + celt_encoder_ctl(celt_enc, OPUS_SET_VBR(0)); + celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(OPUS_BITRATE_MAX)); + err = celt_encode_with_ec(celt_enc, pcm_buf, st->Fs/200, data+nb_compr_bytes, redundancy_bytes, NULL); + if (err < 0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + celt_encoder_ctl(celt_enc, OPUS_GET_FINAL_RANGE(&redundant_rng)); + celt_encoder_ctl(celt_enc, OPUS_RESET_STATE); + } + + celt_encoder_ctl(celt_enc, CELT_SET_START_BAND(start_band)); + + if (st->mode != MODE_SILK_ONLY) + { + if (st->mode != st->prev_mode && st->prev_mode > 0) + { + unsigned char dummy[2]; + celt_encoder_ctl(celt_enc, OPUS_RESET_STATE); + + /* Prefilling */ + celt_encode_with_ec(celt_enc, tmp_prefill, st->Fs/400, dummy, 2, NULL); + celt_encoder_ctl(celt_enc, CELT_SET_PREDICTION(0)); + } + /* If false, we already busted the budget and we'll end up with a "PLC frame" */ + if (ec_tell(&enc) <= 8*nb_compr_bytes) + { + /* Set the bitrate again if it was overridden in the redundancy code above*/ + if (redundancy && celt_to_silk && st->mode==MODE_HYBRID && st->use_vbr) + celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(st->bitrate_bps-st->silk_mode.bitRate)); + celt_encoder_ctl(celt_enc, OPUS_SET_VBR(st->use_vbr)); + ret = celt_encode_with_ec(celt_enc, pcm_buf, frame_size, NULL, nb_compr_bytes, &enc); + if (ret < 0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + /* Put CELT->SILK redundancy data in the right place. */ + if (redundancy && celt_to_silk && st->mode==MODE_HYBRID && st->use_vbr) + { + OPUS_MOVE(data+ret, data+nb_compr_bytes, redundancy_bytes); + nb_compr_bytes = nb_compr_bytes+redundancy_bytes; + } + } + } + + /* 5 ms redundant frame for SILK->CELT */ + if (redundancy && !celt_to_silk) + { + int err; + unsigned char dummy[2]; + int N2, N4; + N2 = st->Fs/200; + N4 = st->Fs/400; + + celt_encoder_ctl(celt_enc, OPUS_RESET_STATE); + celt_encoder_ctl(celt_enc, CELT_SET_START_BAND(0)); + celt_encoder_ctl(celt_enc, CELT_SET_PREDICTION(0)); + celt_encoder_ctl(celt_enc, OPUS_SET_VBR(0)); + celt_encoder_ctl(celt_enc, OPUS_SET_BITRATE(OPUS_BITRATE_MAX)); + + if (st->mode == MODE_HYBRID) + { + /* Shrink packet to what the encoder actually used. */ + nb_compr_bytes = ret; + ec_enc_shrink(&enc, nb_compr_bytes); + } + /* NOTE: We could speed this up slightly (at the expense of code size) by just adding a function that prefills the buffer */ + celt_encode_with_ec(celt_enc, pcm_buf+st->channels*(frame_size-N2-N4), N4, dummy, 2, NULL); + + err = celt_encode_with_ec(celt_enc, pcm_buf+st->channels*(frame_size-N2), N2, data+nb_compr_bytes, redundancy_bytes, NULL); + if (err < 0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + celt_encoder_ctl(celt_enc, OPUS_GET_FINAL_RANGE(&redundant_rng)); + } + + + + /* Signalling the mode in the first byte */ + data--; + data[0] = gen_toc(st->mode, st->Fs/frame_size, curr_bandwidth, st->stream_channels); + + st->rangeFinal = enc.rng ^ redundant_rng; + + if (to_celt) + st->prev_mode = MODE_CELT_ONLY; + else + st->prev_mode = st->mode; + st->prev_channels = st->stream_channels; + st->prev_framesize = frame_size; + + st->first = 0; + + /* DTX decision */ +#ifndef DISABLE_FLOAT_API + if (st->use_dtx && (analysis_info.valid || is_silence)) + { + if (decide_dtx_mode(activity, &st->nb_no_activity_ms_Q1, 2*1000*frame_size/st->Fs)) + { + st->rangeFinal = 0; + data[0] = gen_toc(st->mode, st->Fs/frame_size, curr_bandwidth, st->stream_channels); + RESTORE_STACK; + return 1; + } + } else { + st->nb_no_activity_ms_Q1 = 0; + } +#endif + + /* In the unlikely case that the SILK encoder busted its target, tell + the decoder to call the PLC */ + if (ec_tell(&enc) > (max_data_bytes-1)*8) + { + if (max_data_bytes < 2) + { + RESTORE_STACK; + return OPUS_BUFFER_TOO_SMALL; + } + data[1] = 0; + ret = 1; + st->rangeFinal = 0; + } else if (st->mode==MODE_SILK_ONLY&&!redundancy) + { + /*When in LPC only mode it's perfectly + reasonable to strip off trailing zero bytes as + the required range decoder behavior is to + fill these in. This can't be done when the MDCT + modes are used because the decoder needs to know + the actual length for allocation purposes.*/ + while(ret>2&&data[ret]==0)ret--; + } + /* Count ToC and redundancy */ + ret += 1+redundancy_bytes; + if (!st->use_vbr) + { + if (opus_packet_pad(data, ret, max_data_bytes) != OPUS_OK) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + ret = max_data_bytes; + } + RESTORE_STACK; + return ret; +} + +#ifdef FIXED_POINT + +#ifndef DISABLE_FLOAT_API +opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int analysis_frame_size, + unsigned char *data, opus_int32 max_data_bytes) +{ + int i, ret; + int frame_size; + VARDECL(opus_int16, in); + ALLOC_STACK; + + frame_size = frame_size_select(analysis_frame_size, st->variable_duration, st->Fs); + if (frame_size <= 0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + ALLOC(in, frame_size*st->channels, opus_int16); + + for (i=0;ichannels;i++) + in[i] = FLOAT2INT16(pcm[i]); + ret = opus_encode_native(st, in, frame_size, data, max_data_bytes, 16, + pcm, analysis_frame_size, 0, -2, st->channels, downmix_float, 1); + RESTORE_STACK; + return ret; +} +#endif + +opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int analysis_frame_size, + unsigned char *data, opus_int32 out_data_bytes) +{ + int frame_size; + frame_size = frame_size_select(analysis_frame_size, st->variable_duration, st->Fs); + return opus_encode_native(st, pcm, frame_size, data, out_data_bytes, 16, + pcm, analysis_frame_size, 0, -2, st->channels, downmix_int, 0); +} + +#else +opus_int32 opus_encode(OpusEncoder *st, const opus_int16 *pcm, int analysis_frame_size, + unsigned char *data, opus_int32 max_data_bytes) +{ + int i, ret; + int frame_size; + VARDECL(float, in); + ALLOC_STACK; + + frame_size = frame_size_select(analysis_frame_size, st->variable_duration, st->Fs); + if (frame_size <= 0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + ALLOC(in, frame_size*st->channels, float); + + for (i=0;ichannels;i++) + in[i] = (1.0f/32768)*pcm[i]; + ret = opus_encode_native(st, in, frame_size, data, max_data_bytes, 16, + pcm, analysis_frame_size, 0, -2, st->channels, downmix_int, 0); + RESTORE_STACK; + return ret; +} +opus_int32 opus_encode_float(OpusEncoder *st, const float *pcm, int analysis_frame_size, + unsigned char *data, opus_int32 out_data_bytes) +{ + int frame_size; + frame_size = frame_size_select(analysis_frame_size, st->variable_duration, st->Fs); + return opus_encode_native(st, pcm, frame_size, data, out_data_bytes, 24, + pcm, analysis_frame_size, 0, -2, st->channels, downmix_float, 1); +} +#endif + + +int opus_encoder_ctl(OpusEncoder *st, int request, ...) +{ + int ret; + CELTEncoder *celt_enc; + va_list ap; + + ret = OPUS_OK; + va_start(ap, request); + + celt_enc = (CELTEncoder*)((char*)st+st->celt_enc_offset); + + switch (request) + { + case OPUS_SET_APPLICATION_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if ( (value != OPUS_APPLICATION_VOIP && value != OPUS_APPLICATION_AUDIO + && value != OPUS_APPLICATION_RESTRICTED_LOWDELAY) + || (!st->first && st->application != value)) + { + ret = OPUS_BAD_ARG; + break; + } + st->application = value; +#ifndef DISABLE_FLOAT_API + st->analysis.application = value; +#endif + } + break; + case OPUS_GET_APPLICATION_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->application; + } + break; + case OPUS_SET_BITRATE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value != OPUS_AUTO && value != OPUS_BITRATE_MAX) + { + if (value <= 0) + goto bad_arg; + else if (value <= 500) + value = 500; + else if (value > (opus_int32)300000*st->channels) + value = (opus_int32)300000*st->channels; + } + st->user_bitrate_bps = value; + } + break; + case OPUS_GET_BITRATE_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = user_bitrate_to_bitrate(st, st->prev_framesize, 1276); + } + break; + case OPUS_SET_FORCE_CHANNELS_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if((value<1 || value>st->channels) && value != OPUS_AUTO) + { + goto bad_arg; + } + st->force_channels = value; + } + break; + case OPUS_GET_FORCE_CHANNELS_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->force_channels; + } + break; + case OPUS_SET_MAX_BANDWIDTH_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND) + { + goto bad_arg; + } + st->max_bandwidth = value; + if (st->max_bandwidth == OPUS_BANDWIDTH_NARROWBAND) { + st->silk_mode.maxInternalSampleRate = 8000; + } else if (st->max_bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) { + st->silk_mode.maxInternalSampleRate = 12000; + } else { + st->silk_mode.maxInternalSampleRate = 16000; + } + } + break; + case OPUS_GET_MAX_BANDWIDTH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->max_bandwidth; + } + break; + case OPUS_SET_BANDWIDTH_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if ((value < OPUS_BANDWIDTH_NARROWBAND || value > OPUS_BANDWIDTH_FULLBAND) && value != OPUS_AUTO) + { + goto bad_arg; + } + st->user_bandwidth = value; + if (st->user_bandwidth == OPUS_BANDWIDTH_NARROWBAND) { + st->silk_mode.maxInternalSampleRate = 8000; + } else if (st->user_bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) { + st->silk_mode.maxInternalSampleRate = 12000; + } else { + st->silk_mode.maxInternalSampleRate = 16000; + } + } + break; + case OPUS_GET_BANDWIDTH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->bandwidth; + } + break; + case OPUS_SET_DTX_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>1) + { + goto bad_arg; + } + st->use_dtx = value; + } + break; + case OPUS_GET_DTX_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->use_dtx; + } + break; + case OPUS_SET_COMPLEXITY_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>10) + { + goto bad_arg; + } + st->silk_mode.complexity = value; + celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(value)); + } + break; + case OPUS_GET_COMPLEXITY_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->silk_mode.complexity; + } + break; + case OPUS_SET_INBAND_FEC_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>2) + { + goto bad_arg; + } + st->fec_config = value; + st->silk_mode.useInBandFEC = (value != 0); + } + break; + case OPUS_GET_INBAND_FEC_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->fec_config; + } + break; + case OPUS_SET_PACKET_LOSS_PERC_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value < 0 || value > 100) + { + goto bad_arg; + } + st->silk_mode.packetLossPercentage = value; + celt_encoder_ctl(celt_enc, OPUS_SET_PACKET_LOSS_PERC(value)); + } + break; + case OPUS_GET_PACKET_LOSS_PERC_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->silk_mode.packetLossPercentage; + } + break; + case OPUS_SET_VBR_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>1) + { + goto bad_arg; + } + st->use_vbr = value; + st->silk_mode.useCBR = 1-value; + } + break; + case OPUS_GET_VBR_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->use_vbr; + } + break; + case OPUS_SET_VOICE_RATIO_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<-1 || value>100) + { + goto bad_arg; + } + st->voice_ratio = value; + } + break; + case OPUS_GET_VOICE_RATIO_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->voice_ratio; + } + break; + case OPUS_SET_VBR_CONSTRAINT_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>1) + { + goto bad_arg; + } + st->vbr_constraint = value; + } + break; + case OPUS_GET_VBR_CONSTRAINT_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->vbr_constraint; + } + break; + case OPUS_SET_SIGNAL_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value!=OPUS_AUTO && value!=OPUS_SIGNAL_VOICE && value!=OPUS_SIGNAL_MUSIC) + { + goto bad_arg; + } + st->signal_type = value; + } + break; + case OPUS_GET_SIGNAL_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->signal_type; + } + break; + case OPUS_GET_LOOKAHEAD_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->Fs/400; + if (st->application != OPUS_APPLICATION_RESTRICTED_LOWDELAY) + *value += st->delay_compensation; + } + break; + case OPUS_GET_SAMPLE_RATE_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->Fs; + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + opus_uint32 *value = va_arg(ap, opus_uint32*); + if (!value) + { + goto bad_arg; + } + *value = st->rangeFinal; + } + break; + case OPUS_SET_LSB_DEPTH_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value<8 || value>24) + { + goto bad_arg; + } + st->lsb_depth=value; + } + break; + case OPUS_GET_LSB_DEPTH_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->lsb_depth; + } + break; + case OPUS_SET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value != OPUS_FRAMESIZE_ARG && value != OPUS_FRAMESIZE_2_5_MS && + value != OPUS_FRAMESIZE_5_MS && value != OPUS_FRAMESIZE_10_MS && + value != OPUS_FRAMESIZE_20_MS && value != OPUS_FRAMESIZE_40_MS && + value != OPUS_FRAMESIZE_60_MS && value != OPUS_FRAMESIZE_80_MS && + value != OPUS_FRAMESIZE_100_MS && value != OPUS_FRAMESIZE_120_MS) + { + goto bad_arg; + } + st->variable_duration = value; + } + break; + case OPUS_GET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->variable_duration; + } + break; + case OPUS_SET_PREDICTION_DISABLED_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if (value > 1 || value < 0) + goto bad_arg; + st->silk_mode.reducedDependency = value; + } + break; + case OPUS_GET_PREDICTION_DISABLED_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + goto bad_arg; + *value = st->silk_mode.reducedDependency; + } + break; + case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if(value<0 || value>1) + { + goto bad_arg; + } + celt_encoder_ctl(celt_enc, OPUS_SET_PHASE_INVERSION_DISABLED(value)); + } + break; + case OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + celt_encoder_ctl(celt_enc, OPUS_GET_PHASE_INVERSION_DISABLED(value)); + } + break; + case OPUS_RESET_STATE: + { + void *silk_enc; + silk_EncControlStruct dummy; + char *start; + silk_enc = (char*)st+st->silk_enc_offset; +#ifndef DISABLE_FLOAT_API + tonality_analysis_reset(&st->analysis); +#endif + + start = (char*)&st->OPUS_ENCODER_RESET_START; + OPUS_CLEAR(start, sizeof(OpusEncoder) - (start - (char*)st)); + + celt_encoder_ctl(celt_enc, OPUS_RESET_STATE); + silk_InitEncoder( silk_enc, st->arch, &dummy ); + st->stream_channels = st->channels; + st->hybrid_stereo_width_Q14 = 1 << 14; + st->prev_HB_gain = Q15ONE; + st->first = 1; + st->mode = MODE_HYBRID; + st->bandwidth = OPUS_BANDWIDTH_FULLBAND; + st->variable_HP_smth2_Q15 = silk_LSHIFT( silk_lin2log( VARIABLE_HP_MIN_CUTOFF_HZ ), 8 ); + } + break; + case OPUS_SET_FORCE_MODE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + if ((value < MODE_SILK_ONLY || value > MODE_CELT_ONLY) && value != OPUS_AUTO) + { + goto bad_arg; + } + st->user_forced_mode = value; + } + break; + case OPUS_SET_LFE_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->lfe = value; + ret = celt_encoder_ctl(celt_enc, OPUS_SET_LFE(value)); + } + break; + case OPUS_SET_ENERGY_MASK_REQUEST: + { + opus_val16 *value = va_arg(ap, opus_val16*); + st->energy_masking = value; + ret = celt_encoder_ctl(celt_enc, OPUS_SET_ENERGY_MASK(value)); + } + break; + case OPUS_GET_IN_DTX_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + if (st->silk_mode.useDTX && (st->prev_mode == MODE_SILK_ONLY || st->prev_mode == MODE_HYBRID)) { + /* DTX determined by Silk. */ + silk_encoder *silk_enc = (silk_encoder*)(void *)((char*)st+st->silk_enc_offset); + *value = silk_enc->state_Fxx[0].sCmn.noSpeechCounter >= NB_SPEECH_FRAMES_BEFORE_DTX; + /* Stereo: check second channel unless only the middle channel was encoded. */ + if(*value == 1 && st->silk_mode.nChannelsInternal == 2 && silk_enc->prev_decode_only_middle == 0) { + *value = silk_enc->state_Fxx[1].sCmn.noSpeechCounter >= NB_SPEECH_FRAMES_BEFORE_DTX; + } + } +#ifndef DISABLE_FLOAT_API + else if (st->use_dtx) { + /* DTX determined by Opus. */ + *value = st->nb_no_activity_ms_Q1 >= NB_SPEECH_FRAMES_BEFORE_DTX*20*2; + } +#endif + else { + *value = 0; + } + } + break; + case CELT_GET_MODE_REQUEST: + { + const CELTMode ** value = va_arg(ap, const CELTMode**); + if (!value) + { + goto bad_arg; + } + ret = celt_encoder_ctl(celt_enc, CELT_GET_MODE(value)); + } + break; + default: + /* fprintf(stderr, "unknown opus_encoder_ctl() request: %d", request);*/ + ret = OPUS_UNIMPLEMENTED; + break; + } + va_end(ap); + return ret; +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +} + +void opus_encoder_destroy(OpusEncoder *st) +{ + opus_free(st); +} diff --git a/lib/opus/src/opus_multistream.c b/lib/opus/src/opus_multistream.c new file mode 100644 index 00000000..09c3639b --- /dev/null +++ b/lib/opus/src/opus_multistream.c @@ -0,0 +1,92 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_multistream.h" +#include "opus.h" +#include "opus_private.h" +#include "stack_alloc.h" +#include +#include "float_cast.h" +#include "os_support.h" + + +int validate_layout(const ChannelLayout *layout) +{ + int i, max_channel; + + max_channel = layout->nb_streams+layout->nb_coupled_streams; + if (max_channel>255) + return 0; + for (i=0;inb_channels;i++) + { + if (layout->mapping[i] >= max_channel && layout->mapping[i] != 255) + return 0; + } + return 1; +} + + +int get_left_channel(const ChannelLayout *layout, int stream_id, int prev) +{ + int i; + i = (prev<0) ? 0 : prev+1; + for (;inb_channels;i++) + { + if (layout->mapping[i]==stream_id*2) + return i; + } + return -1; +} + +int get_right_channel(const ChannelLayout *layout, int stream_id, int prev) +{ + int i; + i = (prev<0) ? 0 : prev+1; + for (;inb_channels;i++) + { + if (layout->mapping[i]==stream_id*2+1) + return i; + } + return -1; +} + +int get_mono_channel(const ChannelLayout *layout, int stream_id, int prev) +{ + int i; + i = (prev<0) ? 0 : prev+1; + for (;inb_channels;i++) + { + if (layout->mapping[i]==stream_id+layout->nb_coupled_streams) + return i; + } + return -1; +} + diff --git a/lib/opus/src/opus_multistream_decoder.c b/lib/opus/src/opus_multistream_decoder.c new file mode 100644 index 00000000..a2837c35 --- /dev/null +++ b/lib/opus/src/opus_multistream_decoder.c @@ -0,0 +1,552 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_multistream.h" +#include "opus.h" +#include "opus_private.h" +#include "stack_alloc.h" +#include +#include "float_cast.h" +#include "os_support.h" + +/* DECODER */ + +#if defined(ENABLE_HARDENING) || defined(ENABLE_ASSERTIONS) +static void validate_ms_decoder(OpusMSDecoder *st) +{ + validate_layout(&st->layout); +} +#define VALIDATE_MS_DECODER(st) validate_ms_decoder(st) +#else +#define VALIDATE_MS_DECODER(st) +#endif + + +opus_int32 opus_multistream_decoder_get_size(int nb_streams, int nb_coupled_streams) +{ + int coupled_size; + int mono_size; + + if(nb_streams<1||nb_coupled_streams>nb_streams||nb_coupled_streams<0)return 0; + coupled_size = opus_decoder_get_size(2); + mono_size = opus_decoder_get_size(1); + return align(sizeof(OpusMSDecoder)) + + nb_coupled_streams * align(coupled_size) + + (nb_streams-nb_coupled_streams) * align(mono_size); +} + +int opus_multistream_decoder_init( + OpusMSDecoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping +) +{ + int coupled_size; + int mono_size; + int i, ret; + char *ptr; + + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (streams<1) || (coupled_streams<0) || (streams>255-coupled_streams)) + return OPUS_BAD_ARG; + + st->layout.nb_channels = channels; + st->layout.nb_streams = streams; + st->layout.nb_coupled_streams = coupled_streams; + + for (i=0;ilayout.nb_channels;i++) + st->layout.mapping[i] = mapping[i]; + if (!validate_layout(&st->layout)) + return OPUS_BAD_ARG; + + ptr = (char*)st + align(sizeof(OpusMSDecoder)); + coupled_size = opus_decoder_get_size(2); + mono_size = opus_decoder_get_size(1); + + for (i=0;ilayout.nb_coupled_streams;i++) + { + ret=opus_decoder_init((OpusDecoder*)ptr, Fs, 2); + if(ret!=OPUS_OK)return ret; + ptr += align(coupled_size); + } + for (;ilayout.nb_streams;i++) + { + ret=opus_decoder_init((OpusDecoder*)ptr, Fs, 1); + if(ret!=OPUS_OK)return ret; + ptr += align(mono_size); + } + return OPUS_OK; +} + + +OpusMSDecoder *opus_multistream_decoder_create( + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int *error +) +{ + int ret; + OpusMSDecoder *st; + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (streams<1) || (coupled_streams<0) || (streams>255-coupled_streams)) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + st = (OpusMSDecoder *)opus_alloc(opus_multistream_decoder_get_size(streams, coupled_streams)); + if (st==NULL) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + ret = opus_multistream_decoder_init(st, Fs, channels, streams, coupled_streams, mapping); + if (error) + *error = ret; + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + return st; +} + +static int opus_multistream_packet_validate(const unsigned char *data, + opus_int32 len, int nb_streams, opus_int32 Fs) +{ + int s; + int count; + unsigned char toc; + opus_int16 size[48]; + int samples=0; + opus_int32 packet_offset; + + for (s=0;slayout.nb_streams-1) + { + RESTORE_STACK; + return OPUS_INVALID_PACKET; + } + if (!do_plc) + { + int ret = opus_multistream_packet_validate(data, len, st->layout.nb_streams, Fs); + if (ret < 0) + { + RESTORE_STACK; + return ret; + } else if (ret > frame_size) + { + RESTORE_STACK; + return OPUS_BUFFER_TOO_SMALL; + } + } + for (s=0;slayout.nb_streams;s++) + { + OpusDecoder *dec; + opus_int32 packet_offset; + int ret; + + dec = (OpusDecoder*)ptr; + ptr += (s < st->layout.nb_coupled_streams) ? align(coupled_size) : align(mono_size); + + if (!do_plc && len<=0) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + packet_offset = 0; + ret = opus_decode_native(dec, data, len, buf, frame_size, decode_fec, s!=st->layout.nb_streams-1, &packet_offset, soft_clip); + if (!do_plc) + { + data += packet_offset; + len -= packet_offset; + } + if (ret <= 0) + { + RESTORE_STACK; + return ret; + } + frame_size = ret; + if (s < st->layout.nb_coupled_streams) + { + int chan, prev; + prev = -1; + /* Copy "left" audio to the channel(s) where it belongs */ + while ( (chan = get_left_channel(&st->layout, s, prev)) != -1) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, chan, + buf, 2, frame_size, user_data); + prev = chan; + } + prev = -1; + /* Copy "right" audio to the channel(s) where it belongs */ + while ( (chan = get_right_channel(&st->layout, s, prev)) != -1) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, chan, + buf+1, 2, frame_size, user_data); + prev = chan; + } + } else { + int chan, prev; + prev = -1; + /* Copy audio to the channel(s) where it belongs */ + while ( (chan = get_mono_channel(&st->layout, s, prev)) != -1) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, chan, + buf, 1, frame_size, user_data); + prev = chan; + } + } + } + /* Handle muted channels */ + for (c=0;clayout.nb_channels;c++) + { + if (st->layout.mapping[c] == 255) + { + (*copy_channel_out)(pcm, st->layout.nb_channels, c, + NULL, 0, frame_size, user_data); + } + } + RESTORE_STACK; + return frame_size; +} + +#if !defined(DISABLE_FLOAT_API) +static void opus_copy_channel_out_float( + void *dst, + int dst_stride, + int dst_channel, + const opus_val16 *src, + int src_stride, + int frame_size, + void *user_data +) +{ + float *float_dst; + opus_int32 i; + (void)user_data; + float_dst = (float*)dst; + if (src != NULL) + { + for (i=0;ilayout.nb_streams;s++) + { + OpusDecoder *dec; + dec = (OpusDecoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_decoder_ctl(dec, request, &tmp); + if (ret != OPUS_OK) break; + *value ^= tmp; + } + } + break; + case OPUS_RESET_STATE: + { + int s; + for (s=0;slayout.nb_streams;s++) + { + OpusDecoder *dec; + + dec = (OpusDecoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_decoder_ctl(dec, OPUS_RESET_STATE); + if (ret != OPUS_OK) + break; + } + } + break; + case OPUS_MULTISTREAM_GET_DECODER_STATE_REQUEST: + { + int s; + opus_int32 stream_id; + OpusDecoder **value; + stream_id = va_arg(ap, opus_int32); + if (stream_id<0 || stream_id >= st->layout.nb_streams) + goto bad_arg; + value = va_arg(ap, OpusDecoder**); + if (!value) + { + goto bad_arg; + } + for (s=0;slayout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + *value = (OpusDecoder*)ptr; + } + break; + case OPUS_SET_GAIN_REQUEST: + case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST: + { + int s; + /* This works for int32 params */ + opus_int32 value = va_arg(ap, opus_int32); + for (s=0;slayout.nb_streams;s++) + { + OpusDecoder *dec; + + dec = (OpusDecoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_decoder_ctl(dec, request, value); + if (ret != OPUS_OK) + break; + } + } + break; + default: + ret = OPUS_UNIMPLEMENTED; + break; + } + return ret; +bad_arg: + return OPUS_BAD_ARG; +} + +int opus_multistream_decoder_ctl(OpusMSDecoder *st, int request, ...) +{ + int ret; + va_list ap; + va_start(ap, request); + ret = opus_multistream_decoder_ctl_va_list(st, request, ap); + va_end(ap); + return ret; +} + +void opus_multistream_decoder_destroy(OpusMSDecoder *st) +{ + opus_free(st); +} diff --git a/lib/opus/src/opus_multistream_encoder.c b/lib/opus/src/opus_multistream_encoder.c new file mode 100644 index 00000000..213e3eb2 --- /dev/null +++ b/lib/opus/src/opus_multistream_encoder.c @@ -0,0 +1,1329 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus_multistream.h" +#include "opus.h" +#include "opus_private.h" +#include "stack_alloc.h" +#include +#include "float_cast.h" +#include "os_support.h" +#include "mathops.h" +#include "mdct.h" +#include "modes.h" +#include "bands.h" +#include "quant_bands.h" +#include "pitch.h" + +typedef struct { + int nb_streams; + int nb_coupled_streams; + unsigned char mapping[8]; +} VorbisLayout; + +/* Index is nb_channel-1*/ +static const VorbisLayout vorbis_mappings[8] = { + {1, 0, {0}}, /* 1: mono */ + {1, 1, {0, 1}}, /* 2: stereo */ + {2, 1, {0, 2, 1}}, /* 3: 1-d surround */ + {2, 2, {0, 1, 2, 3}}, /* 4: quadraphonic surround */ + {3, 2, {0, 4, 1, 2, 3}}, /* 5: 5-channel surround */ + {4, 2, {0, 4, 1, 2, 3, 5}}, /* 6: 5.1 surround */ + {4, 3, {0, 4, 1, 2, 3, 5, 6}}, /* 7: 6.1 surround */ + {5, 3, {0, 6, 1, 2, 3, 4, 5, 7}}, /* 8: 7.1 surround */ +}; + +static opus_val32 *ms_get_preemph_mem(OpusMSEncoder *st) +{ + int s; + char *ptr; + int coupled_size, mono_size; + + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + for (s=0;slayout.nb_streams;s++) + { + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + /* void* cast avoids clang -Wcast-align warning */ + return (opus_val32*)(void*)(ptr+st->layout.nb_channels*120*sizeof(opus_val32)); +} + +static opus_val32 *ms_get_window_mem(OpusMSEncoder *st) +{ + int s; + char *ptr; + int coupled_size, mono_size; + + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + for (s=0;slayout.nb_streams;s++) + { + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + /* void* cast avoids clang -Wcast-align warning */ + return (opus_val32*)(void*)ptr; +} + +static int validate_ambisonics(int nb_channels, int *nb_streams, int *nb_coupled_streams) +{ + int order_plus_one; + int acn_channels; + int nondiegetic_channels; + + if (nb_channels < 1 || nb_channels > 227) + return 0; + + order_plus_one = isqrt32(nb_channels); + acn_channels = order_plus_one * order_plus_one; + nondiegetic_channels = nb_channels - acn_channels; + + if (nondiegetic_channels != 0 && nondiegetic_channels != 2) + return 0; + + if (nb_streams) + *nb_streams = acn_channels + (nondiegetic_channels != 0); + if (nb_coupled_streams) + *nb_coupled_streams = nondiegetic_channels != 0; + return 1; +} + +static int validate_encoder_layout(const ChannelLayout *layout) +{ + int s; + for (s=0;snb_streams;s++) + { + if (s < layout->nb_coupled_streams) + { + if (get_left_channel(layout, s, -1)==-1) + return 0; + if (get_right_channel(layout, s, -1)==-1) + return 0; + } else { + if (get_mono_channel(layout, s, -1)==-1) + return 0; + } + } + return 1; +} + +static void channel_pos(int channels, int pos[8]) +{ + /* Position in the mix: 0 don't mix, 1: left, 2: center, 3:right */ + if (channels==4) + { + pos[0]=1; + pos[1]=3; + pos[2]=1; + pos[3]=3; + } else if (channels==3||channels==5||channels==6) + { + pos[0]=1; + pos[1]=2; + pos[2]=3; + pos[3]=1; + pos[4]=3; + pos[5]=0; + } else if (channels==7) + { + pos[0]=1; + pos[1]=2; + pos[2]=3; + pos[3]=1; + pos[4]=3; + pos[5]=2; + pos[6]=0; + } else if (channels==8) + { + pos[0]=1; + pos[1]=2; + pos[2]=3; + pos[3]=1; + pos[4]=3; + pos[5]=1; + pos[6]=3; + pos[7]=0; + } +} + +#if 1 +/* Computes a rough approximation of log2(2^a + 2^b) */ +static opus_val16 logSum(opus_val16 a, opus_val16 b) +{ + opus_val16 max; + opus_val32 diff; + opus_val16 frac; + static const opus_val16 diff_table[17] = { + QCONST16(0.5000000f, DB_SHIFT), QCONST16(0.2924813f, DB_SHIFT), QCONST16(0.1609640f, DB_SHIFT), QCONST16(0.0849625f, DB_SHIFT), + QCONST16(0.0437314f, DB_SHIFT), QCONST16(0.0221971f, DB_SHIFT), QCONST16(0.0111839f, DB_SHIFT), QCONST16(0.0056136f, DB_SHIFT), + QCONST16(0.0028123f, DB_SHIFT) + }; + int low; + if (a>b) + { + max = a; + diff = SUB32(EXTEND32(a),EXTEND32(b)); + } else { + max = b; + diff = SUB32(EXTEND32(b),EXTEND32(a)); + } + if (!(diff < QCONST16(8.f, DB_SHIFT))) /* inverted to catch NaNs */ + return max; +#ifdef FIXED_POINT + low = SHR32(diff, DB_SHIFT-1); + frac = SHL16(diff - SHL16(low, DB_SHIFT-1), 16-DB_SHIFT); +#else + low = (int)floor(2*diff); + frac = 2*diff - low; +#endif + return max + diff_table[low] + MULT16_16_Q15(frac, SUB16(diff_table[low+1], diff_table[low])); +} +#else +opus_val16 logSum(opus_val16 a, opus_val16 b) +{ + return log2(pow(4, a)+ pow(4, b))/2; +} +#endif + +void surround_analysis(const CELTMode *celt_mode, const void *pcm, opus_val16 *bandLogE, opus_val32 *mem, opus_val32 *preemph_mem, + int len, int overlap, int channels, int rate, opus_copy_channel_in_func copy_channel_in, int arch +) +{ + int c; + int i; + int LM; + int pos[8] = {0}; + int upsample; + int frame_size; + int freq_size; + opus_val16 channel_offset; + opus_val32 bandE[21]; + opus_val16 maskLogE[3][21]; + VARDECL(opus_val32, in); + VARDECL(opus_val16, x); + VARDECL(opus_val32, freq); + SAVE_STACK; + + upsample = resampling_factor(rate); + frame_size = len*upsample; + freq_size = IMIN(960, frame_size); + + /* LM = log2(frame_size / 120) */ + for (LM=0;LMmaxLM;LM++) + if (celt_mode->shortMdctSize<preemph, preemph_mem+c, 0); +#ifndef FIXED_POINT + { + opus_val32 sum; + sum = celt_inner_prod(in, in, frame_size+overlap, 0); + /* This should filter out both NaNs and ridiculous signals that could + cause NaNs further down. */ + if (!(sum < 1e18f) || celt_isnan(sum)) + { + OPUS_CLEAR(in, frame_size+overlap); + preemph_mem[c] = 0; + } + } +#endif + OPUS_CLEAR(bandE, 21); + for (frame=0;framemdct, in+960*frame, freq, celt_mode->window, + overlap, celt_mode->maxLM-LM, 1, arch); + if (upsample != 1) + { + int bound = freq_size/upsample; + for (i=0;i=0;i--) + bandLogE[21*c+i] = MAX16(bandLogE[21*c+i], bandLogE[21*c+i+1]-QCONST16(2.f, DB_SHIFT)); + if (pos[c]==1) + { + for (i=0;i<21;i++) + maskLogE[0][i] = logSum(maskLogE[0][i], bandLogE[21*c+i]); + } else if (pos[c]==3) + { + for (i=0;i<21;i++) + maskLogE[2][i] = logSum(maskLogE[2][i], bandLogE[21*c+i]); + } else if (pos[c]==2) + { + for (i=0;i<21;i++) + { + maskLogE[0][i] = logSum(maskLogE[0][i], bandLogE[21*c+i]-QCONST16(.5f, DB_SHIFT)); + maskLogE[2][i] = logSum(maskLogE[2][i], bandLogE[21*c+i]-QCONST16(.5f, DB_SHIFT)); + } + } +#if 0 + for (i=0;i<21;i++) + printf("%f ", bandLogE[21*c+i]); + float sum=0; + for (i=0;i<21;i++) + sum += bandLogE[21*c+i]; + printf("%f ", sum/21); +#endif + OPUS_COPY(mem+c*overlap, in+frame_size, overlap); + } + for (i=0;i<21;i++) + maskLogE[1][i] = MIN32(maskLogE[0][i],maskLogE[2][i]); + channel_offset = HALF16(celt_log2(QCONST32(2.f,14)/(channels-1))); + for (c=0;c<3;c++) + for (i=0;i<21;i++) + maskLogE[c][i] += channel_offset; +#if 0 + for (c=0;c<3;c++) + { + for (i=0;i<21;i++) + printf("%f ", maskLogE[c][i]); + } +#endif + for (c=0;cnb_streams||nb_coupled_streams<0)return 0; + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + return align(sizeof(OpusMSEncoder)) + + nb_coupled_streams * align(coupled_size) + + (nb_streams-nb_coupled_streams) * align(mono_size); +} + +opus_int32 opus_multistream_surround_encoder_get_size(int channels, int mapping_family) +{ + int nb_streams; + int nb_coupled_streams; + opus_int32 size; + + if (mapping_family==0) + { + if (channels==1) + { + nb_streams=1; + nb_coupled_streams=0; + } else if (channels==2) + { + nb_streams=1; + nb_coupled_streams=1; + } else + return 0; + } else if (mapping_family==1 && channels<=8 && channels>=1) + { + nb_streams=vorbis_mappings[channels-1].nb_streams; + nb_coupled_streams=vorbis_mappings[channels-1].nb_coupled_streams; + } else if (mapping_family==255) + { + nb_streams=channels; + nb_coupled_streams=0; + } else if (mapping_family==2) + { + if (!validate_ambisonics(channels, &nb_streams, &nb_coupled_streams)) + return 0; + } else + return 0; + size = opus_multistream_encoder_get_size(nb_streams, nb_coupled_streams); + if (channels>2) + { + size += channels*(120*sizeof(opus_val32) + sizeof(opus_val32)); + } + return size; +} + +static int opus_multistream_encoder_init_impl( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application, + MappingType mapping_type +) +{ + int coupled_size; + int mono_size; + int i, ret; + char *ptr; + + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (streams<1) || (coupled_streams<0) || (streams>255-coupled_streams) || + (streams+coupled_streams>channels)) + return OPUS_BAD_ARG; + + st->arch = opus_select_arch(); + st->layout.nb_channels = channels; + st->layout.nb_streams = streams; + st->layout.nb_coupled_streams = coupled_streams; + if (mapping_type != MAPPING_TYPE_SURROUND) + st->lfe_stream = -1; + st->bitrate_bps = OPUS_AUTO; + st->application = application; + st->variable_duration = OPUS_FRAMESIZE_ARG; + for (i=0;ilayout.nb_channels;i++) + st->layout.mapping[i] = mapping[i]; + if (!validate_layout(&st->layout)) + return OPUS_BAD_ARG; + if (!validate_encoder_layout(&st->layout)) + return OPUS_BAD_ARG; + if (mapping_type == MAPPING_TYPE_AMBISONICS && + !validate_ambisonics(st->layout.nb_channels, NULL, NULL)) + return OPUS_BAD_ARG; + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + + for (i=0;ilayout.nb_coupled_streams;i++) + { + ret = opus_encoder_init((OpusEncoder*)ptr, Fs, 2, application); + if(ret!=OPUS_OK)return ret; + if (i==st->lfe_stream) + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_SET_LFE(1)); + ptr += align(coupled_size); + } + for (;ilayout.nb_streams;i++) + { + ret = opus_encoder_init((OpusEncoder*)ptr, Fs, 1, application); + if (i==st->lfe_stream) + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_SET_LFE(1)); + if(ret!=OPUS_OK)return ret; + ptr += align(mono_size); + } + if (mapping_type == MAPPING_TYPE_SURROUND) + { + OPUS_CLEAR(ms_get_preemph_mem(st), channels); + OPUS_CLEAR(ms_get_window_mem(st), channels*120); + } + st->mapping_type = mapping_type; + return OPUS_OK; +} + +int opus_multistream_encoder_init( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application +) +{ + return opus_multistream_encoder_init_impl(st, Fs, channels, streams, + coupled_streams, mapping, + application, MAPPING_TYPE_NONE); +} + +int opus_multistream_surround_encoder_init( + OpusMSEncoder *st, + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application +) +{ + MappingType mapping_type; + + if ((channels>255) || (channels<1)) + return OPUS_BAD_ARG; + st->lfe_stream = -1; + if (mapping_family==0) + { + if (channels==1) + { + *streams=1; + *coupled_streams=0; + mapping[0]=0; + } else if (channels==2) + { + *streams=1; + *coupled_streams=1; + mapping[0]=0; + mapping[1]=1; + } else + return OPUS_UNIMPLEMENTED; + } else if (mapping_family==1 && channels<=8 && channels>=1) + { + int i; + *streams=vorbis_mappings[channels-1].nb_streams; + *coupled_streams=vorbis_mappings[channels-1].nb_coupled_streams; + for (i=0;i=6) + st->lfe_stream = *streams-1; + } else if (mapping_family==255) + { + int i; + *streams=channels; + *coupled_streams=0; + for(i=0;i2 && mapping_family==1) { + mapping_type = MAPPING_TYPE_SURROUND; + } else if (mapping_family==2) + { + mapping_type = MAPPING_TYPE_AMBISONICS; + } else + { + mapping_type = MAPPING_TYPE_NONE; + } + return opus_multistream_encoder_init_impl(st, Fs, channels, *streams, + *coupled_streams, mapping, + application, mapping_type); +} + +OpusMSEncoder *opus_multistream_encoder_create( + opus_int32 Fs, + int channels, + int streams, + int coupled_streams, + const unsigned char *mapping, + int application, + int *error +) +{ + int ret; + OpusMSEncoder *st; + if ((channels>255) || (channels<1) || (coupled_streams>streams) || + (streams<1) || (coupled_streams<0) || (streams>255-coupled_streams) || + (streams+coupled_streams>channels)) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + st = (OpusMSEncoder *)opus_alloc(opus_multistream_encoder_get_size(streams, coupled_streams)); + if (st==NULL) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + ret = opus_multistream_encoder_init(st, Fs, channels, streams, coupled_streams, mapping, application); + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} + +OpusMSEncoder *opus_multistream_surround_encoder_create( + opus_int32 Fs, + int channels, + int mapping_family, + int *streams, + int *coupled_streams, + unsigned char *mapping, + int application, + int *error +) +{ + int ret; + opus_int32 size; + OpusMSEncoder *st; + if ((channels>255) || (channels<1)) + { + if (error) + *error = OPUS_BAD_ARG; + return NULL; + } + size = opus_multistream_surround_encoder_get_size(channels, mapping_family); + if (!size) + { + if (error) + *error = OPUS_UNIMPLEMENTED; + return NULL; + } + st = (OpusMSEncoder *)opus_alloc(size); + if (st==NULL) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + ret = opus_multistream_surround_encoder_init(st, Fs, channels, mapping_family, streams, coupled_streams, mapping, application); + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} + +static void surround_rate_allocation( + OpusMSEncoder *st, + opus_int32 *rate, + int frame_size, + opus_int32 Fs + ) +{ + int i; + opus_int32 channel_rate; + int stream_offset; + int lfe_offset; + int coupled_ratio; /* Q8 */ + int lfe_ratio; /* Q8 */ + int nb_lfe; + int nb_uncoupled; + int nb_coupled; + int nb_normal; + opus_int32 channel_offset; + opus_int32 bitrate; + int total; + + nb_lfe = (st->lfe_stream!=-1); + nb_coupled = st->layout.nb_coupled_streams; + nb_uncoupled = st->layout.nb_streams-nb_coupled-nb_lfe; + nb_normal = 2*nb_coupled + nb_uncoupled; + + /* Give each non-LFE channel enough bits per channel for coding band energy. */ + channel_offset = 40*IMAX(50, Fs/frame_size); + + if (st->bitrate_bps==OPUS_AUTO) + { + bitrate = nb_normal*(channel_offset + Fs + 10000) + 8000*nb_lfe; + } else if (st->bitrate_bps==OPUS_BITRATE_MAX) + { + bitrate = nb_normal*300000 + nb_lfe*128000; + } else { + bitrate = st->bitrate_bps; + } + + /* Give LFE some basic stream_channel allocation but never exceed 1/20 of the + total rate for the non-energy part to avoid problems at really low rate. */ + lfe_offset = IMIN(bitrate/20, 3000) + 15*IMAX(50, Fs/frame_size); + + /* We give each stream (coupled or uncoupled) a starting bitrate. + This models the main saving of coupled channels over uncoupled. */ + stream_offset = (bitrate - channel_offset*nb_normal - lfe_offset*nb_lfe)/nb_normal/2; + stream_offset = IMAX(0, IMIN(20000, stream_offset)); + + /* Coupled streams get twice the mono rate after the offset is allocated. */ + coupled_ratio = 512; + /* Should depend on the bitrate, for now we assume LFE gets 1/8 the bits of mono */ + lfe_ratio = 32; + + total = (nb_uncoupled<<8) /* mono */ + + coupled_ratio*nb_coupled /* stereo */ + + nb_lfe*lfe_ratio; + channel_rate = 256*(opus_int64)(bitrate - lfe_offset*nb_lfe - stream_offset*(nb_coupled+nb_uncoupled) - channel_offset*nb_normal)/total; + + for (i=0;ilayout.nb_streams;i++) + { + if (ilayout.nb_coupled_streams) + rate[i] = 2*channel_offset + IMAX(0, stream_offset+(channel_rate*coupled_ratio>>8)); + else if (i!=st->lfe_stream) + rate[i] = channel_offset + IMAX(0, stream_offset + channel_rate); + else + rate[i] = IMAX(0, lfe_offset+(channel_rate*lfe_ratio>>8)); + } +} + +static void ambisonics_rate_allocation( + OpusMSEncoder *st, + opus_int32 *rate, + int frame_size, + opus_int32 Fs + ) +{ + int i; + opus_int32 total_rate; + opus_int32 per_stream_rate; + + const int nb_channels = st->layout.nb_streams + st->layout.nb_coupled_streams; + + if (st->bitrate_bps==OPUS_AUTO) + { + total_rate = (st->layout.nb_coupled_streams + st->layout.nb_streams) * + (Fs+60*Fs/frame_size) + st->layout.nb_streams * (opus_int32)15000; + } else if (st->bitrate_bps==OPUS_BITRATE_MAX) + { + total_rate = nb_channels * 320000; + } else + { + total_rate = st->bitrate_bps; + } + + /* Allocate equal number of bits to Ambisonic (uncoupled) and non-diegetic + * (coupled) streams */ + per_stream_rate = total_rate / st->layout.nb_streams; + for (i = 0; i < st->layout.nb_streams; i++) + { + rate[i] = per_stream_rate; + } +} + +static opus_int32 rate_allocation( + OpusMSEncoder *st, + opus_int32 *rate, + int frame_size + ) +{ + int i; + opus_int32 rate_sum=0; + opus_int32 Fs; + char *ptr; + + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_SAMPLE_RATE(&Fs)); + + if (st->mapping_type == MAPPING_TYPE_AMBISONICS) { + ambisonics_rate_allocation(st, rate, frame_size, Fs); + } else + { + surround_rate_allocation(st, rate, frame_size, Fs); + } + + for (i=0;ilayout.nb_streams;i++) + { + rate[i] = IMAX(rate[i], 500); + rate_sum += rate[i]; + } + return rate_sum; +} + +/* Max size in case the encoder decides to return six frames (6 x 20 ms = 120 ms) */ +#define MS_FRAME_TMP (6*1275+12) +int opus_multistream_encode_native +( + OpusMSEncoder *st, + opus_copy_channel_in_func copy_channel_in, + const void *pcm, + int analysis_frame_size, + unsigned char *data, + opus_int32 max_data_bytes, + int lsb_depth, + downmix_func downmix, + int float_api, + void *user_data +) +{ + opus_int32 Fs; + int coupled_size; + int mono_size; + int s; + char *ptr; + int tot_size; + VARDECL(opus_val16, buf); + VARDECL(opus_val16, bandSMR); + unsigned char tmp_data[MS_FRAME_TMP]; + OpusRepacketizer rp; + opus_int32 vbr; + const CELTMode *celt_mode; + opus_int32 bitrates[256]; + opus_val16 bandLogE[42]; + opus_val32 *mem = NULL; + opus_val32 *preemph_mem=NULL; + int frame_size; + opus_int32 rate_sum; + opus_int32 smallest_packet; + ALLOC_STACK; + + if (st->mapping_type == MAPPING_TYPE_SURROUND) + { + preemph_mem = ms_get_preemph_mem(st); + mem = ms_get_window_mem(st); + } + + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_SAMPLE_RATE(&Fs)); + opus_encoder_ctl((OpusEncoder*)ptr, OPUS_GET_VBR(&vbr)); + opus_encoder_ctl((OpusEncoder*)ptr, CELT_GET_MODE(&celt_mode)); + + frame_size = frame_size_select(analysis_frame_size, st->variable_duration, Fs); + if (frame_size <= 0) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + + /* Smallest packet the encoder can produce. */ + smallest_packet = st->layout.nb_streams*2-1; + /* 100 ms needs an extra byte per stream for the ToC. */ + if (Fs/frame_size == 10) + smallest_packet += st->layout.nb_streams; + if (max_data_bytes < smallest_packet) + { + RESTORE_STACK; + return OPUS_BUFFER_TOO_SMALL; + } + ALLOC(buf, 2*frame_size, opus_val16); + coupled_size = opus_encoder_get_size(2); + mono_size = opus_encoder_get_size(1); + + ALLOC(bandSMR, 21*st->layout.nb_channels, opus_val16); + if (st->mapping_type == MAPPING_TYPE_SURROUND) + { + surround_analysis(celt_mode, pcm, bandSMR, mem, preemph_mem, frame_size, 120, st->layout.nb_channels, Fs, copy_channel_in, st->arch); + } + + /* Compute bitrate allocation between streams (this could be a lot better) */ + rate_sum = rate_allocation(st, bitrates, frame_size); + + if (!vbr) + { + if (st->bitrate_bps == OPUS_AUTO) + { + max_data_bytes = IMIN(max_data_bytes, 3*rate_sum/(3*8*Fs/frame_size)); + } else if (st->bitrate_bps != OPUS_BITRATE_MAX) + { + max_data_bytes = IMIN(max_data_bytes, IMAX(smallest_packet, + 3*st->bitrate_bps/(3*8*Fs/frame_size))); + } + } + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(bitrates[s])); + if (st->mapping_type == MAPPING_TYPE_SURROUND) + { + opus_int32 equiv_rate; + equiv_rate = st->bitrate_bps; + if (frame_size*50 < Fs) + equiv_rate -= 60*(Fs/frame_size - 50)*st->layout.nb_channels; + if (equiv_rate > 10000*st->layout.nb_channels) + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + else if (equiv_rate > 7000*st->layout.nb_channels) + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_SUPERWIDEBAND)); + else if (equiv_rate > 5000*st->layout.nb_channels) + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_WIDEBAND)); + else + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + if (s < st->layout.nb_coupled_streams) + { + /* To preserve the spatial image, force stereo CELT on coupled streams */ + opus_encoder_ctl(enc, OPUS_SET_FORCE_MODE(MODE_CELT_ONLY)); + opus_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(2)); + } + } + else if (st->mapping_type == MAPPING_TYPE_AMBISONICS) { + opus_encoder_ctl(enc, OPUS_SET_FORCE_MODE(MODE_CELT_ONLY)); + } + } + + ptr = (char*)st + align(sizeof(OpusMSEncoder)); + /* Counting ToC */ + tot_size = 0; + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + int len; + int curr_max; + int c1, c2; + int ret; + + opus_repacketizer_init(&rp); + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + { + int i; + int left, right; + left = get_left_channel(&st->layout, s, -1); + right = get_right_channel(&st->layout, s, -1); + (*copy_channel_in)(buf, 2, + pcm, st->layout.nb_channels, left, frame_size, user_data); + (*copy_channel_in)(buf+1, 2, + pcm, st->layout.nb_channels, right, frame_size, user_data); + ptr += align(coupled_size); + if (st->mapping_type == MAPPING_TYPE_SURROUND) + { + for (i=0;i<21;i++) + { + bandLogE[i] = bandSMR[21*left+i]; + bandLogE[21+i] = bandSMR[21*right+i]; + } + } + c1 = left; + c2 = right; + } else { + int i; + int chan = get_mono_channel(&st->layout, s, -1); + (*copy_channel_in)(buf, 1, + pcm, st->layout.nb_channels, chan, frame_size, user_data); + ptr += align(mono_size); + if (st->mapping_type == MAPPING_TYPE_SURROUND) + { + for (i=0;i<21;i++) + bandLogE[i] = bandSMR[21*chan+i]; + } + c1 = chan; + c2 = -1; + } + if (st->mapping_type == MAPPING_TYPE_SURROUND) + opus_encoder_ctl(enc, OPUS_SET_ENERGY_MASK(bandLogE)); + /* number of bytes left (+Toc) */ + curr_max = max_data_bytes - tot_size; + /* Reserve one byte for the last stream and two for the others */ + curr_max -= IMAX(0,2*(st->layout.nb_streams-s-1)-1); + /* For 100 ms, reserve an extra byte per stream for the ToC */ + if (Fs/frame_size == 10) + curr_max -= st->layout.nb_streams-s-1; + curr_max = IMIN(curr_max,MS_FRAME_TMP); + /* Repacketizer will add one or two bytes for self-delimited frames */ + if (s != st->layout.nb_streams-1) curr_max -= curr_max>253 ? 2 : 1; + if (!vbr && s == st->layout.nb_streams-1) + opus_encoder_ctl(enc, OPUS_SET_BITRATE(curr_max*(8*Fs/frame_size))); + len = opus_encode_native(enc, buf, frame_size, tmp_data, curr_max, lsb_depth, + pcm, analysis_frame_size, c1, c2, st->layout.nb_channels, downmix, float_api); + if (len<0) + { + RESTORE_STACK; + return len; + } + /* We need to use the repacketizer to add the self-delimiting lengths + while taking into account the fact that the encoder can now return + more than one frame at a time (e.g. 60 ms CELT-only) */ + ret = opus_repacketizer_cat(&rp, tmp_data, len); + /* If the opus_repacketizer_cat() fails, then something's seriously wrong + with the encoder. */ + if (ret != OPUS_OK) + { + RESTORE_STACK; + return OPUS_INTERNAL_ERROR; + } + len = opus_repacketizer_out_range_impl(&rp, 0, opus_repacketizer_get_nb_frames(&rp), + data, max_data_bytes-tot_size, s != st->layout.nb_streams-1, !vbr && s == st->layout.nb_streams-1); + data += len; + tot_size += len; + } + /*printf("\n");*/ + RESTORE_STACK; + return tot_size; +} + +#if !defined(DISABLE_FLOAT_API) +static void opus_copy_channel_in_float( + opus_val16 *dst, + int dst_stride, + const void *src, + int src_stride, + int src_channel, + int frame_size, + void *user_data +) +{ + const float *float_src; + opus_int32 i; + (void)user_data; + float_src = (const float *)src; + for (i=0;ilayout.nb_channels, IMAX(500*st->layout.nb_channels, value)); + } + st->bitrate_bps = value; + } + break; + case OPUS_GET_BITRATE_REQUEST: + { + int s; + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = 0; + for (s=0;slayout.nb_streams;s++) + { + opus_int32 rate; + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + opus_encoder_ctl(enc, request, &rate); + *value += rate; + } + } + break; + case OPUS_GET_LSB_DEPTH_REQUEST: + case OPUS_GET_VBR_REQUEST: + case OPUS_GET_APPLICATION_REQUEST: + case OPUS_GET_BANDWIDTH_REQUEST: + case OPUS_GET_COMPLEXITY_REQUEST: + case OPUS_GET_PACKET_LOSS_PERC_REQUEST: + case OPUS_GET_DTX_REQUEST: + case OPUS_GET_VOICE_RATIO_REQUEST: + case OPUS_GET_VBR_CONSTRAINT_REQUEST: + case OPUS_GET_SIGNAL_REQUEST: + case OPUS_GET_LOOKAHEAD_REQUEST: + case OPUS_GET_SAMPLE_RATE_REQUEST: + case OPUS_GET_INBAND_FEC_REQUEST: + case OPUS_GET_FORCE_CHANNELS_REQUEST: + case OPUS_GET_PREDICTION_DISABLED_REQUEST: + case OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST: + { + OpusEncoder *enc; + /* For int32* GET params, just query the first stream */ + opus_int32 *value = va_arg(ap, opus_int32*); + enc = (OpusEncoder*)ptr; + ret = opus_encoder_ctl(enc, request, value); + } + break; + case OPUS_GET_FINAL_RANGE_REQUEST: + { + int s; + opus_uint32 *value = va_arg(ap, opus_uint32*); + opus_uint32 tmp; + if (!value) + { + goto bad_arg; + } + *value=0; + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_encoder_ctl(enc, request, &tmp); + if (ret != OPUS_OK) break; + *value ^= tmp; + } + } + break; + case OPUS_SET_LSB_DEPTH_REQUEST: + case OPUS_SET_COMPLEXITY_REQUEST: + case OPUS_SET_VBR_REQUEST: + case OPUS_SET_VBR_CONSTRAINT_REQUEST: + case OPUS_SET_MAX_BANDWIDTH_REQUEST: + case OPUS_SET_BANDWIDTH_REQUEST: + case OPUS_SET_SIGNAL_REQUEST: + case OPUS_SET_APPLICATION_REQUEST: + case OPUS_SET_INBAND_FEC_REQUEST: + case OPUS_SET_PACKET_LOSS_PERC_REQUEST: + case OPUS_SET_DTX_REQUEST: + case OPUS_SET_FORCE_MODE_REQUEST: + case OPUS_SET_FORCE_CHANNELS_REQUEST: + case OPUS_SET_PREDICTION_DISABLED_REQUEST: + case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST: + { + int s; + /* This works for int32 params */ + opus_int32 value = va_arg(ap, opus_int32); + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_encoder_ctl(enc, request, value); + if (ret != OPUS_OK) + break; + } + } + break; + case OPUS_MULTISTREAM_GET_ENCODER_STATE_REQUEST: + { + int s; + opus_int32 stream_id; + OpusEncoder **value; + stream_id = va_arg(ap, opus_int32); + if (stream_id<0 || stream_id >= st->layout.nb_streams) + goto bad_arg; + value = va_arg(ap, OpusEncoder**); + if (!value) + { + goto bad_arg; + } + for (s=0;slayout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + } + *value = (OpusEncoder*)ptr; + } + break; + case OPUS_SET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 value = va_arg(ap, opus_int32); + st->variable_duration = value; + } + break; + case OPUS_GET_EXPERT_FRAME_DURATION_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = st->variable_duration; + } + break; + case OPUS_RESET_STATE: + { + int s; + if (st->mapping_type == MAPPING_TYPE_SURROUND) + { + OPUS_CLEAR(ms_get_preemph_mem(st), st->layout.nb_channels); + OPUS_CLEAR(ms_get_window_mem(st), st->layout.nb_channels*120); + } + for (s=0;slayout.nb_streams;s++) + { + OpusEncoder *enc; + enc = (OpusEncoder*)ptr; + if (s < st->layout.nb_coupled_streams) + ptr += align(coupled_size); + else + ptr += align(mono_size); + ret = opus_encoder_ctl(enc, OPUS_RESET_STATE); + if (ret != OPUS_OK) + break; + } + } + break; + default: + ret = OPUS_UNIMPLEMENTED; + break; + } + return ret; +bad_arg: + return OPUS_BAD_ARG; +} + +int opus_multistream_encoder_ctl(OpusMSEncoder *st, int request, ...) +{ + int ret; + va_list ap; + va_start(ap, request); + ret = opus_multistream_encoder_ctl_va_list(st, request, ap); + va_end(ap); + return ret; +} + +void opus_multistream_encoder_destroy(OpusMSEncoder *st) +{ + opus_free(st); +} diff --git a/lib/opus/src/opus_private.h b/lib/opus/src/opus_private.h new file mode 100644 index 00000000..5e2463f5 --- /dev/null +++ b/lib/opus/src/opus_private.h @@ -0,0 +1,201 @@ +/* Copyright (c) 2012 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + + +#ifndef OPUS_PRIVATE_H +#define OPUS_PRIVATE_H + +#include "arch.h" +#include "opus.h" +#include "celt.h" + +#include /* va_list */ +#include /* offsetof */ + +struct OpusRepacketizer { + unsigned char toc; + int nb_frames; + const unsigned char *frames[48]; + opus_int16 len[48]; + int framesize; +}; + +typedef struct ChannelLayout { + int nb_channels; + int nb_streams; + int nb_coupled_streams; + unsigned char mapping[256]; +} ChannelLayout; + +typedef enum { + MAPPING_TYPE_NONE, + MAPPING_TYPE_SURROUND, + MAPPING_TYPE_AMBISONICS +} MappingType; + +struct OpusMSEncoder { + ChannelLayout layout; + int arch; + int lfe_stream; + int application; + int variable_duration; + MappingType mapping_type; + opus_int32 bitrate_bps; + /* Encoder states go here */ + /* then opus_val32 window_mem[channels*120]; */ + /* then opus_val32 preemph_mem[channels]; */ +}; + +struct OpusMSDecoder { + ChannelLayout layout; + /* Decoder states go here */ +}; + +int opus_multistream_encoder_ctl_va_list(struct OpusMSEncoder *st, int request, + va_list ap); +int opus_multistream_decoder_ctl_va_list(struct OpusMSDecoder *st, int request, + va_list ap); + +int validate_layout(const ChannelLayout *layout); +int get_left_channel(const ChannelLayout *layout, int stream_id, int prev); +int get_right_channel(const ChannelLayout *layout, int stream_id, int prev); +int get_mono_channel(const ChannelLayout *layout, int stream_id, int prev); + +typedef void (*opus_copy_channel_in_func)( + opus_val16 *dst, + int dst_stride, + const void *src, + int src_stride, + int src_channel, + int frame_size, + void *user_data +); + +typedef void (*opus_copy_channel_out_func)( + void *dst, + int dst_stride, + int dst_channel, + const opus_val16 *src, + int src_stride, + int frame_size, + void *user_data +); + +#define MODE_SILK_ONLY 1000 +#define MODE_HYBRID 1001 +#define MODE_CELT_ONLY 1002 + +#define OPUS_SET_VOICE_RATIO_REQUEST 11018 +#define OPUS_GET_VOICE_RATIO_REQUEST 11019 + +/** Configures the encoder's expected percentage of voice + * opposed to music or other signals. + * + * @note This interface is currently more aspiration than actuality. It's + * ultimately expected to bias an automatic signal classifier, but it currently + * just shifts the static bitrate to mode mapping around a little bit. + * + * @param[in] x int: Voice percentage in the range 0-100, inclusive. + * @hideinitializer */ +#define OPUS_SET_VOICE_RATIO(x) OPUS_SET_VOICE_RATIO_REQUEST, __opus_check_int(x) +/** Gets the encoder's configured voice ratio value, @see OPUS_SET_VOICE_RATIO + * + * @param[out] x int*: Voice percentage in the range 0-100, inclusive. + * @hideinitializer */ +#define OPUS_GET_VOICE_RATIO(x) OPUS_GET_VOICE_RATIO_REQUEST, __opus_check_int_ptr(x) + + +#define OPUS_SET_FORCE_MODE_REQUEST 11002 +#define OPUS_SET_FORCE_MODE(x) OPUS_SET_FORCE_MODE_REQUEST, __opus_check_int(x) + +typedef void (*downmix_func)(const void *, opus_val32 *, int, int, int, int, int); +void downmix_float(const void *_x, opus_val32 *sub, int subframe, int offset, int c1, int c2, int C); +void downmix_int(const void *_x, opus_val32 *sub, int subframe, int offset, int c1, int c2, int C); +int is_digital_silence(const opus_val16* pcm, int frame_size, int channels, int lsb_depth); + +int encode_size(int size, unsigned char *data); + +opus_int32 frame_size_select(opus_int32 frame_size, int variable_duration, opus_int32 Fs); + +opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_size, + unsigned char *data, opus_int32 out_data_bytes, int lsb_depth, + const void *analysis_pcm, opus_int32 analysis_size, int c1, int c2, + int analysis_channels, downmix_func downmix, int float_api); + +int opus_decode_native(OpusDecoder *st, const unsigned char *data, opus_int32 len, + opus_val16 *pcm, int frame_size, int decode_fec, int self_delimited, + opus_int32 *packet_offset, int soft_clip); + +/* Make sure everything is properly aligned. */ +static OPUS_INLINE int align(int i) +{ + struct foo {char c; union { void* p; opus_int32 i; opus_val32 v; } u;}; + + unsigned int alignment = offsetof(struct foo, u); + + /* Optimizing compilers should optimize div and multiply into and + for all sensible alignment values. */ + return ((i + alignment - 1) / alignment) * alignment; +} + +int opus_packet_parse_impl(const unsigned char *data, opus_int32 len, + int self_delimited, unsigned char *out_toc, + const unsigned char *frames[48], opus_int16 size[48], + int *payload_offset, opus_int32 *packet_offset); + +opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int end, + unsigned char *data, opus_int32 maxlen, int self_delimited, int pad); + +int pad_frame(unsigned char *data, opus_int32 len, opus_int32 new_len); + +int opus_multistream_encode_native +( + struct OpusMSEncoder *st, + opus_copy_channel_in_func copy_channel_in, + const void *pcm, + int analysis_frame_size, + unsigned char *data, + opus_int32 max_data_bytes, + int lsb_depth, + downmix_func downmix, + int float_api, + void *user_data +); + +int opus_multistream_decode_native( + struct OpusMSDecoder *st, + const unsigned char *data, + opus_int32 len, + void *pcm, + opus_copy_channel_out_func copy_channel_out, + int frame_size, + int decode_fec, + int soft_clip, + void *user_data +); + +#endif /* OPUS_PRIVATE_H */ diff --git a/lib/opus/src/opus_projection_decoder.c b/lib/opus/src/opus_projection_decoder.c new file mode 100644 index 00000000..c2e07d5b --- /dev/null +++ b/lib/opus/src/opus_projection_decoder.c @@ -0,0 +1,258 @@ +/* Copyright (c) 2017 Google Inc. + Written by Andrew Allen */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "mathops.h" +#include "os_support.h" +#include "opus_private.h" +#include "opus_defines.h" +#include "opus_projection.h" +#include "opus_multistream.h" +#include "mapping_matrix.h" +#include "stack_alloc.h" + +struct OpusProjectionDecoder +{ + opus_int32 demixing_matrix_size_in_bytes; + /* Encoder states go here */ +}; + +#if !defined(DISABLE_FLOAT_API) +static void opus_projection_copy_channel_out_float( + void *dst, + int dst_stride, + int dst_channel, + const opus_val16 *src, + int src_stride, + int frame_size, + void *user_data) +{ + float *float_dst; + const MappingMatrix *matrix; + float_dst = (float *)dst; + matrix = (const MappingMatrix *)user_data; + + if (dst_channel == 0) + OPUS_CLEAR(float_dst, frame_size * dst_stride); + + if (src != NULL) + mapping_matrix_multiply_channel_out_float(matrix, src, dst_channel, + src_stride, float_dst, dst_stride, frame_size); +} +#endif + +static void opus_projection_copy_channel_out_short( + void *dst, + int dst_stride, + int dst_channel, + const opus_val16 *src, + int src_stride, + int frame_size, + void *user_data) +{ + opus_int16 *short_dst; + const MappingMatrix *matrix; + short_dst = (opus_int16 *)dst; + matrix = (const MappingMatrix *)user_data; + if (dst_channel == 0) + OPUS_CLEAR(short_dst, frame_size * dst_stride); + + if (src != NULL) + mapping_matrix_multiply_channel_out_short(matrix, src, dst_channel, + src_stride, short_dst, dst_stride, frame_size); +} + +static MappingMatrix *get_dec_demixing_matrix(OpusProjectionDecoder *st) +{ + /* void* cast avoids clang -Wcast-align warning */ + return (MappingMatrix*)(void*)((char*)st + + align(sizeof(OpusProjectionDecoder))); +} + +static OpusMSDecoder *get_multistream_decoder(OpusProjectionDecoder *st) +{ + /* void* cast avoids clang -Wcast-align warning */ + return (OpusMSDecoder*)(void*)((char*)st + + align(sizeof(OpusProjectionDecoder) + + st->demixing_matrix_size_in_bytes)); +} + +opus_int32 opus_projection_decoder_get_size(int channels, int streams, + int coupled_streams) +{ + opus_int32 matrix_size; + opus_int32 decoder_size; + + matrix_size = + mapping_matrix_get_size(streams + coupled_streams, channels); + if (!matrix_size) + return 0; + + decoder_size = opus_multistream_decoder_get_size(streams, coupled_streams); + if (!decoder_size) + return 0; + + return align(sizeof(OpusProjectionDecoder)) + matrix_size + decoder_size; +} + +int opus_projection_decoder_init(OpusProjectionDecoder *st, opus_int32 Fs, + int channels, int streams, int coupled_streams, + unsigned char *demixing_matrix, opus_int32 demixing_matrix_size) +{ + int nb_input_streams; + opus_int32 expected_matrix_size; + int i, ret; + unsigned char mapping[255]; + VARDECL(opus_int16, buf); + ALLOC_STACK; + + /* Verify supplied matrix size. */ + nb_input_streams = streams + coupled_streams; + expected_matrix_size = nb_input_streams * channels * sizeof(opus_int16); + if (expected_matrix_size != demixing_matrix_size) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + + /* Convert demixing matrix input into internal format. */ + ALLOC(buf, nb_input_streams * channels, opus_int16); + for (i = 0; i < nb_input_streams * channels; i++) + { + int s = demixing_matrix[2*i + 1] << 8 | demixing_matrix[2*i]; + s = ((s & 0xFFFF) ^ 0x8000) - 0x8000; + buf[i] = (opus_int16)s; + } + + /* Assign demixing matrix. */ + st->demixing_matrix_size_in_bytes = + mapping_matrix_get_size(channels, nb_input_streams); + if (!st->demixing_matrix_size_in_bytes) + { + RESTORE_STACK; + return OPUS_BAD_ARG; + } + + mapping_matrix_init(get_dec_demixing_matrix(st), channels, nb_input_streams, 0, + buf, demixing_matrix_size); + + /* Set trivial mapping so each input channel pairs with a matrix column. */ + for (i = 0; i < channels; i++) + mapping[i] = i; + + ret = opus_multistream_decoder_init( + get_multistream_decoder(st), Fs, channels, streams, coupled_streams, mapping); + RESTORE_STACK; + return ret; +} + +OpusProjectionDecoder *opus_projection_decoder_create( + opus_int32 Fs, int channels, int streams, int coupled_streams, + unsigned char *demixing_matrix, opus_int32 demixing_matrix_size, int *error) +{ + int size; + int ret; + OpusProjectionDecoder *st; + + /* Allocate space for the projection decoder. */ + size = opus_projection_decoder_get_size(channels, streams, coupled_streams); + if (!size) { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + st = (OpusProjectionDecoder *)opus_alloc(size); + if (!st) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + + /* Initialize projection decoder with provided settings. */ + ret = opus_projection_decoder_init(st, Fs, channels, streams, coupled_streams, + demixing_matrix, demixing_matrix_size); + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} + +#ifdef FIXED_POINT +int opus_projection_decode(OpusProjectionDecoder *st, const unsigned char *data, + opus_int32 len, opus_int16 *pcm, int frame_size, + int decode_fec) +{ + return opus_multistream_decode_native(get_multistream_decoder(st), data, len, + pcm, opus_projection_copy_channel_out_short, frame_size, decode_fec, 0, + get_dec_demixing_matrix(st)); +} +#else +int opus_projection_decode(OpusProjectionDecoder *st, const unsigned char *data, + opus_int32 len, opus_int16 *pcm, int frame_size, + int decode_fec) +{ + return opus_multistream_decode_native(get_multistream_decoder(st), data, len, + pcm, opus_projection_copy_channel_out_short, frame_size, decode_fec, 1, + get_dec_demixing_matrix(st)); +} +#endif + +#ifndef DISABLE_FLOAT_API +int opus_projection_decode_float(OpusProjectionDecoder *st, const unsigned char *data, + opus_int32 len, float *pcm, int frame_size, int decode_fec) +{ + return opus_multistream_decode_native(get_multistream_decoder(st), data, len, + pcm, opus_projection_copy_channel_out_float, frame_size, decode_fec, 0, + get_dec_demixing_matrix(st)); +} +#endif + +int opus_projection_decoder_ctl(OpusProjectionDecoder *st, int request, ...) +{ + va_list ap; + int ret = OPUS_OK; + + va_start(ap, request); + ret = opus_multistream_decoder_ctl_va_list(get_multistream_decoder(st), + request, ap); + va_end(ap); + return ret; +} + +void opus_projection_decoder_destroy(OpusProjectionDecoder *st) +{ + opus_free(st); +} + diff --git a/lib/opus/src/opus_projection_encoder.c b/lib/opus/src/opus_projection_encoder.c new file mode 100644 index 00000000..06fb2d25 --- /dev/null +++ b/lib/opus/src/opus_projection_encoder.c @@ -0,0 +1,468 @@ +/* Copyright (c) 2017 Google Inc. + Written by Andrew Allen */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "mathops.h" +#include "os_support.h" +#include "opus_private.h" +#include "opus_defines.h" +#include "opus_projection.h" +#include "opus_multistream.h" +#include "stack_alloc.h" +#include "mapping_matrix.h" + +struct OpusProjectionEncoder +{ + opus_int32 mixing_matrix_size_in_bytes; + opus_int32 demixing_matrix_size_in_bytes; + /* Encoder states go here */ +}; + +#if !defined(DISABLE_FLOAT_API) +static void opus_projection_copy_channel_in_float( + opus_val16 *dst, + int dst_stride, + const void *src, + int src_stride, + int src_channel, + int frame_size, + void *user_data +) +{ + mapping_matrix_multiply_channel_in_float((const MappingMatrix*)user_data, + (const float*)src, src_stride, dst, src_channel, dst_stride, frame_size); +} +#endif + +static void opus_projection_copy_channel_in_short( + opus_val16 *dst, + int dst_stride, + const void *src, + int src_stride, + int src_channel, + int frame_size, + void *user_data +) +{ + mapping_matrix_multiply_channel_in_short((const MappingMatrix*)user_data, + (const opus_int16*)src, src_stride, dst, src_channel, dst_stride, frame_size); +} + +static int get_order_plus_one_from_channels(int channels, int *order_plus_one) +{ + int order_plus_one_; + int acn_channels; + int nondiegetic_channels; + + /* Allowed numbers of channels: + * (1 + n)^2 + 2j, for n = 0...14 and j = 0 or 1. + */ + if (channels < 1 || channels > 227) + return OPUS_BAD_ARG; + + order_plus_one_ = isqrt32(channels); + acn_channels = order_plus_one_ * order_plus_one_; + nondiegetic_channels = channels - acn_channels; + if (nondiegetic_channels != 0 && nondiegetic_channels != 2) + return OPUS_BAD_ARG; + + if (order_plus_one) + *order_plus_one = order_plus_one_; + return OPUS_OK; +} + +static int get_streams_from_channels(int channels, int mapping_family, + int *streams, int *coupled_streams, + int *order_plus_one) +{ + if (mapping_family == 3) + { + if (get_order_plus_one_from_channels(channels, order_plus_one) != OPUS_OK) + return OPUS_BAD_ARG; + if (streams) + *streams = (channels + 1) / 2; + if (coupled_streams) + *coupled_streams = channels / 2; + return OPUS_OK; + } + return OPUS_BAD_ARG; +} + +static MappingMatrix *get_mixing_matrix(OpusProjectionEncoder *st) +{ + /* void* cast avoids clang -Wcast-align warning */ + return (MappingMatrix *)(void*)((char*)st + + align(sizeof(OpusProjectionEncoder))); +} + +static MappingMatrix *get_enc_demixing_matrix(OpusProjectionEncoder *st) +{ + /* void* cast avoids clang -Wcast-align warning */ + return (MappingMatrix *)(void*)((char*)st + + align(sizeof(OpusProjectionEncoder) + + st->mixing_matrix_size_in_bytes)); +} + +static OpusMSEncoder *get_multistream_encoder(OpusProjectionEncoder *st) +{ + /* void* cast avoids clang -Wcast-align warning */ + return (OpusMSEncoder *)(void*)((char*)st + + align(sizeof(OpusProjectionEncoder) + + st->mixing_matrix_size_in_bytes + + st->demixing_matrix_size_in_bytes)); +} + +opus_int32 opus_projection_ambisonics_encoder_get_size(int channels, + int mapping_family) +{ + int nb_streams; + int nb_coupled_streams; + int order_plus_one; + int mixing_matrix_rows, mixing_matrix_cols; + int demixing_matrix_rows, demixing_matrix_cols; + opus_int32 mixing_matrix_size, demixing_matrix_size; + opus_int32 encoder_size; + int ret; + + ret = get_streams_from_channels(channels, mapping_family, &nb_streams, + &nb_coupled_streams, &order_plus_one); + if (ret != OPUS_OK) + return 0; + + if (order_plus_one == 2) + { + mixing_matrix_rows = mapping_matrix_foa_mixing.rows; + mixing_matrix_cols = mapping_matrix_foa_mixing.cols; + demixing_matrix_rows = mapping_matrix_foa_demixing.rows; + demixing_matrix_cols = mapping_matrix_foa_demixing.cols; + } + else if (order_plus_one == 3) + { + mixing_matrix_rows = mapping_matrix_soa_mixing.rows; + mixing_matrix_cols = mapping_matrix_soa_mixing.cols; + demixing_matrix_rows = mapping_matrix_soa_demixing.rows; + demixing_matrix_cols = mapping_matrix_soa_demixing.cols; + } + else if (order_plus_one == 4) + { + mixing_matrix_rows = mapping_matrix_toa_mixing.rows; + mixing_matrix_cols = mapping_matrix_toa_mixing.cols; + demixing_matrix_rows = mapping_matrix_toa_demixing.rows; + demixing_matrix_cols = mapping_matrix_toa_demixing.cols; + } + else + return 0; + + mixing_matrix_size = + mapping_matrix_get_size(mixing_matrix_rows, mixing_matrix_cols); + if (!mixing_matrix_size) + return 0; + + demixing_matrix_size = + mapping_matrix_get_size(demixing_matrix_rows, demixing_matrix_cols); + if (!demixing_matrix_size) + return 0; + + encoder_size = + opus_multistream_encoder_get_size(nb_streams, nb_coupled_streams); + if (!encoder_size) + return 0; + + return align(sizeof(OpusProjectionEncoder)) + + mixing_matrix_size + demixing_matrix_size + encoder_size; +} + +int opus_projection_ambisonics_encoder_init(OpusProjectionEncoder *st, opus_int32 Fs, + int channels, int mapping_family, + int *streams, int *coupled_streams, + int application) +{ + MappingMatrix *mixing_matrix; + MappingMatrix *demixing_matrix; + OpusMSEncoder *ms_encoder; + int i; + int ret; + int order_plus_one; + unsigned char mapping[255]; + + if (streams == NULL || coupled_streams == NULL) { + return OPUS_BAD_ARG; + } + + if (get_streams_from_channels(channels, mapping_family, streams, + coupled_streams, &order_plus_one) != OPUS_OK) + return OPUS_BAD_ARG; + + if (mapping_family == 3) + { + /* Assign mixing matrix based on available pre-computed matrices. */ + mixing_matrix = get_mixing_matrix(st); + if (order_plus_one == 2) + { + mapping_matrix_init(mixing_matrix, mapping_matrix_foa_mixing.rows, + mapping_matrix_foa_mixing.cols, mapping_matrix_foa_mixing.gain, + mapping_matrix_foa_mixing_data, + sizeof(mapping_matrix_foa_mixing_data)); + } + else if (order_plus_one == 3) + { + mapping_matrix_init(mixing_matrix, mapping_matrix_soa_mixing.rows, + mapping_matrix_soa_mixing.cols, mapping_matrix_soa_mixing.gain, + mapping_matrix_soa_mixing_data, + sizeof(mapping_matrix_soa_mixing_data)); + } + else if (order_plus_one == 4) + { + mapping_matrix_init(mixing_matrix, mapping_matrix_toa_mixing.rows, + mapping_matrix_toa_mixing.cols, mapping_matrix_toa_mixing.gain, + mapping_matrix_toa_mixing_data, + sizeof(mapping_matrix_toa_mixing_data)); + } + else + return OPUS_BAD_ARG; + + st->mixing_matrix_size_in_bytes = mapping_matrix_get_size( + mixing_matrix->rows, mixing_matrix->cols); + if (!st->mixing_matrix_size_in_bytes) + return OPUS_BAD_ARG; + + /* Assign demixing matrix based on available pre-computed matrices. */ + demixing_matrix = get_enc_demixing_matrix(st); + if (order_plus_one == 2) + { + mapping_matrix_init(demixing_matrix, mapping_matrix_foa_demixing.rows, + mapping_matrix_foa_demixing.cols, mapping_matrix_foa_demixing.gain, + mapping_matrix_foa_demixing_data, + sizeof(mapping_matrix_foa_demixing_data)); + } + else if (order_plus_one == 3) + { + mapping_matrix_init(demixing_matrix, mapping_matrix_soa_demixing.rows, + mapping_matrix_soa_demixing.cols, mapping_matrix_soa_demixing.gain, + mapping_matrix_soa_demixing_data, + sizeof(mapping_matrix_soa_demixing_data)); + } + else if (order_plus_one == 4) + { + mapping_matrix_init(demixing_matrix, mapping_matrix_toa_demixing.rows, + mapping_matrix_toa_demixing.cols, mapping_matrix_toa_demixing.gain, + mapping_matrix_toa_demixing_data, + sizeof(mapping_matrix_toa_demixing_data)); + } + else + return OPUS_BAD_ARG; + + st->demixing_matrix_size_in_bytes = mapping_matrix_get_size( + demixing_matrix->rows, demixing_matrix->cols); + if (!st->demixing_matrix_size_in_bytes) + return OPUS_BAD_ARG; + } + else + return OPUS_UNIMPLEMENTED; + + /* Ensure matrices are large enough for desired coding scheme. */ + if (*streams + *coupled_streams > mixing_matrix->rows || + channels > mixing_matrix->cols || + channels > demixing_matrix->rows || + *streams + *coupled_streams > demixing_matrix->cols) + return OPUS_BAD_ARG; + + /* Set trivial mapping so each input channel pairs with a matrix column. */ + for (i = 0; i < channels; i++) + mapping[i] = i; + + /* Initialize multistream encoder with provided settings. */ + ms_encoder = get_multistream_encoder(st); + ret = opus_multistream_encoder_init(ms_encoder, Fs, channels, *streams, + *coupled_streams, mapping, application); + return ret; +} + +OpusProjectionEncoder *opus_projection_ambisonics_encoder_create( + opus_int32 Fs, int channels, int mapping_family, int *streams, + int *coupled_streams, int application, int *error) +{ + int size; + int ret; + OpusProjectionEncoder *st; + + /* Allocate space for the projection encoder. */ + size = opus_projection_ambisonics_encoder_get_size(channels, mapping_family); + if (!size) { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + st = (OpusProjectionEncoder *)opus_alloc(size); + if (!st) + { + if (error) + *error = OPUS_ALLOC_FAIL; + return NULL; + } + + /* Initialize projection encoder with provided settings. */ + ret = opus_projection_ambisonics_encoder_init(st, Fs, channels, + mapping_family, streams, coupled_streams, application); + if (ret != OPUS_OK) + { + opus_free(st); + st = NULL; + } + if (error) + *error = ret; + return st; +} + +int opus_projection_encode(OpusProjectionEncoder *st, const opus_int16 *pcm, + int frame_size, unsigned char *data, + opus_int32 max_data_bytes) +{ + return opus_multistream_encode_native(get_multistream_encoder(st), + opus_projection_copy_channel_in_short, pcm, frame_size, data, + max_data_bytes, 16, downmix_int, 0, get_mixing_matrix(st)); +} + +#ifndef DISABLE_FLOAT_API +#ifdef FIXED_POINT +int opus_projection_encode_float(OpusProjectionEncoder *st, const float *pcm, + int frame_size, unsigned char *data, + opus_int32 max_data_bytes) +{ + return opus_multistream_encode_native(get_multistream_encoder(st), + opus_projection_copy_channel_in_float, pcm, frame_size, data, + max_data_bytes, 16, downmix_float, 1, get_mixing_matrix(st)); +} +#else +int opus_projection_encode_float(OpusProjectionEncoder *st, const float *pcm, + int frame_size, unsigned char *data, + opus_int32 max_data_bytes) +{ + return opus_multistream_encode_native(get_multistream_encoder(st), + opus_projection_copy_channel_in_float, pcm, frame_size, data, + max_data_bytes, 24, downmix_float, 1, get_mixing_matrix(st)); +} +#endif +#endif + +void opus_projection_encoder_destroy(OpusProjectionEncoder *st) +{ + opus_free(st); +} + +int opus_projection_encoder_ctl(OpusProjectionEncoder *st, int request, ...) +{ + va_list ap; + MappingMatrix *demixing_matrix; + OpusMSEncoder *ms_encoder; + int ret = OPUS_OK; + + ms_encoder = get_multistream_encoder(st); + demixing_matrix = get_enc_demixing_matrix(st); + + va_start(ap, request); + switch(request) + { + case OPUS_PROJECTION_GET_DEMIXING_MATRIX_SIZE_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = + ms_encoder->layout.nb_channels * (ms_encoder->layout.nb_streams + + ms_encoder->layout.nb_coupled_streams) * sizeof(opus_int16); + } + break; + case OPUS_PROJECTION_GET_DEMIXING_MATRIX_GAIN_REQUEST: + { + opus_int32 *value = va_arg(ap, opus_int32*); + if (!value) + { + goto bad_arg; + } + *value = demixing_matrix->gain; + } + break; + case OPUS_PROJECTION_GET_DEMIXING_MATRIX_REQUEST: + { + int i, j, k, l; + int nb_input_streams; + int nb_output_streams; + unsigned char *external_char; + opus_int16 *internal_short; + opus_int32 external_size; + opus_int32 internal_size; + + /* (I/O is in relation to the decoder's perspective). */ + nb_input_streams = ms_encoder->layout.nb_streams + + ms_encoder->layout.nb_coupled_streams; + nb_output_streams = ms_encoder->layout.nb_channels; + + external_char = va_arg(ap, unsigned char *); + external_size = va_arg(ap, opus_int32); + if (!external_char) + { + goto bad_arg; + } + internal_short = mapping_matrix_get_data(demixing_matrix); + internal_size = nb_input_streams * nb_output_streams * sizeof(opus_int16); + if (external_size != internal_size) + { + goto bad_arg; + } + + /* Copy demixing matrix subset to output destination. */ + l = 0; + for (i = 0; i < nb_input_streams; i++) { + for (j = 0; j < nb_output_streams; j++) { + k = demixing_matrix->rows * i + j; + external_char[2*l] = (unsigned char)internal_short[k]; + external_char[2*l+1] = (unsigned char)(internal_short[k] >> 8); + l++; + } + } + } + break; + default: + { + ret = opus_multistream_encoder_ctl_va_list(ms_encoder, request, ap); + } + break; + } + va_end(ap); + return ret; + +bad_arg: + va_end(ap); + return OPUS_BAD_ARG; +} + diff --git a/lib/opus/src/repacketizer.c b/lib/opus/src/repacketizer.c new file mode 100644 index 00000000..bda44a14 --- /dev/null +++ b/lib/opus/src/repacketizer.c @@ -0,0 +1,349 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include "opus.h" +#include "opus_private.h" +#include "os_support.h" + + +int opus_repacketizer_get_size(void) +{ + return sizeof(OpusRepacketizer); +} + +OpusRepacketizer *opus_repacketizer_init(OpusRepacketizer *rp) +{ + rp->nb_frames = 0; + return rp; +} + +OpusRepacketizer *opus_repacketizer_create(void) +{ + OpusRepacketizer *rp; + rp=(OpusRepacketizer *)opus_alloc(opus_repacketizer_get_size()); + if(rp==NULL)return NULL; + return opus_repacketizer_init(rp); +} + +void opus_repacketizer_destroy(OpusRepacketizer *rp) +{ + opus_free(rp); +} + +static int opus_repacketizer_cat_impl(OpusRepacketizer *rp, const unsigned char *data, opus_int32 len, int self_delimited) +{ + unsigned char tmp_toc; + int curr_nb_frames,ret; + /* Set of check ToC */ + if (len<1) return OPUS_INVALID_PACKET; + if (rp->nb_frames == 0) + { + rp->toc = data[0]; + rp->framesize = opus_packet_get_samples_per_frame(data, 8000); + } else if ((rp->toc&0xFC) != (data[0]&0xFC)) + { + /*fprintf(stderr, "toc mismatch: 0x%x vs 0x%x\n", rp->toc, data[0]);*/ + return OPUS_INVALID_PACKET; + } + curr_nb_frames = opus_packet_get_nb_frames(data, len); + if(curr_nb_frames<1) return OPUS_INVALID_PACKET; + + /* Check the 120 ms maximum packet size */ + if ((curr_nb_frames+rp->nb_frames)*rp->framesize > 960) + { + return OPUS_INVALID_PACKET; + } + + ret=opus_packet_parse_impl(data, len, self_delimited, &tmp_toc, &rp->frames[rp->nb_frames], &rp->len[rp->nb_frames], NULL, NULL); + if(ret<1)return ret; + + rp->nb_frames += curr_nb_frames; + return OPUS_OK; +} + +int opus_repacketizer_cat(OpusRepacketizer *rp, const unsigned char *data, opus_int32 len) +{ + return opus_repacketizer_cat_impl(rp, data, len, 0); +} + +int opus_repacketizer_get_nb_frames(OpusRepacketizer *rp) +{ + return rp->nb_frames; +} + +opus_int32 opus_repacketizer_out_range_impl(OpusRepacketizer *rp, int begin, int end, + unsigned char *data, opus_int32 maxlen, int self_delimited, int pad) +{ + int i, count; + opus_int32 tot_size; + opus_int16 *len; + const unsigned char **frames; + unsigned char * ptr; + + if (begin<0 || begin>=end || end>rp->nb_frames) + { + /*fprintf(stderr, "%d %d %d\n", begin, end, rp->nb_frames);*/ + return OPUS_BAD_ARG; + } + count = end-begin; + + len = rp->len+begin; + frames = rp->frames+begin; + if (self_delimited) + tot_size = 1 + (len[count-1]>=252); + else + tot_size = 0; + + ptr = data; + if (count==1) + { + /* Code 0 */ + tot_size += len[0]+1; + if (tot_size > maxlen) + return OPUS_BUFFER_TOO_SMALL; + *ptr++ = rp->toc&0xFC; + } else if (count==2) + { + if (len[1] == len[0]) + { + /* Code 1 */ + tot_size += 2*len[0]+1; + if (tot_size > maxlen) + return OPUS_BUFFER_TOO_SMALL; + *ptr++ = (rp->toc&0xFC) | 0x1; + } else { + /* Code 2 */ + tot_size += len[0]+len[1]+2+(len[0]>=252); + if (tot_size > maxlen) + return OPUS_BUFFER_TOO_SMALL; + *ptr++ = (rp->toc&0xFC) | 0x2; + ptr += encode_size(len[0], ptr); + } + } + if (count > 2 || (pad && tot_size < maxlen)) + { + /* Code 3 */ + int vbr; + int pad_amount=0; + + /* Restart the process for the padding case */ + ptr = data; + if (self_delimited) + tot_size = 1 + (len[count-1]>=252); + else + tot_size = 0; + vbr = 0; + for (i=1;i=252) + len[i]; + tot_size += len[count-1]; + + if (tot_size > maxlen) + return OPUS_BUFFER_TOO_SMALL; + *ptr++ = (rp->toc&0xFC) | 0x3; + *ptr++ = count | 0x80; + } else { + tot_size += count*len[0]+2; + if (tot_size > maxlen) + return OPUS_BUFFER_TOO_SMALL; + *ptr++ = (rp->toc&0xFC) | 0x3; + *ptr++ = count; + } + pad_amount = pad ? (maxlen-tot_size) : 0; + if (pad_amount != 0) + { + int nb_255s; + data[1] |= 0x40; + nb_255s = (pad_amount-1)/255; + for (i=0;inb_frames, data, maxlen, 0, 0); +} + +int opus_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len) +{ + OpusRepacketizer rp; + opus_int32 ret; + if (len < 1) + return OPUS_BAD_ARG; + if (len==new_len) + return OPUS_OK; + else if (len > new_len) + return OPUS_BAD_ARG; + opus_repacketizer_init(&rp); + /* Moving payload to the end of the packet so we can do in-place padding */ + OPUS_MOVE(data+new_len-len, data, len); + ret = opus_repacketizer_cat(&rp, data+new_len-len, len); + if (ret != OPUS_OK) + return ret; + ret = opus_repacketizer_out_range_impl(&rp, 0, rp.nb_frames, data, new_len, 0, 1); + if (ret > 0) + return OPUS_OK; + else + return ret; +} + +opus_int32 opus_packet_unpad(unsigned char *data, opus_int32 len) +{ + OpusRepacketizer rp; + opus_int32 ret; + if (len < 1) + return OPUS_BAD_ARG; + opus_repacketizer_init(&rp); + ret = opus_repacketizer_cat(&rp, data, len); + if (ret < 0) + return ret; + ret = opus_repacketizer_out_range_impl(&rp, 0, rp.nb_frames, data, len, 0, 0); + celt_assert(ret > 0 && ret <= len); + return ret; +} + +int opus_multistream_packet_pad(unsigned char *data, opus_int32 len, opus_int32 new_len, int nb_streams) +{ + int s; + int count; + unsigned char toc; + opus_int16 size[48]; + opus_int32 packet_offset; + opus_int32 amount; + + if (len < 1) + return OPUS_BAD_ARG; + if (len==new_len) + return OPUS_OK; + else if (len > new_len) + return OPUS_BAD_ARG; + amount = new_len - len; + /* Seek to last stream */ + for (s=0;s +#include +#include + +#define MAX_PACKETOUT 32000 + +void usage(char *argv0) +{ + fprintf(stderr, "usage: %s [options] input_file output_file\n", argv0); +} + +static void int_to_char(opus_uint32 i, unsigned char ch[4]) +{ + ch[0] = i>>24; + ch[1] = (i>>16)&0xFF; + ch[2] = (i>>8)&0xFF; + ch[3] = i&0xFF; +} + +static opus_uint32 char_to_int(unsigned char ch[4]) +{ + return ((opus_uint32)ch[0]<<24) | ((opus_uint32)ch[1]<<16) + | ((opus_uint32)ch[2]<< 8) | (opus_uint32)ch[3]; +} + +int main(int argc, char *argv[]) +{ + int i, eof=0; + FILE *fin, *fout; + unsigned char packets[48][1500]; + int len[48]; + int rng[48]; + OpusRepacketizer *rp; + unsigned char output_packet[MAX_PACKETOUT]; + int merge = 1, split=0; + + if (argc < 3) + { + usage(argv[0]); + return EXIT_FAILURE; + } + for (i=1;i48) + { + fprintf(stderr, "-merge parameter must be less than 48.\n"); + return EXIT_FAILURE; + } + i++; + } else if (strcmp(argv[i], "-split")==0) + split = 1; + else + { + fprintf(stderr, "Unknown option: %s\n", argv[i]); + usage(argv[0]); + return EXIT_FAILURE; + } + } + fin = fopen(argv[argc-2], "r"); + if(fin==NULL) + { + fprintf(stderr, "Error opening input file: %s\n", argv[argc-2]); + return EXIT_FAILURE; + } + fout = fopen(argv[argc-1], "w"); + if(fout==NULL) + { + fprintf(stderr, "Error opening output file: %s\n", argv[argc-1]); + fclose(fin); + return EXIT_FAILURE; + } + + rp = opus_repacketizer_create(); + while (!eof) + { + int err; + int nb_packets=merge; + opus_repacketizer_init(rp); + for (i=0;i1500 || len[i]<0) + { + if (feof(fin)) + { + eof = 1; + } else { + fprintf(stderr, "Invalid payload length\n"); + fclose(fin); + fclose(fout); + return EXIT_FAILURE; + } + break; + } + if (fread(ch, 1, 4, fin)!=4) + { + if (feof(fin)) + { + eof = 1; + } else { + fprintf(stderr, "Error reading.\n"); + fclose(fin); + fclose(fout); + return EXIT_FAILURE; + } + break; + } + rng[i] = char_to_int(ch); + if (fread(packets[i], len[i], 1, fin)!=1) { + if (feof(fin)) + { + eof = 1; + } else { + fprintf(stderr, "Error reading packet of %u bytes.\n", len[i]); + fclose(fin); + fclose(fout); + return EXIT_FAILURE; + } + break; + } + err = opus_repacketizer_cat(rp, packets[i], len[i]); + if (err!=OPUS_OK) + { + fprintf(stderr, "opus_repacketizer_cat() failed: %s\n", opus_strerror(err)); + break; + } + } + nb_packets = i; + + if (eof) + break; + + if (!split) + { + err = opus_repacketizer_out(rp, output_packet, MAX_PACKETOUT); + if (err>0) { + unsigned char int_field[4]; + int_to_char(err, int_field); + if(fwrite(int_field, 1, 4, fout)!=4){ + fprintf(stderr, "Error writing.\n"); + return EXIT_FAILURE; + } + int_to_char(rng[nb_packets-1], int_field); + if (fwrite(int_field, 1, 4, fout)!=4) { + fprintf(stderr, "Error writing.\n"); + return EXIT_FAILURE; + } + if (fwrite(output_packet, 1, err, fout)!=(unsigned)err) { + fprintf(stderr, "Error writing.\n"); + return EXIT_FAILURE; + } + /*fprintf(stderr, "out len = %d\n", err);*/ + } else { + fprintf(stderr, "opus_repacketizer_out() failed: %s\n", opus_strerror(err)); + } + } else { + int nb_frames = opus_repacketizer_get_nb_frames(rp); + for (i=0;i0) { + unsigned char int_field[4]; + int_to_char(err, int_field); + if (fwrite(int_field, 1, 4, fout)!=4) { + fprintf(stderr, "Error writing.\n"); + return EXIT_FAILURE; + } + if (i==nb_frames-1) + int_to_char(rng[nb_packets-1], int_field); + else + int_to_char(0, int_field); + if (fwrite(int_field, 1, 4, fout)!=4) { + fprintf(stderr, "Error writing.\n"); + return EXIT_FAILURE; + } + if (fwrite(output_packet, 1, err, fout)!=(unsigned)err) { + fprintf(stderr, "Error writing.\n"); + return EXIT_FAILURE; + } + /*fprintf(stderr, "out len = %d\n", err);*/ + } else { + fprintf(stderr, "opus_repacketizer_out() failed: %s\n", opus_strerror(err)); + } + + } + } + } + + fclose(fin); + fclose(fout); + return EXIT_SUCCESS; +} diff --git a/lib/opus/src/tansig_table.h b/lib/opus/src/tansig_table.h new file mode 100644 index 00000000..c76f844a --- /dev/null +++ b/lib/opus/src/tansig_table.h @@ -0,0 +1,45 @@ +/* This file is auto-generated by gen_tables */ + +static const float tansig_table[201] = { +0.000000f, 0.039979f, 0.079830f, 0.119427f, 0.158649f, +0.197375f, 0.235496f, 0.272905f, 0.309507f, 0.345214f, +0.379949f, 0.413644f, 0.446244f, 0.477700f, 0.507977f, +0.537050f, 0.564900f, 0.591519f, 0.616909f, 0.641077f, +0.664037f, 0.685809f, 0.706419f, 0.725897f, 0.744277f, +0.761594f, 0.777888f, 0.793199f, 0.807569f, 0.821040f, +0.833655f, 0.845456f, 0.856485f, 0.866784f, 0.876393f, +0.885352f, 0.893698f, 0.901468f, 0.908698f, 0.915420f, +0.921669f, 0.927473f, 0.932862f, 0.937863f, 0.942503f, +0.946806f, 0.950795f, 0.954492f, 0.957917f, 0.961090f, +0.964028f, 0.966747f, 0.969265f, 0.971594f, 0.973749f, +0.975743f, 0.977587f, 0.979293f, 0.980869f, 0.982327f, +0.983675f, 0.984921f, 0.986072f, 0.987136f, 0.988119f, +0.989027f, 0.989867f, 0.990642f, 0.991359f, 0.992020f, +0.992631f, 0.993196f, 0.993718f, 0.994199f, 0.994644f, +0.995055f, 0.995434f, 0.995784f, 0.996108f, 0.996407f, +0.996682f, 0.996937f, 0.997172f, 0.997389f, 0.997590f, +0.997775f, 0.997946f, 0.998104f, 0.998249f, 0.998384f, +0.998508f, 0.998623f, 0.998728f, 0.998826f, 0.998916f, +0.999000f, 0.999076f, 0.999147f, 0.999213f, 0.999273f, +0.999329f, 0.999381f, 0.999428f, 0.999472f, 0.999513f, +0.999550f, 0.999585f, 0.999617f, 0.999646f, 0.999673f, +0.999699f, 0.999722f, 0.999743f, 0.999763f, 0.999781f, +0.999798f, 0.999813f, 0.999828f, 0.999841f, 0.999853f, +0.999865f, 0.999875f, 0.999885f, 0.999893f, 0.999902f, +0.999909f, 0.999916f, 0.999923f, 0.999929f, 0.999934f, +0.999939f, 0.999944f, 0.999948f, 0.999952f, 0.999956f, +0.999959f, 0.999962f, 0.999965f, 0.999968f, 0.999970f, +0.999973f, 0.999975f, 0.999977f, 0.999978f, 0.999980f, +0.999982f, 0.999983f, 0.999984f, 0.999986f, 0.999987f, +0.999988f, 0.999989f, 0.999990f, 0.999990f, 0.999991f, +0.999992f, 0.999992f, 0.999993f, 0.999994f, 0.999994f, +0.999994f, 0.999995f, 0.999995f, 0.999996f, 0.999996f, +0.999996f, 0.999997f, 0.999997f, 0.999997f, 0.999997f, +0.999997f, 0.999998f, 0.999998f, 0.999998f, 0.999998f, +0.999998f, 0.999998f, 0.999999f, 0.999999f, 0.999999f, +0.999999f, 0.999999f, 0.999999f, 0.999999f, 0.999999f, +0.999999f, 0.999999f, 0.999999f, 0.999999f, 0.999999f, +1.000000f, 1.000000f, 1.000000f, 1.000000f, 1.000000f, +1.000000f, 1.000000f, 1.000000f, 1.000000f, 1.000000f, +1.000000f, +}; diff --git a/lib/opus/tests/meson.build b/lib/opus/tests/meson.build new file mode 100644 index 00000000..5f3ac9df --- /dev/null +++ b/lib/opus/tests/meson.build @@ -0,0 +1,34 @@ +# Tests that link to libopus +opus_tests = [ + ['test_opus_api'], + ['test_opus_decode', [], 60], + ['test_opus_encode', 'opus_encode_regressions.c', 120], + ['test_opus_padding'], + ['test_opus_projection'], +] + +foreach t : opus_tests + test_name = t.get(0) + extra_srcs = t.get(1, []) + + test_kwargs = {} + if t.length() > 2 + test_kwargs += {'timeout': t[2]} + endif + + exe_kwargs = {} + # This test uses private symbols + if test_name == 'test_opus_projection' + exe_kwargs = { + 'link_with': [celt_lib, silk_lib], + 'objects': opus_lib.extract_all_objects(), + } + endif + + exe = executable(test_name, '@0@.c'.format(test_name), extra_srcs, + include_directories: opus_includes, + dependencies: [libm, opus_dep], + install: false, + kwargs: exe_kwargs) + test(test_name, exe, kwargs: test_kwargs) +endforeach diff --git a/lib/opus/tests/opus_build_test.sh b/lib/opus/tests/opus_build_test.sh new file mode 100755 index 00000000..573f4473 --- /dev/null +++ b/lib/opus/tests/opus_build_test.sh @@ -0,0 +1,29 @@ +#!/bin/sh + +tarball=`realpath "$1"` +nb_tests="$2" +oldvectors=`realpath "$3"` +newvectors=`realpath "$4"` +base=`basename "$tarball" .tar.gz` + +tar xvf "$tarball" > /dev/null 2>&1 +cd "$base" + +if [ $? -ne 0 ] +then + echo cannot go to "$base" + exit 1 +fi + +mkdir build_tests + +configure_dir=`pwd` +seq -w "$nb_tests" | parallel --halt now,fail=10 -j +2 -q ../random_config.sh "build_tests/run_{}" "$configure_dir" "$oldvectors" "$newvectors" + +if [ $? -ne 0 ] +then + echo Check found errors + exit 1 +else + echo No error found +fi diff --git a/lib/opus/tests/opus_decode_fuzzer.c b/lib/opus/tests/opus_decode_fuzzer.c new file mode 100644 index 00000000..ea6ec4fd --- /dev/null +++ b/lib/opus/tests/opus_decode_fuzzer.c @@ -0,0 +1,124 @@ +/* Copyright (c) 2017 Google Inc. */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include "opus.h" +#include "opus_types.h" + +#define MAX_FRAME_SAMP 5760 +#define MAX_PACKET 1500 + +/* 4 bytes: packet length, 4 bytes: encoder final range */ +#define SETUP_BYTE_COUNT 8 + +#define MAX_DECODES 12 + +typedef struct { + int fs; + int channels; +} TocInfo; + +static void ParseToc(const uint8_t *toc, TocInfo *const info) { + const int samp_freqs[5] = {8000, 12000, 16000, 24000, 48000}; + const int bandwidth = opus_packet_get_bandwidth(toc); + + info->fs = samp_freqs[bandwidth - OPUS_BANDWIDTH_NARROWBAND]; + info->channels = opus_packet_get_nb_channels(toc); +} + +/* Treats the input data as concatenated packets encoded by opus_demo, + * structured as + * bytes 0..3: packet length + * bytes 4..7: encoder final range + * bytes 8+ : Opus packet, including ToC + */ +int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) { + OpusDecoder *dec; + opus_int16 *pcm; + uint8_t *temp_data; + TocInfo toc; + int i = 0; + int err = OPUS_OK; + int num_decodes = 0; + + /* Not enough data to setup the decoder (+1 for the ToC) */ + if (size < SETUP_BYTE_COUNT + 1) { + return 0; + } + + /* Create decoder based on info from the first ToC available */ + ParseToc(&data[SETUP_BYTE_COUNT], &toc); + + dec = opus_decoder_create(toc.fs, toc.channels, &err); + if (err != OPUS_OK || dec == NULL) { + return 0; + } + + pcm = (opus_int16*) malloc(sizeof(*pcm) * MAX_FRAME_SAMP * toc.channels); + + while (i + SETUP_BYTE_COUNT < size && num_decodes++ < MAX_DECODES) { + int len, fec; + + len = (opus_uint32) data[i ] << 24 | + (opus_uint32) data[i + 1] << 16 | + (opus_uint32) data[i + 2] << 8 | + (opus_uint32) data[i + 3]; + if (len > MAX_PACKET || len < 0 || i + SETUP_BYTE_COUNT + len > size) { + break; + } + + /* Bytes 4..7 represent encoder final range, but are unused here. + * Instead, byte 4 is repurposed to determine if FEC is used. */ + fec = data[i + 4] & 1; + + if (len == 0) { + /* Lost packet */ + int frame_size; + opus_decoder_ctl(dec, OPUS_GET_LAST_PACKET_DURATION(&frame_size)); + (void) opus_decode(dec, NULL, len, pcm, frame_size, fec); + } else { + temp_data = (uint8_t*) malloc(len); + memcpy(temp_data, &data[i + SETUP_BYTE_COUNT], len); + + (void) opus_decode(dec, temp_data, len, pcm, MAX_FRAME_SAMP, fec); + + free(temp_data); + } + + i += SETUP_BYTE_COUNT + len; + } + + opus_decoder_destroy(dec); + free(pcm); + + return 0; +} diff --git a/lib/opus/tests/opus_decode_fuzzer.options b/lib/opus/tests/opus_decode_fuzzer.options new file mode 100644 index 00000000..e5ae71b9 --- /dev/null +++ b/lib/opus/tests/opus_decode_fuzzer.options @@ -0,0 +1,2 @@ +[libfuzzer] +max_len = 1000000 diff --git a/lib/opus/tests/opus_encode_regressions.c b/lib/opus/tests/opus_encode_regressions.c new file mode 100644 index 00000000..4d506eb6 --- /dev/null +++ b/lib/opus/tests/opus_encode_regressions.c @@ -0,0 +1,1034 @@ +/* Copyright (c) 2016 Mark Harris, Jean-Marc Valin */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include +#include +#include +#include "opus_multistream.h" +#include "opus.h" +#include "test_opus_common.h" + + +static int celt_ec_internal_error(void) +{ + OpusMSEncoder *enc; + int err; + unsigned char data[2460]; + int streams; + int coupled_streams; + unsigned char mapping[1]; + + enc = opus_multistream_surround_encoder_create(16000, 1, 1, &streams, + &coupled_streams, mapping, OPUS_APPLICATION_VOIP, &err); + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(8)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(OPUS_AUTO)); + { + static const short pcm[320] = + { + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 1792, 1799, 1799, + 1799, 1799, 1799, 1799, 1799, 1799, 1799, 1799, 1799, + 1799, 1799, 1799, 1799, 1799, 0, 25600, 1799, 1799, + 1799, 1799, 1799, 1799, 1799, 1799, 1799, 1799, 1799, + 1799, 1799, 1799, 1799, 7, 0, 255, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 32767, -1, + 0, 0, 0, 100, 0, 16384, 0, 0, 0, + 0, 0, 0, 4, 0, 0, -256, 255, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0,-32768, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 128, 0, 0, 0, 0, + 0, 0, 0, 0, -256, 0, 0, 32, 0, + 0, 0, 0, 0, 0, 0, 4352, 4, 228, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 5632, 0, 0, + 0, 0,-32768, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 256, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + -3944, 10500, 4285, 10459, -6474, 10204, -6539, 11601, -6824, + 13385, -7142, 13872,-11553, 13670, -7725, 13463, -6887, 7874, + -5580, 12600, -4964, 12480, 3254, 11741, -4210, 9741, -3155, + 7558, -5468, 5431, -1073, 3641, -1304, 0, -1, 343, + 26, 0, 0, 150, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 1799, 1799, 1799, 1799, 1799, -2553, + 7, 1792, 1799, 1799, 1799, 1799, 1799, 1799, 1799, + 1799, 1799, 1799, 1799, -9721 + }; + err = opus_multistream_encode(enc, pcm, 320, data, 2460); + opus_test_assert(err > 0); + } + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(10)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(18)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(90)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(280130)); + { + static const short pcm[160] = + { + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9526, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, 25600, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510 + }; + err = opus_multistream_encode(enc, pcm, 160, data, 2460); + opus_test_assert(err > 0); + } + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(10)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(18)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(90)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(280130)); + { + static const short pcm[160] = + { + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9494, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510 + }; + err = opus_multistream_encode(enc, pcm, 160, data, 2460); + opus_test_assert(err > 0); + } + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(10)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(18)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(90)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(280130)); + { + static const short pcm[160] = + { + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9479, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, -9510, + -9510, -9510, -9510, -9510, -9510, -9510, -9510 + }; + err = opus_multistream_encode(enc, pcm, 160, data, 2460); + opus_test_assert(err > 0); + } + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(10)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(18)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(90)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(280130)); + { + static const short pcm[160] = + { + -9510, -9510, 1799, 1799, 1799, 1799, 1799, 1799, 1799, + 1799, 1799, 1799, 1799, 1799, 1799, 1799, 1799, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + -256, 255, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 128, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 32, 0, 0, 0, 0, 0, 0, 0, + 4352, 4, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 148, 0, 0, 0, 0, + 5632 + }; + err = opus_multistream_encode(enc, pcm, 160, data, 2460); + opus_test_assert(err > 0); + } + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(12)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(41)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(21425)); + { + static const short pcm[40] = + { + 10459, -6474, 10204, -6539, 11601, -6824, 13385, -7142, 13872, + -11553, 13670, -7725, 13463, -6887, 12482, -5580, 12600, -4964, + 12480, 3254, 11741, -4210, 9741, -3155, 7558, -5468, 5431, + -1073, 3641, -1304, 0, -1, 343, 26, 0, 0, + 0, 0, -256, 226 + }; + err = opus_multistream_encode(enc, pcm, 40, data, 2460); + opus_test_assert(err > 0); + /* returns -3 */ + } + opus_multistream_encoder_destroy(enc); + return 0; +} + +static int mscbr_encode_fail10(void) +{ + OpusMSEncoder *enc; + int err; + unsigned char data[627300]; + static const unsigned char mapping[255] = + { + 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, + 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, + 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, + 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, + 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, + 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99,100,101, + 102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118, + 119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135, + 136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152, + 153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169, + 170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186, + 187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203, + 204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219,220, + 221,222,223,224,225,226,227,228,229,230,231,232,233,234,235,236,237, + 238,239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254 + }; + + enc = opus_multistream_encoder_create(8000, 255, 254, 1, mapping, + OPUS_APPLICATION_RESTRICTED_LOWDELAY, &err); + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(2)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(2)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(14)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(57)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(3642675)); + { + static const short pcm[20*255] = + { + 0 + }; + err = opus_multistream_encode(enc, pcm, 20, data, 627300); + opus_test_assert(err > 0); + /* returns -1 */ + } + opus_multistream_encoder_destroy(enc); + return 0; +} + +static int mscbr_encode_fail(void) +{ + OpusMSEncoder *enc; + int err; + unsigned char data[472320]; + static const unsigned char mapping[192] = + { + 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, + 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, + 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, + 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, + 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, + 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99,100,101, + 102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118, + 119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135, + 136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152, + 153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169, + 170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186, + 187,188,189,190,191 + }; + + enc = opus_multistream_encoder_create(8000, 192, 189, 3, mapping, + OPUS_APPLICATION_RESTRICTED_LOWDELAY, &err); + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_MEDIUMBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(8)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(15360)); + { + static const short pcm[20*192] = + { + 0 + }; + err = opus_multistream_encode(enc, pcm, 20, data, 472320); + opus_test_assert(err > 0); + /* returns -1 */ + } + opus_multistream_encoder_destroy(enc); + return 0; +} + +static int surround_analysis_uninit(void) +{ + OpusMSEncoder *enc; + int err; + unsigned char data[7380]; + int streams; + int coupled_streams; + unsigned char mapping[3]; + + enc = opus_multistream_surround_encoder_create(24000, 3, 1, &streams, + &coupled_streams, mapping, OPUS_APPLICATION_AUDIO, &err); + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(8)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(84315)); + { + static const short pcm[960*3] = + { + -6896, 4901, -6158, 4120, -5164, 3631, -4442, 3153, -4070, + 3349, -4577, 4474, -5541, 5058, -6701, 3881, -7933, 1863, + -8041, 697, -6738,-31464, 14330,-12523, 4096, -6130, 29178, + -250,-21252, 10467, 16907, -3359, -6644, 31965, 14607,-21544, + -32497, 24020, 12557,-26926,-18421, -1842, 24587, 19659, 4878, + 10954, 23060, 8907,-10215,-16179, 31772,-11825,-15590,-23089, + 17173,-25903,-17387, 11733, 4884, 10204,-16476,-14367, 516, + 20453,-16898, 20967,-23813, -20, 22011,-17167, 9459, 32499, + -25855, -523, -3883, -390, -4206, 634, -3767, 2325, -2751, + 3115, -2392, 2746, -2173, 2317, -1147, 2326, 23142, 11314, + -15350,-24529, 3026, 6146, 2150, 2476, 1105, -830, 1775, + -3425, 3674,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + 4293,-14023, 3879,-15553, 3158,-16161, 2629, 18433,-12535, + -6645,-20735,-32763,-13824,-20992, 25859, 13052, -8731, 2292, + -3860, 24049, 10225,-19220, 10478,-22294, 22773, 28137, 13816, + 30953,-25863,-24598, 16888,-14612,-28942, 20974,-27397,-18944, + -18690, 20991,-16638, 5632,-14330, 28911,-25594, 17408, 29958, + -517,-20984, -1800, 11281, 9977,-21221,-14854, 23840, -9477, + 3362,-12805,-22493, 32507, 156, 16384, -1163, 2301, -1874, + 4600, -1748, 6950, 16557, 8192, -7372, -1033, -3278, 2806, + 20275, 3317, -717, 9792, -767, 9099, -613, 8362, 5027, + 7774, 2597, 8549, 5278, 8743, 9343, 6940, 13038, 4826, + 14086, 2964, 13215, 1355, 11596, 455, 9850, -519, 10680, + -2287, 12551, -3736, 13639, -4291, 13790, -2722, 14544, -866, + 15050, -304, 22833, -1196, 13520, -2063, 13051, -2317, 13066, + -2737, 13773, -2664, 14105, -3447, 13854, 24589, 24672, -5280, + 10388, -4933, 7543, -4149, 3654, -1552, 1726, 661, 57, + 2922, -751, 3917, 8419, 3840, -5218, 3435, 5540, -1073, + 4153, -6656, 1649, -769, -7276,-13072, 6380, -7948, 20717, + 18425, 17392, 14335,-18190, -1842, 24587, 19659, 11790, 10954, + 23060, 8907,-10215,-16179, 31772,-11825,-15590,-23101, 17173, + -25903,-17387, 11733, 4884, 10192,-16627,-14367, 516, 20453, + -16898, 20967,-23813, -20, 22011,-17167, 9468, 32499,-25607, + -523, -3883, -390, -4206, 634, -3767, 2325, -2751, 3115, + -2392, 2746, -2161, 2317, -1147, 2326, 23142, 11314,-15350, + -29137, 3026,-15056, -491,-15170, -386,-16015, -641,-16505, + -930,-16206, -717,-16175, -2839,-16374, -4558,-16237, -5207, + -15903, -6421, 6373, -1403, 5431, -1073, 3641, -1304, -4495, + -769, -7276, 2856, -7870, 3314, -8730, 3964,-10183, 4011, + -11135, 3421,-11727, 2966,-12360, 2818,-13472, 3660,-13805, + 5162,-13478, 6434,-12840, 7335,-12420, 6865,-12349, 5541, + -11965, 5530,-10820, 5132, -9197, 3367, -7745, 1223, -6910, + -433, -6211, -1711, -4958, -1025, -3755, -836, -3292, -1666, + -2661,-10755, 31472,-27906, 31471, 18690, 5617, 16649, 11253, + -22516,-17674,-31990, 3575,-31479, 5883, 26121, 12890, -6612, + 12228,-11634, 523, 26136,-21496, 20745,-15868, -4100,-24826, + 23282, 22798, 491, -1774, 15075,-27373,-13094, 6417,-29487, + 14608, 10185, 16143, 22211, -8436, 4288, -8694, 2375, 3023, + 486, 1455, 128, 202, 942, -923, 2068, -1233, -717, + -1042, -2167, 32255, -4632, 310, -4458, -3639, -5258, 2106, + -6857, 2681, -7497, 2765, -6601, 1945, -5219, 19154, -4877, + 619, -5719, -1928, -6208, -121, 593, 188, 1558, -4152, + 1648, 156, 1604, -3664, -6862, -2851, -5112, -3600, -3747, + -5081, -4428, -5592, 20974,-27397,-18944,-18690, 20991,-17406, + 5632,-14330, 28911, 15934, 15934, 15934, 15934, 15934, 15934, + 15934, 15934, 15934, 15934, 15934, 15934,-25594, 17408, 29958, + -7173,-16888, 9098, -613, 8362, 675, 7774, 2597, 8549, + 5278, 8743, 9375, 6940, 13038, 4826, 14598, 7721,-24308, + -29905,-19703,-17106,-16124, -3287,-26118,-19709,-10769, 24353, + 28648, 6946, -1363, 12485, -1187, 26074,-25055, 10004,-24798, + 7204, -4581, -9678, 1554, 10553, 3102, 12193, 2443, 11955, + 1213, 10689, -1293, 921, -4173, 10709, -6049, 8815, -7128, + 8147, -8308, 6847, -2977, 4920,-11447,-22426,-11794, 3514, + -10220, 3430, -7993, 1926, -7072, 327, -7569, -608, -7605, + 3695, -6271, -1579, -4877, -1419, -3103, -2197, 128, -3904, + 3760, -5401, 4906, -6051, 4250, -6272, 3492, -6343, 3197, + -6397, 4041, -6341, 6255, -6381, 7905, 16504, 0, -6144, + 8062, -5606, 8622, -5555, -9, -1, 7423, 0, 1, + 238, 5148, 1309, 4700, 2218, 4403, 2573, 3568, 28303, + 1758, 3454, -1247, 3434, -4912, 2862, -7844, 1718,-10095, + 369,-12631, 128, -3904, 3632, -5401, 4906, -6051, 4250, + -6272, 3492, -6343, 3197, -6397, 4041, -6341, 6255, -6381, + 7905, 16504, 0, -6144, 8062, -5606, 8622, -5555, 8439, + -3382, 7398, -1170, 6132, 238, 5148, 1309, 4700, 2218, + 4403, 2573, 3568, 2703, 1758, 3454, -1247, 3434, -4912, + 2862, -7844, 1718,-10095, 369,-12631, -259,-14632, 234, + -15056, -521,-15170, -386,-16015, -641,-16505, -930,-16206, + -1209,-16146, -2839,-16374, -4558,-16218, -5207,-15903, -6421, + -15615, -6925,-14871, -6149,-13759, -5233,-12844, 18313, -4357, + -5696, 2804, 12992,-22802, -6720, -9770, -7088, -8998, 14330, + -12523, 14843, -6130, 29178, -250,-27396, 10467, 16907, -3359, + -6644, 31965, 14607,-21544,-32497, 24020, 12557,-26926, -173, + -129, -6401, -130,-25089, -3841, -4916, -3048, 224, -237, + -3969, -189, -3529, -535, -3464,-14863,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14395,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907, 0, 32512,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907, 9925, -718, 9753, -767, + 9099, -613, 8362, 675, 7774, 2597, 8549, 5278, 8743, + 9375, 6940, 13038, 4826, 14598, 7721,-24308,-29905,-19703, + -17106,-16124, -3287,-26118,-19709, 0, 24353, 28648, 10274, + -11292,-29665,-16417, 24346, 14553, 18707, 26323, -4596,-17711, + 5133, 26328, 13,-31168, 24583, 18404,-28927,-24350, 19453, + 28642, 1019,-10777, -3079, 30188, -7686, 27635,-32521,-16384, + 12528, -6386, 10986, 23827,-25880,-32752,-23321, 14605, 32231, + 780,-13849, 15119, 28647, 4888, -7705, 30750, 64, 0, + 32488, 6687,-20758, 19745, -2070,-13792, -6414, 28188, -2821, + -4585, 7168, 7444, 23557,-21998, 13064, 3345, -4086,-28915, + -8694, 32262, 8461, 27387,-12275, 12012, 23563,-18719,-28410, + 29144,-22271, -562, -9986, -5434, 12288, 5573,-16642, 32448, + 29182, 32705,-30723, 24255,-19716, 18368, -4357, -5696, 2804, + 12992,-22802,-22080, -7701, -5183, 486, -3133, -5660, -1083, + 16871,-28726,-11029,-30259, -1209,-16146, -2839,-16374, -4558, + -16218,-10523, 20697, -9500, -1316, 5431, -1073, 3641, -1304, + 1649, -769, -7276, 2856, -7870, 3314, -8730, 3964,-10183, + 4011,-11135, 3421,-11727, 21398, 32767, -1, 32486, -1, + 6301,-13071, 6380, -7948, -1, 32767, 240, 14081, -5646, + 30973, -3598,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907, 32767,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907, 8901, 9375, 6940, 13038, 4826, 14598, 7721,-24308, + -29905,-19703,-17106,-16124, -3287,-26118,-19709,-10769, 24361, + 28648, 10274,-11292,-29665,-16417, 24346, 14580, 18707, 26323, + -4440,-17711, 5133, 26328,-14579,-31008, 24583, 18404, 28417, + -24350, 19453, 28642,-32513,-10777, -3079, 30188, -7686, 27635, + -32521,-16384,-20240, -6386, 10986, 23827,-25880,-32752,-23321, + 14605, 32231, 780,-13849, 15119, 28647, 4888, -7705,-15074, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907, 8192,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14897,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -15931,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907, 26121, 12890, 2604, + 12228,-11634, 12299, 5573,-16642, 32452, 29182, 32705,-30723, + 24255,-19716, 13248,-11779, -5696, 2804, 12992,-27666,-22080, + -7701, -5183, -6682,-31464, 14330,-12523, 14843, -6130, 29178, + -18,-27396, 10467, 16907, -3359, -6644, 31965, 14607,-21544, + -32497, 24020, 12557,-26926,-18421, 706, 24587, 19659, 4878, + 10954, 23060, 8907,-10215,-22579, 31772,-11825,-15590,-23089, + 17173,-25903,-17387, 3285, 4884, 10204,-16627,-14367, 516, + 20453,-16898, 20967,-23815, -20, 22011,-17167, 9468, 32499, + -25607, -523, -3883, -390, -4206, 634, -3767, 2325, -2751, + 3115, -2392, 2746, -2173, 2317, -1147, 2326, 23142, 11314, + -15130,-29137, 3026, 6146, 2150, 2476, 1105, -830, 1775, + -3425, 3674, -5287, 4609, -7175, 4922, -9579, 4556,-12007, + 4236,-14023, 3879,-15553, 3158,-16161, 2576, 18398,-12535, + -6645,-20735,-32763,-13824,-20992, 25859, 5372, 12040, 13307, + -4355,-30213, -9, -6019, 14061,-31487,-13842, 30449, 15083, + 14088, 31205,-18678,-12830, 14090,-26138,-25337,-11541, -3254, + 27628,-22270, 30953,-16136,-30745, 20991,-17406, 5632,-14330, + 28911,-25594, 17408,-20474, 13041, -8731, 2292, -3860, 24049, + 10225,-19220, 10478, -4374, -1199, 148, -330, -74, 593, + 188, 1558, -4152, 15984, 15934, 15934, 15934, 15934, 15934, + 15934, 15934, 15934, 15934, 15934, 15934, 1598, 156, 1604, + -1163, 2278,-30018,-25821,-21763,-23776, 24066, 9502, 25866, + -25055, 10004,-24798, 7204, -4581, -9678, 1554, 10553, 3102, + 12193, 2443, 11955, 1213, 10689, -1293, 921, -4173, 8661, + -6049, 8815,-21221,-14854, 23840, -9477, 8549, 5278, 8743, + 9375, 6940, 13038, 4826, 14598, 7721,-24308,-29905,-19703, + -17106,-16124, -3287,-26118,-19709,-10769, 24361, 28648, 10274, + -11292,-29665,-16417, 24346, 14580, 18707, 26323, -4410,-17711, + 5133, 26328,-14579,-31008, 24583, 18404, 28417,-24350, 19453, + 28642,-32513,-10777, -3079, 30188, -7686, 27635,-32521,-16384, + -20240, -6386, 10986, 23827,-25880,-32752,-23321, 14605, 32231, + 780,-13849, 15119, 28647, 4888, -7705, 30750, 64, 0, + 32488, 6687,-20758, 19745, -2070, -1, -1, 28, 256, + -4608, 7168, 7444, 23557,-21998, 13064, 3345, -4086,-28915, + -8594, 32262, 8461, 27387,-12275, 12012, 23563,-18719,-28410, + 29144,-22271,-32562,-16384, 12528, -6386, 10986, 23827,-25880, + -32752,-23321, 14605, 32231, 780,-13849, 15119, 28647, 4888, + -7705, 30750, 64, 0, 32488, 6687,-20758, 19745, -2070, + -13792, -6414, 28188, -2821, -4585, 7168, 7444, 23557,-21998, + 13064, 3345, -4086,-28915, -8694, 32262, 8461, 27387,-12275, + 12012, 23563,-18719,-28410, 29144,-22271, -562, -9986, -5434, + 12288, -2107,-16643, 32452, 29182, 32705,-30723, 24255,-19716, + 18368, -4357, -5696, 2804, 12992,-22802,-22080, -7701, -5183, + 486, -3133, -5660, -1083, 16871,-28726,-11029,-30259, -1209, + -16146, -2839,-16374, -4558,-16218,-10523, 20697, -9500, -1316, + 5431, -1073, 3641, -1304, 1649, -769, -7276, 2856, -7870, + 3314, -8730, 3964,-10183, 4011,-11135, 3421,-11727, 21398, + 32767, -1, 32486, -1, -99,-13072, 6380, -7948, 4864, + 32767, 17392, 14335, -5646, 30973, -3598,-10299,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14905,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-19771,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-16443,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-15931,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907, -1,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907, 7877, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, -994, -7276, 2856, -7870, + 3314, -8730, 3964,-10183, 4011,-11135, 3421,-11727, 21398, + 32767, -1, 32486, -1, -99,-13072, 6380, -7948, 4864, + 32767, 17392, 14335, -5646, 30973, -3598,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14905,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907, 197, 0,-14977,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907, 12838, 6653, 294, + -29699,-25821,-21763,-23776, 24066, 9502, 25866,-25055, 10004, + -24798, 7204, -4581, -9678, 1554, 10553, 3102, 12193, 2443, + 11955, 1213, 10689, -1293, 921, 179, 8448, -6049, 8815, + -7128, 8147, -8308, 6847, -9889, 4920,-11447, 3174,-11794, + 3514,-10220, 3430, 16384, 1926, -7072, 327, -7537, -608, + -7605, -1169, -6397, -1579, -4877, -1419, -3103, -2197, 128, + -3904, 3632, -5401, 4906, -6051, 4250, -6272, 3492, -6343, + 3197, -6397, 4041, -6341, 6255, -6381, 7905, 16504, 0, + -6144, 8062, -5606, 8622, -5555, 8439, -3382, 7398, -1170, + 6132, 238, 5148, 1309, 4700, 2218, 4403, 2573, 3568, + 2703, 1758, 3454, -1247, 3434, -4912, 2862, -7844, 1718, + -10095, 369,-12631, -259,-14632, 234,-15056, -491,-16194, + -386,-16015, -641,-16505, -930,-16206, -1209,-16146, -2839, + -16374, -4558,-16218, -5207,-15903, -6421,-15615, -6925,-14871, + -6149,-13759, -5233,-12844, 18313, -4357, -5696, 2804, 12992, + -22802, -6720, -9770, -7088, -8998, 14330,-12523, 14843, -6130, + 29178, -250,-27396, 10467, 16907, -3359, -6644, 31965, 14607, + -21544,-32497, 24020, 12557,-26926, -173, -129, -6401, -130, + -25089, -3816, -4916, -3048, -32, -1, -3969, 256, -3529, + -535, -3464,-14863,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -1209,-16146, -2839,-16374, -4558,-16218,-10523, 20697, -9500, + -1316, 5431, -1073, 3641, -1304, 1649, -769, -7276, 2856, + -7870, 3314, -8730, 3964,-10183, 4011,-11135, 3421,-11727, + 21398, 32767, -1, 32486, -1, 6301,-13071, 6380, -7948, + -1, 32767, 240, 14081, -5646, 30973, -3598,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + 32767,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907, 8901, 9375, 6940, + 13038, 4826, 14598, 7721,-24308,-29905,-19703,-17106,-16124, + -3287,-26118,-19709,-10769, 24361, 28648, 10274,-11292,-29665, + -16417, 24346, 14580, 18707, 26323, -4440,-17711, 5133, 26328, + -14579,-31008, 24583, 18404, 28417,-24350, 19453, 28642,-32513, + -10777, -3079, 30188, -7686, 27635,-32521,-16384,-20240, -6386, + 10986, 23827,-25880,-32752,-23321, 14605, 32231, 780,-13849, + 15119, 28647, 4888, -7705,-15074,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, 8192, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14897, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-15931,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907, 26121, 12890, 2604, 12228,-11634, 12299, 5573, + -16642, 32452, 29182, 32705,-30723, 24255,-19716, 13248,-11779, + -5696, 2804, 12992,-27666,-22080, -7701, -5183, -6682,-31464, + 14330,-12523, 14843, -6130, 29178, -18,-27396, 10467, 16907, + -3359, -6644, 31965, 14607,-21544,-32497, 24020, 12557,-26926, + -18421, 706, 24587, 19659, 4878, 10954, 23060, 8907,-10215, + -22579, 31772,-11825,-15590,-23089, 17173,-25903,-17387, 3285, + 4884, 10204,-16627,-14367, 516, 20453,-16898, 20967,-23815, + -20, 22011,-17167, 9468, 32499,-25607, -523, -3883, -390, + -4206, 634, -3767, 2325, -2751, 3115, -2392, 2746, -2173, + 2317, -1147, 2326, 23142, 11314,-15130,-29137, 3026, 6146, + 2150, 2476, 1105, -830, 1775, -3425, 3674, -5287, 4609, + -7175, 4922, -9579, 4556,-12007, 4236,-14023, 3879,-15553, + 3158,-16161, 2576, 18398,-12535, -6645,-20735,-32763,-13824, + -20992, 25859, 5372, 12040, 13307, -4355,-30213, -9, -6019 + }; + err = opus_multistream_encode(enc, pcm, 960, data, 7380); + opus_test_assert(err > 0); + } + opus_multistream_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_VBR_CONSTRAINT(0)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PHASE_INVERSION_DISABLED(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_DTX(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_COMPLEXITY(6)); + opus_multistream_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_AUTO)); + opus_multistream_encoder_ctl(enc, OPUS_SET_LSB_DEPTH(9)); + opus_multistream_encoder_ctl(enc, OPUS_SET_INBAND_FEC(1)); + opus_multistream_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(5)); + opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(775410)); + { + static const short pcm[1440*3] = + { + 30449, 15083, 14088, 31205,-18678,-12830, 14090,-26138,-25337, + -11541, -3254, 27628,-22270, 30953,-16136,-30745, 20991,-17406, + 5632,-14330, 28911,-25594, 17408,-20474, 13041, -8731, 2292, + -3860, 24049, 10225,-19220, 10478, -4374, -1199, 148, -330, + -74, 593, 188, 1558, -4152, 15984, 15934, 15934, 15934, + 15934, 15934, 15934, 15934, 15934, 15934, 15934, 15934, 1598, + 156, 1604, -1163, 2278,-30018,-25821,-21763,-23776, 24066, + 9502, 25866,-25055, 10004,-24798, 7204, -4581, -9678, 1554, + 10553, 3102, 12193, 2443, 11955, 1213, 10689, -1293, 921, + -4173, 8661, -6049, 8815,-21221,-14854, 23840, -9477, 8549, + 5278, 8743, 9375, 6940, 13038, 4826, 14598, 7721,-24308, + -29905,-19703,-17106,-16124, -3287,-26118,-19709,-10769, 24361, + 28648, 10274,-11292,-29665,-16417, 24346, 14580, 18707, 26323, + -4410,-17711, 5133, 26328,-14579,-31008, 24583, 18404, 28417, + -24350, 19453, 28642,-32513,-10777, -3079, 30188, -7686, 27635, + -32521,-16384,-20240, -6386, 10986, 23827,-25880,-32752,-23321, + 14605, 32231, 780,-13849, 15119, 28647, 4888, -7705, 30750, + 64, 0, 32488, 6687,-20758, 19745, -2070, -1, -1, + 28, 256, -4608, 7168, 7444, 23557,-21998, 13064, 3345, + -4086,-28915, -8594, 32262, 8461, 27387,-12275, 12012, 23563, + -18719,-28410, 29144,-22271,-32562,-16384, 12528, -6386, 10986, + 23827,-25880,-32752,-23321, 14605, 32231, 780,-13849, 15119, + 28647, 4888, -7705, 30750, 64, 0, 32488, 6687,-20758, + 19745, -2070,-13792, -6414, 28188, -2821, -4585, 7168, 7444, + 23557,-21998, 13064, 3345, -4086,-28915, -8694, 32262, 8461, + -14853,-14907,-14907,-14907,-14907, 32767,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14891,-14907,-14907,-14907, + -14907,-14907, 8901, 9375, 6940, 13038, 4826, 14598, 7721, + -24308,-29905,-19703,-17106,-16124, -3287,-26118,-19709,-10769, + 24361, 28648, 10274,-11292,-29665,-16417, 24346, 14580, 18707, + 26323, -4440,-17711, 5133, 26328,-14579,-31008, 24583, 18404, + 28417,-24350, 19453, 28642,-32513,-10777, -3079, 30188, -7686, + 27635,-32521,-16384,-20240, -6386, 10986, 23827,-25880,-32752, + -23321, 14605, 32231, 780,-13849, 15119, 28647, 4888, -7705, + -15074,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907, 8192,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14897,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-15931,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907, 26121, 12890, + 2604, 12228,-11634, 12299, 5573,-16642, 32452, 29182, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, 7710, + 7710,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-10811,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14917,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14938,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907,-14907, + -14907,-14907,-14907,-14907, -571, -9986, -58, 12542,-18491, + 32507, 12838, 6653, 294, -1, 0,-19968, 18368, -4357, + -5696, 2804, 12998,-22802,-22080, -7701, -5183, 486, -3133, + -5660, -1083, 13799,-28726,-11029, 205,-14848, 32464, -1, + -129,-13072, 6380, -7948, 20717, 18425, 17392, 14335, -5646, + 30973, -3598, 7188, -3867, 3055, -4247, 5597, -4011,-26427, + -11,-30418, 7922, 2614, 237, -5839,-27413,-17624,-29716, + -13539, 239, 20991, 18164, -4082,-16647,-27386, 19458, 20224, + 4619, 19728, -7409,-18186,-25073, 27627,-23539, -7945,-31464, + 14330,-12523,-22021, -7701, -5183, 486, -3133, -5660, -1083, + 13799,-28726,-11029, 205,-14848, 32464, -1, -129,-13072, + 6380, -7948, 20717, 18425, 17392, 14093, -5646, 30973, -3598, + 7188, -3867, 3055, 3689, -5401, 4906, -6051, 4250, -6272, + 3492, -6343, 3197, -6397, 4041, -6341, 6255, -6381, 239, + 20991, 18164, -4082,-16647,-27386, 19458, 20224, 4619, 19728, + -7409,-18186,-25073, 27627,-23539, -7945,-31464, 14330,-12523, + 14843, -6130, 30202, -250,-28420, 10467, 16907, -3359, -6644, + 31965, 3343,-11727, 2966,-12616, 3064,-13472, 6732,-12349, + 5541,-11965, 5530,-10820, -1912, -3637, 32285, -4607, 310, + -32768, 0, -5258, 2106, -6857, 2681, -5449, -3606, -6717, + -5482, -3606, -1853, 4082, -7631, -9808, -1742, -2851, -5112, + 64, -868,-13546,-13365,-13365,-13365,-13365,-13365,-13365, + -13365,-13365,-13365,-13365,-13365,-13365,-13365,-13365,-13365, + -13365,-13365,-13365,-13365,-13365,-13365,-13365,-13365,-13365, + -13365,-13365,-13365,-13365,-13365,-13365,-13365,-13365,-13365, + -13365,-13365,-13365,-13365,-13365,-13365,-13365, 7883, -2316, + 9086, -3944, 10500, 4285, 10459, -6474, 10204, -6539, 11601, + -6824, 13385, -7142, 13872, -7457, 13670, -7725, 13463, -6887, + 12482, -5580, 12600, -4964, 12480, 3254, 11741, -4210,-24819, + 23282, 22798, 491, -1774, -1073, 3641, -1304, 28928, -250, + -27396, 6657, -8961, 22524, 19987, 10231, 1791, 8947,-32763, + -26385,-31227, -792,-30461, 8926, 4866, 27863, 27756, 27756, + 27756, 27756, 27756, 27756, 27756, 27756, 5630,-11070,-16136, + 20671,-11530, 27328, 8179, 5059,-31503,-24379,-19472, 17863, + -29202, 22986, -23, 8909, 8422, 10450 + }; + err = opus_multistream_encode(enc, pcm, 1440, data, 7380); + /* reads uninitialized data at src/opus_multistream_encoder.c:293 */ + opus_test_assert(err > 0); + } + opus_multistream_encoder_destroy(enc); + return 0; +} + +static int ec_enc_shrink_assert(void) +{ + OpusEncoder *enc; + int err; + int data_len; + unsigned char data[2000]; + static const short pcm1[960] = { 5140 }; + static const short pcm2[2880] = + { + -256,-12033, 0, -2817, 6912, 0, -5359, 5200, 3061, + 0, -2903, 5652, -1281,-24656,-14433,-24678, 32,-29793, + 2870, 0, 4096, 5120, 5140, -234,-20230,-24673,-24633, + -24673,-24705, 0,-32768,-25444,-25444, 0,-25444,-25444, + 156,-20480, -7948, -5920, -7968, -7968, 224, 0, 20480, + 11, 20496, 13, 20496, 11,-20480, 2292,-20240, 244, + 20480, 11, 20496, 11,-20480, 244,-20240, 7156, 20456, + -246,-20243, 244, 128, 244, 20480, 11, 20496, 11, + -20480, 244,-20256, 244, 20480, 256, 0, -246, 16609, + -176, 0, 29872, -4096, -2888, 516, 2896, 4096, 2896, + -20480, -3852, -2896, -1025,-31056,-14433, 244, 1792, -256, + -12033, 0, -2817, 0, 0, -5359, 5200, 3061, 16, + -2903, 5652, -1281,-24656,-14433,-24678, 32,-29793, 2870, + 0, 4096, 5120, 5140, -234,-20230,-24673,-24633,-24673, + -24705, 0,-32768,-25444,-25444, 0,-25444,-25444, 156, + -20480, -7973, -5920, -7968, -7968, 224, 0, 20480, 11, + 20496, 11, 20496, 11,-20480, 2292,-20213, 244, 20480, + 11, 20496, 11,-24698, -2873, 0, 7, -1, 208, + -256, 244, 0, 4352, 20715, -2796, 11,-22272, 5364, + -234,-20230,-24673,-25913, 8351,-24832, 13963, 11, 0, + 16, 5140, 5652, -1281,-24656,-14433,-24673, 32671, 159, + 0,-25472,-25444, 156,-25600,-25444,-25444, 0, -2896, + -7968, -7960, -7968, -7968, 0, 0, 2896, 4096, 2896, + 4096, 2896, 0, -2896, -4088, -2896, 0, 2896, 0, + -2896, -4096, -2896, 11, 2640, -4609, -2896,-32768, -3072, + 0, 2896, 4096, 2896, 0, -2896, -4096, -2896, 0, + 80, 1, 2816, 0, 20656, 255,-20480, 116,-18192 + }; + static const short pcm3[2880] = { 0 }; + + enc = opus_encoder_create(48000, 1, OPUS_APPLICATION_AUDIO, &err); + opus_encoder_ctl(enc, OPUS_SET_COMPLEXITY(10)); + opus_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(6)); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(6000)); + data_len = opus_encode(enc, pcm1, 960, data, 2000); + opus_test_assert(data_len > 0); + + opus_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE)); + opus_encoder_ctl(enc, OPUS_SET_PREDICTION_DISABLED(1)); + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_SUPERWIDEBAND)); + opus_encoder_ctl(enc, OPUS_SET_INBAND_FEC(1)); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(15600)); + data_len = opus_encode(enc, pcm2, 2880, data, 122); + opus_test_assert(data_len > 0); + + opus_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(27000)); + data_len = opus_encode(enc, pcm3, 2880, data, 122); /* assertion failure */ + opus_test_assert(data_len > 0); + + opus_encoder_destroy(enc); + return 0; +} + +static int ec_enc_shrink_assert2(void) +{ + OpusEncoder *enc; + int err; + int data_len; + unsigned char data[2000]; + + enc = opus_encoder_create(48000, 1, OPUS_APPLICATION_AUDIO, &err); + opus_encoder_ctl(enc, OPUS_SET_COMPLEXITY(6)); + opus_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE)); + opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(OPUS_BANDWIDTH_FULLBAND)); + opus_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC(26)); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(27000)); + { + static const short pcm[960] = { 0 }; + data_len = opus_encode(enc, pcm, 960, data, 2000); + opus_test_assert(data_len > 0); + } + opus_encoder_ctl(enc, OPUS_SET_SIGNAL(OPUS_SIGNAL_MUSIC)); + { + static const short pcm[480] = + { + 32767, 32767, 0, 0, 32767, 32767, 0, 0, 32767, 32767, + -32768, -32768, 0, 0, -32768, -32768, 0, 0, -32768, -32768 + }; + data_len = opus_encode(enc, pcm, 480, data, 19); + opus_test_assert(data_len > 0); + } + opus_encoder_destroy(enc); + return 0; +} + +static int silk_gain_assert(void) +{ + OpusEncoder *enc; + int err; + int data_len; + unsigned char data[1000]; + static const short pcm1[160] = { 0 }; + static const short pcm2[960] = + { + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 32767, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 32767 + }; + + enc = opus_encoder_create(8000, 1, OPUS_APPLICATION_AUDIO, &err); + opus_encoder_ctl(enc, OPUS_SET_COMPLEXITY(3)); + opus_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_NARROWBAND)); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(6000)); + data_len = opus_encode(enc, pcm1, 160, data, 1000); + opus_test_assert(data_len > 0); + + opus_encoder_ctl(enc, OPUS_SET_VBR(0)); + opus_encoder_ctl(enc, OPUS_SET_COMPLEXITY(0)); + opus_encoder_ctl(enc, OPUS_SET_MAX_BANDWIDTH(OPUS_BANDWIDTH_MEDIUMBAND)); + opus_encoder_ctl(enc, OPUS_SET_BITRATE(2867)); + data_len = opus_encode(enc, pcm2, 960, data, 1000); + opus_test_assert(data_len > 0); + + opus_encoder_destroy(enc); + return 0; +} + +void regression_test(void) +{ + fprintf(stderr, "Running simple tests for bugs that have been fixed previously\n"); + celt_ec_internal_error(); + mscbr_encode_fail10(); + mscbr_encode_fail(); + surround_analysis_uninit(); + ec_enc_shrink_assert(); + ec_enc_shrink_assert2(); + silk_gain_assert(); +} diff --git a/lib/opus/tests/random_config.sh b/lib/opus/tests/random_config.sh new file mode 100755 index 00000000..0cdd855f --- /dev/null +++ b/lib/opus/tests/random_config.sh @@ -0,0 +1,126 @@ +#!/bin/bash + +dir="$1" +mkdir "$dir" +if [ $? -ne 0 ] +then + exit 1 +fi + +cd "$dir" +if [ $? -ne 0 ] +then + exit 1 +fi + + +configure_path="$2" +config="random_config.txt" + +case `seq 3 | shuf -n1` in +1) +approx=--enable-float-approx +math=-ffast-math +;; +2) +approx=--enable-float-approx +;; +*) +approx= +math= +;; +esac + +CFLAGS='-g' + +opt=`echo -e "-O1\n-O2\n-O3" | shuf -n1` + +#arch=-march=`echo -e "core2\nsandybridge\nbroadwell\nskylake" | shuf -n1` +arch=`echo -e "\n-march=core2\n-march=sandybridge\n-march=broadwell\n-march=skylake\n-march=native" | shuf -n1` + +footprint=`echo -e "\n-DSMALL_FOOTPRINT" | shuf -n1` +std=`echo -e "\n-std=c90\n-std=c99\n-std=c11\n-std=c17" | shuf -n1` +sanitize=`echo -e "\n-fsanitize=address -fno-sanitize-recover=all\n-fsanitize=undefined -fno-sanitize-recover=all -fsanitize-recover=signed-integer-overflow" | shuf -n1` + + +CFLAGS="$CFLAGS $std $opt $arch $footprint $math $sanitize" + +echo "CFLAGS=$CFLAGS" > "$config" + +lib=`echo -e "\n--disable-static\n--disable-shared" | shuf -n1` + +arithmetic=`echo -e "\n--enable-fixed-point\n--enable-fixed-point --enable-fixed-point-debug\n--enable-fixed-point --disable-float-api\n--enable-fixed-point --enable-fixed-point-debug --disable-float-api" | shuf -n1` + +custom=`echo -e "\n--enable-custom-modes" | shuf -n1` + +asm=`echo -e "\n--disable-asm\n--disable-rtcd\n--disable-intrinsics" | shuf -n1` +#asm=`echo -e "\n--disable-asm\n--disable-intrinsics" | shuf -n1` + +assert=`echo -e "\n--enable-assertions" | shuf -n1` +harden=`echo -e "\n--enable-hardening" | shuf -n1` +fuzz=`echo -e "\n--enable-fuzzing" | shuf -n1` +checkasm=`echo -e "\n--enable-check-asm" | shuf -n1` +rfc8251=`echo -e "\n--disable-rfc8251" | shuf -n1` + +if [ "$rfc8251" = --disable-rfc8251 ] +then + vectors="$3" +else + vectors="$4" +fi +echo using testvectors at "$vectors" >> "$config" + + +config_opt="$lib $arithmetic $custom $asm $assert $harden $fuzz $checkasm $rfc8251 $approx" + +echo configure $config_opt >> "$config" + +export CFLAGS +"$configure_path/configure" $config_opt > configure_output.txt 2>&1 + +if [ $? -ne 0 ] +then + echo configure FAIL >> "$config" + exit 1 +fi + +make > make_output.txt 2>&1 + +if [ $? -ne 0 ] +then + echo make FAIL >> "$config" + exit 1 +fi + +#Run valgrind 5% of the time (minus the asan cases) +if [ "`seq 20 | shuf -n1`" -ne 1 -o "$sanitize" = "-fsanitize=address -fno-sanitize-recover=all" ] +then + make check > makecheck_output.txt 2>&1 +else + echo valgrind enabled >> "$config" + valgrind --trace-children=yes --error-exitcode=128 make check > makecheck_output.txt 2>&1 +fi + +if [ $? -ne 0 ] +then + echo check FAIL >> "$config" + exit 1 +fi + + +rate=`echo -e "8000\n12000\n16000\n24000\n48000" | shuf -n1` +echo testvectors for "$rate" Hz > testvectors_output.txt +../../../run_vectors.sh . "$vectors" "$rate" >> testvectors_output.txt 2>&1 + +if [ $? -ne 0 ] +then + echo testvectors FAIL >> "$config" + exit 1 +fi + +echo all tests PASS >> "$config" + +#When everything's good, do some cleaning up to save space +make distclean > /dev/null 2>&1 +rm -f tmp.out +gzip make_output.txt diff --git a/lib/opus/tests/run_vectors.sh b/lib/opus/tests/run_vectors.sh new file mode 100755 index 00000000..dcb76cf1 --- /dev/null +++ b/lib/opus/tests/run_vectors.sh @@ -0,0 +1,143 @@ +#!/bin/sh + +# Copyright (c) 2011-2012 Jean-Marc Valin +# +# This file is extracted from RFC6716. Please see that RFC for additional +# information. +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions +# are met: +# +# - Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# - Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in the +# documentation and/or other materials provided with the distribution. +# +# - Neither the name of Internet Society, IETF or IETF Trust, nor the +# names of specific contributors, may be used to endorse or promote +# products derived from this software without specific prior written +# permission. +# +# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +# ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER +# OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + +rm -f logs_mono.txt logs_mono2.txt +rm -f logs_stereo.txt logs_stereo2.txt + +if [ "$#" -ne "3" ]; then + echo "usage: run_vectors.sh " + exit 1 +fi + +CMD_PATH=$1 +VECTOR_PATH=$2 +RATE=$3 + +: ${OPUS_DEMO:=$CMD_PATH/opus_demo} +: ${OPUS_COMPARE:=$CMD_PATH/opus_compare} + +if [ -d "$VECTOR_PATH" ]; then + echo "Test vectors found in $VECTOR_PATH" +else + echo "No test vectors found" + #Don't make the test fail here because the test vectors + #will be distributed separately + exit 0 +fi + +if [ ! -x "$OPUS_COMPARE" ]; then + echo "ERROR: Compare program not found: $OPUS_COMPARE" + exit 1 +fi + +if [ -x "$OPUS_DEMO" ]; then + echo "Decoding with $OPUS_DEMO" +else + echo "ERROR: Decoder not found: $OPUS_DEMO" + exit 1 +fi + +echo "==============" +echo "Testing mono" +echo "==============" +echo + +for file in 01 02 03 04 05 06 07 08 09 10 11 12 +do + if [ -e "$VECTOR_PATH/testvector$file.bit" ]; then + echo "Testing testvector$file" + else + echo "Bitstream file not found: testvector$file.bit" + fi + if "$OPUS_DEMO" -d "$RATE" 1 "$VECTOR_PATH/testvector$file.bit" tmp.out >> logs_mono.txt 2>&1; then + echo "successfully decoded" + else + echo "ERROR: decoding failed" + exit 1 + fi + "$OPUS_COMPARE" -r "$RATE" "$VECTOR_PATH/testvector${file}.dec" tmp.out >> logs_mono.txt 2>&1 + float_ret=$? + "$OPUS_COMPARE" -r "$RATE" "$VECTOR_PATH/testvector${file}m.dec" tmp.out >> logs_mono2.txt 2>&1 + float_ret2=$? + if [ "$float_ret" -eq "0" ] || [ "$float_ret2" -eq "0" ]; then + echo "output matches reference" + else + echo "ERROR: output does not match reference" + exit 1 + fi + echo +done + +echo "==============" +echo Testing stereo +echo "==============" +echo + +for file in 01 02 03 04 05 06 07 08 09 10 11 12 +do + if [ -e "$VECTOR_PATH/testvector$file.bit" ]; then + echo "Testing testvector$file" + else + echo "Bitstream file not found: testvector$file" + fi + if "$OPUS_DEMO" -d "$RATE" 2 "$VECTOR_PATH/testvector$file.bit" tmp.out >> logs_stereo.txt 2>&1; then + echo "successfully decoded" + else + echo "ERROR: decoding failed" + exit 1 + fi + "$OPUS_COMPARE" -s -r "$RATE" "$VECTOR_PATH/testvector${file}.dec" tmp.out >> logs_stereo.txt 2>&1 + float_ret=$? + "$OPUS_COMPARE" -s -r "$RATE" "$VECTOR_PATH/testvector${file}m.dec" tmp.out >> logs_stereo2.txt 2>&1 + float_ret2=$? + if [ "$float_ret" -eq "0" ] || [ "$float_ret2" -eq "0" ]; then + echo "output matches reference" + else + echo "ERROR: output does not match reference" + exit 1 + fi + echo +done + + + +echo "All tests have passed successfully" +mono1=`grep quality logs_mono.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` +mono2=`grep quality logs_mono2.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` +echo $mono1 $mono2 | awk '{if ($2 > $1) $1 = $2; print "Average mono quality is", $1, "%"}' + +stereo1=`grep quality logs_stereo.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` +stereo2=`grep quality logs_stereo2.txt | awk '{sum+=$4}END{if (NR == 12) sum /= 12; else sum = 0; print sum}'` +echo $stereo1 $stereo2 | awk '{if ($2 > $1) $1 = $2; print "Average stereo quality is", $1, "%"}' diff --git a/lib/opus/tests/test_opus_api.c b/lib/opus/tests/test_opus_api.c new file mode 100644 index 00000000..0e7ed2cc --- /dev/null +++ b/lib/opus/tests/test_opus_api.c @@ -0,0 +1,1904 @@ +/* Copyright (c) 2011-2013 Xiph.Org Foundation + Written by Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* This tests the API presented by the libopus system. + It does not attempt to extensively exercise the codec internals. + The strategy here is to simply the API interface invariants: + That sane options are accepted, insane options are rejected, + and that nothing blows up. In particular we don't actually test + that settings are heeded by the codec (though we do check that + get after set returns a sane value when it should). Other + tests check the actual codec behavior. + In cases where its reasonable to do so we test exhaustively, + but its not reasonable to do so in all cases. + Although these tests are simple they found several library bugs + when they were initially developed. */ + +/* These tests are more sensitive if compiled with -DVALGRIND and + run inside valgrind. Malloc failure testing requires glibc. */ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include +#include "arch.h" +#include "opus_multistream.h" +#include "opus.h" +#include "test_opus_common.h" + +#ifdef VALGRIND +#include +#define VG_UNDEF(x,y) VALGRIND_MAKE_MEM_UNDEFINED((x),(y)) +#define VG_CHECK(x,y) VALGRIND_CHECK_MEM_IS_DEFINED((x),(y)) +#else +#define VG_UNDEF(x,y) +#define VG_CHECK(x,y) +#endif + +#if defined(HAVE___MALLOC_HOOK) +#define MALLOC_FAIL +#include "os_support.h" +#include + +static const opus_int32 opus_apps[3] = {OPUS_APPLICATION_VOIP, + OPUS_APPLICATION_AUDIO,OPUS_APPLICATION_RESTRICTED_LOWDELAY}; + +void *malloc_hook(__attribute__((unused)) size_t size, + __attribute__((unused)) const void *caller) +{ + return 0; +} +#endif + +opus_int32 *null_int_ptr = (opus_int32 *)NULL; +opus_uint32 *null_uint_ptr = (opus_uint32 *)NULL; + +static const opus_int32 opus_rates[5] = {48000,24000,16000,12000,8000}; + +opus_int32 test_dec_api(void) +{ + opus_uint32 dec_final_range; + OpusDecoder *dec; + OpusDecoder *dec2; + opus_int32 i,j,cfgs; + unsigned char packet[1276]; +#ifndef DISABLE_FLOAT_API + float fbuf[960*2]; +#endif + short sbuf[960*2]; + int c,err; + + cfgs=0; + /*First test invalid configurations which should fail*/ + fprintf(stdout,"\n Decoder basic API tests\n"); + fprintf(stdout," ---------------------------------------------------\n"); + for(c=0;c<4;c++) + { + i=opus_decoder_get_size(c); + if(((c==1||c==2)&&(i<=2048||i>1<<16))||((c!=1&&c!=2)&&i!=0))test_failed(); + fprintf(stdout," opus_decoder_get_size(%d)=%d ...............%s OK.\n",c,i,i>0?"":"...."); + cfgs++; + } + + /*Test with unsupported sample rates*/ + for(c=0;c<4;c++) + { + for(i=-7;i<=96000;i++) + { + int fs; + if((i==8000||i==12000||i==16000||i==24000||i==48000)&&(c==1||c==2))continue; + switch(i) + { + case(-5):fs=-8000;break; + case(-6):fs=INT32_MAX;break; + case(-7):fs=INT32_MIN;break; + default:fs=i; + } + err = OPUS_OK; + VG_UNDEF(&err,sizeof(err)); + dec = opus_decoder_create(fs, c, &err); + if(err!=OPUS_BAD_ARG || dec!=NULL)test_failed(); + cfgs++; + dec = opus_decoder_create(fs, c, 0); + if(dec!=NULL)test_failed(); + cfgs++; + dec=malloc(opus_decoder_get_size(2)); + if(dec==NULL)test_failed(); + err = opus_decoder_init(dec,fs,c); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + free(dec); + } + } + + VG_UNDEF(&err,sizeof(err)); + dec = opus_decoder_create(48000, 2, &err); + if(err!=OPUS_OK || dec==NULL)test_failed(); + VG_CHECK(dec,opus_decoder_get_size(2)); + cfgs++; + + fprintf(stdout," opus_decoder_create() ........................ OK.\n"); + fprintf(stdout," opus_decoder_init() .......................... OK.\n"); + + err=opus_decoder_ctl(dec, OPUS_GET_FINAL_RANGE(null_uint_ptr)); + if(err != OPUS_BAD_ARG)test_failed(); + VG_UNDEF(&dec_final_range,sizeof(dec_final_range)); + err=opus_decoder_ctl(dec, OPUS_GET_FINAL_RANGE(&dec_final_range)); + if(err!=OPUS_OK)test_failed(); + VG_CHECK(&dec_final_range,sizeof(dec_final_range)); + fprintf(stdout," OPUS_GET_FINAL_RANGE ......................... OK.\n"); + cfgs++; + + err=opus_decoder_ctl(dec,OPUS_UNIMPLEMENTED); + if(err!=OPUS_UNIMPLEMENTED)test_failed(); + fprintf(stdout," OPUS_UNIMPLEMENTED ........................... OK.\n"); + cfgs++; + + err=opus_decoder_ctl(dec, OPUS_GET_BANDWIDTH(null_int_ptr)); + if(err != OPUS_BAD_ARG)test_failed(); + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_BANDWIDTH(&i)); + if(err != OPUS_OK || i!=0)test_failed(); + fprintf(stdout," OPUS_GET_BANDWIDTH ........................... OK.\n"); + cfgs++; + + err=opus_decoder_ctl(dec, OPUS_GET_SAMPLE_RATE(null_int_ptr)); + if(err != OPUS_BAD_ARG)test_failed(); + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_SAMPLE_RATE(&i)); + if(err != OPUS_OK || i!=48000)test_failed(); + fprintf(stdout," OPUS_GET_SAMPLE_RATE ......................... OK.\n"); + cfgs++; + + /*GET_PITCH has different execution paths depending on the previously decoded frame.*/ + err=opus_decoder_ctl(dec, OPUS_GET_PITCH(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_PITCH(&i)); + if(err != OPUS_OK || i>0 || i<-1)test_failed(); + cfgs++; + VG_UNDEF(packet,sizeof(packet)); + packet[0]=63<<2;packet[1]=packet[2]=0; + if(opus_decode(dec, packet, 3, sbuf, 960, 0)!=960)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_PITCH(&i)); + if(err != OPUS_OK || i>0 || i<-1)test_failed(); + cfgs++; + packet[0]=1; + if(opus_decode(dec, packet, 1, sbuf, 960, 0)!=960)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_PITCH(&i)); + if(err != OPUS_OK || i>0 || i<-1)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_PITCH ............................... OK.\n"); + + err=opus_decoder_ctl(dec, OPUS_GET_LAST_PACKET_DURATION(null_int_ptr)); + if(err != OPUS_BAD_ARG)test_failed(); + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_LAST_PACKET_DURATION(&i)); + if(err != OPUS_OK || i!=960)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_LAST_PACKET_DURATION ................ OK.\n"); + + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_GAIN(&i)); + VG_CHECK(&i,sizeof(i)); + if(err != OPUS_OK || i!=0)test_failed(); + cfgs++; + err=opus_decoder_ctl(dec, OPUS_GET_GAIN(null_int_ptr)); + if(err != OPUS_BAD_ARG)test_failed(); + cfgs++; + err=opus_decoder_ctl(dec, OPUS_SET_GAIN(-32769)); + if(err != OPUS_BAD_ARG)test_failed(); + cfgs++; + err=opus_decoder_ctl(dec, OPUS_SET_GAIN(32768)); + if(err != OPUS_BAD_ARG)test_failed(); + cfgs++; + err=opus_decoder_ctl(dec, OPUS_SET_GAIN(-15)); + if(err != OPUS_OK)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(dec, OPUS_GET_GAIN(&i)); + VG_CHECK(&i,sizeof(i)); + if(err != OPUS_OK || i!=-15)test_failed(); + cfgs++; + fprintf(stdout," OPUS_SET_GAIN ................................ OK.\n"); + fprintf(stdout," OPUS_GET_GAIN ................................ OK.\n"); + + /*Reset the decoder*/ + dec2=malloc(opus_decoder_get_size(2)); + memcpy(dec2,dec,opus_decoder_get_size(2)); + if(opus_decoder_ctl(dec, OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + if(memcmp(dec2,dec,opus_decoder_get_size(2))==0)test_failed(); + free(dec2); + fprintf(stdout," OPUS_RESET_STATE ............................. OK.\n"); + cfgs++; + + VG_UNDEF(packet,sizeof(packet)); + packet[0]=0; + if(opus_decoder_get_nb_samples(dec,packet,1)!=480)test_failed(); + if(opus_packet_get_nb_samples(packet,1,48000)!=480)test_failed(); + if(opus_packet_get_nb_samples(packet,1,96000)!=960)test_failed(); + if(opus_packet_get_nb_samples(packet,1,32000)!=320)test_failed(); + if(opus_packet_get_nb_samples(packet,1,8000)!=80)test_failed(); + packet[0]=3; + if(opus_packet_get_nb_samples(packet,1,24000)!=OPUS_INVALID_PACKET)test_failed(); + packet[0]=(63<<2)|3; + packet[1]=63; + if(opus_packet_get_nb_samples(packet,0,24000)!=OPUS_BAD_ARG)test_failed(); + if(opus_packet_get_nb_samples(packet,2,48000)!=OPUS_INVALID_PACKET)test_failed(); + if(opus_decoder_get_nb_samples(dec,packet,2)!=OPUS_INVALID_PACKET)test_failed(); + fprintf(stdout," opus_{packet,decoder}_get_nb_samples() ....... OK.\n"); + cfgs+=9; + + if(OPUS_BAD_ARG!=opus_packet_get_nb_frames(packet,0))test_failed(); + for(i=0;i<256;i++) { + int l1res[4]={1,2,2,OPUS_INVALID_PACKET}; + packet[0]=i; + if(l1res[packet[0]&3]!=opus_packet_get_nb_frames(packet,1))test_failed(); + cfgs++; + for(j=0;j<256;j++) { + packet[1]=j; + if(((packet[0]&3)!=3?l1res[packet[0]&3]:packet[1]&63)!=opus_packet_get_nb_frames(packet,2))test_failed(); + cfgs++; + } + } + fprintf(stdout," opus_packet_get_nb_frames() .................. OK.\n"); + + for(i=0;i<256;i++) { + int bw; + packet[0]=i; + bw=packet[0]>>4; + bw=OPUS_BANDWIDTH_NARROWBAND+(((((bw&7)*9)&(63-(bw&8)))+2+12*((bw&8)!=0))>>4); + if(bw!=opus_packet_get_bandwidth(packet))test_failed(); + cfgs++; + } + fprintf(stdout," opus_packet_get_bandwidth() .................. OK.\n"); + + for(i=0;i<256;i++) { + int fp3s,rate; + packet[0]=i; + fp3s=packet[0]>>3; + fp3s=((((3-(fp3s&3))*13&119)+9)>>2)*((fp3s>13)*(3-((fp3s&3)==3))+1)*25; + for(rate=0;rate<5;rate++) { + if((opus_rates[rate]*3/fp3s)!=opus_packet_get_samples_per_frame(packet,opus_rates[rate]))test_failed(); + cfgs++; + } + } + fprintf(stdout," opus_packet_get_samples_per_frame() .......... OK.\n"); + + packet[0]=(63<<2)+3; + packet[1]=49; + for(j=2;j<51;j++)packet[j]=0; + VG_UNDEF(sbuf,sizeof(sbuf)); + if(opus_decode(dec, packet, 51, sbuf, 960, 0)!=OPUS_INVALID_PACKET)test_failed(); + cfgs++; + packet[0]=(63<<2); + packet[1]=packet[2]=0; + if(opus_decode(dec, packet, -1, sbuf, 960, 0)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_decode(dec, packet, 3, sbuf, 60, 0)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + if(opus_decode(dec, packet, 3, sbuf, 480, 0)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + if(opus_decode(dec, packet, 3, sbuf, 960, 0)!=960)test_failed(); + cfgs++; + fprintf(stdout," opus_decode() ................................ OK.\n"); +#ifndef DISABLE_FLOAT_API + VG_UNDEF(fbuf,sizeof(fbuf)); + if(opus_decode_float(dec, packet, 3, fbuf, 960, 0)!=960)test_failed(); + cfgs++; + fprintf(stdout," opus_decode_float() .......................... OK.\n"); +#endif + +#if 0 + /*These tests are disabled because the library crashes with null states*/ + if(opus_decoder_ctl(0,OPUS_RESET_STATE) !=OPUS_INVALID_STATE)test_failed(); + if(opus_decoder_init(0,48000,1) !=OPUS_INVALID_STATE)test_failed(); + if(opus_decode(0,packet,1,outbuf,2880,0) !=OPUS_INVALID_STATE)test_failed(); + if(opus_decode_float(0,packet,1,0,2880,0) !=OPUS_INVALID_STATE)test_failed(); + if(opus_decoder_get_nb_samples(0,packet,1) !=OPUS_INVALID_STATE)test_failed(); + if(opus_packet_get_nb_frames(NULL,1) !=OPUS_BAD_ARG)test_failed(); + if(opus_packet_get_bandwidth(NULL) !=OPUS_BAD_ARG)test_failed(); + if(opus_packet_get_samples_per_frame(NULL,48000)!=OPUS_BAD_ARG)test_failed(); +#endif + opus_decoder_destroy(dec); + cfgs++; + fprintf(stdout," All decoder interface tests passed\n"); + fprintf(stdout," (%6d API invocations)\n",cfgs); + return cfgs; +} + +opus_int32 test_msdec_api(void) +{ + opus_uint32 dec_final_range; + OpusMSDecoder *dec; + OpusDecoder *streamdec; + opus_int32 i,j,cfgs; + unsigned char packet[1276]; + unsigned char mapping[256]; +#ifndef DISABLE_FLOAT_API + float fbuf[960*2]; +#endif + short sbuf[960*2]; + int a,b,c,err; + + mapping[0]=0; + mapping[1]=1; + for(i=2;i<256;i++)VG_UNDEF(&mapping[i],sizeof(unsigned char)); + + cfgs=0; + /*First test invalid configurations which should fail*/ + fprintf(stdout,"\n Multistream decoder basic API tests\n"); + fprintf(stdout," ---------------------------------------------------\n"); + for(a=-1;a<4;a++) + { + for(b=-1;b<4;b++) + { + i=opus_multistream_decoder_get_size(a,b); + if(((a>0&&b<=a&&b>=0)&&(i<=2048||i>((1<<16)*a)))||((a<1||b>a||b<0)&&i!=0))test_failed(); + fprintf(stdout," opus_multistream_decoder_get_size(%2d,%2d)=%d %sOK.\n",a,b,i,i>0?"":"... "); + cfgs++; + } + } + + /*Test with unsupported sample rates*/ + for(c=1;c<3;c++) + { + for(i=-7;i<=96000;i++) + { + int fs; + if((i==8000||i==12000||i==16000||i==24000||i==48000)&&(c==1||c==2))continue; + switch(i) + { + case(-5):fs=-8000;break; + case(-6):fs=INT32_MAX;break; + case(-7):fs=INT32_MIN;break; + default:fs=i; + } + err = OPUS_OK; + VG_UNDEF(&err,sizeof(err)); + dec = opus_multistream_decoder_create(fs, c, 1, c-1, mapping, &err); + if(err!=OPUS_BAD_ARG || dec!=NULL)test_failed(); + cfgs++; + dec = opus_multistream_decoder_create(fs, c, 1, c-1, mapping, 0); + if(dec!=NULL)test_failed(); + cfgs++; + dec=malloc(opus_multistream_decoder_get_size(1,1)); + if(dec==NULL)test_failed(); + err = opus_multistream_decoder_init(dec,fs,c,1,c-1, mapping); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + free(dec); + } + } + + for(c=0;c<2;c++) + { + int *ret_err; + ret_err = c?0:&err; + + mapping[0]=0; + mapping[1]=1; + for(i=2;i<256;i++)VG_UNDEF(&mapping[i],sizeof(unsigned char)); + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 2, 1, 0, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + mapping[0]=mapping[1]=0; + dec = opus_multistream_decoder_create(48000, 2, 1, 0, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_OK) || dec==NULL)test_failed(); + cfgs++; + opus_multistream_decoder_destroy(dec); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 1, 4, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_OK) || dec==NULL)test_failed(); + cfgs++; + + err = opus_multistream_decoder_init(dec,48000, 1, 0, 0, mapping); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + + err = opus_multistream_decoder_init(dec,48000, 1, 1, -1, mapping); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + + opus_multistream_decoder_destroy(dec); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 2, 1, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_OK) || dec==NULL)test_failed(); + cfgs++; + opus_multistream_decoder_destroy(dec); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 255, 255, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, -1, 1, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 0, 1, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 1, -1, 2, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 1, -1, -1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 256, 255, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + dec = opus_multistream_decoder_create(48000, 256, 255, 0, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + mapping[0]=255; + mapping[1]=1; + mapping[2]=2; + dec = opus_multistream_decoder_create(48000, 3, 2, 0, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + mapping[0]=0; + mapping[1]=0; + mapping[2]=0; + dec = opus_multistream_decoder_create(48000, 3, 2, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_OK) || dec==NULL)test_failed(); + cfgs++; + opus_multistream_decoder_destroy(dec); + cfgs++; + + VG_UNDEF(ret_err,sizeof(*ret_err)); + mapping[0]=0; + mapping[1]=255; + mapping[2]=1; + mapping[3]=2; + mapping[4]=3; + dec = opus_multistream_decoder_create(48001, 5, 4, 1, mapping, ret_err); + if(ret_err){VG_CHECK(ret_err,sizeof(*ret_err));} + if((ret_err && *ret_err!=OPUS_BAD_ARG) || dec!=NULL)test_failed(); + cfgs++; + } + + VG_UNDEF(&err,sizeof(err)); + mapping[0]=0; + mapping[1]=255; + mapping[2]=1; + mapping[3]=2; + dec = opus_multistream_decoder_create(48000, 4, 2, 1, mapping, &err); + VG_CHECK(&err,sizeof(err)); + if(err!=OPUS_OK || dec==NULL)test_failed(); + cfgs++; + + fprintf(stdout," opus_multistream_decoder_create() ............ OK.\n"); + fprintf(stdout," opus_multistream_decoder_init() .............. OK.\n"); + + VG_UNDEF(&dec_final_range,sizeof(dec_final_range)); + err=opus_multistream_decoder_ctl(dec, OPUS_GET_FINAL_RANGE(&dec_final_range)); + if(err!=OPUS_OK)test_failed(); + VG_CHECK(&dec_final_range,sizeof(dec_final_range)); + fprintf(stdout," OPUS_GET_FINAL_RANGE ......................... OK.\n"); + cfgs++; + + streamdec=0; + VG_UNDEF(&streamdec,sizeof(streamdec)); + err=opus_multistream_decoder_ctl(dec, OPUS_MULTISTREAM_GET_DECODER_STATE(-1,&streamdec)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + err=opus_multistream_decoder_ctl(dec, OPUS_MULTISTREAM_GET_DECODER_STATE(1,&streamdec)); + if(err!=OPUS_OK||streamdec==NULL)test_failed(); + VG_CHECK(streamdec,opus_decoder_get_size(1)); + cfgs++; + err=opus_multistream_decoder_ctl(dec, OPUS_MULTISTREAM_GET_DECODER_STATE(2,&streamdec)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + err=opus_multistream_decoder_ctl(dec, OPUS_MULTISTREAM_GET_DECODER_STATE(0,&streamdec)); + if(err!=OPUS_OK||streamdec==NULL)test_failed(); + VG_CHECK(streamdec,opus_decoder_get_size(1)); + fprintf(stdout," OPUS_MULTISTREAM_GET_DECODER_STATE ........... OK.\n"); + cfgs++; + + for(j=0;j<2;j++) + { + OpusDecoder *od; + err=opus_multistream_decoder_ctl(dec,OPUS_MULTISTREAM_GET_DECODER_STATE(j,&od)); + if(err != OPUS_OK)test_failed(); + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(od, OPUS_GET_GAIN(&i)); + VG_CHECK(&i,sizeof(i)); + if(err != OPUS_OK || i!=0)test_failed(); + cfgs++; + } + err=opus_multistream_decoder_ctl(dec,OPUS_SET_GAIN(15)); + if(err!=OPUS_OK)test_failed(); + fprintf(stdout," OPUS_SET_GAIN ................................ OK.\n"); + for(j=0;j<2;j++) + { + OpusDecoder *od; + err=opus_multistream_decoder_ctl(dec,OPUS_MULTISTREAM_GET_DECODER_STATE(j,&od)); + if(err != OPUS_OK)test_failed(); + VG_UNDEF(&i,sizeof(i)); + err=opus_decoder_ctl(od, OPUS_GET_GAIN(&i)); + VG_CHECK(&i,sizeof(i)); + if(err != OPUS_OK || i!=15)test_failed(); + cfgs++; + } + fprintf(stdout," OPUS_GET_GAIN ................................ OK.\n"); + + VG_UNDEF(&i,sizeof(i)); + err=opus_multistream_decoder_ctl(dec, OPUS_GET_BANDWIDTH(&i)); + if(err != OPUS_OK || i!=0)test_failed(); + fprintf(stdout," OPUS_GET_BANDWIDTH ........................... OK.\n"); + cfgs++; + + err=opus_multistream_decoder_ctl(dec,OPUS_UNIMPLEMENTED); + if(err!=OPUS_UNIMPLEMENTED)test_failed(); + fprintf(stdout," OPUS_UNIMPLEMENTED ........................... OK.\n"); + cfgs++; + +#if 0 + /*Currently unimplemented for multistream*/ + /*GET_PITCH has different execution paths depending on the previously decoded frame.*/ + err=opus_multistream_decoder_ctl(dec, OPUS_GET_PITCH(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + err=opus_multistream_decoder_ctl(dec, OPUS_GET_PITCH(&i)); + if(err != OPUS_OK || i>0 || i<-1)test_failed(); + cfgs++; + VG_UNDEF(packet,sizeof(packet)); + packet[0]=63<<2;packet[1]=packet[2]=0; + if(opus_multistream_decode(dec, packet, 3, sbuf, 960, 0)!=960)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + err=opus_multistream_decoder_ctl(dec, OPUS_GET_PITCH(&i)); + if(err != OPUS_OK || i>0 || i<-1)test_failed(); + cfgs++; + packet[0]=1; + if(opus_multistream_decode(dec, packet, 1, sbuf, 960, 0)!=960)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + err=opus_multistream_decoder_ctl(dec, OPUS_GET_PITCH(&i)); + if(err != OPUS_OK || i>0 || i<-1)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_PITCH ............................... OK.\n"); +#endif + + /*Reset the decoder*/ + if(opus_multistream_decoder_ctl(dec, OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + fprintf(stdout," OPUS_RESET_STATE ............................. OK.\n"); + cfgs++; + + opus_multistream_decoder_destroy(dec); + cfgs++; + VG_UNDEF(&err,sizeof(err)); + dec = opus_multistream_decoder_create(48000, 2, 1, 1, mapping, &err); + if(err!=OPUS_OK || dec==NULL)test_failed(); + cfgs++; + + packet[0]=(63<<2)+3; + packet[1]=49; + for(j=2;j<51;j++)packet[j]=0; + VG_UNDEF(sbuf,sizeof(sbuf)); + if(opus_multistream_decode(dec, packet, 51, sbuf, 960, 0)!=OPUS_INVALID_PACKET)test_failed(); + cfgs++; + packet[0]=(63<<2); + packet[1]=packet[2]=0; + if(opus_multistream_decode(dec, packet, -1, sbuf, 960, 0)!=OPUS_BAD_ARG){printf("%d\n",opus_multistream_decode(dec, packet, -1, sbuf, 960, 0));test_failed();} + cfgs++; + if(opus_multistream_decode(dec, packet, 3, sbuf, -960, 0)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_multistream_decode(dec, packet, 3, sbuf, 60, 0)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + if(opus_multistream_decode(dec, packet, 3, sbuf, 480, 0)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + if(opus_multistream_decode(dec, packet, 3, sbuf, 960, 0)!=960)test_failed(); + cfgs++; + fprintf(stdout," opus_multistream_decode() .................... OK.\n"); +#ifndef DISABLE_FLOAT_API + VG_UNDEF(fbuf,sizeof(fbuf)); + if(opus_multistream_decode_float(dec, packet, 3, fbuf, 960, 0)!=960)test_failed(); + cfgs++; + fprintf(stdout," opus_multistream_decode_float() .............. OK.\n"); +#endif + +#if 0 + /*These tests are disabled because the library crashes with null states*/ + if(opus_multistream_decoder_ctl(0,OPUS_RESET_STATE) !=OPUS_INVALID_STATE)test_failed(); + if(opus_multistream_decoder_init(0,48000,1) !=OPUS_INVALID_STATE)test_failed(); + if(opus_multistream_decode(0,packet,1,outbuf,2880,0) !=OPUS_INVALID_STATE)test_failed(); + if(opus_multistream_decode_float(0,packet,1,0,2880,0) !=OPUS_INVALID_STATE)test_failed(); + if(opus_multistream_decoder_get_nb_samples(0,packet,1) !=OPUS_INVALID_STATE)test_failed(); +#endif + opus_multistream_decoder_destroy(dec); + cfgs++; + fprintf(stdout," All multistream decoder interface tests passed\n"); + fprintf(stdout," (%6d API invocations)\n",cfgs); + return cfgs; +} + +#ifdef VALGRIND +#define UNDEFINE_FOR_PARSE toc=-1; \ + frames[0]=(unsigned char *)0; \ + frames[1]=(unsigned char *)0; \ + payload_offset=-1; \ + VG_UNDEF(&toc,sizeof(toc)); \ + VG_UNDEF(frames,sizeof(frames));\ + VG_UNDEF(&payload_offset,sizeof(payload_offset)); +#else +#define UNDEFINE_FOR_PARSE toc=-1; \ + frames[0]=(unsigned char *)0; \ + frames[1]=(unsigned char *)0; \ + payload_offset=-1; +#endif + +/* This test exercises the heck out of the libopus parser. + It is much larger than the parser itself in part because + it tries to hit a lot of corner cases that could never + fail with the libopus code, but might be problematic for + other implementations. */ +opus_int32 test_parse(void) +{ + opus_int32 i,j,jj,sz; + unsigned char packet[1276]; + opus_int32 cfgs,cfgs_total; + unsigned char toc; + const unsigned char *frames[48]; + short size[48]; + int payload_offset, ret; + fprintf(stdout,"\n Packet header parsing tests\n"); + fprintf(stdout," ---------------------------------------------------\n"); + memset(packet,0,sizeof(char)*1276); + packet[0]=63<<2; + if(opus_packet_parse(packet,1,&toc,frames,0,&payload_offset)!=OPUS_BAD_ARG)test_failed(); + cfgs_total=cfgs=1; + /*code 0*/ + for(i=0;i<64;i++) + { + packet[0]=i<<2; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,4,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=1)test_failed(); + if(size[0]!=3)test_failed(); + if(frames[0]!=packet+1)test_failed(); + } + fprintf(stdout," code 0 (%2d cases) ............................ OK.\n",cfgs); + cfgs_total+=cfgs;cfgs=0; + + /*code 1, two frames of the same size*/ + for(i=0;i<64;i++) + { + packet[0]=(i<<2)+1; + for(jj=0;jj<=1275*2+3;jj++) + { + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,jj,&toc,frames,size,&payload_offset); + cfgs++; + if((jj&1)==1 && jj<=2551) + { + /* Must pass if payload length even (packet length odd) and + size<=2551, must fail otherwise. */ + if(ret!=2)test_failed(); + if(size[0]!=size[1] || size[0]!=((jj-1)>>1))test_failed(); + if(frames[0]!=packet+1)test_failed(); + if(frames[1]!=frames[0]+size[0])test_failed(); + if((toc>>2)!=i)test_failed(); + } else if(ret!=OPUS_INVALID_PACKET)test_failed(); + } + } + fprintf(stdout," code 1 (%6d cases) ........................ OK.\n",cfgs); + cfgs_total+=cfgs;cfgs=0; + + for(i=0;i<64;i++) + { + /*code 2, length code overflow*/ + packet[0]=(i<<2)+2; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + packet[1]=252; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,2,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + for(j=0;j<1275;j++) + { + if(j<252)packet[1]=j; + else{packet[1]=252+(j&3);packet[2]=(j-252)>>2;} + /*Code 2, one too short*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,j+(j<252?2:3)-1,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + /*Code 2, one too long*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,j+(j<252?2:3)+1276,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + /*Code 2, second zero*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,j+(j<252?2:3),&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=2)test_failed(); + if(size[0]!=j||size[1]!=0)test_failed(); + if(frames[1]!=frames[0]+size[0])test_failed(); + if((toc>>2)!=i)test_failed(); + /*Code 2, normal*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,(j<<1)+4,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=2)test_failed(); + if(size[0]!=j||size[1]!=(j<<1)+3-j-(j<252?1:2))test_failed(); + if(frames[1]!=frames[0]+size[0])test_failed(); + if((toc>>2)!=i)test_failed(); + } + } + fprintf(stdout," code 2 (%6d cases) ........................ OK.\n",cfgs); + cfgs_total+=cfgs;cfgs=0; + + for(i=0;i<64;i++) + { + packet[0]=(i<<2)+3; + /*code 3, length code overflow*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + } + fprintf(stdout," code 3 m-truncation (%2d cases) ............... OK.\n",cfgs); + cfgs_total+=cfgs;cfgs=0; + + for(i=0;i<64;i++) + { + /*code 3, m is zero or 49-63*/ + packet[0]=(i<<2)+3; + for(jj=49;jj<=64;jj++) + { + packet[1]=0+(jj&63); /*CBR, no padding*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1275,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + packet[1]=128+(jj&63); /*VBR, no padding*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1275,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + packet[1]=64+(jj&63); /*CBR, padding*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1275,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + packet[1]=128+64+(jj&63); /*VBR, padding*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1275,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + } + } + fprintf(stdout," code 3 m=0,49-64 (%2d cases) ................ OK.\n",cfgs); + cfgs_total+=cfgs;cfgs=0; + + for(i=0;i<64;i++) + { + packet[0]=(i<<2)+3; + /*code 3, m is one, cbr*/ + packet[1]=1; + for(j=0;j<1276;j++) + { + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,j+2,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=1)test_failed(); + if(size[0]!=j)test_failed(); + if((toc>>2)!=i)test_failed(); + } + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1276+2,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + } + fprintf(stdout," code 3 m=1 CBR (%2d cases) ................. OK.\n",cfgs); + cfgs_total+=cfgs;cfgs=0; + + for(i=0;i<64;i++) + { + int frame_samp; + /*code 3, m>1 CBR*/ + packet[0]=(i<<2)+3; + frame_samp=opus_packet_get_samples_per_frame(packet,48000); + for(j=2;j<49;j++) + { + packet[1]=j; + for(sz=2;sz<((j+2)*1275);sz++) + { + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,sz,&toc,frames,size,&payload_offset); + cfgs++; + /*Must be <=120ms, must be evenly divisible, can't have frames>1275 bytes*/ + if(frame_samp*j<=5760 && (sz-2)%j==0 && (sz-2)/j<1276) + { + if(ret!=j)test_failed(); + for(jj=1;jj>2)!=i)test_failed(); + } else if(ret!=OPUS_INVALID_PACKET)test_failed(); + } + } + /*Super jumbo packets*/ + packet[1]=5760/frame_samp; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,1275*packet[1]+2,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=packet[1])test_failed(); + for(jj=0;jj>2)!=i)test_failed(); + } + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,2+1276,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + for(j=2;j<49;j++) + { + packet[1]=128+j; + /*Length code overflow*/ + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,2+j-2,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + packet[2]=252; + packet[3]=0; + for(jj=4;jj<2+j;jj++)packet[jj]=0; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,2+j,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + /*One byte too short*/ + for(jj=2;jj<2+j;jj++)packet[jj]=0; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,2+j-2,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + /*One byte too short thanks to length coding*/ + packet[2]=252; + packet[3]=0; + for(jj=4;jj<2+j;jj++)packet[jj]=0; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,2+j+252-1,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + /*Most expensive way of coding zeros*/ + for(jj=2;jj<2+j;jj++)packet[jj]=0; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,2+j-1,&toc,frames,size,&payload_offset); + cfgs++; + if(frame_samp*j<=5760){ + if(ret!=j)test_failed(); + for(jj=0;jj>2)!=i)test_failed(); + } else if(ret!=OPUS_INVALID_PACKET)test_failed(); + /*Quasi-CBR use of mode 3*/ + for(sz=0;sz<8;sz++) + { + const int tsz[8]={50,201,403,700,1472,5110,20400,61298}; + int pos=0; + int as=(tsz[sz]+i-j-2)/j; + for(jj=0;jj>2;pos+=2;} + } + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,tsz[sz]+i,&toc,frames,size,&payload_offset); + cfgs++; + if(frame_samp*j<=5760 && as<1276 && (tsz[sz]+i-2-pos-as*(j-1))<1276){ + if(ret!=j)test_failed(); + for(jj=0;jj>2)!=i)test_failed(); + } else if(ret!=OPUS_INVALID_PACKET)test_failed(); + } + } + } + fprintf(stdout," code 3 m=1-48 VBR (%2d cases) ............. OK.\n",cfgs); + cfgs_total+=cfgs;cfgs=0; + + for(i=0;i<64;i++) + { + packet[0]=(i<<2)+3; + /*Padding*/ + packet[1]=128+1+64; + /*Overflow the length coding*/ + for(jj=2;jj<127;jj++)packet[jj]=255; + UNDEFINE_FOR_PARSE + ret=opus_packet_parse(packet,127,&toc,frames,size,&payload_offset); + cfgs++; + if(ret!=OPUS_INVALID_PACKET)test_failed(); + + for(sz=0;sz<4;sz++) + { + const int tsz[4]={0,72,512,1275}; + for(jj=sz;jj<65025;jj+=11) + { + int pos; + for(pos=0;pos>2)!=i)test_failed(); + } else if (ret!=OPUS_INVALID_PACKET)test_failed(); + } + } + } + fprintf(stdout," code 3 padding (%2d cases) ............... OK.\n",cfgs); + cfgs_total+=cfgs; + fprintf(stdout," opus_packet_parse ............................ OK.\n"); + fprintf(stdout," All packet parsing tests passed\n"); + fprintf(stdout," (%d API invocations)\n",cfgs_total); + return cfgs_total; +} + +/* This is a helper macro for the encoder tests. + The encoder api tests all have a pattern of set-must-fail, set-must-fail, + set-must-pass, get-and-compare, set-must-pass, get-and-compare. */ +#define CHECK_SETGET(setcall,getcall,badv,badv2,goodv,goodv2,sok,gok) \ + i=(badv);\ + if(opus_encoder_ctl(enc,setcall)==OPUS_OK)test_failed();\ + i=(badv2);\ + if(opus_encoder_ctl(enc,setcall)==OPUS_OK)test_failed();\ + j=i=(goodv);\ + if(opus_encoder_ctl(enc,setcall)!=OPUS_OK)test_failed();\ + i=-12345;\ + VG_UNDEF(&i,sizeof(i)); \ + err=opus_encoder_ctl(enc,getcall);\ + if(err!=OPUS_OK || i!=j)test_failed();\ + j=i=(goodv2);\ + if(opus_encoder_ctl(enc,setcall)!=OPUS_OK)test_failed();\ + fprintf(stdout,sok);\ + i=-12345;\ + VG_UNDEF(&i,sizeof(i)); \ + err=opus_encoder_ctl(enc,getcall);\ + if(err!=OPUS_OK || i!=j)test_failed();\ + fprintf(stdout,gok);\ + cfgs+=6; + +opus_int32 test_enc_api(void) +{ + opus_uint32 enc_final_range; + OpusEncoder *enc; + opus_int32 i,j; + unsigned char packet[1276]; +#ifndef DISABLE_FLOAT_API + float fbuf[960*2]; +#endif + short sbuf[960*2]; + int c,err,cfgs; + + cfgs=0; + /*First test invalid configurations which should fail*/ + fprintf(stdout,"\n Encoder basic API tests\n"); + fprintf(stdout," ---------------------------------------------------\n"); + for(c=0;c<4;c++) + { + i=opus_encoder_get_size(c); + if(((c==1||c==2)&&(i<=2048||i>1<<17))||((c!=1&&c!=2)&&i!=0))test_failed(); + fprintf(stdout," opus_encoder_get_size(%d)=%d ...............%s OK.\n",c,i,i>0?"":"...."); + cfgs++; + } + + /*Test with unsupported sample rates, channel counts*/ + for(c=0;c<4;c++) + { + for(i=-7;i<=96000;i++) + { + int fs; + if((i==8000||i==12000||i==16000||i==24000||i==48000)&&(c==1||c==2))continue; + switch(i) + { + case(-5):fs=-8000;break; + case(-6):fs=INT32_MAX;break; + case(-7):fs=INT32_MIN;break; + default:fs=i; + } + err = OPUS_OK; + VG_UNDEF(&err,sizeof(err)); + enc = opus_encoder_create(fs, c, OPUS_APPLICATION_VOIP, &err); + if(err!=OPUS_BAD_ARG || enc!=NULL)test_failed(); + cfgs++; + enc = opus_encoder_create(fs, c, OPUS_APPLICATION_VOIP, 0); + if(enc!=NULL)test_failed(); + cfgs++; + opus_encoder_destroy(enc); + enc=malloc(opus_encoder_get_size(2)); + if(enc==NULL)test_failed(); + err = opus_encoder_init(enc, fs, c, OPUS_APPLICATION_VOIP); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + free(enc); + } + } + + enc = opus_encoder_create(48000, 2, OPUS_AUTO, NULL); + if(enc!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(&err,sizeof(err)); + enc = opus_encoder_create(48000, 2, OPUS_AUTO, &err); + if(err!=OPUS_BAD_ARG || enc!=NULL)test_failed(); + cfgs++; + + VG_UNDEF(&err,sizeof(err)); + enc = opus_encoder_create(48000, 2, OPUS_APPLICATION_VOIP, NULL); + if(enc==NULL)test_failed(); + opus_encoder_destroy(enc); + cfgs++; + + VG_UNDEF(&err,sizeof(err)); + enc = opus_encoder_create(48000, 2, OPUS_APPLICATION_RESTRICTED_LOWDELAY, &err); + if(err!=OPUS_OK || enc==NULL)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_GET_LOOKAHEAD(&i)); + if(err!=OPUS_OK || i<0 || i>32766)test_failed(); + cfgs++; + opus_encoder_destroy(enc); + + VG_UNDEF(&err,sizeof(err)); + enc = opus_encoder_create(48000, 2, OPUS_APPLICATION_AUDIO, &err); + if(err!=OPUS_OK || enc==NULL)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_GET_LOOKAHEAD(&i)); + if(err!=OPUS_OK || i<0 || i>32766)test_failed(); + opus_encoder_destroy(enc); + cfgs++; + + VG_UNDEF(&err,sizeof(err)); + enc = opus_encoder_create(48000, 2, OPUS_APPLICATION_VOIP, &err); + if(err!=OPUS_OK || enc==NULL)test_failed(); + cfgs++; + + fprintf(stdout," opus_encoder_create() ........................ OK.\n"); + fprintf(stdout," opus_encoder_init() .......................... OK.\n"); + + i=-12345; + VG_UNDEF(&i,sizeof(i)); + err=opus_encoder_ctl(enc,OPUS_GET_LOOKAHEAD(&i)); + if(err!=OPUS_OK || i<0 || i>32766)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_GET_LOOKAHEAD(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_LOOKAHEAD ........................... OK.\n"); + + err=opus_encoder_ctl(enc,OPUS_GET_SAMPLE_RATE(&i)); + if(err!=OPUS_OK || i!=48000)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_GET_SAMPLE_RATE(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_SAMPLE_RATE ......................... OK.\n"); + + if(opus_encoder_ctl(enc,OPUS_UNIMPLEMENTED)!=OPUS_UNIMPLEMENTED)test_failed(); + fprintf(stdout," OPUS_UNIMPLEMENTED ........................... OK.\n"); + cfgs++; + + err=opus_encoder_ctl(enc,OPUS_GET_APPLICATION(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_APPLICATION(i),OPUS_GET_APPLICATION(&i),-1,OPUS_AUTO, + OPUS_APPLICATION_AUDIO,OPUS_APPLICATION_RESTRICTED_LOWDELAY, + " OPUS_SET_APPLICATION ......................... OK.\n", + " OPUS_GET_APPLICATION ......................... OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_BITRATE(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_encoder_ctl(enc,OPUS_SET_BITRATE(1073741832))!=OPUS_OK)test_failed(); + cfgs++; + VG_UNDEF(&i,sizeof(i)); + if(opus_encoder_ctl(enc,OPUS_GET_BITRATE(&i))!=OPUS_OK)test_failed(); + if(i>700000||i<256000)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_BITRATE(i),OPUS_GET_BITRATE(&i),-12345,0, + 500,256000, + " OPUS_SET_BITRATE ............................. OK.\n", + " OPUS_GET_BITRATE ............................. OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_FORCE_CHANNELS(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_FORCE_CHANNELS(i),OPUS_GET_FORCE_CHANNELS(&i),-1,3, + 1,OPUS_AUTO, + " OPUS_SET_FORCE_CHANNELS ...................... OK.\n", + " OPUS_GET_FORCE_CHANNELS ...................... OK.\n") + + i=-2; + if(opus_encoder_ctl(enc,OPUS_SET_BANDWIDTH(i))==OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_FULLBAND+1; + if(opus_encoder_ctl(enc,OPUS_SET_BANDWIDTH(i))==OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_NARROWBAND; + if(opus_encoder_ctl(enc,OPUS_SET_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_FULLBAND; + if(opus_encoder_ctl(enc,OPUS_SET_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_WIDEBAND; + if(opus_encoder_ctl(enc,OPUS_SET_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_MEDIUMBAND; + if(opus_encoder_ctl(enc,OPUS_SET_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + fprintf(stdout," OPUS_SET_BANDWIDTH ........................... OK.\n"); + /*We don't test if the bandwidth has actually changed. + because the change may be delayed until the encoder is advanced.*/ + i=-12345; + VG_UNDEF(&i,sizeof(i)); + err=opus_encoder_ctl(enc,OPUS_GET_BANDWIDTH(&i)); + if(err!=OPUS_OK || (i!=OPUS_BANDWIDTH_NARROWBAND&& + i!=OPUS_BANDWIDTH_MEDIUMBAND&&i!=OPUS_BANDWIDTH_WIDEBAND&& + i!=OPUS_BANDWIDTH_FULLBAND&&i!=OPUS_AUTO))test_failed(); + cfgs++; + if(opus_encoder_ctl(enc,OPUS_SET_BANDWIDTH(OPUS_AUTO))!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_GET_BANDWIDTH(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_BANDWIDTH ........................... OK.\n"); + + i=-2; + if(opus_encoder_ctl(enc,OPUS_SET_MAX_BANDWIDTH(i))==OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_FULLBAND+1; + if(opus_encoder_ctl(enc,OPUS_SET_MAX_BANDWIDTH(i))==OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_NARROWBAND; + if(opus_encoder_ctl(enc,OPUS_SET_MAX_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_FULLBAND; + if(opus_encoder_ctl(enc,OPUS_SET_MAX_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_WIDEBAND; + if(opus_encoder_ctl(enc,OPUS_SET_MAX_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + i=OPUS_BANDWIDTH_MEDIUMBAND; + if(opus_encoder_ctl(enc,OPUS_SET_MAX_BANDWIDTH(i))!=OPUS_OK)test_failed(); + cfgs++; + fprintf(stdout," OPUS_SET_MAX_BANDWIDTH ....................... OK.\n"); + /*We don't test if the bandwidth has actually changed. + because the change may be delayed until the encoder is advanced.*/ + i=-12345; + VG_UNDEF(&i,sizeof(i)); + err=opus_encoder_ctl(enc,OPUS_GET_MAX_BANDWIDTH(&i)); + if(err!=OPUS_OK || (i!=OPUS_BANDWIDTH_NARROWBAND&& + i!=OPUS_BANDWIDTH_MEDIUMBAND&&i!=OPUS_BANDWIDTH_WIDEBAND&& + i!=OPUS_BANDWIDTH_FULLBAND))test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_GET_MAX_BANDWIDTH(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_MAX_BANDWIDTH ....................... OK.\n"); + + err=opus_encoder_ctl(enc,OPUS_GET_DTX(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_DTX(i),OPUS_GET_DTX(&i),-1,2, + 1,0, + " OPUS_SET_DTX ................................. OK.\n", + " OPUS_GET_DTX ................................. OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_COMPLEXITY(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_COMPLEXITY(i),OPUS_GET_COMPLEXITY(&i),-1,11, + 0,10, + " OPUS_SET_COMPLEXITY .......................... OK.\n", + " OPUS_GET_COMPLEXITY .......................... OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_INBAND_FEC(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_INBAND_FEC(i),OPUS_GET_INBAND_FEC(&i),-1,3, + 1,0, + " OPUS_SET_INBAND_FEC .......................... OK.\n", + " OPUS_GET_INBAND_FEC .......................... OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_PACKET_LOSS_PERC(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_PACKET_LOSS_PERC(i),OPUS_GET_PACKET_LOSS_PERC(&i),-1,101, + 100,0, + " OPUS_SET_PACKET_LOSS_PERC .................... OK.\n", + " OPUS_GET_PACKET_LOSS_PERC .................... OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_VBR(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_VBR(i),OPUS_GET_VBR(&i),-1,2, + 1,0, + " OPUS_SET_VBR ................................. OK.\n", + " OPUS_GET_VBR ................................. OK.\n") + +/* err=opus_encoder_ctl(enc,OPUS_GET_VOICE_RATIO(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_VOICE_RATIO(i),OPUS_GET_VOICE_RATIO(&i),-2,101, + 0,50, + " OPUS_SET_VOICE_RATIO ......................... OK.\n", + " OPUS_GET_VOICE_RATIO ......................... OK.\n")*/ + + err=opus_encoder_ctl(enc,OPUS_GET_VBR_CONSTRAINT(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_VBR_CONSTRAINT(i),OPUS_GET_VBR_CONSTRAINT(&i),-1,2, + 1,0, + " OPUS_SET_VBR_CONSTRAINT ...................... OK.\n", + " OPUS_GET_VBR_CONSTRAINT ...................... OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_SIGNAL(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_SIGNAL(i),OPUS_GET_SIGNAL(&i),-12345,0x7FFFFFFF, + OPUS_SIGNAL_MUSIC,OPUS_AUTO, + " OPUS_SET_SIGNAL .............................. OK.\n", + " OPUS_GET_SIGNAL .............................. OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_LSB_DEPTH(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_LSB_DEPTH(i),OPUS_GET_LSB_DEPTH(&i),7,25,16,24, + " OPUS_SET_LSB_DEPTH ........................... OK.\n", + " OPUS_GET_LSB_DEPTH ........................... OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_PREDICTION_DISABLED(&i)); + if(i!=0)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_GET_PREDICTION_DISABLED(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_PREDICTION_DISABLED(i),OPUS_GET_PREDICTION_DISABLED(&i),-1,2,1,0, + " OPUS_SET_PREDICTION_DISABLED ................. OK.\n", + " OPUS_GET_PREDICTION_DISABLED ................. OK.\n") + + err=opus_encoder_ctl(enc,OPUS_GET_EXPERT_FRAME_DURATION(null_int_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_2_5_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_5_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_10_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_20_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_40_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_60_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_80_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_100_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + err=opus_encoder_ctl(enc,OPUS_SET_EXPERT_FRAME_DURATION(OPUS_FRAMESIZE_120_MS)); + if(err!=OPUS_OK)test_failed(); + cfgs++; + CHECK_SETGET(OPUS_SET_EXPERT_FRAME_DURATION(i),OPUS_GET_EXPERT_FRAME_DURATION(&i),0,-1, + OPUS_FRAMESIZE_60_MS,OPUS_FRAMESIZE_ARG, + " OPUS_SET_EXPERT_FRAME_DURATION ............... OK.\n", + " OPUS_GET_EXPERT_FRAME_DURATION ............... OK.\n") + + /*OPUS_SET_FORCE_MODE is not tested here because it's not a public API, however the encoder tests use it*/ + + err=opus_encoder_ctl(enc,OPUS_GET_FINAL_RANGE(null_uint_ptr)); + if(err!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_encoder_ctl(enc,OPUS_GET_FINAL_RANGE(&enc_final_range))!=OPUS_OK)test_failed(); + cfgs++; + fprintf(stdout," OPUS_GET_FINAL_RANGE ......................... OK.\n"); + + /*Reset the encoder*/ + if(opus_encoder_ctl(enc, OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + cfgs++; + fprintf(stdout," OPUS_RESET_STATE ............................. OK.\n"); + + memset(sbuf,0,sizeof(short)*2*960); + VG_UNDEF(packet,sizeof(packet)); + i=opus_encode(enc, sbuf, 960, packet, sizeof(packet)); + if(i<1 || (i>(opus_int32)sizeof(packet)))test_failed(); + VG_CHECK(packet,i); + cfgs++; + fprintf(stdout," opus_encode() ................................ OK.\n"); +#ifndef DISABLE_FLOAT_API + memset(fbuf,0,sizeof(float)*2*960); + VG_UNDEF(packet,sizeof(packet)); + i=opus_encode_float(enc, fbuf, 960, packet, sizeof(packet)); + if(i<1 || (i>(opus_int32)sizeof(packet)))test_failed(); + VG_CHECK(packet,i); + cfgs++; + fprintf(stdout," opus_encode_float() .......................... OK.\n"); +#endif + +#if 0 + /*These tests are disabled because the library crashes with null states*/ + if(opus_encoder_ctl(0,OPUS_RESET_STATE) !=OPUS_INVALID_STATE)test_failed(); + if(opus_encoder_init(0,48000,1,OPUS_APPLICATION_VOIP) !=OPUS_INVALID_STATE)test_failed(); + if(opus_encode(0,sbuf,960,packet,sizeof(packet)) !=OPUS_INVALID_STATE)test_failed(); + if(opus_encode_float(0,fbuf,960,packet,sizeof(packet))!=OPUS_INVALID_STATE)test_failed(); +#endif + opus_encoder_destroy(enc); + cfgs++; + fprintf(stdout," All encoder interface tests passed\n"); + fprintf(stdout," (%d API invocations)\n",cfgs); + return cfgs; +} + +#define max_out (1276*48+48*2+2) +int test_repacketizer_api(void) +{ + int ret,cfgs,i,j,k; + OpusRepacketizer *rp; + unsigned char *packet; + unsigned char *po; + cfgs=0; + fprintf(stdout,"\n Repacketizer tests\n"); + fprintf(stdout," ---------------------------------------------------\n"); + + packet=malloc(max_out); + if(packet==NULL)test_failed(); + memset(packet,0,max_out); + po=malloc(max_out+256); + if(po==NULL)test_failed(); + + i=opus_repacketizer_get_size(); + if(i<=0)test_failed(); + cfgs++; + fprintf(stdout," opus_repacketizer_get_size()=%d ............. OK.\n",i); + + rp=malloc(i); + rp=opus_repacketizer_init(rp); + if(rp==NULL)test_failed(); + cfgs++; + free(rp); + fprintf(stdout," opus_repacketizer_init ....................... OK.\n"); + + rp=opus_repacketizer_create(); + if(rp==NULL)test_failed(); + cfgs++; + fprintf(stdout," opus_repacketizer_create ..................... OK.\n"); + + if(opus_repacketizer_get_nb_frames(rp)!=0)test_failed(); + cfgs++; + fprintf(stdout," opus_repacketizer_get_nb_frames .............. OK.\n"); + + /*Length overflows*/ + VG_UNDEF(packet,4); + if(opus_repacketizer_cat(rp,packet,0)!=OPUS_INVALID_PACKET)test_failed(); /* Zero len */ + cfgs++; + packet[0]=1; + if(opus_repacketizer_cat(rp,packet,2)!=OPUS_INVALID_PACKET)test_failed(); /* Odd payload code 1 */ + cfgs++; + packet[0]=2; + if(opus_repacketizer_cat(rp,packet,1)!=OPUS_INVALID_PACKET)test_failed(); /* Code 2 overflow one */ + cfgs++; + packet[0]=3; + if(opus_repacketizer_cat(rp,packet,1)!=OPUS_INVALID_PACKET)test_failed(); /* Code 3 no count */ + cfgs++; + packet[0]=2; + packet[1]=255; + if(opus_repacketizer_cat(rp,packet,2)!=OPUS_INVALID_PACKET)test_failed(); /* Code 2 overflow two */ + cfgs++; + packet[0]=2; + packet[1]=250; + if(opus_repacketizer_cat(rp,packet,251)!=OPUS_INVALID_PACKET)test_failed(); /* Code 2 overflow three */ + cfgs++; + packet[0]=3; + packet[1]=0; + if(opus_repacketizer_cat(rp,packet,2)!=OPUS_INVALID_PACKET)test_failed(); /* Code 3 m=0 */ + cfgs++; + packet[1]=49; + if(opus_repacketizer_cat(rp,packet,100)!=OPUS_INVALID_PACKET)test_failed(); /* Code 3 m=49 */ + cfgs++; + packet[0]=0; + if(opus_repacketizer_cat(rp,packet,3)!=OPUS_OK)test_failed(); + cfgs++; + packet[0]=1<<2; + if(opus_repacketizer_cat(rp,packet,3)!=OPUS_INVALID_PACKET)test_failed(); /* Change in TOC */ + cfgs++; + + /* Code 0,1,3 CBR -> Code 0,1,3 CBR */ + opus_repacketizer_init(rp); + for(j=0;j<32;j++) + { + /* TOC types, test half with stereo */ + int maxi; + packet[0]=((j<<1)+(j&1))<<2; + maxi=960/opus_packet_get_samples_per_frame(packet,8000); + for(i=1;i<=maxi;i++) + { + /* Number of CBR frames in the input packets */ + int maxp; + packet[0]=((j<<1)+(j&1))<<2; + if(i>1)packet[0]+=i==2?1:3; + packet[1]=i>2?i:0; + maxp=960/(i*opus_packet_get_samples_per_frame(packet,8000)); + for(k=0;k<=(1275+75);k+=3) + { + /*Payload size*/ + opus_int32 cnt,rcnt; + if(k%i!=0)continue; /* Only testing CBR here, payload must be a multiple of the count */ + for(cnt=0;cnt0) + { + ret=opus_repacketizer_cat(rp,packet,k+(i>2?2:1)); + if((cnt<=maxp&&k<=(1275*i))?ret!=OPUS_OK:ret!=OPUS_INVALID_PACKET)test_failed(); + cfgs++; + } + rcnt=k<=(1275*i)?(cnt0) + { + int len; + len=k*rcnt+((rcnt*i)>2?2:1); + if(ret!=len)test_failed(); + if((rcnt*i)<2&&(po[0]&3)!=0)test_failed(); /* Code 0 */ + if((rcnt*i)==2&&(po[0]&3)!=1)test_failed(); /* Code 1 */ + if((rcnt*i)>2&&(((po[0]&3)!=3)||(po[1]!=rcnt*i)))test_failed(); /* Code 3 CBR */ + cfgs++; + if(opus_repacketizer_out(rp,po,len)!=len)test_failed(); + cfgs++; + if(opus_packet_unpad(po,len)!=len)test_failed(); + cfgs++; + if(opus_packet_pad(po,len,len+1)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_packet_pad(po,len+1,len+256)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_packet_unpad(po,len+256)!=len)test_failed(); + cfgs++; + if(opus_multistream_packet_unpad(po,len,1)!=len)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,len,len+1,1)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,len+1,len+256,1)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_multistream_packet_unpad(po,len+256,1)!=len)test_failed(); + cfgs++; + if(opus_repacketizer_out(rp,po,len-1)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + if(len>1) + { + if(opus_repacketizer_out(rp,po,1)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + } + if(opus_repacketizer_out(rp,po,0)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + } else if (ret!=OPUS_BAD_ARG)test_failed(); /* M must not be 0 */ + } + opus_repacketizer_init(rp); + } + } + } + + /*Change in input count code, CBR out*/ + opus_repacketizer_init(rp); + packet[0]=0; + if(opus_repacketizer_cat(rp,packet,5)!=OPUS_OK)test_failed(); + cfgs++; + packet[0]+=1; + if(opus_repacketizer_cat(rp,packet,9)!=OPUS_OK)test_failed(); + cfgs++; + i=opus_repacketizer_out(rp,po,max_out); + if((i!=(4+8+2))||((po[0]&3)!=3)||((po[1]&63)!=3)||((po[1]>>7)!=0))test_failed(); + cfgs++; + i=opus_repacketizer_out_range(rp,0,1,po,max_out); + if(i!=5||(po[0]&3)!=0)test_failed(); + cfgs++; + i=opus_repacketizer_out_range(rp,1,2,po,max_out); + if(i!=5||(po[0]&3)!=0)test_failed(); + cfgs++; + + /*Change in input count code, VBR out*/ + opus_repacketizer_init(rp); + packet[0]=1; + if(opus_repacketizer_cat(rp,packet,9)!=OPUS_OK)test_failed(); + cfgs++; + packet[0]=0; + if(opus_repacketizer_cat(rp,packet,3)!=OPUS_OK)test_failed(); + cfgs++; + i=opus_repacketizer_out(rp,po,max_out); + if((i!=(2+8+2+2))||((po[0]&3)!=3)||((po[1]&63)!=3)||((po[1]>>7)!=1))test_failed(); + cfgs++; + + /*VBR in, VBR out*/ + opus_repacketizer_init(rp); + packet[0]=2; + packet[1]=4; + if(opus_repacketizer_cat(rp,packet,8)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_repacketizer_cat(rp,packet,8)!=OPUS_OK)test_failed(); + cfgs++; + i=opus_repacketizer_out(rp,po,max_out); + if((i!=(2+1+1+1+4+2+4+2))||((po[0]&3)!=3)||((po[1]&63)!=4)||((po[1]>>7)!=1))test_failed(); + cfgs++; + + /*VBR in, CBR out*/ + opus_repacketizer_init(rp); + packet[0]=2; + packet[1]=4; + if(opus_repacketizer_cat(rp,packet,10)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_repacketizer_cat(rp,packet,10)!=OPUS_OK)test_failed(); + cfgs++; + i=opus_repacketizer_out(rp,po,max_out); + if((i!=(2+4+4+4+4))||((po[0]&3)!=3)||((po[1]&63)!=4)||((po[1]>>7)!=0))test_failed(); + cfgs++; + + /*Count 0 in, VBR out*/ + for(j=0;j<32;j++) + { + /* TOC types, test half with stereo */ + int maxi,sum,rcnt; + packet[0]=((j<<1)+(j&1))<<2; + maxi=960/opus_packet_get_samples_per_frame(packet,8000); + sum=0; + rcnt=0; + opus_repacketizer_init(rp); + for(i=1;i<=maxi+2;i++) + { + int len; + ret=opus_repacketizer_cat(rp,packet,i); + if(rcnt2&&(po[1]&63)!=rcnt)test_failed(); + if(rcnt==2&&(po[0]&3)!=2)test_failed(); + if(rcnt==1&&(po[0]&3)!=0)test_failed(); + cfgs++; + if(opus_repacketizer_out(rp,po,len)!=len)test_failed(); + cfgs++; + if(opus_packet_unpad(po,len)!=len)test_failed(); + cfgs++; + if(opus_packet_pad(po,len,len+1)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_packet_pad(po,len+1,len+256)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_packet_unpad(po,len+256)!=len)test_failed(); + cfgs++; + if(opus_multistream_packet_unpad(po,len,1)!=len)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,len,len+1,1)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,len+1,len+256,1)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_multistream_packet_unpad(po,len+256,1)!=len)test_failed(); + cfgs++; + if(opus_repacketizer_out(rp,po,len-1)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + if(len>1) + { + if(opus_repacketizer_out(rp,po,1)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + } + if(opus_repacketizer_out(rp,po,0)!=OPUS_BUFFER_TOO_SMALL)test_failed(); + cfgs++; + } + } + + po[0]='O'; + po[1]='p'; + if(opus_packet_pad(po,4,4)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,4,4,1)!=OPUS_OK)test_failed(); + cfgs++; + if(opus_packet_pad(po,4,5)!=OPUS_INVALID_PACKET)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,4,5,1)!=OPUS_INVALID_PACKET)test_failed(); + cfgs++; + if(opus_packet_pad(po,0,5)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,0,5,1)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_packet_unpad(po,0)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_multistream_packet_unpad(po,0,1)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_packet_unpad(po,4)!=OPUS_INVALID_PACKET)test_failed(); + cfgs++; + if(opus_multistream_packet_unpad(po,4,1)!=OPUS_INVALID_PACKET)test_failed(); + cfgs++; + po[0]=0; + po[1]=0; + po[2]=0; + if(opus_packet_pad(po,5,4)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + if(opus_multistream_packet_pad(po,5,4,1)!=OPUS_BAD_ARG)test_failed(); + cfgs++; + + fprintf(stdout," opus_repacketizer_cat ........................ OK.\n"); + fprintf(stdout," opus_repacketizer_out ........................ OK.\n"); + fprintf(stdout," opus_repacketizer_out_range .................. OK.\n"); + fprintf(stdout," opus_packet_pad .............................. OK.\n"); + fprintf(stdout," opus_packet_unpad ............................ OK.\n"); + fprintf(stdout," opus_multistream_packet_pad .................. OK.\n"); + fprintf(stdout," opus_multistream_packet_unpad ................ OK.\n"); + + opus_repacketizer_destroy(rp); + cfgs++; + free(packet); + free(po); + fprintf(stdout," All repacketizer tests passed\n"); + fprintf(stdout," (%7d API invocations)\n",cfgs); + + return cfgs; +} + +#ifdef MALLOC_FAIL +/* GLIBC 2.14 declares __malloc_hook as deprecated, generating a warning + * under GCC. However, this is the cleanest way to test malloc failure + * handling in our codebase, and the lack of thread safety isn't an + * issue here. We therefore disable the warning for this function. + */ +#if OPUS_GNUC_PREREQ(4,6) +/* Save the current warning settings */ +#pragma GCC diagnostic push +#endif +#pragma GCC diagnostic ignored "-Wdeprecated-declarations" + +typedef void *(*mhook)(size_t __size, __const void *); +#endif + +int test_malloc_fail(void) +{ +#ifdef MALLOC_FAIL + OpusDecoder *dec; + OpusEncoder *enc; + OpusRepacketizer *rp; + unsigned char mapping[256] = {0,1}; + OpusMSDecoder *msdec; + OpusMSEncoder *msenc; + int rate,c,app,cfgs,err,useerr; + int *ep; + mhook orig_malloc; + cfgs=0; +#endif + fprintf(stdout,"\n malloc() failure tests\n"); + fprintf(stdout," ---------------------------------------------------\n"); +#ifdef MALLOC_FAIL + orig_malloc=__malloc_hook; + __malloc_hook=malloc_hook; + ep=(int *)opus_alloc(sizeof(int)); + if(ep!=NULL) + { + if(ep)free(ep); + __malloc_hook=orig_malloc; +#endif + fprintf(stdout," opus_decoder_create() ................... SKIPPED.\n"); + fprintf(stdout," opus_encoder_create() ................... SKIPPED.\n"); + fprintf(stdout," opus_repacketizer_create() .............. SKIPPED.\n"); + fprintf(stdout," opus_multistream_decoder_create() ....... SKIPPED.\n"); + fprintf(stdout," opus_multistream_encoder_create() ....... SKIPPED.\n"); + fprintf(stdout,"(Test only supported with GLIBC and without valgrind)\n"); + return 0; +#ifdef MALLOC_FAIL + } + for(useerr=0;useerr<2;useerr++) + { + ep=useerr?&err:0; + for(rate=0;rate<5;rate++) + { + for(c=1;c<3;c++) + { + err=1; + if(useerr) + { + VG_UNDEF(&err,sizeof(err)); + } + dec=opus_decoder_create(opus_rates[rate], c, ep); + if(dec!=NULL||(useerr&&err!=OPUS_ALLOC_FAIL)) + { + __malloc_hook=orig_malloc; + test_failed(); + } + cfgs++; + msdec=opus_multistream_decoder_create(opus_rates[rate], c, 1, c-1, mapping, ep); + if(msdec!=NULL||(useerr&&err!=OPUS_ALLOC_FAIL)) + { + __malloc_hook=orig_malloc; + test_failed(); + } + cfgs++; + for(app=0;app<3;app++) + { + if(useerr) + { + VG_UNDEF(&err,sizeof(err)); + } + enc=opus_encoder_create(opus_rates[rate], c, opus_apps[app],ep); + if(enc!=NULL||(useerr&&err!=OPUS_ALLOC_FAIL)) + { + __malloc_hook=orig_malloc; + test_failed(); + } + cfgs++; + msenc=opus_multistream_encoder_create(opus_rates[rate], c, 1, c-1, mapping, opus_apps[app],ep); + if(msenc!=NULL||(useerr&&err!=OPUS_ALLOC_FAIL)) + { + __malloc_hook=orig_malloc; + test_failed(); + } + cfgs++; + } + } + } + } + rp=opus_repacketizer_create(); + if(rp!=NULL) + { + __malloc_hook=orig_malloc; + test_failed(); + } + cfgs++; + __malloc_hook=orig_malloc; + fprintf(stdout," opus_decoder_create() ........................ OK.\n"); + fprintf(stdout," opus_encoder_create() ........................ OK.\n"); + fprintf(stdout," opus_repacketizer_create() ................... OK.\n"); + fprintf(stdout," opus_multistream_decoder_create() ............ OK.\n"); + fprintf(stdout," opus_multistream_encoder_create() ............ OK.\n"); + fprintf(stdout," All malloc failure tests passed\n"); + fprintf(stdout," (%2d API invocations)\n",cfgs); + return cfgs; +#endif +} + +#ifdef MALLOC_FAIL +#if __GNUC_PREREQ(4,6) +#pragma GCC diagnostic pop /* restore -Wdeprecated-declarations */ +#endif +#endif + +int main(int _argc, char **_argv) +{ + opus_int32 total; + const char * oversion; + if(_argc>1) + { + fprintf(stderr,"Usage: %s\n",_argv[0]); + return 1; + } + iseed=0; + + oversion=opus_get_version_string(); + if(!oversion)test_failed(); + fprintf(stderr,"Testing the %s API deterministically\n", oversion); + if(opus_strerror(-32768)==NULL)test_failed(); + if(opus_strerror(32767)==NULL)test_failed(); + if(strlen(opus_strerror(0))<1)test_failed(); + total=4; + + total+=test_dec_api(); + total+=test_msdec_api(); + total+=test_parse(); + total+=test_enc_api(); + total+=test_repacketizer_api(); + total+=test_malloc_fail(); + + fprintf(stderr,"\nAll API tests passed.\nThe libopus API was invoked %d times.\n",total); + + return 0; +} diff --git a/lib/opus/tests/test_opus_common.h b/lib/opus/tests/test_opus_common.h new file mode 100644 index 00000000..5fb924f4 --- /dev/null +++ b/lib/opus/tests/test_opus_common.h @@ -0,0 +1,85 @@ +/* Copyright (c) 2011 Xiph.Org Foundation + Written by Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +static OPUS_INLINE void deb2_impl(unsigned char *_t,unsigned char **_p,int _k,int _x,int _y) +{ + int i; + if(_x>2){ + if(_y<3)for(i=0;i<_y;i++)*(--*_p)=_t[i+1]; + }else{ + _t[_x]=_t[_x-_y]; + deb2_impl(_t,_p,_k,_x+1,_y); + for(i=_t[_x-_y]+1;i<_k;i++){ + _t[_x]=i; + deb2_impl(_t,_p,_k,_x+1,_x); + } + } +} + +/*Generates a De Bruijn sequence (k,2) with length k^2*/ +static OPUS_INLINE void debruijn2(int _k, unsigned char *_res) +{ + unsigned char *p; + unsigned char *t; + t=malloc(sizeof(unsigned char)*_k*2); + memset(t,0,sizeof(unsigned char)*_k*2); + p=&_res[_k*_k]; + deb2_impl(t,&p,_k,1,1); + free(t); +} + +/*MWC RNG of George Marsaglia*/ +static opus_uint32 Rz, Rw; +static OPUS_INLINE opus_uint32 fast_rand(void) +{ + Rz=36969*(Rz&65535)+(Rz>>16); + Rw=18000*(Rw&65535)+(Rw>>16); + return (Rz<<16)+Rw; +} +static opus_uint32 iseed; + +#ifdef __GNUC__ +__attribute__((noreturn)) +#elif defined(_MSC_VER) +__declspec(noreturn) +#endif +static OPUS_INLINE void _test_failed(const char *file, int line) +{ + fprintf(stderr,"\n ***************************************************\n"); + fprintf(stderr," *** A fatal error was detected. ***\n"); + fprintf(stderr," ***************************************************\n"); + fprintf(stderr,"Please report this failure and include\n"); + fprintf(stderr,"'make check SEED=%u fails %s at line %d for %s'\n",iseed,file,line,opus_get_version_string()); + fprintf(stderr,"and any relevant details about your system.\n\n"); +#if defined(_MSC_VER) + _set_abort_behavior( 0, _WRITE_ABORT_MSG); +#endif + abort(); +} +#define test_failed() _test_failed(__FILE__, __LINE__); +#define opus_test_assert(cond) {if (!(cond)) {test_failed();}} +void regression_test(void); diff --git a/lib/opus/tests/test_opus_decode.c b/lib/opus/tests/test_opus_decode.c new file mode 100644 index 00000000..2d1e2d41 --- /dev/null +++ b/lib/opus/tests/test_opus_decode.c @@ -0,0 +1,463 @@ +/* Copyright (c) 2011-2013 Xiph.Org Foundation + Written by Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include +#include +#include +#include +#ifndef _WIN32 +#include +#else +#include +#define getpid _getpid +#endif +#include "opus.h" +#include "test_opus_common.h" + +#define MAX_PACKET (1500) +#define MAX_FRAME_SAMP (5760) + +int test_decoder_code0(int no_fuzz) +{ + static const opus_int32 fsv[5]={48000,24000,16000,12000,8000}; + int err,skip,plen; + int out_samples,fec; + int t; + opus_int32 i; + OpusDecoder *dec[5*2]; + opus_int32 decsize; + OpusDecoder *decbak; + opus_uint32 dec_final_range1,dec_final_range2,dec_final_acc; + unsigned char *packet; + unsigned char modes[4096]; + short *outbuf_int; + short *outbuf; + + dec_final_range1=dec_final_range2=2; + + packet=malloc(sizeof(unsigned char)*MAX_PACKET); + if(packet==NULL)test_failed(); + + outbuf_int=malloc(sizeof(short)*(MAX_FRAME_SAMP+16)*2); + for(i=0;i<(MAX_FRAME_SAMP+16)*2;i++)outbuf_int[i]=32749; + outbuf=&outbuf_int[8*2]; + + fprintf(stdout," Starting %d decoders...\n",5*2); + for(t=0;t<5*2;t++) + { + int fs=fsv[t>>1]; + int c=(t&1)+1; + err=OPUS_INTERNAL_ERROR; + dec[t] = opus_decoder_create(fs, c, &err); + if(err!=OPUS_OK || dec[t]==NULL)test_failed(); + fprintf(stdout," opus_decoder_create(%5d,%d) OK. Copy ",fs,c); + { + OpusDecoder *dec2; + /*The opus state structures contain no pointers and can be freely copied*/ + dec2=(OpusDecoder *)malloc(opus_decoder_get_size(c)); + if(dec2==NULL)test_failed(); + memcpy(dec2,dec[t],opus_decoder_get_size(c)); + memset(dec[t],255,opus_decoder_get_size(c)); + opus_decoder_destroy(dec[t]); + printf("OK.\n"); + dec[t]=dec2; + } + } + + decsize=opus_decoder_get_size(1); + decbak=(OpusDecoder *)malloc(decsize); + if(decbak==NULL)test_failed(); + + for(t=0;t<5*2;t++) + { + int factor=48000/fsv[t>>1]; + for(fec=0;fec<2;fec++) + { + opus_int32 dur; + /*Test PLC on a fresh decoder*/ + out_samples = opus_decode(dec[t], 0, 0, outbuf, 120/factor, fec); + if(out_samples!=120/factor)test_failed(); + if(opus_decoder_ctl(dec[t], OPUS_GET_LAST_PACKET_DURATION(&dur))!=OPUS_OK)test_failed(); + if(dur!=120/factor)test_failed(); + + /*Test on a size which isn't a multiple of 2.5ms*/ + out_samples = opus_decode(dec[t], 0, 0, outbuf, 120/factor+2, fec); + if(out_samples!=OPUS_BAD_ARG)test_failed(); + + /*Test null pointer input*/ + out_samples = opus_decode(dec[t], 0, -1, outbuf, 120/factor, fec); + if(out_samples!=120/factor)test_failed(); + out_samples = opus_decode(dec[t], 0, 1, outbuf, 120/factor, fec); + if(out_samples!=120/factor)test_failed(); + out_samples = opus_decode(dec[t], 0, 10, outbuf, 120/factor, fec); + if(out_samples!=120/factor)test_failed(); + out_samples = opus_decode(dec[t], 0, fast_rand(), outbuf, 120/factor, fec); + if(out_samples!=120/factor)test_failed(); + if(opus_decoder_ctl(dec[t], OPUS_GET_LAST_PACKET_DURATION(&dur))!=OPUS_OK)test_failed(); + if(dur!=120/factor)test_failed(); + + /*Zero lengths*/ + out_samples = opus_decode(dec[t], packet, 0, outbuf, 120/factor, fec); + if(out_samples!=120/factor)test_failed(); + + /*Zero buffer*/ + outbuf[0]=32749; + out_samples = opus_decode(dec[t], packet, 0, outbuf, 0, fec); + if(out_samples>0)test_failed(); +#if !defined(OPUS_BUILD) && (OPUS_GNUC_PREREQ(4, 6) || (defined(__clang_major__) && __clang_major__ >= 3)) +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wnonnull" +#endif + out_samples = opus_decode(dec[t], packet, 0, 0, 0, fec); +#if !defined(OPUS_BUILD) && (OPUS_GNUC_PREREQ(4, 6) || (defined(__clang_major__) && __clang_major__ >= 3)) +#pragma GCC diagnostic pop +#endif + if(out_samples>0)test_failed(); + if(outbuf[0]!=32749)test_failed(); + + /*Invalid lengths*/ + out_samples = opus_decode(dec[t], packet, -1, outbuf, MAX_FRAME_SAMP, fec); + if(out_samples>=0)test_failed(); + out_samples = opus_decode(dec[t], packet, INT_MIN, outbuf, MAX_FRAME_SAMP, fec); + if(out_samples>=0)test_failed(); + out_samples = opus_decode(dec[t], packet, -1, outbuf, -1, fec); + if(out_samples>=0)test_failed(); + + /*Crazy FEC values*/ + out_samples = opus_decode(dec[t], packet, 1, outbuf, MAX_FRAME_SAMP, fec?-1:2); + if(out_samples>=0)test_failed(); + + /*Reset the decoder*/ + if(opus_decoder_ctl(dec[t], OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + } + } + fprintf(stdout," dec[all] initial frame PLC OK.\n"); + + /*Count code 0 tests*/ + for(i=0;i<64;i++) + { + opus_int32 dur; + int j,expected[5*2]; + packet[0]=i<<2; + packet[1]=255; + packet[2]=255; + err=opus_packet_get_nb_channels(packet); + if(err!=(i&1)+1)test_failed(); + + for(t=0;t<5*2;t++){ + expected[t]=opus_decoder_get_nb_samples(dec[t],packet,1); + if(expected[t]>2880)test_failed(); + } + + for(j=0;j<256;j++) + { + packet[1]=j; + for(t=0;t<5*2;t++) + { + out_samples = opus_decode(dec[t], packet, 3, outbuf, MAX_FRAME_SAMP, 0); + if(out_samples!=expected[t])test_failed(); + if(opus_decoder_ctl(dec[t], OPUS_GET_LAST_PACKET_DURATION(&dur))!=OPUS_OK)test_failed(); + if(dur!=out_samples)test_failed(); + opus_decoder_ctl(dec[t], OPUS_GET_FINAL_RANGE(&dec_final_range1)); + if(t==0)dec_final_range2=dec_final_range1; + else if(dec_final_range1!=dec_final_range2)test_failed(); + } + } + + for(t=0;t<5*2;t++){ + int factor=48000/fsv[t>>1]; + /* The PLC is run for 6 frames in order to get better PLC coverage. */ + for(j=0;j<6;j++) + { + out_samples = opus_decode(dec[t], 0, 0, outbuf, expected[t], 0); + if(out_samples!=expected[t])test_failed(); + if(opus_decoder_ctl(dec[t], OPUS_GET_LAST_PACKET_DURATION(&dur))!=OPUS_OK)test_failed(); + if(dur!=out_samples)test_failed(); + } + /* Run the PLC once at 2.5ms, as a simulation of someone trying to + do small drift corrections. */ + if(expected[t]!=120/factor) + { + out_samples = opus_decode(dec[t], 0, 0, outbuf, 120/factor, 0); + if(out_samples!=120/factor)test_failed(); + if(opus_decoder_ctl(dec[t], OPUS_GET_LAST_PACKET_DURATION(&dur))!=OPUS_OK)test_failed(); + if(dur!=out_samples)test_failed(); + } + out_samples = opus_decode(dec[t], packet, 2, outbuf, expected[t]-1, 0); + if(out_samples>0)test_failed(); + } + } + fprintf(stdout," dec[all] all 2-byte prefix for length 3 and PLC, all modes (64) OK.\n"); + + if(no_fuzz) + { + fprintf(stdout," Skipping many tests which fuzz the decoder as requested.\n"); + free(decbak); + for(t=0;t<5*2;t++)opus_decoder_destroy(dec[t]); + printf(" Decoders stopped.\n"); + + err=0; + for(i=0;i<8*2;i++)err|=outbuf_int[i]!=32749; + for(i=MAX_FRAME_SAMP*2;i<(MAX_FRAME_SAMP+8)*2;i++)err|=outbuf[i]!=32749; + if(err)test_failed(); + + free(outbuf_int); + free(packet); + return 0; + } + + { + /*We only test a subset of the modes here simply because the longer + durations end up taking a long time.*/ + static const int cmodes[4]={16,20,24,28}; + static const opus_uint32 cres[4]={116290185,2172123586u,2172123586u,2172123586u}; + static const opus_uint32 lres[3]={3285687739u,1481572662,694350475}; + static const int lmodes[3]={0,4,8}; + int mode=fast_rand()%4; + + packet[0]=cmodes[mode]<<3; + dec_final_acc=0; + t=fast_rand()%10; + + for(i=0;i<65536;i++) + { + int factor=48000/fsv[t>>1]; + packet[1]=i>>8; + packet[2]=i&255; + packet[3]=255; + out_samples = opus_decode(dec[t], packet, 4, outbuf, MAX_FRAME_SAMP, 0); + if(out_samples!=120/factor)test_failed(); + opus_decoder_ctl(dec[t], OPUS_GET_FINAL_RANGE(&dec_final_range1)); + dec_final_acc+=dec_final_range1; + } + if(dec_final_acc!=cres[mode])test_failed(); + fprintf(stdout," dec[%3d] all 3-byte prefix for length 4, mode %2d OK.\n",t,cmodes[mode]); + + mode=fast_rand()%3; + packet[0]=lmodes[mode]<<3; + dec_final_acc=0; + t=fast_rand()%10; + for(i=0;i<65536;i++) + { + int factor=48000/fsv[t>>1]; + packet[1]=i>>8; + packet[2]=i&255; + packet[3]=255; + out_samples = opus_decode(dec[t], packet, 4, outbuf, MAX_FRAME_SAMP, 0); + if(out_samples!=480/factor)test_failed(); + opus_decoder_ctl(dec[t], OPUS_GET_FINAL_RANGE(&dec_final_range1)); + dec_final_acc+=dec_final_range1; + } + if(dec_final_acc!=lres[mode])test_failed(); + fprintf(stdout," dec[%3d] all 3-byte prefix for length 4, mode %2d OK.\n",t,lmodes[mode]); + } + + skip=fast_rand()%7; + for(i=0;i<64;i++) + { + int j,expected[5*2]; + packet[0]=i<<2; + for(t=0;t<5*2;t++)expected[t]=opus_decoder_get_nb_samples(dec[t],packet,1); + for(j=2+skip;j<1275;j+=4) + { + int jj; + for(jj=0;jj1.f)test_failed(); + if(x[j]<-1.f)test_failed(); + } + } + for(i=1;i<9;i++) + { + for (j=0;j<1024;j++) + { + x[j]=(j&255)*(1/32.f)-4.f; + } + opus_pcm_soft_clip(x,1024/i,i,s); + for (j=0;j<(1024/i)*i;j++) + { + if(x[j]>1.f)test_failed(); + if(x[j]<-1.f)test_failed(); + } + } + opus_pcm_soft_clip(x,0,1,s); + opus_pcm_soft_clip(x,1,0,s); + opus_pcm_soft_clip(x,1,1,0); + opus_pcm_soft_clip(x,1,-1,s); + opus_pcm_soft_clip(x,-1,1,s); + opus_pcm_soft_clip(0,1,1,s); + printf("OK.\n"); +} +#endif + +int main(int _argc, char **_argv) +{ + const char * oversion; + const char * env_seed; + int env_used; + + if(_argc>2) + { + fprintf(stderr,"Usage: %s []\n",_argv[0]); + return 1; + } + + env_used=0; + env_seed=getenv("SEED"); + if(_argc>1)iseed=atoi(_argv[1]); + else if(env_seed) + { + iseed=atoi(env_seed); + env_used=1; + } + else iseed=(opus_uint32)time(NULL)^(((opus_uint32)getpid()&65535)<<16); + Rw=Rz=iseed; + + oversion=opus_get_version_string(); + if(!oversion)test_failed(); + fprintf(stderr,"Testing %s decoder. Random seed: %u (%.4X)\n", oversion, iseed, fast_rand() % 65535); + if(env_used)fprintf(stderr," Random seed set from the environment (SEED=%s).\n", env_seed); + + /*Setting TEST_OPUS_NOFUZZ tells the tool not to send garbage data + into the decoders. This is helpful because garbage data + may cause the decoders to clip, which angers CLANG IOC.*/ + test_decoder_code0(getenv("TEST_OPUS_NOFUZZ")!=NULL); +#ifndef DISABLE_FLOAT_API + test_soft_clip(); +#endif + + return 0; +} diff --git a/lib/opus/tests/test_opus_encode.c b/lib/opus/tests/test_opus_encode.c new file mode 100644 index 00000000..d6e8e2d3 --- /dev/null +++ b/lib/opus/tests/test_opus_encode.c @@ -0,0 +1,707 @@ +/* Copyright (c) 2011-2013 Xiph.Org Foundation + Written by Gregory Maxwell */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include +#include +#include +#include +#if (!defined WIN32 && !defined _WIN32) || defined(__MINGW32__) +#include +#else +#include +#define getpid _getpid +#endif +#include "opus_multistream.h" +#include "opus.h" +#include "../src/opus_private.h" +#include "test_opus_common.h" + +#define MAX_PACKET (1500) +#define SAMPLES (48000*30) +#define SSAMPLES (SAMPLES/3) +#define MAX_FRAME_SAMP (5760) +#define PI (3.141592653589793238462643f) +#define RAND_SAMPLE(a) (a[fast_rand() % sizeof(a)/sizeof(a[0])]) + +void generate_music(short *buf, opus_int32 len) +{ + opus_int32 a1,b1,a2,b2; + opus_int32 c1,c2,d1,d2; + opus_int32 i,j; + a1=b1=a2=b2=0; + c1=c2=d1=d2=0; + j=0; + /*60ms silence*/ + for(i=0;i<2880;i++)buf[i*2]=buf[i*2+1]=0; + for(i=2880;i>12)^((j>>10|j>>12)&26&j>>7)))&128)+128)<<15; + r=fast_rand();v1+=r&65535;v1-=r>>16; + r=fast_rand();v2+=r&65535;v2-=r>>16; + b1=v1-a1+((b1*61+32)>>6);a1=v1; + b2=v2-a2+((b2*61+32)>>6);a2=v2; + c1=(30*(c1+b1+d1)+32)>>6;d1=b1; + c2=(30*(c2+b2+d2)+32)>>6;d2=b2; + v1=(c1+128)>>8; + v2=(c2+128)>>8; + buf[i*2]=v1>32767?32767:(v1<-32768?-32768:v1); + buf[i*2+1]=v2>32767?32767:(v2<-32768?-32768:v2); + if(i%6==0)j++; + } +} + +#if 0 +static int save_ctr = 0; +static void int_to_char(opus_uint32 i, unsigned char ch[4]) +{ + ch[0] = i>>24; + ch[1] = (i>>16)&0xFF; + ch[2] = (i>>8)&0xFF; + ch[3] = i&0xFF; +} + +static OPUS_INLINE void save_packet(unsigned char* p, int len, opus_uint32 rng) +{ + FILE *fout; + unsigned char int_field[4]; + char name[256]; + snprintf(name,255,"test_opus_encode.%llu.%d.bit",(unsigned long long)iseed,save_ctr); + fprintf(stdout,"writing %d byte packet to %s\n",len,name); + fout=fopen(name, "wb+"); + if(fout==NULL)test_failed(); + int_to_char(len, int_field); + fwrite(int_field, 1, 4, fout); + int_to_char(rng, int_field); + fwrite(int_field, 1, 4, fout); + fwrite(p, 1, len, fout); + fclose(fout); + save_ctr++; +} +#endif + +int get_frame_size_enum(int frame_size, int sampling_rate) +{ + int frame_size_enum; + + if(frame_size==sampling_rate/400) + frame_size_enum = OPUS_FRAMESIZE_2_5_MS; + else if(frame_size==sampling_rate/200) + frame_size_enum = OPUS_FRAMESIZE_5_MS; + else if(frame_size==sampling_rate/100) + frame_size_enum = OPUS_FRAMESIZE_10_MS; + else if(frame_size==sampling_rate/50) + frame_size_enum = OPUS_FRAMESIZE_20_MS; + else if(frame_size==sampling_rate/25) + frame_size_enum = OPUS_FRAMESIZE_40_MS; + else if(frame_size==3*sampling_rate/50) + frame_size_enum = OPUS_FRAMESIZE_60_MS; + else if(frame_size==4*sampling_rate/50) + frame_size_enum = OPUS_FRAMESIZE_80_MS; + else if(frame_size==5*sampling_rate/50) + frame_size_enum = OPUS_FRAMESIZE_100_MS; + else if(frame_size==6*sampling_rate/50) + frame_size_enum = OPUS_FRAMESIZE_120_MS; + else + test_failed(); + + return frame_size_enum; +} + +int test_encode(OpusEncoder *enc, int channels, int frame_size, OpusDecoder *dec) +{ + int samp_count = 0; + opus_int16 *inbuf; + unsigned char packet[MAX_PACKET+257]; + int len; + opus_int16 *outbuf; + int out_samples; + int ret = 0; + + /* Generate input data */ + inbuf = (opus_int16*)malloc(sizeof(*inbuf)*SSAMPLES); + generate_music(inbuf, SSAMPLES/2); + + /* Allocate memory for output data */ + outbuf = (opus_int16*)malloc(sizeof(*outbuf)*MAX_FRAME_SAMP*3); + + /* Encode data, then decode for sanity check */ + do { + len = opus_encode(enc, &inbuf[samp_count*channels], frame_size, packet, MAX_PACKET); + if(len<0 || len>MAX_PACKET) { + fprintf(stderr,"opus_encode() returned %d\n",len); + ret = -1; + break; + } + + out_samples = opus_decode(dec, packet, len, outbuf, MAX_FRAME_SAMP, 0); + if(out_samples!=frame_size) { + fprintf(stderr,"opus_decode() returned %d\n",out_samples); + ret = -1; + break; + } + + samp_count += frame_size; + } while (samp_count < ((SSAMPLES/2)-MAX_FRAME_SAMP)); + + /* Clean up */ + free(inbuf); + free(outbuf); + return ret; +} + +void fuzz_encoder_settings(const int num_encoders, const int num_setting_changes) +{ + OpusEncoder *enc; + OpusDecoder *dec; + int i,j,err; + + /* Parameters to fuzz. Some values are duplicated to increase their probability of being tested. */ + int sampling_rates[5] = {8000, 12000, 16000, 24000, 48000}; + int channels[2] = {1, 2}; + int applications[3] = {OPUS_APPLICATION_AUDIO, OPUS_APPLICATION_VOIP, OPUS_APPLICATION_RESTRICTED_LOWDELAY}; + int bitrates[11] = {6000, 12000, 16000, 24000, 32000, 48000, 64000, 96000, 510000, OPUS_AUTO, OPUS_BITRATE_MAX}; + int force_channels[4] = {OPUS_AUTO, OPUS_AUTO, 1, 2}; + int use_vbr[3] = {0, 1, 1}; + int vbr_constraints[3] = {0, 1, 1}; + int complexities[11] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}; + int max_bandwidths[6] = {OPUS_BANDWIDTH_NARROWBAND, OPUS_BANDWIDTH_MEDIUMBAND, + OPUS_BANDWIDTH_WIDEBAND, OPUS_BANDWIDTH_SUPERWIDEBAND, + OPUS_BANDWIDTH_FULLBAND, OPUS_BANDWIDTH_FULLBAND}; + int signals[4] = {OPUS_AUTO, OPUS_AUTO, OPUS_SIGNAL_VOICE, OPUS_SIGNAL_MUSIC}; + int inband_fecs[3] = {0, 0, 1}; + int packet_loss_perc[4] = {0, 1, 2, 5}; + int lsb_depths[2] = {8, 24}; + int prediction_disabled[3] = {0, 0, 1}; + int use_dtx[2] = {0, 1}; + int frame_sizes_ms_x2[9] = {5, 10, 20, 40, 80, 120, 160, 200, 240}; /* x2 to avoid 2.5 ms */ + + for (i=0; i=64000?2:1)))!=OPUS_OK)test_failed(); + if(opus_encoder_ctl(enc, OPUS_SET_COMPLEXITY((count>>2)%11))!=OPUS_OK)test_failed(); + if(opus_encoder_ctl(enc, OPUS_SET_PACKET_LOSS_PERC((fast_rand()&15)&(fast_rand()%15)))!=OPUS_OK)test_failed(); + bw=modes[j]==0?OPUS_BANDWIDTH_NARROWBAND+(fast_rand()%3): + modes[j]==1?OPUS_BANDWIDTH_SUPERWIDEBAND+(fast_rand()&1): + OPUS_BANDWIDTH_NARROWBAND+(fast_rand()%5); + if(modes[j]==2&&bw==OPUS_BANDWIDTH_MEDIUMBAND)bw+=3; + if(opus_encoder_ctl(enc, OPUS_SET_BANDWIDTH(bw))!=OPUS_OK)test_failed(); + len = opus_encode(enc, &inbuf[i<<1], frame_size, packet, MAX_PACKET); + if(len<0 || len>MAX_PACKET)test_failed(); + if(opus_encoder_ctl(enc, OPUS_GET_FINAL_RANGE(&enc_final_range))!=OPUS_OK)test_failed(); + if((fast_rand()&3)==0) + { + if(opus_packet_pad(packet,len,len+1)!=OPUS_OK)test_failed(); + len++; + } + if((fast_rand()&7)==0) + { + if(opus_packet_pad(packet,len,len+256)!=OPUS_OK)test_failed(); + len+=256; + } + if((fast_rand()&3)==0) + { + len=opus_packet_unpad(packet,len); + if(len<1)test_failed(); + } + out_samples = opus_decode(dec, packet, len, &outbuf[i<<1], MAX_FRAME_SAMP, 0); + if(out_samples!=frame_size)test_failed(); + if(opus_decoder_ctl(dec, OPUS_GET_FINAL_RANGE(&dec_final_range))!=OPUS_OK)test_failed(); + if(enc_final_range!=dec_final_range)test_failed(); + /*LBRR decode*/ + out_samples = opus_decode(dec_err[0], packet, len, out2buf, frame_size, (fast_rand()&3)!=0); + if(out_samples!=frame_size)test_failed(); + out_samples = opus_decode(dec_err[1], packet, (fast_rand()&3)==0?0:len, out2buf, MAX_FRAME_SAMP, (fast_rand()&7)!=0); + if(out_samples<120)test_failed(); + i+=frame_size; + count++; + }while(i<(SSAMPLES-MAX_FRAME_SAMP)); + fprintf(stdout," Mode %s FB encode %s, %6d bps OK.\n",mstrings[modes[j]],rc==0?" VBR":rc==1?"CVBR":" CBR",rate); + fflush(stdout); + } + } + + if(opus_encoder_ctl(enc, OPUS_SET_FORCE_MODE(OPUS_AUTO))!=OPUS_OK)test_failed(); + if(opus_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(OPUS_AUTO))!=OPUS_OK)test_failed(); + if(opus_encoder_ctl(enc, OPUS_SET_INBAND_FEC(0))!=OPUS_OK)test_failed(); + if(opus_encoder_ctl(enc, OPUS_SET_DTX(0))!=OPUS_OK)test_failed(); + + for(rc=0;rc<3;rc++) + { + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_VBR(rc<2))!=OPUS_OK)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_VBR_CONSTRAINT(rc==1))!=OPUS_OK)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_VBR_CONSTRAINT(rc==1))!=OPUS_OK)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_INBAND_FEC(rc==0))!=OPUS_OK)test_failed(); + for(j=0;j<16;j++) + { + int rate; + int modes[16]={0,0,0,0,0,0,0,0,2,2,2,2,2,2,2,2}; + int rates[16]={4000,12000,32000,8000,16000,32000,48000,88000,4000,12000,32000,8000,16000,32000,48000,88000}; + int frame[16]={160*1,160,80,160,160,80,40,20,160*1,160,80,160,160,80,40,20}; + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_INBAND_FEC(rc==0&&j==1))!=OPUS_OK)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_FORCE_MODE(MODE_SILK_ONLY+modes[j]))!=OPUS_OK)test_failed(); + rate=rates[j]+fast_rand()%rates[j]; + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_DTX(fast_rand()&1))!=OPUS_OK)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_BITRATE(rate))!=OPUS_OK)test_failed(); + count=i=0; + do { + int len,out_samples,frame_size,loss; + opus_int32 pred; + if(opus_multistream_encoder_ctl(MSenc, OPUS_GET_PREDICTION_DISABLED(&pred))!=OPUS_OK)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_PREDICTION_DISABLED((int)(fast_rand()&15)<(pred?11:4)))!=OPUS_OK)test_failed(); + frame_size=frame[j]; + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_COMPLEXITY((count>>2)%11))!=OPUS_OK)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_SET_PACKET_LOSS_PERC((fast_rand()&15)&(fast_rand()%15)))!=OPUS_OK)test_failed(); + if((fast_rand()&255)==0) + { + if(opus_multistream_encoder_ctl(MSenc, OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + if(opus_multistream_decoder_ctl(MSdec, OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + if((fast_rand()&3)!=0) + { + if(opus_multistream_decoder_ctl(MSdec_err, OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + } + } + if((fast_rand()&255)==0) + { + if(opus_multistream_decoder_ctl(MSdec_err, OPUS_RESET_STATE)!=OPUS_OK)test_failed(); + } + len = opus_multistream_encode(MSenc, &inbuf[i<<1], frame_size, packet, MAX_PACKET); + if(len<0 || len>MAX_PACKET)test_failed(); + if(opus_multistream_encoder_ctl(MSenc, OPUS_GET_FINAL_RANGE(&enc_final_range))!=OPUS_OK)test_failed(); + if((fast_rand()&3)==0) + { + if(opus_multistream_packet_pad(packet,len,len+1,2)!=OPUS_OK)test_failed(); + len++; + } + if((fast_rand()&7)==0) + { + if(opus_multistream_packet_pad(packet,len,len+256,2)!=OPUS_OK)test_failed(); + len+=256; + } + if((fast_rand()&3)==0) + { + len=opus_multistream_packet_unpad(packet,len,2); + if(len<1)test_failed(); + } + out_samples = opus_multistream_decode(MSdec, packet, len, out2buf, MAX_FRAME_SAMP, 0); + if(out_samples!=frame_size*6)test_failed(); + if(opus_multistream_decoder_ctl(MSdec, OPUS_GET_FINAL_RANGE(&dec_final_range))!=OPUS_OK)test_failed(); + if(enc_final_range!=dec_final_range)test_failed(); + /*LBRR decode*/ + loss=(fast_rand()&63)==0; + out_samples = opus_multistream_decode(MSdec_err, packet, loss?0:len, out2buf, frame_size*6, (fast_rand()&3)!=0); + if(out_samples!=(frame_size*6))test_failed(); + i+=frame_size; + count++; + }while(i<(SSAMPLES/12-MAX_FRAME_SAMP)); + fprintf(stdout," Mode %s NB dual-mono MS encode %s, %6d bps OK.\n",mstrings[modes[j]],rc==0?" VBR":rc==1?"CVBR":" CBR",rate); + fflush(stdout); + } + } + + bitrate_bps=512000; + fsize=fast_rand()%31; + fswitch=100; + + debruijn2(6,db62); + count=i=0; + do { + unsigned char toc; + const unsigned char *frames[48]; + short size[48]; + int payload_offset; + opus_uint32 dec_final_range2; + int jj,dec2; + int len,out_samples; + int frame_size=fsizes[db62[fsize]]; + opus_int32 offset=i%(SAMPLES-MAX_FRAME_SAMP); + + opus_encoder_ctl(enc, OPUS_SET_BITRATE(bitrate_bps)); + + len = opus_encode(enc, &inbuf[offset<<1], frame_size, packet, MAX_PACKET); + if(len<0 || len>MAX_PACKET)test_failed(); + count++; + + opus_encoder_ctl(enc, OPUS_GET_FINAL_RANGE(&enc_final_range)); + + out_samples = opus_decode(dec, packet, len, &outbuf[offset<<1], MAX_FRAME_SAMP, 0); + if(out_samples!=frame_size)test_failed(); + + opus_decoder_ctl(dec, OPUS_GET_FINAL_RANGE(&dec_final_range)); + + /* compare final range encoder rng values of encoder and decoder */ + if(dec_final_range!=enc_final_range)test_failed(); + + /* We fuzz the packet, but take care not to only corrupt the payload + Corrupted headers are tested elsewhere and we need to actually run + the decoders in order to compare them. */ + if(opus_packet_parse(packet,len,&toc,frames,size,&payload_offset)<=0)test_failed(); + if((fast_rand()&1023)==0)len=0; + for(j=(opus_int32)(frames[0]-packet);j0?packet:NULL, len, out2buf, MAX_FRAME_SAMP, 0); + if(out_samples<0||out_samples>MAX_FRAME_SAMP)test_failed(); + if((len>0&&out_samples!=frame_size))test_failed(); /*FIXME use lastframe*/ + + opus_decoder_ctl(dec_err[0], OPUS_GET_FINAL_RANGE(&dec_final_range)); + + /*randomly select one of the decoders to compare with*/ + dec2=fast_rand()%9+1; + out_samples = opus_decode(dec_err[dec2], len>0?packet:NULL, len, out2buf, MAX_FRAME_SAMP, 0); + if(out_samples<0||out_samples>MAX_FRAME_SAMP)test_failed(); /*FIXME, use factor, lastframe for loss*/ + + opus_decoder_ctl(dec_err[dec2], OPUS_GET_FINAL_RANGE(&dec_final_range2)); + if(len>0&&dec_final_range!=dec_final_range2)test_failed(); + + fswitch--; + if(fswitch<1) + { + int new_size; + fsize=(fsize+1)%36; + new_size=fsizes[db62[fsize]]; + if(new_size==960||new_size==480)fswitch=2880/new_size*(fast_rand()%19+1); + else fswitch=(fast_rand()%(2880/new_size))+1; + } + bitrate_bps=((fast_rand()%508000+4000)+bitrate_bps)>>1; + i+=frame_size; + }while(i] [-fuzz ]\n",_argv[0]); +} + +int main(int _argc, char **_argv) +{ + int args=1; + char * strtol_str=NULL; + const char * oversion; + const char * env_seed; + int env_used; + int num_encoders_to_fuzz=5; + int num_setting_changes=40; + + env_used=0; + env_seed=getenv("SEED"); + if(_argc>1) + iseed=strtol(_argv[1], &strtol_str, 10); /* the first input argument might be the seed */ + if(strtol_str!=NULL && strtol_str[0]=='\0') /* iseed is a valid number */ + args++; + else if(env_seed) { + iseed=atoi(env_seed); + env_used=1; + } + else iseed=(opus_uint32)time(NULL)^(((opus_uint32)getpid()&65535)<<16); + Rw=Rz=iseed; + + while(args<_argc) + { + if(strcmp(_argv[args], "-fuzz")==0 && _argc==(args+3)) { + num_encoders_to_fuzz=strtol(_argv[args+1], &strtol_str, 10); + if(strtol_str[0]!='\0' || num_encoders_to_fuzz<=0) { + print_usage(_argv); + return EXIT_FAILURE; + } + num_setting_changes=strtol(_argv[args+2], &strtol_str, 10); + if(strtol_str[0]!='\0' || num_setting_changes<=0) { + print_usage(_argv); + return EXIT_FAILURE; + } + args+=3; + } + else { + print_usage(_argv); + return EXIT_FAILURE; + } + } + + oversion=opus_get_version_string(); + if(!oversion)test_failed(); + fprintf(stderr,"Testing %s encoder. Random seed: %u (%.4X)\n", oversion, iseed, fast_rand() % 65535); + if(env_used)fprintf(stderr," Random seed set from the environment (SEED=%s).\n", env_seed); + + regression_test(); + + /*Setting TEST_OPUS_NOFUZZ tells the tool not to send garbage data + into the decoders. This is helpful because garbage data + may cause the decoders to clip, which angers CLANG IOC.*/ + run_test1(getenv("TEST_OPUS_NOFUZZ")!=NULL); + + /* Fuzz encoder settings online */ + if(getenv("TEST_OPUS_NOFUZZ")==NULL) { + fprintf(stderr,"Running fuzz_encoder_settings with %d encoder(s) and %d setting change(s) each.\n", + num_encoders_to_fuzz, num_setting_changes); + fuzz_encoder_settings(num_encoders_to_fuzz, num_setting_changes); + } + + fprintf(stderr,"Tests completed successfully.\n"); + + return 0; +} diff --git a/lib/opus/tests/test_opus_padding.c b/lib/opus/tests/test_opus_padding.c new file mode 100644 index 00000000..c9ef7375 --- /dev/null +++ b/lib/opus/tests/test_opus_padding.c @@ -0,0 +1,90 @@ +/* Copyright (c) 2012 Xiph.Org Foundation + Written by Jüri Aedla and Ralph Giles */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +/* Check for overflow in reading the padding length. + * http://lists.xiph.org/pipermail/opus/2012-November/001834.html + */ + +#include +#include +#include +#include "opus.h" +#include "test_opus_common.h" + +#define PACKETSIZE 16909318 +#define CHANNELS 2 +#define FRAMESIZE 5760 + +void test_overflow(void) +{ + OpusDecoder *decoder; + int result; + int error; + + unsigned char *in = malloc(PACKETSIZE); + opus_int16 *out = malloc(FRAMESIZE*CHANNELS*sizeof(*out)); + + fprintf(stderr, " Checking for padding overflow... "); + if (!in || !out) { + fprintf(stderr, "FAIL (out of memory)\n"); + test_failed(); + } + in[0] = 0xff; + in[1] = 0x41; + memset(in + 2, 0xff, PACKETSIZE - 3); + in[PACKETSIZE-1] = 0x0b; + + decoder = opus_decoder_create(48000, CHANNELS, &error); + result = opus_decode(decoder, in, PACKETSIZE, out, FRAMESIZE, 0); + opus_decoder_destroy(decoder); + + free(in); + free(out); + + if (result != OPUS_INVALID_PACKET) { + fprintf(stderr, "FAIL!\n"); + test_failed(); + } + + fprintf(stderr, "OK.\n"); +} + +int main(void) +{ + const char *oversion; + + iseed = 0; + oversion = opus_get_version_string(); + if (!oversion) test_failed(); + fprintf(stderr, "Testing %s padding.\n", oversion); + + test_overflow(); + + fprintf(stderr, "All padding tests passed.\n"); + + return 0; +} diff --git a/lib/opus/tests/test_opus_projection.c b/lib/opus/tests/test_opus_projection.c new file mode 100644 index 00000000..4e06613e --- /dev/null +++ b/lib/opus/tests/test_opus_projection.c @@ -0,0 +1,393 @@ +/* Copyright (c) 2017 Google Inc. + Written by Andrew Allen */ +/* + Redistribution and use in source and binary forms, with or without + modification, are permitted provided that the following conditions + are met: + + - Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + + - Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + + THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER + OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, + EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, + PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR + PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF + LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING + NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS + SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +*/ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include +#include "float_cast.h" +#include "opus.h" +#include "test_opus_common.h" +#include "opus_projection.h" +#include "mathops.h" +#include "../src/mapping_matrix.h" +#include "mathops.h" + +#define BUFFER_SIZE 960 +#define MAX_DATA_BYTES 32768 +#define MAX_FRAME_SAMPLES 5760 +#define ERROR_TOLERANCE 1 + +#define SIMPLE_MATRIX_SIZE 12 +#define SIMPLE_MATRIX_FRAME_SIZE 10 +#define SIMPLE_MATRIX_INPUT_SIZE 30 +#define SIMPLE_MATRIX_OUTPUT_SIZE 40 + +int assert_is_equal( + const opus_val16 *a, const opus_int16 *b, int size, opus_int16 tolerance) +{ + int i; + for (i = 0; i < size; i++) + { +#ifdef FIXED_POINT + opus_int16 val = a[i]; +#else + opus_int16 val = FLOAT2INT16(a[i]); +#endif + if (abs(val - b[i]) > tolerance) + return 1; + } + return 0; +} + +int assert_is_equal_short( + const opus_int16 *a, const opus_int16 *b, int size, opus_int16 tolerance) +{ + int i; + for (i = 0; i < size; i++) + if (abs(a[i] - b[i]) > tolerance) + return 1; + return 0; +} + +void test_simple_matrix(void) +{ + const MappingMatrix simple_matrix_params = {4, 3, 0}; + const opus_int16 simple_matrix_data[SIMPLE_MATRIX_SIZE] = {0, 32767, 0, 0, 32767, 0, 0, 0, 0, 0, 0, 32767}; + const opus_int16 input_int16[SIMPLE_MATRIX_INPUT_SIZE] = { + 32767, 0, -32768, 29491, -3277, -29491, 26214, -6554, -26214, 22938, -9830, + -22938, 19661, -13107, -19661, 16384, -16384, -16384, 13107, -19661, -13107, + 9830, -22938, -9830, 6554, -26214, -6554, 3277, -29491, -3277}; + const opus_int16 expected_output_int16[SIMPLE_MATRIX_OUTPUT_SIZE] = { + 0, 32767, 0, -32768, -3277, 29491, 0, -29491, -6554, 26214, 0, -26214, + -9830, 22938, 0, -22938, -13107, 19661, 0, -19661, -16384, 16384, 0, -16384, + -19661, 13107, 0, -13107, -22938, 9830, 0, -9830, -26214, 6554, 0, -6554, + -29491, 3277, 0, -3277}; + + int i, ret; + opus_int32 simple_matrix_size; + opus_val16 *input_val16; + opus_val16 *output_val16; + opus_int16 *output_int16; + MappingMatrix *simple_matrix; + + /* Allocate input/output buffers. */ + input_val16 = (opus_val16 *)opus_alloc(sizeof(opus_val16) * SIMPLE_MATRIX_INPUT_SIZE); + output_int16 = (opus_int16 *)opus_alloc(sizeof(opus_int16) * SIMPLE_MATRIX_OUTPUT_SIZE); + output_val16 = (opus_val16 *)opus_alloc(sizeof(opus_val16) * SIMPLE_MATRIX_OUTPUT_SIZE); + + /* Initialize matrix */ + simple_matrix_size = mapping_matrix_get_size(simple_matrix_params.rows, + simple_matrix_params.cols); + if (!simple_matrix_size) + test_failed(); + + simple_matrix = (MappingMatrix *)opus_alloc(simple_matrix_size); + mapping_matrix_init(simple_matrix, simple_matrix_params.rows, + simple_matrix_params.cols, simple_matrix_params.gain, simple_matrix_data, + sizeof(simple_matrix_data)); + + /* Copy inputs. */ + for (i = 0; i < SIMPLE_MATRIX_INPUT_SIZE; i++) + { +#ifdef FIXED_POINT + input_val16[i] = input_int16[i]; +#else + input_val16[i] = (1/32768.f)*input_int16[i]; +#endif + } + + /* _in_short */ + for (i = 0; i < SIMPLE_MATRIX_OUTPUT_SIZE; i++) + output_val16[i] = 0; + for (i = 0; i < simple_matrix->rows; i++) + { + mapping_matrix_multiply_channel_in_short(simple_matrix, + input_int16, simple_matrix->cols, &output_val16[i], i, + simple_matrix->rows, SIMPLE_MATRIX_FRAME_SIZE); + } + ret = assert_is_equal(output_val16, expected_output_int16, SIMPLE_MATRIX_OUTPUT_SIZE, ERROR_TOLERANCE); + if (ret) + test_failed(); + + /* _out_short */ + for (i = 0; i < SIMPLE_MATRIX_OUTPUT_SIZE; i++) + output_int16[i] = 0; + for (i = 0; i < simple_matrix->cols; i++) + { + mapping_matrix_multiply_channel_out_short(simple_matrix, + &input_val16[i], i, simple_matrix->cols, output_int16, + simple_matrix->rows, SIMPLE_MATRIX_FRAME_SIZE); + } + ret = assert_is_equal_short(output_int16, expected_output_int16, SIMPLE_MATRIX_OUTPUT_SIZE, ERROR_TOLERANCE); + if (ret) + test_failed(); + +#if !defined(DISABLE_FLOAT_API) && !defined(FIXED_POINT) + /* _in_float */ + for (i = 0; i < SIMPLE_MATRIX_OUTPUT_SIZE; i++) + output_val16[i] = 0; + for (i = 0; i < simple_matrix->rows; i++) + { + mapping_matrix_multiply_channel_in_float(simple_matrix, + input_val16, simple_matrix->cols, &output_val16[i], i, + simple_matrix->rows, SIMPLE_MATRIX_FRAME_SIZE); + } + ret = assert_is_equal(output_val16, expected_output_int16, SIMPLE_MATRIX_OUTPUT_SIZE, ERROR_TOLERANCE); + if (ret) + test_failed(); + + /* _out_float */ + for (i = 0; i < SIMPLE_MATRIX_OUTPUT_SIZE; i++) + output_val16[i] = 0; + for (i = 0; i < simple_matrix->cols; i++) + { + mapping_matrix_multiply_channel_out_float(simple_matrix, + &input_val16[i], i, simple_matrix->cols, output_val16, + simple_matrix->rows, SIMPLE_MATRIX_FRAME_SIZE); + } + ret = assert_is_equal(output_val16, expected_output_int16, SIMPLE_MATRIX_OUTPUT_SIZE, ERROR_TOLERANCE); + if (ret) + test_failed(); +#endif + + opus_free(input_val16); + opus_free(output_int16); + opus_free(output_val16); + opus_free(simple_matrix); +} + +void test_creation_arguments(const int channels, const int mapping_family) +{ + int streams; + int coupled_streams; + int enc_error; + int dec_error; + int ret; + OpusProjectionEncoder *st_enc = NULL; + OpusProjectionDecoder *st_dec = NULL; + + const opus_int32 Fs = 48000; + const int application = OPUS_APPLICATION_AUDIO; + + int order_plus_one = (int)floor(sqrt((float)channels)); + int nondiegetic_channels = channels - order_plus_one * order_plus_one; + + int is_channels_valid = 0; + int is_projection_valid = 0; + + st_enc = opus_projection_ambisonics_encoder_create(Fs, channels, + mapping_family, &streams, &coupled_streams, application, &enc_error); + if (st_enc != NULL) + { + opus_int32 matrix_size; + unsigned char *matrix; + + ret = opus_projection_encoder_ctl(st_enc, + OPUS_PROJECTION_GET_DEMIXING_MATRIX_SIZE_REQUEST, &matrix_size); + if (ret != OPUS_OK || !matrix_size) + test_failed(); + + matrix = (unsigned char *)opus_alloc(matrix_size); + ret = opus_projection_encoder_ctl(st_enc, + OPUS_PROJECTION_GET_DEMIXING_MATRIX_REQUEST, matrix, matrix_size); + + opus_projection_encoder_destroy(st_enc); + + st_dec = opus_projection_decoder_create(Fs, channels, streams, + coupled_streams, matrix, matrix_size, &dec_error); + if (st_dec != NULL) + { + opus_projection_decoder_destroy(st_dec); + } + opus_free(matrix); + } + + is_channels_valid = (order_plus_one >= 2 && order_plus_one <= 4) && + (nondiegetic_channels == 0 || nondiegetic_channels == 2); + is_projection_valid = (enc_error == OPUS_OK && dec_error == OPUS_OK); + if (is_channels_valid ^ is_projection_valid) + { + fprintf(stderr, "Channels: %d, Family: %d\n", channels, mapping_family); + fprintf(stderr, "Order+1: %d, Non-diegetic Channels: %d\n", + order_plus_one, nondiegetic_channels); + fprintf(stderr, "Streams: %d, Coupled Streams: %d\n", + streams, coupled_streams); + test_failed(); + } +} + +void generate_music(short *buf, opus_int32 len, opus_int32 channels) +{ + opus_int32 i,j,k; + opus_int32 *a,*b,*c,*d; + a = (opus_int32 *)malloc(sizeof(opus_int32) * channels); + b = (opus_int32 *)malloc(sizeof(opus_int32) * channels); + c = (opus_int32 *)malloc(sizeof(opus_int32) * channels); + d = (opus_int32 *)malloc(sizeof(opus_int32) * channels); + memset(a, 0, sizeof(opus_int32) * channels); + memset(b, 0, sizeof(opus_int32) * channels); + memset(c, 0, sizeof(opus_int32) * channels); + memset(d, 0, sizeof(opus_int32) * channels); + j=0; + + for(i=0;i>12)^((j>>10|j>>12)&26&j>>7)))&128)+128)<<15; + r=fast_rand();v+=r&65535;v-=r>>16; + b[k]=v-a[k]+((b[k]*61+32)>>6);a[k]=v; + c[k]=(30*(c[k]+b[k]+d[k])+32)>>6;d[k]=b[k]; + v=(c[k]+128)>>8; + buf[i*channels+k]=v>32767?32767:(v<-32768?-32768:v); + if(i%6==0)j++; + } + } + + free(a); + free(b); + free(c); + free(d); +} + +void test_encode_decode(opus_int32 bitrate, opus_int32 channels, + const int mapping_family) +{ + const opus_int32 Fs = 48000; + const int application = OPUS_APPLICATION_AUDIO; + + OpusProjectionEncoder *st_enc; + OpusProjectionDecoder *st_dec; + int streams; + int coupled; + int error; + short *buffer_in; + short *buffer_out; + unsigned char data[MAX_DATA_BYTES] = { 0 }; + int len; + int out_samples; + opus_int32 matrix_size = 0; + unsigned char *matrix = NULL; + + buffer_in = (short *)malloc(sizeof(short) * BUFFER_SIZE * channels); + buffer_out = (short *)malloc(sizeof(short) * BUFFER_SIZE * channels); + + st_enc = opus_projection_ambisonics_encoder_create(Fs, channels, + mapping_family, &streams, &coupled, application, &error); + if (error != OPUS_OK) { + fprintf(stderr, + "Couldn\'t create encoder with %d channels and mapping family %d.\n", + channels, mapping_family); + free(buffer_in); + free(buffer_out); + test_failed(); + } + + error = opus_projection_encoder_ctl(st_enc, + OPUS_SET_BITRATE(bitrate * 1000 * (streams + coupled))); + if (error != OPUS_OK) + { + goto bad_cleanup; + } + + error = opus_projection_encoder_ctl(st_enc, + OPUS_PROJECTION_GET_DEMIXING_MATRIX_SIZE_REQUEST, &matrix_size); + if (error != OPUS_OK || !matrix_size) + { + goto bad_cleanup; + } + + matrix = (unsigned char *)opus_alloc(matrix_size); + error = opus_projection_encoder_ctl(st_enc, + OPUS_PROJECTION_GET_DEMIXING_MATRIX_REQUEST, matrix, matrix_size); + + st_dec = opus_projection_decoder_create(Fs, channels, streams, coupled, + matrix, matrix_size, &error); + opus_free(matrix); + + if (error != OPUS_OK) { + fprintf(stderr, + "Couldn\'t create decoder with %d channels, %d streams " + "and %d coupled streams.\n", channels, streams, coupled); + goto bad_cleanup; + } + + generate_music(buffer_in, BUFFER_SIZE, channels); + + len = opus_projection_encode( + st_enc, buffer_in, BUFFER_SIZE, data, MAX_DATA_BYTES); + if(len<0 || len>MAX_DATA_BYTES) { + fprintf(stderr,"opus_encode() returned %d\n", len); + goto bad_cleanup; + } + + out_samples = opus_projection_decode( + st_dec, data, len, buffer_out, MAX_FRAME_SAMPLES, 0); + if(out_samples!=BUFFER_SIZE) { + fprintf(stderr,"opus_decode() returned %d\n", out_samples); + goto bad_cleanup; + } + + opus_projection_decoder_destroy(st_dec); + opus_projection_encoder_destroy(st_enc); + free(buffer_in); + free(buffer_out); + return; +bad_cleanup: + free(buffer_in); + free(buffer_out); + test_failed(); +} + +int main(int _argc, char **_argv) +{ + unsigned int i; + + (void)_argc; + (void)_argv; + + /* Test simple matrix multiplication routines. */ + test_simple_matrix(); + + /* Test full range of channels in creation arguments. */ + for (i = 0; i < 255; i++) + test_creation_arguments(i, 3); + + /* Test encode/decode pipeline. */ + test_encode_decode(64 * 18, 18, 3); + + fprintf(stderr, "All projection tests passed.\n"); + return 0; +} + diff --git a/lib/opus/training/rnn_dump.py b/lib/opus/training/rnn_dump.py new file mode 100755 index 00000000..c312088e --- /dev/null +++ b/lib/opus/training/rnn_dump.py @@ -0,0 +1,66 @@ +#!/usr/bin/python + +from __future__ import print_function + +from keras.models import Sequential +from keras.models import Model +from keras.layers import Input +from keras.layers import Dense +from keras.layers import LSTM +from keras.layers import GRU +from keras.models import load_model +from keras import backend as K +import sys + +import numpy as np + +def printVector(f, vector, name): + v = np.reshape(vector, (-1)); + #print('static const float ', name, '[', len(v), '] = \n', file=f) + f.write('static const opus_int8 {}[{}] = {{\n '.format(name, len(v))) + for i in range(0, len(v)): + f.write('{}'.format(max(-128,min(127,int(round(128*v[i])))))) + if (i!=len(v)-1): + f.write(',') + else: + break; + if (i%8==7): + f.write("\n ") + else: + f.write(" ") + #print(v, file=f) + f.write('\n};\n\n') + return; + +def binary_crossentrop2(y_true, y_pred): + return K.mean(2*K.abs(y_true-0.5) * K.binary_crossentropy(y_pred, y_true), axis=-1) + + +#model = load_model(sys.argv[1], custom_objects={'binary_crossentrop2': binary_crossentrop2}) +main_input = Input(shape=(None, 25), name='main_input') +x = Dense(32, activation='tanh')(main_input) +x = GRU(24, activation='tanh', recurrent_activation='sigmoid', return_sequences=True)(x) +x = Dense(2, activation='sigmoid')(x) +model = Model(inputs=main_input, outputs=x) +model.load_weights(sys.argv[1]) + +weights = model.get_weights() + +f = open(sys.argv[2], 'w') + +f.write('/*This file is automatically generated from a Keras model*/\n\n') +f.write('#ifdef HAVE_CONFIG_H\n#include "config.h"\n#endif\n\n#include "mlp.h"\n\n') + +printVector(f, weights[0], 'layer0_weights') +printVector(f, weights[1], 'layer0_bias') +printVector(f, weights[2], 'layer1_weights') +printVector(f, weights[3], 'layer1_recur_weights') +printVector(f, weights[4], 'layer1_bias') +printVector(f, weights[5], 'layer2_weights') +printVector(f, weights[6], 'layer2_bias') + +f.write('const DenseLayer layer0 = {\n layer0_bias,\n layer0_weights,\n 25, 32, 0\n};\n\n') +f.write('const GRULayer layer1 = {\n layer1_bias,\n layer1_weights,\n layer1_recur_weights,\n 32, 24\n};\n\n') +f.write('const DenseLayer layer2 = {\n layer2_bias,\n layer2_weights,\n 24, 2, 1\n};\n\n') + +f.close() diff --git a/lib/opus/training/rnn_train.py b/lib/opus/training/rnn_train.py new file mode 100755 index 00000000..29bcb034 --- /dev/null +++ b/lib/opus/training/rnn_train.py @@ -0,0 +1,177 @@ +#!/usr/bin/python3 + +from __future__ import print_function + +from keras.models import Sequential +from keras.models import Model +from keras.layers import Input +from keras.layers import Dense +from keras.layers import LSTM +from keras.layers import GRU +from keras.layers import CuDNNGRU +from keras.layers import SimpleRNN +from keras.layers import Dropout +from keras import losses +import h5py +from keras.optimizers import Adam + +from keras.constraints import Constraint +from keras import backend as K +import numpy as np + +import tensorflow as tf +from keras.backend.tensorflow_backend import set_session +config = tf.ConfigProto() +config.gpu_options.per_process_gpu_memory_fraction = 0.44 +set_session(tf.Session(config=config)) + +def binary_crossentrop2(y_true, y_pred): + return K.mean(2*K.abs(y_true-0.5) * K.binary_crossentropy(y_true, y_pred), axis=-1) + +def binary_accuracy2(y_true, y_pred): + return K.mean(K.cast(K.equal(y_true, K.round(y_pred)), 'float32') + K.cast(K.equal(y_true, 0.5), 'float32'), axis=-1) + +def quant_model(model): + weights = model.get_weights() + for k in range(len(weights)): + weights[k] = np.maximum(-128, np.minimum(127, np.round(128*weights[k])*0.0078125)) + model.set_weights(weights) + +class WeightClip(Constraint): + '''Clips the weights incident to each hidden unit to be inside a range + ''' + def __init__(self, c=2): + self.c = c + + def __call__(self, p): + return K.clip(p, -self.c, self.c) + + def get_config(self): + return {'name': self.__class__.__name__, + 'c': self.c} + +reg = 0.000001 +constraint = WeightClip(.998) + +print('Build model...') + +main_input = Input(shape=(None, 25), name='main_input') +x = Dense(32, activation='tanh', kernel_constraint=constraint, bias_constraint=constraint)(main_input) +#x = CuDNNGRU(24, return_sequences=True, kernel_constraint=constraint, recurrent_constraint=constraint, bias_constraint=constraint)(x) +x = GRU(24, recurrent_activation='sigmoid', activation='tanh', return_sequences=True, kernel_constraint=constraint, recurrent_constraint=constraint, bias_constraint=constraint)(x) +x = Dense(2, activation='sigmoid', kernel_constraint=constraint, bias_constraint=constraint)(x) +model = Model(inputs=main_input, outputs=x) + +batch_size = 2048 + +print('Loading data...') +with h5py.File('features10b.h5', 'r') as hf: + all_data = hf['data'][:] +print('done.') + +window_size = 1500 + +nb_sequences = len(all_data)//window_size +print(nb_sequences, ' sequences') +x_train = all_data[:nb_sequences*window_size, :-2] +x_train = np.reshape(x_train, (nb_sequences, window_size, 25)) + +y_train = np.copy(all_data[:nb_sequences*window_size, -2:]) +y_train = np.reshape(y_train, (nb_sequences, window_size, 2)) + +print("Marking ignores") +for s in y_train: + for e in s: + if (e[1] >= 1): + break + e[0] = 0.5 + +all_data = 0; +x_train = x_train.astype('float32') +y_train = y_train.astype('float32') + +print(len(x_train), 'train sequences. x shape =', x_train.shape, 'y shape = ', y_train.shape) + +model.load_weights('newweights10a1b_ep206.hdf5') + +#weights = model.get_weights() +#for k in range(len(weights)): +# weights[k] = np.round(128*weights[k])*0.0078125 +#model.set_weights(weights) + +# try using different optimizers and different optimizer configs +model.compile(loss=binary_crossentrop2, + optimizer=Adam(0.0001), + metrics=[binary_accuracy2]) + +print('Train...') +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=10, validation_data=(x_train, y_train)) +model.save("newweights10a1c_ep10.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=50, initial_epoch=10) +model.save("newweights10a1c_ep50.hdf5") + +model.compile(loss=binary_crossentrop2, + optimizer=Adam(0.0001), + metrics=[binary_accuracy2]) + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=100, initial_epoch=50) +model.save("newweights10a1c_ep100.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=150, initial_epoch=100) +model.save("newweights10a1c_ep150.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=200, initial_epoch=150) +model.save("newweights10a1c_ep200.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=201, initial_epoch=200) +model.save("newweights10a1c_ep201.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=202, initial_epoch=201, validation_data=(x_train, y_train)) +model.save("newweights10a1c_ep202.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=203, initial_epoch=202, validation_data=(x_train, y_train)) +model.save("newweights10a1c_ep203.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=204, initial_epoch=203, validation_data=(x_train, y_train)) +model.save("newweights10a1c_ep204.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=205, initial_epoch=204, validation_data=(x_train, y_train)) +model.save("newweights10a1c_ep205.hdf5") + +quant_model(model) +model.fit(x_train, y_train, + batch_size=batch_size, + epochs=206, initial_epoch=205, validation_data=(x_train, y_train)) +model.save("newweights10a1c_ep206.hdf5") + diff --git a/lib/opus/training/txt2hdf5.py b/lib/opus/training/txt2hdf5.py new file mode 100755 index 00000000..9c602877 --- /dev/null +++ b/lib/opus/training/txt2hdf5.py @@ -0,0 +1,12 @@ +#!/usr/bin/python + +from __future__ import print_function + +import numpy as np +import h5py +import sys + +data = np.loadtxt(sys.argv[1], dtype='float32') +h5f = h5py.File(sys.argv[2], 'w'); +h5f.create_dataset('data', data=data) +h5f.close() diff --git a/lib/opus/win32/VS2015/common.props b/lib/opus/win32/VS2015/common.props new file mode 100644 index 00000000..f6e920b6 --- /dev/null +++ b/lib/opus/win32/VS2015/common.props @@ -0,0 +1,82 @@ + + + + + + $(Platform)\$(Configuration)\ + $(Platform)\$(Configuration)\$(ProjectName)\ + Unicode + + + true + true + false + + + false + false + true + + + + Level3 + false + false + ..\..;..\..\include;..\..\silk;..\..\celt;..\..\win32;%(AdditionalIncludeDirectories) + HAVE_CONFIG_H;_CRT_SECURE_NO_WARNINGS;%(PreprocessorDefinitions) + false + false + + + Console + + + true + Console + + + + + Guard + ProgramDatabase + NoExtensions + false + true + false + Disabled + false + false + Disabled + MultiThreadedDebug + MultiThreadedDebugDLL + true + false + + + true + + + + + false + None + true + true + false + Speed + Fast + Precise + true + true + true + MaxSpeed + MultiThreaded + MultiThreadedDLL + 16Bytes + + + false + + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/opus.sln b/lib/opus/win32/VS2015/opus.sln new file mode 100644 index 00000000..7b678e7f --- /dev/null +++ b/lib/opus/win32/VS2015/opus.sln @@ -0,0 +1,168 @@ + +Microsoft Visual Studio Solution File, Format Version 12.00 +# Visual Studio 14 +VisualStudioVersion = 14.0.25420.1 +MinimumVisualStudioVersion = 10.0.40219.1 +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "opus", "opus.vcxproj", "{219EC965-228A-1824-174D-96449D05F88A}" +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "opus_demo", "opus_demo.vcxproj", "{016C739D-6389-43BF-8D88-24B2BF6F620F}" + ProjectSection(ProjectDependencies) = postProject + {219EC965-228A-1824-174D-96449D05F88A} = {219EC965-228A-1824-174D-96449D05F88A} + EndProjectSection +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "test_opus_api", "test_opus_api.vcxproj", "{1D257A17-D254-42E5-82D6-1C87A6EC775A}" + ProjectSection(ProjectDependencies) = postProject + {219EC965-228A-1824-174D-96449D05F88A} = {219EC965-228A-1824-174D-96449D05F88A} + EndProjectSection +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "test_opus_decode", "test_opus_decode.vcxproj", "{8578322A-1883-486B-B6FA-E0094B65C9F2}" + ProjectSection(ProjectDependencies) = postProject + {219EC965-228A-1824-174D-96449D05F88A} = {219EC965-228A-1824-174D-96449D05F88A} + EndProjectSection +EndProject +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "test_opus_encode", "test_opus_encode.vcxproj", "{84DAA768-1A38-4312-BB61-4C78BB59E5B8}" + ProjectSection(ProjectDependencies) = postProject + {219EC965-228A-1824-174D-96449D05F88A} = {219EC965-228A-1824-174D-96449D05F88A} + EndProjectSection +EndProject +Global + GlobalSection(SolutionConfigurationPlatforms) = preSolution + Debug|Win32 = Debug|Win32 + Debug|x64 = Debug|x64 + DebugDLL_fixed|Win32 = DebugDLL_fixed|Win32 + DebugDLL_fixed|x64 = DebugDLL_fixed|x64 + DebugDLL|Win32 = DebugDLL|Win32 + DebugDLL|x64 = DebugDLL|x64 + Release|Win32 = Release|Win32 + Release|x64 = Release|x64 + ReleaseDLL_fixed|Win32 = ReleaseDLL_fixed|Win32 + ReleaseDLL_fixed|x64 = ReleaseDLL_fixed|x64 + ReleaseDLL|Win32 = ReleaseDLL|Win32 + ReleaseDLL|x64 = ReleaseDLL|x64 + EndGlobalSection + GlobalSection(ProjectConfigurationPlatforms) = postSolution + {219EC965-228A-1824-174D-96449D05F88A}.Debug|Win32.ActiveCfg = Debug|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.Debug|Win32.Build.0 = Debug|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.Debug|x64.ActiveCfg = Debug|x64 + {219EC965-228A-1824-174D-96449D05F88A}.Debug|x64.Build.0 = Debug|x64 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL_fixed|Win32.ActiveCfg = DebugDLL_fixed|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL_fixed|Win32.Build.0 = DebugDLL_fixed|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL_fixed|x64.ActiveCfg = DebugDLL_fixed|x64 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL_fixed|x64.Build.0 = DebugDLL_fixed|x64 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL|Win32.ActiveCfg = DebugDLL|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL|Win32.Build.0 = DebugDLL|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL|x64.ActiveCfg = DebugDLL|x64 + {219EC965-228A-1824-174D-96449D05F88A}.DebugDLL|x64.Build.0 = DebugDLL|x64 + {219EC965-228A-1824-174D-96449D05F88A}.Release|Win32.ActiveCfg = Release|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.Release|Win32.Build.0 = Release|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.Release|x64.ActiveCfg = Release|x64 + {219EC965-228A-1824-174D-96449D05F88A}.Release|x64.Build.0 = Release|x64 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL_fixed|Win32.ActiveCfg = ReleaseDLL_fixed|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL_fixed|Win32.Build.0 = ReleaseDLL_fixed|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL_fixed|x64.ActiveCfg = ReleaseDLL_fixed|x64 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL_fixed|x64.Build.0 = ReleaseDLL_fixed|x64 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL|Win32.ActiveCfg = ReleaseDLL|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL|Win32.Build.0 = ReleaseDLL|Win32 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL|x64.ActiveCfg = ReleaseDLL|x64 + {219EC965-228A-1824-174D-96449D05F88A}.ReleaseDLL|x64.Build.0 = ReleaseDLL|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Debug|Win32.ActiveCfg = Debug|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Debug|Win32.Build.0 = Debug|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Debug|x64.ActiveCfg = Debug|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Debug|x64.Build.0 = Debug|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL_fixed|Win32.ActiveCfg = DebugDLL_fixed|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL_fixed|Win32.Build.0 = DebugDLL_fixed|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL_fixed|x64.ActiveCfg = DebugDLL_fixed|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL_fixed|x64.Build.0 = DebugDLL_fixed|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL|Win32.ActiveCfg = DebugDLL|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL|Win32.Build.0 = DebugDLL|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL|x64.ActiveCfg = DebugDLL|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.DebugDLL|x64.Build.0 = DebugDLL|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Release|Win32.ActiveCfg = Release|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Release|Win32.Build.0 = Release|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Release|x64.ActiveCfg = Release|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.Release|x64.Build.0 = Release|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL_fixed|Win32.ActiveCfg = ReleaseDLL_fixed|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL_fixed|Win32.Build.0 = ReleaseDLL_fixed|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL_fixed|x64.ActiveCfg = ReleaseDLL_fixed|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL_fixed|x64.Build.0 = ReleaseDLL_fixed|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL|Win32.ActiveCfg = ReleaseDLL|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL|Win32.Build.0 = ReleaseDLL|Win32 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL|x64.ActiveCfg = ReleaseDLL|x64 + {016C739D-6389-43BF-8D88-24B2BF6F620F}.ReleaseDLL|x64.Build.0 = ReleaseDLL|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Debug|Win32.ActiveCfg = Debug|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Debug|Win32.Build.0 = Debug|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Debug|x64.ActiveCfg = Debug|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Debug|x64.Build.0 = Debug|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL_fixed|Win32.ActiveCfg = DebugDLL_fixed|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL_fixed|Win32.Build.0 = DebugDLL_fixed|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL_fixed|x64.ActiveCfg = DebugDLL_fixed|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL_fixed|x64.Build.0 = DebugDLL_fixed|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL|Win32.ActiveCfg = DebugDLL|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL|Win32.Build.0 = DebugDLL|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL|x64.ActiveCfg = DebugDLL|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.DebugDLL|x64.Build.0 = DebugDLL|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Release|Win32.ActiveCfg = Release|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Release|Win32.Build.0 = Release|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Release|x64.ActiveCfg = Release|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.Release|x64.Build.0 = Release|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL_fixed|Win32.ActiveCfg = ReleaseDLL_fixed|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL_fixed|Win32.Build.0 = ReleaseDLL_fixed|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL_fixed|x64.ActiveCfg = ReleaseDLL_fixed|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL_fixed|x64.Build.0 = ReleaseDLL_fixed|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL|Win32.ActiveCfg = ReleaseDLL|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL|Win32.Build.0 = ReleaseDLL|Win32 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL|x64.ActiveCfg = ReleaseDLL|x64 + {1D257A17-D254-42E5-82D6-1C87A6EC775A}.ReleaseDLL|x64.Build.0 = ReleaseDLL|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Debug|Win32.ActiveCfg = Debug|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Debug|Win32.Build.0 = Debug|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Debug|x64.ActiveCfg = Debug|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Debug|x64.Build.0 = Debug|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL_fixed|Win32.ActiveCfg = DebugDLL_fixed|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL_fixed|Win32.Build.0 = DebugDLL_fixed|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL_fixed|x64.ActiveCfg = DebugDLL_fixed|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL_fixed|x64.Build.0 = DebugDLL_fixed|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL|Win32.ActiveCfg = DebugDLL|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL|Win32.Build.0 = DebugDLL|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL|x64.ActiveCfg = DebugDLL|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.DebugDLL|x64.Build.0 = DebugDLL|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Release|Win32.ActiveCfg = Release|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Release|Win32.Build.0 = Release|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Release|x64.ActiveCfg = Release|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.Release|x64.Build.0 = Release|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL_fixed|Win32.ActiveCfg = ReleaseDLL_fixed|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL_fixed|Win32.Build.0 = ReleaseDLL_fixed|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL_fixed|x64.ActiveCfg = ReleaseDLL_fixed|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL_fixed|x64.Build.0 = ReleaseDLL_fixed|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL|Win32.ActiveCfg = ReleaseDLL|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL|Win32.Build.0 = ReleaseDLL|Win32 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL|x64.ActiveCfg = ReleaseDLL|x64 + {8578322A-1883-486B-B6FA-E0094B65C9F2}.ReleaseDLL|x64.Build.0 = ReleaseDLL|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Debug|Win32.ActiveCfg = Debug|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Debug|Win32.Build.0 = Debug|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Debug|x64.ActiveCfg = Debug|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Debug|x64.Build.0 = Debug|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL_fixed|Win32.ActiveCfg = DebugDLL_fixed|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL_fixed|Win32.Build.0 = DebugDLL_fixed|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL_fixed|x64.ActiveCfg = DebugDLL_fixed|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL_fixed|x64.Build.0 = DebugDLL_fixed|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL|Win32.ActiveCfg = DebugDLL|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL|Win32.Build.0 = DebugDLL|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL|x64.ActiveCfg = DebugDLL|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.DebugDLL|x64.Build.0 = DebugDLL|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Release|Win32.ActiveCfg = Release|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Release|Win32.Build.0 = Release|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Release|x64.ActiveCfg = Release|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.Release|x64.Build.0 = Release|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL_fixed|Win32.ActiveCfg = ReleaseDLL_fixed|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL_fixed|Win32.Build.0 = ReleaseDLL_fixed|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL_fixed|x64.ActiveCfg = ReleaseDLL_fixed|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL_fixed|x64.Build.0 = ReleaseDLL_fixed|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL|Win32.ActiveCfg = ReleaseDLL|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL|Win32.Build.0 = ReleaseDLL|Win32 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL|x64.ActiveCfg = ReleaseDLL|x64 + {84DAA768-1A38-4312-BB61-4C78BB59E5B8}.ReleaseDLL|x64.Build.0 = ReleaseDLL|x64 + EndGlobalSection + GlobalSection(SolutionProperties) = preSolution + HideSolutionNode = FALSE + EndGlobalSection +EndGlobal diff --git a/lib/opus/win32/VS2015/opus.vcxproj b/lib/opus/win32/VS2015/opus.vcxproj new file mode 100644 index 00000000..34b1233c --- /dev/null +++ b/lib/opus/win32/VS2015/opus.vcxproj @@ -0,0 +1,399 @@ + + + + + DebugDLL_fixed + Win32 + + + DebugDLL_fixed + x64 + + + DebugDLL + Win32 + + + DebugDLL + x64 + + + Debug + Win32 + + + Debug + x64 + + + ReleaseDLL_fixed + Win32 + + + ReleaseDLL_fixed + x64 + + + ReleaseDLL + Win32 + + + ReleaseDLL + x64 + + + Release + Win32 + + + Release + x64 + + + + Win32Proj + opus + {219EC965-228A-1824-174D-96449D05F88A} + + + + StaticLibrary + v140 + + + DynamicLibrary + v140 + + + DynamicLibrary + v140 + + + StaticLibrary + v140 + + + DynamicLibrary + v140 + + + DynamicLibrary + v140 + + + StaticLibrary + v140 + + + DynamicLibrary + v140 + + + DynamicLibrary + v140 + + + StaticLibrary + v140 + + + DynamicLibrary + v140 + + + DynamicLibrary + v140 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + ..\..\silk\fixed;..\..\silk\float;%(AdditionalIncludeDirectories) + DLL_EXPORT;%(PreprocessorDefinitions) + FIXED_POINT;%(PreprocessorDefinitions) + /arch:IA32 %(AdditionalOptions) + + + /ignore:4221 %(AdditionalOptions) + + + "$(ProjectDir)..\..\win32\genversion.bat" "$(ProjectDir)..\..\win32\version.h" PACKAGE_VERSION + Generating version.h + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + 4244;%(DisableSpecificWarnings) + + + + + + + + + + + + + + + false + + + false + + + true + + + + + + + true + + + true + + + false + + + + + + + + diff --git a/lib/opus/win32/VS2015/opus.vcxproj.filters b/lib/opus/win32/VS2015/opus.vcxproj.filters new file mode 100644 index 00000000..8c61d6a0 --- /dev/null +++ b/lib/opus/win32/VS2015/opus.vcxproj.filters @@ -0,0 +1,585 @@ + + + + + {4FC737F1-C7A5-4376-A066-2A32D752A2FF} + cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx + + + {93995380-89BD-4b04-88EB-625FBE52EBFB} + h;hpp;hxx;hm;inl;inc;xsd + + + {67DA6AB6-F800-4c08-8B7A-83BB121AAD01} + rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe;resx;tiff;tif;png;wav + + + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + Header Files + + + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + Source Files + + + diff --git a/lib/opus/win32/VS2015/opus_demo.vcxproj b/lib/opus/win32/VS2015/opus_demo.vcxproj new file mode 100644 index 00000000..f4344e53 --- /dev/null +++ b/lib/opus/win32/VS2015/opus_demo.vcxproj @@ -0,0 +1,171 @@ + + + + + DebugDLL_fixed + Win32 + + + DebugDLL_fixed + x64 + + + DebugDLL + Win32 + + + DebugDLL + x64 + + + Debug + Win32 + + + Debug + x64 + + + ReleaseDLL_fixed + Win32 + + + ReleaseDLL_fixed + x64 + + + ReleaseDLL + Win32 + + + ReleaseDLL + x64 + + + Release + Win32 + + + Release + x64 + + + + + {219ec965-228a-1824-174d-96449d05f88a} + + + + + + + {016C739D-6389-43BF-8D88-24B2BF6F620F} + Win32Proj + opus_demo + + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/opus_demo.vcxproj.filters b/lib/opus/win32/VS2015/opus_demo.vcxproj.filters new file mode 100644 index 00000000..2eb113ac --- /dev/null +++ b/lib/opus/win32/VS2015/opus_demo.vcxproj.filters @@ -0,0 +1,22 @@ + + + + + {4FC737F1-C7A5-4376-A066-2A32D752A2FF} + cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx + + + {93995380-89BD-4b04-88EB-625FBE52EBFB} + h;hpp;hxx;hm;inl;inc;xsd + + + {67DA6AB6-F800-4c08-8B7A-83BB121AAD01} + rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe;resx;tiff;tif;png;wav;mfcribbon-ms + + + + + Source Files + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/test_opus_api.vcxproj b/lib/opus/win32/VS2015/test_opus_api.vcxproj new file mode 100644 index 00000000..7cae131b --- /dev/null +++ b/lib/opus/win32/VS2015/test_opus_api.vcxproj @@ -0,0 +1,171 @@ + + + + + DebugDLL_fixed + Win32 + + + DebugDLL_fixed + x64 + + + DebugDLL + Win32 + + + DebugDLL + x64 + + + Debug + Win32 + + + Debug + x64 + + + ReleaseDLL_fixed + Win32 + + + ReleaseDLL_fixed + x64 + + + ReleaseDLL + Win32 + + + ReleaseDLL + x64 + + + Release + Win32 + + + Release + x64 + + + + + + + + {219ec965-228a-1824-174d-96449d05f88a} + + + + {1D257A17-D254-42E5-82D6-1C87A6EC775A} + Win32Proj + test_opus_api + + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/test_opus_api.vcxproj.filters b/lib/opus/win32/VS2015/test_opus_api.vcxproj.filters new file mode 100644 index 00000000..383d19f7 --- /dev/null +++ b/lib/opus/win32/VS2015/test_opus_api.vcxproj.filters @@ -0,0 +1,14 @@ + + + + + {4FC737F1-C7A5-4376-A066-2A32D752A2FF} + cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx + + + + + Source Files + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/test_opus_decode.vcxproj b/lib/opus/win32/VS2015/test_opus_decode.vcxproj new file mode 100644 index 00000000..df01dca1 --- /dev/null +++ b/lib/opus/win32/VS2015/test_opus_decode.vcxproj @@ -0,0 +1,171 @@ + + + + + DebugDLL_fixed + Win32 + + + DebugDLL_fixed + x64 + + + DebugDLL + Win32 + + + DebugDLL + x64 + + + Debug + Win32 + + + Debug + x64 + + + ReleaseDLL_fixed + Win32 + + + ReleaseDLL_fixed + x64 + + + ReleaseDLL + Win32 + + + ReleaseDLL + x64 + + + Release + Win32 + + + Release + x64 + + + + + + + + {219ec965-228a-1824-174d-96449d05f88a} + + + + {8578322A-1883-486B-B6FA-E0094B65C9F2} + Win32Proj + test_opus_api + + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/test_opus_decode.vcxproj.filters b/lib/opus/win32/VS2015/test_opus_decode.vcxproj.filters new file mode 100644 index 00000000..3036a4e7 --- /dev/null +++ b/lib/opus/win32/VS2015/test_opus_decode.vcxproj.filters @@ -0,0 +1,14 @@ + + + + + {4a0dd677-931f-4728-afe5-b761149fc7eb} + cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx + + + + + Source Files + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/test_opus_encode.vcxproj b/lib/opus/win32/VS2015/test_opus_encode.vcxproj new file mode 100644 index 00000000..405efee9 --- /dev/null +++ b/lib/opus/win32/VS2015/test_opus_encode.vcxproj @@ -0,0 +1,172 @@ + + + + + DebugDLL_fixed + Win32 + + + DebugDLL_fixed + x64 + + + DebugDLL + Win32 + + + DebugDLL + x64 + + + Debug + Win32 + + + Debug + x64 + + + ReleaseDLL_fixed + Win32 + + + ReleaseDLL_fixed + x64 + + + ReleaseDLL + Win32 + + + ReleaseDLL + x64 + + + Release + Win32 + + + Release + x64 + + + + + + + + + {219ec965-228a-1824-174d-96449d05f88a} + + + + {84DAA768-1A38-4312-BB61-4C78BB59E5B8} + Win32Proj + test_opus_api + + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + Application + v140 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/lib/opus/win32/VS2015/test_opus_encode.vcxproj.filters b/lib/opus/win32/VS2015/test_opus_encode.vcxproj.filters new file mode 100644 index 00000000..4ed3bb9e --- /dev/null +++ b/lib/opus/win32/VS2015/test_opus_encode.vcxproj.filters @@ -0,0 +1,17 @@ + + + + + {546c8d9a-103e-4f78-972b-b44e8d3c8aba} + cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx + + + + + Source Files + + + Source Files + + + \ No newline at end of file diff --git a/lib/opus/win32/config.h b/lib/opus/win32/config.h new file mode 100644 index 00000000..3e54bcbb --- /dev/null +++ b/lib/opus/win32/config.h @@ -0,0 +1,64 @@ +/*********************************************************************** +Copyright (c) 2011, Skype Limited. All rights reserved. +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions +are met: +- Redistributions of source code must retain the above copyright notice, +this list of conditions and the following disclaimer. +- Redistributions in binary form must reproduce the above copyright +notice, this list of conditions and the following disclaimer in the +documentation and/or other materials provided with the distribution. +- Neither the name of Internet Society, IETF or IETF Trust, nor the +names of specific contributors, may be used to endorse or promote +products derived from this software without specific prior written +permission. +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE +ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR +CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF +SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS +INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN +CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) +ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. +***********************************************************************/ + +#ifndef CONFIG_H +#define CONFIG_H + +#define USE_ALLOCA 1 + +/* Comment out the next line for floating-point code */ +/*#define FIXED_POINT 1 */ + +#define OPUS_BUILD 1 + +#if defined(_M_IX86) || defined(_M_X64) +/* Can always compile SSE intrinsics (no special compiler flags necessary) */ +#define OPUS_X86_MAY_HAVE_SSE +#define OPUS_X86_MAY_HAVE_SSE2 +#define OPUS_X86_MAY_HAVE_SSE4_1 + +/* Presume SSE functions, if compiled to use SSE/SSE2/AVX (note that AMD64 implies SSE2, and AVX + implies SSE4.1) */ +#if defined(_M_X64) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 1)) || defined(__AVX__) +#define OPUS_X86_PRESUME_SSE 1 +#endif +#if defined(_M_X64) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 2)) || defined(__AVX__) +#define OPUS_X86_PRESUME_SSE2 1 +#endif +#if defined(__AVX__) +#define OPUS_X86_PRESUME_SSE4_1 1 +#endif + +#if !defined(OPUS_X86_PRESUME_SSE4_1) || !defined(OPUS_X86_PRESUME_SSE2) || !defined(OPUS_X86_PRESUME_SSE) +#define OPUS_HAVE_RTCD 1 +#endif + +#endif + +#include "version.h" + +#endif /* CONFIG_H */ diff --git a/lib/opus/win32/genversion.bat b/lib/opus/win32/genversion.bat new file mode 100644 index 00000000..1def7460 --- /dev/null +++ b/lib/opus/win32/genversion.bat @@ -0,0 +1,37 @@ +@echo off + +setlocal enableextensions enabledelayedexpansion + +for /f %%v in ('cd "%~dp0.." ^&^& git status ^>NUL 2^>NUL ^&^& git describe --tags --match "v*" --dirty 2^>NUL') do set version=%%v + +if not "%version%"=="" set version=!version:~1! && goto :gotversion + +if exist "%~dp0..\package_version" goto :getversion + +echo Git cannot be found, nor can package_version. Generating unknown version. + +set version=unknown + +goto :gotversion + +:getversion + +for /f "delims== tokens=2" %%v in (%~dps0..\package_version) do set version=%%v +set version=!version:"=! + +:gotversion + +set version=!version: =! +set version_out=#define %~2 "%version%" + +echo %version_out%> "%~1_temp" + +echo n | comp "%~1_temp" "%~1" > NUL 2> NUL + +if not errorlevel 1 goto exit + +copy /y "%~1_temp" "%~1" + +:exit + +del "%~1_temp"