From 52217b637d09fc8fe19dcaa16378e39b6034a54b Mon Sep 17 00:00:00 2001 From: ayumi Date: Thu, 30 Jan 2025 07:58:36 +0100 Subject: [PATCH 1/3] =?UTF-8?q?Don=E2=80=99t=20assume=20that=20each=20tagg?= =?UTF-8?q?ing=20format=20is=20only=20used=20by=20one=20file=20format?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- lib/libtags/tags.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/lib/libtags/tags.c b/lib/libtags/tags.c index 750d9077..b1d6ac33 100644 --- a/lib/libtags/tags.c +++ b/lib/libtags/tags.c @@ -71,7 +71,9 @@ tagsget(Tagctx *ctx) for(i = 0; i < nelem(g); i++){ ctx->num = 0; if(g[i].f(ctx) == 0){ - ctx->format = g[i].format; + if(ctx->format == Funknown){ + ctx->format = g[i].format; + } res = 0; } ctx->seek(ctx, ctx->restart, 0); From a3639860761dcdb5ef9c31bb34497f32cadd9ff3 Mon Sep 17 00:00:00 2001 From: ayumi Date: Thu, 30 Jan 2025 10:08:37 +0100 Subject: [PATCH 2/3] Add support for APEv2 tags and detecting WavPack files --- lib/libtags/CMakeLists.txt | 2 +- lib/libtags/ape.c | 233 +++++++++++++++++++++++++++++++++++++ lib/libtags/tags.c | 2 + lib/libtags/tags.h | 1 + 4 files changed, 237 insertions(+), 1 deletion(-) create mode 100644 lib/libtags/ape.c diff --git a/lib/libtags/CMakeLists.txt b/lib/libtags/CMakeLists.txt index d8dce988..db5cd0c2 100644 --- a/lib/libtags/CMakeLists.txt +++ b/lib/libtags/CMakeLists.txt @@ -1,5 +1,5 @@ idf_component_register( - SRCS 437.c 8859.c flac.c id3genres.c id3v1.c id3v2.c it.c m4a.c mod.c opus.c + SRCS 437.c 8859.c ape.c flac.c id3genres.c id3v1.c id3v2.c it.c m4a.c mod.c opus.c s3m.c tags.c utf16.c vorbis.c wav.c xm.c INCLUDE_DIRS . ) diff --git a/lib/libtags/ape.c b/lib/libtags/ape.c new file mode 100644 index 00000000..7ba30649 --- /dev/null +++ b/lib/libtags/ape.c @@ -0,0 +1,233 @@ +#include +#include "tagspriv.h" + +#define leu16int(d) (u16int)(((uchar*)(d))[1]<<8 | ((uchar*)(d))[0]<<0) + +enum +{ + HeaderSize = 32, + FooterSize = HeaderSize, + + MagicOffset = 0, + VersionOffset = 8, + SizeOffset = 12, + CountOffset = 16, + FlagsOffset = 20, + + WvHeaderSize = 32, + WvMagicOffset = 0, + WvVersionOffset = 8, + WvSamplesHighOffset = 11, + WvSamplesLowOffset = 12, + WvFlagsOffset = 24, + WvSampleRateMask = 0xf << 23, + WvSampleRateShift = 23, + WvCustomSampleRate = 16, + WvMonoMask = 4, +}; + +typedef enum +{ + TagUTF8, + TagBinary, + TagExternal, + TagReserved, + + TagInvalid, +} TagType; + +static int +isWavpack(Tagctx *ctx) +{ + uchar header[WvHeaderSize]; + int size; + u16int version; + u32int flags, samplerate; + uvlong samples; + + if(ctx->seek(ctx, 0, 0) < 0) + return 0; + if(ctx->read(ctx, header, WvHeaderSize) != WvHeaderSize) + return 0; + if(memcmp(header+WvMagicOffset, "wvpk", 4)) + return 0; + version = leu16int(header+WvVersionOffset); + if(version<0x402 || version>0x410) + return 0; + samples = (uvlong)(*(uchar*)(header+WvSamplesHighOffset))<<32 | leuint(header+WvSamplesLowOffset); + flags = leuint(header+WvFlagsOffset); + if((flags&WvSampleRateMask)>>WvSampleRateShift != WvCustomSampleRate){ + const u32int samplerates[] = {6000, 8000, 9600, 11025, 12000, 16000, 22050, 24000, 32000, 44100, 48000, 64000, 88200, 96000, 192000}; + samplerate = samplerates[(flags&WvSampleRateMask)>>WvSampleRateShift]; + ctx->samplerate = samplerate; + uvlong duration = round((double)samples/samplerate*1000); + ctx->duration = duration; + } + ctx->channels = flags&WvMonoMask ? 1 : 2; + if(ctx->seek(ctx, 0, 0) < 0) + return 0; + if(size = ctx->seek(ctx, 0, 2), size < 0) + return 0; + ctx->bitrate = (double)size*8.0/(samples/samplerate)/1000; + return 1; +} + +static int +detectFormat(Tagctx *ctx) +{ + if(isWavpack(ctx)) + return Fwavpack; + return Funknown; +} + +static int +tagHasHeader(u32int tags) +{ + return (tags&(1<<31)) >> 31; +} + +static int +tagIsHeader(u32int tags) +{ + return (tags&(1<<29)) >> 29; +} + +static TagType +tagGetType(u32int tags) +{ + switch((tags&(1<<1))>>1 | (tags&(1<<2))>>1 | (tags&(1<<3))>>1 | (tags&(1<<4))>>1){ + case 0: + return TagUTF8; + case 1: + return TagBinary; + case 2: + return TagExternal; + case 3: + return TagReserved; + default: + return TagInvalid; + } +} + +static int +tagTagType(char *name) +{ + if(!strcmp(name, "Album")) + return Talbum; + else if(!strcmp(name, "Album Artist")) + return Talbumartist; + else if(!strcmp(name, "Artist")) + return Tartist; + else if(!strcmp(name, "Comment")) + return Tcomment; + else if(!strcmp(name, "Composer")) + return Tcomposer; + else if(!strncmp(name, "Cover Art (", 11)){ + if(name[strlen(name)-1] == ')') + return Timage; + }else if(!strcmp(name, "Genre")) + return Tgenre; + else if(!strcmp(name, "Replaygain_Album_Gain")) + return Talbumgain; + else if(!strcmp(name, "Replaygain_Album_Peak")) + return Talbumpeak; + else if(!strcmp(name, "Replaygain_Track_Gain")) + return Ttrackgain; + else if(!strcmp(name, "Replaygain_Track_Peak")) + return Ttrackpeak; + else if(!strcmp(name, "Title")) + return Ttitle; + else if(!strcmp(name, "Track")) + return Ttrack; + else if(!strcmp(name, "Year")) + return Tdate; + return Tunknown; +} + +int +tagape(Tagctx *ctx) +{ + uchar footer[FooterSize]; + u32int i, count; + + ctx->format = detectFormat(ctx); + + if(ctx->seek(ctx, -FooterSize, 2) < 0) + return -1; + if(ctx->read(ctx, footer, FooterSize) != FooterSize) + return -1; + if(memcmp(footer+MagicOffset, "APETAGEX", 8)) + return -1; + if(leuint(footer+VersionOffset) != 2000) + return -1; + if(tagIsHeader(leuint(footer+FlagsOffset))) + return -1; + + if(ctx->seek(ctx, -FooterSize-leuint(footer+SizeOffset), 2) < 0) + return -1; + if(tagHasHeader(leuint(footer+FlagsOffset))){ + uchar header[HeaderSize]; + if(ctx->read(ctx, header, HeaderSize) != HeaderSize) + return -1; + if(memcmp(header, footer, 23)) + return -1; + if(!tagHasHeader(leuint(header+FlagsOffset))) + return -1; + if(!tagIsHeader(leuint(header+FlagsOffset))) + return -1; + }else if(ctx->seek(ctx, HeaderSize, 1) < 0) + return -1; + + for(i = 0, count = leuint(footer+CountOffset); i < count; i++){ + int valueOffset = 0; + char c; + u32int d, length, flags; + + if(ctx->read(ctx, &d, 4) != 4) + return -1; + length = leuint(&d); + if(ctx->read(ctx, &d, 4) != 4) + return -1; + flags = leuint(&d); + + do{ + if(valueOffset == ctx->bufsz) + return -1; + if(ctx->read(ctx, &c, 1) != 1) + return -1; + if(c<' ' || c>'~') + if(c != '\0') + return -1; + ctx->buf[valueOffset++] = c; + }while(c != '\0'); + if(valueOffset+1+(int)length>ctx->bufsz && tagTagType(ctx->buf)!=Timage){ + if(ctx->seek(ctx, length, 1) < 0) + return -1; + continue; + } + + switch(tagGetType(flags)){ + u32int keyOffset; + case TagUTF8: + if(ctx->read(ctx, ctx->buf+valueOffset, length) != (int)length) + return -1; + (ctx->buf+valueOffset)[length] = '\0'; + for(keyOffset = 0; keyOffset != length;){ + txtcb(ctx, tagTagType(ctx->buf), ctx->buf, ctx->buf+valueOffset+keyOffset); + if(keyOffset += strlen(ctx->buf+valueOffset+keyOffset), keyOffset != length) + keyOffset++; + } + break; + case TagBinary: + if(tagTagType(ctx->buf) == Timage) + tagscallcb(ctx, Timage, ctx->buf, ctx->buf, ctx->seek(ctx, 0, 1), length, NULL); + if(ctx->seek(ctx, length, 1) < 0) + return -1; + break; + default: + if(ctx->seek(ctx, length, 1) < 0) + return -1; + } + } + return 0; +} diff --git a/lib/libtags/tags.c b/lib/libtags/tags.c index b1d6ac33..d3c577dd 100644 --- a/lib/libtags/tags.c +++ b/lib/libtags/tags.c @@ -8,6 +8,7 @@ struct Getter int format; }; +extern int tagape(Tagctx *ctx); extern int tagflac(Tagctx *ctx); extern int tagid3v1(Tagctx *ctx); extern int tagid3v2(Tagctx *ctx); @@ -22,6 +23,7 @@ extern int tagmod(Tagctx *ctx); static const Getter g[] = { + {tagape, Funknown}, {tagid3v2, Fmp3}, {tagid3v1, Fmp3}, {tagvorbis, Fogg}, diff --git a/lib/libtags/tags.h b/lib/libtags/tags.h index b2aa2dfb..0b54936a 100644 --- a/lib/libtags/tags.h +++ b/lib/libtags/tags.h @@ -37,6 +37,7 @@ enum Fm4a, Fopus, Fwav, + Fwavpack, Fit, Fxm, Fs3m, From 885eb1812c15263ad759741ad138cf7188fdf739 Mon Sep 17 00:00:00 2001 From: ayumi Date: Fri, 31 Jan 2025 19:08:39 +0100 Subject: [PATCH 3/3] Add WavPack support --- REUSE.toml | 6 + lib/wavpack/CMakeLists.txt | 4 + lib/wavpack/bits.c | 141 ++++ lib/wavpack/float.c | 50 ++ lib/wavpack/license.txt | 25 + lib/wavpack/metadata.c | 105 +++ lib/wavpack/readme.txt | 68 ++ lib/wavpack/unpack.c | 785 +++++++++++++++++++++ lib/wavpack/wavpack.h | 394 +++++++++++ lib/wavpack/words.c | 560 +++++++++++++++ lib/wavpack/wputils.c | 350 +++++++++ lib/wavpack/wvfilter.c | 200 ++++++ src/codecs/CMakeLists.txt | 4 +- src/codecs/codec.cpp | 5 + src/codecs/include/types.hpp | 1 + src/codecs/include/wavpack.hpp | 46 ++ src/codecs/wavpack.cpp | 161 +++++ src/tangara/audio/fatfs_stream_factory.cpp | 2 + src/tangara/database/tag_parser.cpp | 3 + src/tangara/database/tag_parser.hpp | 3 +- src/tangara/database/track.hpp | 1 + tools/cmake/common.cmake | 1 + 22 files changed, 2912 insertions(+), 3 deletions(-) create mode 100644 lib/wavpack/CMakeLists.txt create mode 100644 lib/wavpack/bits.c create mode 100644 lib/wavpack/float.c create mode 100644 lib/wavpack/license.txt create mode 100644 lib/wavpack/metadata.c create mode 100644 lib/wavpack/readme.txt create mode 100644 lib/wavpack/unpack.c create mode 100644 lib/wavpack/wavpack.h create mode 100644 lib/wavpack/words.c create mode 100644 lib/wavpack/wputils.c create mode 100644 lib/wavpack/wvfilter.c create mode 100644 src/codecs/include/wavpack.hpp create mode 100644 src/codecs/wavpack.cpp diff --git a/REUSE.toml b/REUSE.toml index dd406c80..b09f1085 100644 --- a/REUSE.toml +++ b/REUSE.toml @@ -181,6 +181,12 @@ precedence = "aggregate" SPDX-FileCopyrightText = "2002, Xiph.org Foundation" SPDX-License-Identifier = "BSD-3-Clause" +[[annotations]] +path = "lib/wavpack/**" +precedence = "aggregate" +SPDX-FileCopyrightText = "1998 - 2006 Conifer Software" +SPDX-License-Identifier = "BSD-3-Clause" + [[annotations]] path = "lua/fonts/fusion*" precedence = "aggregate" diff --git a/lib/wavpack/CMakeLists.txt b/lib/wavpack/CMakeLists.txt new file mode 100644 index 00000000..98fcda95 --- /dev/null +++ b/lib/wavpack/CMakeLists.txt @@ -0,0 +1,4 @@ +idf_component_register( + SRCS bits.c float.c wputils.c metadata.c unpack.c words.c + INCLUDE_DIRS . +) diff --git a/lib/wavpack/bits.c b/lib/wavpack/bits.c new file mode 100644 index 00000000..d69cd288 --- /dev/null +++ b/lib/wavpack/bits.c @@ -0,0 +1,141 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// bits.c + +// This module provides utilities to support the BitStream structure which is +// used to read and write all WavPack audio data streams. It also contains a +// wrapper for the stream I/O functions and a set of functions dealing with +// endian-ness, both for enhancing portability. Finally, a debug wrapper for +// the malloc() system is provided. + +#include "wavpack.h" + +#include +#include + +////////////////////////// Bitstream functions //////////////////////////////// + +// Open the specified BitStream and associate with the specified buffer. + +static void bs_read (Bitstream *bs); + +void bs_open_read (Bitstream *bs, uchar *buffer_start, uchar *buffer_end, read_stream file, void *user_data, uint32_t file_bytes) +{ + CLEAR (*bs); + bs->buf = buffer_start; + bs->end = buffer_end; + + if (file) { + bs->ptr = bs->end - 1; + bs->file_bytes = file_bytes; + bs->file = file; + bs->user_data = user_data; + } + else + bs->ptr = bs->buf - 1; + + bs->wrap = bs_read; +} + +// This function is only called from the getbit() and getbits() macros when +// the BitStream has been exhausted and more data is required. Sinve these +// bistreams no longer access files, this function simple sets an error and +// resets the buffer. + +static void bs_read (Bitstream *bs) +{ + if (bs->file && bs->file_bytes) { + uint32_t bytes_read, bytes_to_read = bs->end - bs->buf; + + if (bytes_to_read > bs->file_bytes) + bytes_to_read = bs->file_bytes; + + bytes_read = bs->file (bs->user_data, bs->buf, bytes_to_read); + + if (bytes_read) { + bs->end = bs->buf + bytes_read; + bs->file_bytes -= bytes_read; + } + else { + memset (bs->buf, -1, bs->end - bs->buf); + bs->error = 1; + } + } + else + bs->error = 1; + + if (bs->error) + memset (bs->buf, -1, bs->end - bs->buf); + + bs->ptr = bs->buf; +} + +/////////////////////// Endian Correction Routines //////////////////////////// + +void little_endian_to_native (void *data, char *format) +{ + uchar *cp = (uchar *) data; + int32_t temp; + + while (*format) { + switch (*format) { + case 'L': + temp = cp [0] + ((int32_t) cp [1] << 8) + ((int32_t) cp [2] << 16) + ((int32_t) cp [3] << 24); + * (int32_t *) cp = temp; + cp += 4; + break; + + case 'S': + temp = cp [0] + (cp [1] << 8); + * (short *) cp = (short) temp; + cp += 2; + break; + + default: + if (isdigit ((unsigned char) *format)) + cp += *format - '0'; + + break; + } + + format++; + } +} + +void native_to_little_endian (void *data, char *format) +{ + uchar *cp = (uchar *) data; + int32_t temp; + + while (*format) { + switch (*format) { + case 'L': + temp = * (int32_t *) cp; + *cp++ = (uchar) temp; + *cp++ = (uchar) (temp >> 8); + *cp++ = (uchar) (temp >> 16); + *cp++ = (uchar) (temp >> 24); + break; + + case 'S': + temp = * (short *) cp; + *cp++ = (uchar) temp; + *cp++ = (uchar) (temp >> 8); + break; + + default: + if (isdigit ((unsigned char) *format)) + cp += *format - '0'; + + break; + } + + format++; + } +} diff --git a/lib/wavpack/float.c b/lib/wavpack/float.c new file mode 100644 index 00000000..4b9b44ee --- /dev/null +++ b/lib/wavpack/float.c @@ -0,0 +1,50 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// float.c + +#include "wavpack.h" + +int read_float_info (WavpackStream *wps, WavpackMetadata *wpmd) +{ + int bytecnt = wpmd->byte_length; + char *byteptr = wpmd->data; + + if (bytecnt != 4) + return FALSE; + + wps->float_flags = *byteptr++; + wps->float_shift = *byteptr++; + wps->float_max_exp = *byteptr++; + wps->float_norm_exp = *byteptr; + return TRUE; +} + +void float_values (WavpackStream *wps, int32_t *values, int32_t num_values) +{ + int shift = wps->float_max_exp - wps->float_norm_exp + wps->float_shift; + + if (shift > 32) + shift = 32; + else if (shift < -32) + shift = -32; + + while (num_values--) { + if (shift > 0) + *values <<= shift; + else if (shift < 0) + *values >>= -shift; + + if (*values > 8388607L) + *values = 8388607L; + else if (*values < -8388608L) + *values = -8388608L; + + values++; + } +} diff --git a/lib/wavpack/license.txt b/lib/wavpack/license.txt new file mode 100644 index 00000000..98f6e6b1 --- /dev/null +++ b/lib/wavpack/license.txt @@ -0,0 +1,25 @@ + Copyright (c) 1998 - 2006 Conifer Software + 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 Conifer Software nor the names of its 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 REGENTS OR CONTRIBUTORS BE LIABLE FOR +ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL +DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, +OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/lib/wavpack/metadata.c b/lib/wavpack/metadata.c new file mode 100644 index 00000000..b80e905a --- /dev/null +++ b/lib/wavpack/metadata.c @@ -0,0 +1,105 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// metadata.c + +// This module handles the metadata structure introduced in WavPack 4.0 + +#include "wavpack.h" + +int read_metadata_buff (WavpackContext *wpc, WavpackMetadata *wpmd) +{ + uchar tchar; + + if (!wpc->infile (wpc->user_data, &wpmd->id, 1) || !wpc->infile (wpc->user_data, &tchar, 1)) + return FALSE; + + wpmd->byte_length = tchar << 1; + + if (wpmd->id & ID_LARGE) { + wpmd->id &= ~ID_LARGE; + + if (!wpc->infile (wpc->user_data, &tchar, 1)) + return FALSE; + + wpmd->byte_length += (int32_t) tchar << 9; + + if (!wpc->infile (wpc->user_data, &tchar, 1)) + return FALSE; + + wpmd->byte_length += (int32_t) tchar << 17; + } + + if (wpmd->id & ID_ODD_SIZE) { + wpmd->id &= ~ID_ODD_SIZE; + wpmd->byte_length--; + } + + if (wpmd->byte_length && wpmd->byte_length <= sizeof (wpc->read_buffer)) { + uint32_t bytes_to_read = wpmd->byte_length + (wpmd->byte_length & 1); + + if (wpc->infile (wpc->user_data, wpc->read_buffer, bytes_to_read) != (int32_t) bytes_to_read) { + wpmd->data = NULL; + return FALSE; + } + + wpmd->data = wpc->read_buffer; + } + else + wpmd->data = NULL; + + return TRUE; +} + +int process_metadata (WavpackContext *wpc, WavpackMetadata *wpmd) +{ + WavpackStream *wps = &wpc->stream; + + switch (wpmd->id) { + case ID_DUMMY: + return TRUE; + + case ID_DECORR_TERMS: + return read_decorr_terms (wps, wpmd); + + case ID_DECORR_WEIGHTS: + return read_decorr_weights (wps, wpmd); + + case ID_DECORR_SAMPLES: + return read_decorr_samples (wps, wpmd); + + case ID_ENTROPY_VARS: + return read_entropy_vars (wps, wpmd); + + case ID_HYBRID_PROFILE: + return read_hybrid_profile (wps, wpmd); + + case ID_FLOAT_INFO: + return read_float_info (wps, wpmd); + + case ID_INT32_INFO: + return read_int32_info (wps, wpmd); + + case ID_CHANNEL_INFO: + return read_channel_info (wpc, wpmd); + + case ID_CONFIG_BLOCK: + return read_config_info (wpc, wpmd); + + case ID_WV_BITSTREAM: + return init_wv_bitstream (wpc, wpmd); + + case ID_SHAPING_WEIGHTS: + case ID_WVC_BITSTREAM: + case ID_WVX_BITSTREAM: + return TRUE; + + default: + return (wpmd->id & ID_OPTIONAL_DATA) ? TRUE : FALSE; + } +} diff --git a/lib/wavpack/readme.txt b/lib/wavpack/readme.txt new file mode 100644 index 00000000..4ccbdf42 --- /dev/null +++ b/lib/wavpack/readme.txt @@ -0,0 +1,68 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +This package contains a tiny version of the WavPack 4.40 decoder that might +be used in a "resource limited" CPU environment or form the basis for a +hardware decoding implementation. It is packaged with a demo command-line +program that accepts a WavPack audio file on stdin and outputs a RIFF wav +file to stdout. The program is standard C, and a win32 executable is +included which was compiled under MS Visual C++ 6.0 using this command: + +cl /O1 /DWIN32 wvfilter.c wputils.c unpack.c float.c metadata.c words.c bits.c + +WavPack data is read with a stream reading callback. No direct seeking is +provided for, but it is possible to start decoding anywhere in a WavPack +stream. In this case, WavPack will be able to provide the sample-accurate +position when it synchs with the data and begins decoding. The WIN32 macro +is used for Windows to force the stdin and stdout streams to be binary mode. + +Compared to the previous version, this library has been optimized somewhat +for improved performance in exchange for slightly larger code size. The +library also now includes hand-optimized assembly language versions of the +decorrelation functions for both the ColdFire (w/EMAC) and ARM processors. + +For demonstration purposes this uses a single static copy of the +WavpackContext structure, so obviously it cannot be used for more than one +file at a time. Also, this decoder will not handle "correction" files, plays +only the first two channels of multi-channel files, and is limited in +resolution in some large integer or floating point files (but always +provides at least 24 bits of resolution). It also will not accept WavPack +files from before version 4.0. + +The previous version of this library would handle float files by returning +32-bit floating-point data (even though no floating point math was used). +Because this library would normally be used for simply playing WavPack +files where lossless performance (beyond 24-bits) is not relevant, I have +changed this behavior. Now, these files will generate clipped 24-bit data. +The MODE_FLOAT flag will still be returned by WavpackGetMode(), but the +BitsPerSample and BytesPerSample queries will be 24 and 3, respectfully. +What this means is that an application that can handle 24-bit data will +now be able to handle floating point data (assuming that the MODE_FLOAT +flag is ignored). + +To make this code viable on the greatest number of hardware platforms, the +following are true: + + speed is about 5x realtime on an AMD K6 300 MHz + ("high" mode 16/44 stereo; normal mode is about twice that fast) + + no floating-point math required; just 32b * 32b = 32b int multiply + + large data areas are static and less than 4K total + executable code and tables are less than 40K + no malloc / free usage + +To maintain compatibility on various platforms, the following conventions +are used: + + a "char" must be exactly 8-bits + a "short" must be exactly 16-bits + an "int" must be at least 16-bits, but may be larger + the "long" type is not used to avoid problems with 64-bit compilers + +Questions or comments should be directed to david@wavpack.com diff --git a/lib/wavpack/unpack.c b/lib/wavpack/unpack.c new file mode 100644 index 00000000..e169c47f --- /dev/null +++ b/lib/wavpack/unpack.c @@ -0,0 +1,785 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// unpack.c + +// This module actually handles the decompression of the audio data, except +// for the entropy decoding which is handled by the words.c module. For +// maximum efficiency, the conversion is isolated to tight loops that handle +// an entire buffer. + +#include "wavpack.h" + +#include +#include + +#define LOSSY_MUTE + +///////////////////////////// executable code //////////////////////////////// + +// This function initializes everything required to unpack a WavPack block +// and must be called before unpack_samples() is called to obtain audio data. +// It is assumed that the WavpackHeader has been read into the wps->wphdr +// (in the current WavpackStream). This is where all the metadata blocks are +// scanned up to the one containing the audio bitstream. + +int unpack_init (WavpackContext *wpc) +{ + WavpackStream *wps = &wpc->stream; + WavpackMetadata wpmd; + + if (wps->wphdr.block_samples && wps->wphdr.block_index != (uint32_t) -1) + wps->sample_index = wps->wphdr.block_index; + + wps->mute_error = FALSE; + wps->crc = 0xffffffff; + CLEAR (wps->wvbits); + CLEAR (wps->decorr_passes); + CLEAR (wps->w); + + while (read_metadata_buff (wpc, &wpmd)) { + if (!process_metadata (wpc, &wpmd)) { + strcpy (wpc->error_message, "invalid metadata!"); + return FALSE; + } + + if (wpmd.id == ID_WV_BITSTREAM) + break; + } + + if (wps->wphdr.block_samples && !bs_is_open (&wps->wvbits)) { + strcpy (wpc->error_message, "invalid WavPack file!"); + return FALSE; + } + + if (wps->wphdr.block_samples) { + if ((wps->wphdr.flags & INT32_DATA) && wps->int32_sent_bits) + wpc->lossy_blocks = TRUE; + + if ((wps->wphdr.flags & FLOAT_DATA) && + wps->float_flags & (FLOAT_EXCEPTIONS | FLOAT_ZEROS_SENT | FLOAT_SHIFT_SENT | FLOAT_SHIFT_SAME)) + wpc->lossy_blocks = TRUE; + } + + return TRUE; +} + +// This function initialzes the main bitstream for audio samples, which must +// be in the "wv" file. + +int init_wv_bitstream (WavpackContext *wpc, WavpackMetadata *wpmd) +{ + WavpackStream *wps = &wpc->stream; + + if (wpmd->data) + bs_open_read (&wps->wvbits, wpmd->data, (unsigned char *) wpmd->data + wpmd->byte_length, NULL, NULL, 0); + else if (wpmd->byte_length) + bs_open_read (&wps->wvbits, wpc->read_buffer, wpc->read_buffer + sizeof (wpc->read_buffer), + wpc->infile, wpc->user_data, wpmd->byte_length + (wpmd->byte_length & 1)); + + return TRUE; +} + +// Read decorrelation terms from specified metadata block into the +// decorr_passes array. The terms range from -3 to 8, plus 17 & 18; +// other values are reserved and generate errors for now. The delta +// ranges from 0 to 7 with all values valid. Note that the terms are +// stored in the opposite order in the decorr_passes array compared +// to packing. + +int read_decorr_terms (WavpackStream *wps, WavpackMetadata *wpmd) +{ + int termcnt = wpmd->byte_length; + uchar *byteptr = wpmd->data; + struct decorr_pass *dpp; + + if (termcnt > MAX_NTERMS) + return FALSE; + + wps->num_terms = termcnt; + + for (dpp = wps->decorr_passes + termcnt - 1; termcnt--; dpp--) { + dpp->term = (int)(*byteptr & 0x1f) - 5; + dpp->delta = (*byteptr++ >> 5) & 0x7; + + if (!dpp->term || dpp->term < -3 || (dpp->term > MAX_TERM && dpp->term < 17) || dpp->term > 18) + return FALSE; + } + + return TRUE; +} + +// Read decorrelation weights from specified metadata block into the +// decorr_passes array. The weights range +/-1024, but are rounded and +// truncated to fit in signed chars for metadata storage. Weights are +// separate for the two channels and are specified from the "last" term +// (first during encode). Unspecified weights are set to zero. + +int read_decorr_weights (WavpackStream *wps, WavpackMetadata *wpmd) +{ + int termcnt = wpmd->byte_length, tcount; + signed char *byteptr = wpmd->data; + struct decorr_pass *dpp; + + if (!(wps->wphdr.flags & MONO_DATA)) + termcnt /= 2; + + if (termcnt > wps->num_terms) + return FALSE; + + for (tcount = wps->num_terms, dpp = wps->decorr_passes; tcount--; dpp++) + dpp->weight_A = dpp->weight_B = 0; + + while (--dpp >= wps->decorr_passes && termcnt--) { + dpp->weight_A = restore_weight (*byteptr++); + + if (!(wps->wphdr.flags & MONO_DATA)) + dpp->weight_B = restore_weight (*byteptr++); + } + + return TRUE; +} + +// Read decorrelation samples from specified metadata block into the +// decorr_passes array. The samples are signed 32-bit values, but are +// converted to signed log2 values for storage in metadata. Values are +// stored for both channels and are specified from the "last" term +// (first during encode) with unspecified samples set to zero. The +// number of samples stored varies with the actual term value, so +// those must obviously come first in the metadata. + +int read_decorr_samples (WavpackStream *wps, WavpackMetadata *wpmd) +{ + uchar *byteptr = wpmd->data; + uchar *endptr = byteptr + wpmd->byte_length; + struct decorr_pass *dpp; + int tcount; + + for (tcount = wps->num_terms, dpp = wps->decorr_passes; tcount--; dpp++) { + CLEAR (dpp->samples_A); + CLEAR (dpp->samples_B); + } + + if (wps->wphdr.version == 0x402 && (wps->wphdr.flags & HYBRID_FLAG)) { + byteptr += 2; + + if (!(wps->wphdr.flags & MONO_DATA)) + byteptr += 2; + } + + while (dpp-- > wps->decorr_passes && byteptr < endptr) + if (dpp->term > MAX_TERM) { + dpp->samples_A [0] = exp2s ((short)(byteptr [0] + (byteptr [1] << 8))); + dpp->samples_A [1] = exp2s ((short)(byteptr [2] + (byteptr [3] << 8))); + byteptr += 4; + + if (!(wps->wphdr.flags & MONO_DATA)) { + dpp->samples_B [0] = exp2s ((short)(byteptr [0] + (byteptr [1] << 8))); + dpp->samples_B [1] = exp2s ((short)(byteptr [2] + (byteptr [3] << 8))); + byteptr += 4; + } + } + else if (dpp->term < 0) { + dpp->samples_A [0] = exp2s ((short)(byteptr [0] + (byteptr [1] << 8))); + dpp->samples_B [0] = exp2s ((short)(byteptr [2] + (byteptr [3] << 8))); + byteptr += 4; + } + else { + int m = 0, cnt = dpp->term; + + while (cnt--) { + dpp->samples_A [m] = exp2s ((short)(byteptr [0] + (byteptr [1] << 8))); + byteptr += 2; + + if (!(wps->wphdr.flags & MONO_DATA)) { + dpp->samples_B [m] = exp2s ((short)(byteptr [0] + (byteptr [1] << 8))); + byteptr += 2; + } + + m++; + } + } + + return byteptr == endptr; +} + +// Read the int32 data from the specified metadata into the specified stream. +// This data is used for integer data that has more than 24 bits of magnitude +// or, in some cases, used to eliminate redundant bits from any audio stream. + +int read_int32_info (WavpackStream *wps, WavpackMetadata *wpmd) +{ + int bytecnt = wpmd->byte_length; + char *byteptr = wpmd->data; + + if (bytecnt != 4) + return FALSE; + + wps->int32_sent_bits = *byteptr++; + wps->int32_zeros = *byteptr++; + wps->int32_ones = *byteptr++; + wps->int32_dups = *byteptr; + return TRUE; +} + +// Read multichannel information from metadata. The first byte is the total +// number of channels and the following bytes represent the channel_mask +// as described for Microsoft WAVEFORMATEX. + +int read_channel_info (WavpackContext *wpc, WavpackMetadata *wpmd) +{ + int bytecnt = wpmd->byte_length, shift = 0; + char *byteptr = wpmd->data; + uint32_t mask = 0; + + if (!bytecnt || bytecnt > 5) + return FALSE; + + wpc->config.num_channels = *byteptr++; + + while (--bytecnt) { + mask |= (uint32_t) *byteptr++ << shift; + shift += 8; + } + + wpc->config.channel_mask = mask; + return TRUE; +} + +// Read configuration information from metadata. + +int read_config_info (WavpackContext *wpc, WavpackMetadata *wpmd) +{ + int bytecnt = wpmd->byte_length; + uchar *byteptr = wpmd->data; + + if (bytecnt >= 3) { + wpc->config.flags &= 0xff; + wpc->config.flags |= (int32_t) *byteptr++ << 8; + wpc->config.flags |= (int32_t) *byteptr++ << 16; + wpc->config.flags |= (int32_t) *byteptr << 24; + } + + return TRUE; +} + +// This monster actually unpacks the WavPack bitstream(s) into the specified +// buffer as 32-bit integers or floats (depending on orignal data). Lossy +// samples will be clipped to their original limits (i.e. 8-bit samples are +// clipped to -128/+127) but are still returned in int32_ts. It is up to the +// caller to potentially reformat this for the final output including any +// multichannel distribution, block alignment or endian compensation. The +// function unpack_init() must have been called and the entire WavPack block +// must still be visible (although wps->blockbuff will not be accessed again). +// For maximum clarity, the function is broken up into segments that handle +// various modes. This makes for a few extra infrequent flag checks, but +// makes the code easier to follow because the nesting does not become so +// deep. For maximum efficiency, the conversion is isolated to tight loops +// that handle an entire buffer. The function returns the total number of +// samples unpacked, which can be less than the number requested if an error +// occurs or the end of the block is reached. + +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) +extern void decorr_stereo_pass_cont_mcf5249 (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count); +#elif defined(CPU_ARM) && !defined(SIMULATOR) +extern void decorr_stereo_pass_cont_arm (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count); +extern void decorr_stereo_pass_cont_arml (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count); +#else +static void decorr_stereo_pass_cont (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count); +#endif + +static void decorr_mono_pass (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count); +static void decorr_stereo_pass (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count); +static void fixup_samples (WavpackStream *wps, int32_t *buffer, uint32_t sample_count); + +int32_t unpack_samples (WavpackContext *wpc, int32_t *buffer, uint32_t sample_count) +{ + WavpackStream *wps = &wpc->stream; + uint32_t flags = wps->wphdr.flags, crc = wps->crc, i; + int32_t mute_limit = (1L << ((flags & MAG_MASK) >> MAG_LSB)) + 2; + struct decorr_pass *dpp; + int32_t *bptr, *eptr; + int tcount; + + if (wps->sample_index + sample_count > wps->wphdr.block_index + wps->wphdr.block_samples) + sample_count = wps->wphdr.block_index + wps->wphdr.block_samples - wps->sample_index; + + if (wps->mute_error) { + memset (buffer, 0, sample_count * (flags & MONO_FLAG ? 4 : 8)); + wps->sample_index += sample_count; + return sample_count; + } + + if (flags & HYBRID_FLAG) + mute_limit *= 2; + + ///////////////////// handle version 4 mono data ///////////////////////// + + if (flags & MONO_DATA) { + eptr = buffer + sample_count; + i = get_words (buffer, sample_count, flags, &wps->w, &wps->wvbits); + + for (tcount = wps->num_terms, dpp = wps->decorr_passes; tcount--; dpp++) + decorr_mono_pass (dpp, buffer, sample_count); + + for (bptr = buffer; bptr < eptr; ++bptr) { + if (labs (bptr [0]) > mute_limit) { + i = bptr - buffer; + break; + } + + crc = crc * 3 + bptr [0]; + } + } + + //////////////////// handle version 4 stereo data //////////////////////// + + else { + eptr = buffer + (sample_count * 2); + i = get_words (buffer, sample_count, flags, &wps->w, &wps->wvbits); + + if (sample_count < 16) + for (tcount = wps->num_terms, dpp = wps->decorr_passes; tcount--; dpp++) + decorr_stereo_pass (dpp, buffer, sample_count); + else + for (tcount = wps->num_terms, dpp = wps->decorr_passes; tcount--; dpp++) { + decorr_stereo_pass (dpp, buffer, 8); +#if defined(CPU_COLDFIRE) && !defined(SIMULATOR) + decorr_stereo_pass_cont_mcf5249 (dpp, buffer + 16, sample_count - 8); +#elif defined(CPU_ARM) && !defined(SIMULATOR) + if (((flags & MAG_MASK) >> MAG_LSB) > 15) + decorr_stereo_pass_cont_arml (dpp, buffer + 16, sample_count - 8); + else + decorr_stereo_pass_cont_arm (dpp, buffer + 16, sample_count - 8); +#else + decorr_stereo_pass_cont (dpp, buffer + 16, sample_count - 8); +#endif + } + + if (flags & JOINT_STEREO) + for (bptr = buffer; bptr < eptr; bptr += 2) { + bptr [0] += (bptr [1] -= (bptr [0] >> 1)); + + if (labs (bptr [0]) > mute_limit || labs (bptr [1]) > mute_limit) { + i = (bptr - buffer) / 2; + break; + } + + crc = (crc * 3 + bptr [0]) * 3 + bptr [1]; + } + else + for (bptr = buffer; bptr < eptr; bptr += 2) { + if (labs (bptr [0]) > mute_limit || labs (bptr [1]) > mute_limit) { + i = (bptr - buffer) / 2; + break; + } + + crc = (crc * 3 + bptr [0]) * 3 + bptr [1]; + } + } + + if (i != sample_count) { + memset (buffer, 0, sample_count * (flags & MONO_FLAG ? 4 : 8)); + wps->mute_error = TRUE; + i = sample_count; + } + + fixup_samples (wps, buffer, i); + + if (flags & FALSE_STEREO) { + int32_t *dptr = buffer + i * 2; + int32_t *sptr = buffer + i; + int32_t c = i; + + while (c--) { + *--dptr = *--sptr; + *--dptr = *sptr; + } + } + + wps->sample_index += i; + wps->crc = crc; + + return i; +} + +static void decorr_stereo_pass (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count) +{ + int32_t delta = dpp->delta, weight_A = dpp->weight_A, weight_B = dpp->weight_B; + int32_t *bptr, *eptr = buffer + (sample_count * 2), sam_A, sam_B; + int m, k; + + switch (dpp->term) { + + case 17: + for (bptr = buffer; bptr < eptr; bptr += 2) { + sam_A = 2 * dpp->samples_A [0] - dpp->samples_A [1]; + dpp->samples_A [1] = dpp->samples_A [0]; + dpp->samples_A [0] = apply_weight (weight_A, sam_A) + bptr [0]; + update_weight (weight_A, delta, sam_A, bptr [0]); + bptr [0] = dpp->samples_A [0]; + + sam_A = 2 * dpp->samples_B [0] - dpp->samples_B [1]; + dpp->samples_B [1] = dpp->samples_B [0]; + dpp->samples_B [0] = apply_weight (weight_B, sam_A) + bptr [1]; + update_weight (weight_B, delta, sam_A, bptr [1]); + bptr [1] = dpp->samples_B [0]; + } + + break; + + case 18: + for (bptr = buffer; bptr < eptr; bptr += 2) { + sam_A = (3 * dpp->samples_A [0] - dpp->samples_A [1]) >> 1; + dpp->samples_A [1] = dpp->samples_A [0]; + dpp->samples_A [0] = apply_weight (weight_A, sam_A) + bptr [0]; + update_weight (weight_A, delta, sam_A, bptr [0]); + bptr [0] = dpp->samples_A [0]; + + sam_A = (3 * dpp->samples_B [0] - dpp->samples_B [1]) >> 1; + dpp->samples_B [1] = dpp->samples_B [0]; + dpp->samples_B [0] = apply_weight (weight_B, sam_A) + bptr [1]; + update_weight (weight_B, delta, sam_A, bptr [1]); + bptr [1] = dpp->samples_B [0]; + } + + break; + + default: + for (m = 0, k = dpp->term & (MAX_TERM - 1), bptr = buffer; bptr < eptr; bptr += 2) { + sam_A = dpp->samples_A [m]; + dpp->samples_A [k] = apply_weight (weight_A, sam_A) + bptr [0]; + update_weight (weight_A, delta, sam_A, bptr [0]); + bptr [0] = dpp->samples_A [k]; + + sam_A = dpp->samples_B [m]; + dpp->samples_B [k] = apply_weight (weight_B, sam_A) + bptr [1]; + update_weight (weight_B, delta, sam_A, bptr [1]); + bptr [1] = dpp->samples_B [k]; + + m = (m + 1) & (MAX_TERM - 1); + k = (k + 1) & (MAX_TERM - 1); + } + + if (m) { + int32_t temp_samples [MAX_TERM]; + + memcpy (temp_samples, dpp->samples_A, sizeof (dpp->samples_A)); + + for (k = 0; k < MAX_TERM; k++, m++) + dpp->samples_A [k] = temp_samples [m & (MAX_TERM - 1)]; + + memcpy (temp_samples, dpp->samples_B, sizeof (dpp->samples_B)); + + for (k = 0; k < MAX_TERM; k++, m++) + dpp->samples_B [k] = temp_samples [m & (MAX_TERM - 1)]; + } + + break; + + case -1: + for (bptr = buffer; bptr < eptr; bptr += 2) { + sam_A = bptr [0] + apply_weight (weight_A, dpp->samples_A [0]); + update_weight_clip (weight_A, delta, dpp->samples_A [0], bptr [0]); + bptr [0] = sam_A; + dpp->samples_A [0] = bptr [1] + apply_weight (weight_B, sam_A); + update_weight_clip (weight_B, delta, sam_A, bptr [1]); + bptr [1] = dpp->samples_A [0]; + } + + break; + + case -2: + for (bptr = buffer; bptr < eptr; bptr += 2) { + sam_B = bptr [1] + apply_weight (weight_B, dpp->samples_B [0]); + update_weight_clip (weight_B, delta, dpp->samples_B [0], bptr [1]); + bptr [1] = sam_B; + dpp->samples_B [0] = bptr [0] + apply_weight (weight_A, sam_B); + update_weight_clip (weight_A, delta, sam_B, bptr [0]); + bptr [0] = dpp->samples_B [0]; + } + + break; + + case -3: + for (bptr = buffer; bptr < eptr; bptr += 2) { + sam_A = bptr [0] + apply_weight (weight_A, dpp->samples_A [0]); + update_weight_clip (weight_A, delta, dpp->samples_A [0], bptr [0]); + sam_B = bptr [1] + apply_weight (weight_B, dpp->samples_B [0]); + update_weight_clip (weight_B, delta, dpp->samples_B [0], bptr [1]); + bptr [0] = dpp->samples_B [0] = sam_A; + bptr [1] = dpp->samples_A [0] = sam_B; + } + + break; + } + + dpp->weight_A = weight_A; + dpp->weight_B = weight_B; +} + +#if (!defined(CPU_COLDFIRE) && !defined(CPU_ARM)) || defined(SIMULATOR) + +static void decorr_stereo_pass_cont (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count) +{ + int32_t delta = dpp->delta, weight_A = dpp->weight_A, weight_B = dpp->weight_B; + int32_t *bptr, *tptr, *eptr = buffer + (sample_count * 2), sam_A, sam_B; + int k, i; + + switch (dpp->term) { + + case 17: + for (bptr = buffer; bptr < eptr; bptr += 2) { + sam_A = 2 * bptr [-2] - bptr [-4]; + bptr [0] = apply_weight (weight_A, sam_A) + (sam_B = bptr [0]); + update_weight (weight_A, delta, sam_A, sam_B); + + sam_A = 2 * bptr [-1] - bptr [-3]; + bptr [1] = apply_weight (weight_B, sam_A) + (sam_B = bptr [1]); + update_weight (weight_B, delta, sam_A, sam_B); + } + + dpp->samples_B [0] = bptr [-1]; + dpp->samples_A [0] = bptr [-2]; + dpp->samples_B [1] = bptr [-3]; + dpp->samples_A [1] = bptr [-4]; + break; + + case 18: + for (bptr = buffer; bptr < eptr; bptr += 2) { + sam_A = (3 * bptr [-2] - bptr [-4]) >> 1; + bptr [0] = apply_weight (weight_A, sam_A) + (sam_B = bptr [0]); + update_weight (weight_A, delta, sam_A, sam_B); + + sam_A = (3 * bptr [-1] - bptr [-3]) >> 1; + bptr [1] = apply_weight (weight_B, sam_A) + (sam_B = bptr [1]); + update_weight (weight_B, delta, sam_A, sam_B); + } + + dpp->samples_B [0] = bptr [-1]; + dpp->samples_A [0] = bptr [-2]; + dpp->samples_B [1] = bptr [-3]; + dpp->samples_A [1] = bptr [-4]; + break; + + default: + for (bptr = buffer, tptr = buffer - (dpp->term * 2); bptr < eptr; bptr += 2, tptr += 2) { + bptr [0] = apply_weight (weight_A, tptr [0]) + (sam_A = bptr [0]); + update_weight (weight_A, delta, tptr [0], sam_A); + + bptr [1] = apply_weight (weight_B, tptr [1]) + (sam_A = bptr [1]); + update_weight (weight_B, delta, tptr [1], sam_A); + } + + for (k = dpp->term - 1, i = 8; i--; k--) { + dpp->samples_B [k & (MAX_TERM - 1)] = *--bptr; + dpp->samples_A [k & (MAX_TERM - 1)] = *--bptr; + } + + break; + + case -1: + for (bptr = buffer; bptr < eptr; bptr += 2) { + bptr [0] = apply_weight (weight_A, bptr [-1]) + (sam_A = bptr [0]); + update_weight_clip (weight_A, delta, bptr [-1], sam_A); + bptr [1] = apply_weight (weight_B, bptr [0]) + (sam_A = bptr [1]); + update_weight_clip (weight_B, delta, bptr [0], sam_A); + } + + dpp->samples_A [0] = bptr [-1]; + break; + + case -2: + for (bptr = buffer; bptr < eptr; bptr += 2) { + bptr [1] = apply_weight (weight_B, bptr [-2]) + (sam_A = bptr [1]); + update_weight_clip (weight_B, delta, bptr [-2], sam_A); + bptr [0] = apply_weight (weight_A, bptr [1]) + (sam_A = bptr [0]); + update_weight_clip (weight_A, delta, bptr [1], sam_A); + } + + dpp->samples_B [0] = bptr [-2]; + break; + + case -3: + for (bptr = buffer; bptr < eptr; bptr += 2) { + bptr [0] = apply_weight (weight_A, bptr [-1]) + (sam_A = bptr [0]); + update_weight_clip (weight_A, delta, bptr [-1], sam_A); + bptr [1] = apply_weight (weight_B, bptr [-2]) + (sam_A = bptr [1]); + update_weight_clip (weight_B, delta, bptr [-2], sam_A); + } + + dpp->samples_A [0] = bptr [-1]; + dpp->samples_B [0] = bptr [-2]; + break; + } + + dpp->weight_A = weight_A; + dpp->weight_B = weight_B; +} + +#endif + +static void decorr_mono_pass (struct decorr_pass *dpp, int32_t *buffer, int32_t sample_count) +{ + int32_t delta = dpp->delta, weight_A = dpp->weight_A; + int32_t *bptr, *eptr = buffer + sample_count, sam_A; + int m, k; + + switch (dpp->term) { + + case 17: + for (bptr = buffer; bptr < eptr; bptr++) { + sam_A = 2 * dpp->samples_A [0] - dpp->samples_A [1]; + dpp->samples_A [1] = dpp->samples_A [0]; + dpp->samples_A [0] = apply_weight (weight_A, sam_A) + bptr [0]; + update_weight (weight_A, delta, sam_A, bptr [0]); + bptr [0] = dpp->samples_A [0]; + } + + break; + + case 18: + for (bptr = buffer; bptr < eptr; bptr++) { + sam_A = (3 * dpp->samples_A [0] - dpp->samples_A [1]) >> 1; + dpp->samples_A [1] = dpp->samples_A [0]; + dpp->samples_A [0] = apply_weight (weight_A, sam_A) + bptr [0]; + update_weight (weight_A, delta, sam_A, bptr [0]); + bptr [0] = dpp->samples_A [0]; + } + + break; + + default: + for (m = 0, k = dpp->term & (MAX_TERM - 1), bptr = buffer; bptr < eptr; bptr++) { + sam_A = dpp->samples_A [m]; + dpp->samples_A [k] = apply_weight (weight_A, sam_A) + bptr [0]; + update_weight (weight_A, delta, sam_A, bptr [0]); + bptr [0] = dpp->samples_A [k]; + m = (m + 1) & (MAX_TERM - 1); + k = (k + 1) & (MAX_TERM - 1); + } + + if (m) { + int32_t temp_samples [MAX_TERM]; + + memcpy (temp_samples, dpp->samples_A, sizeof (dpp->samples_A)); + + for (k = 0; k < MAX_TERM; k++, m++) + dpp->samples_A [k] = temp_samples [m & (MAX_TERM - 1)]; + } + + break; + } + + dpp->weight_A = weight_A; +} + + +// This is a helper function for unpack_samples() that applies several final +// operations. First, if the data is 32-bit float data, then that conversion +// is done in the float.c module (whether lossy or lossless) and we return. +// Otherwise, if the extended integer data applies, then that operation is +// executed first. If the unpacked data is lossy (and not corrected) then +// it is clipped and shifted in a single operation. Otherwise, if it's +// lossless then the last step is to apply the final shift (if any). + +static void fixup_samples (WavpackStream *wps, int32_t *buffer, uint32_t sample_count) +{ + uint32_t flags = wps->wphdr.flags; + int shift = (flags & SHIFT_MASK) >> SHIFT_LSB; + + if (flags & FLOAT_DATA) { + float_values (wps, buffer, (flags & MONO_FLAG) ? sample_count : sample_count * 2); + return; + } + + if (flags & INT32_DATA) { + uint32_t count = (flags & MONO_FLAG) ? sample_count : sample_count * 2; + int sent_bits = wps->int32_sent_bits, zeros = wps->int32_zeros; + int ones = wps->int32_ones, dups = wps->int32_dups; + int32_t *dptr = buffer; + + if (!(flags & HYBRID_FLAG) && !sent_bits && (zeros + ones + dups)) + while (count--) { + if (zeros) + *dptr <<= zeros; + else if (ones) + *dptr = ((*dptr + 1) << ones) - 1; + else if (dups) + *dptr = ((*dptr + (*dptr & 1)) << dups) - (*dptr & 1); + + dptr++; + } + else + shift += zeros + sent_bits + ones + dups; + } + + if (flags & HYBRID_FLAG) { + int32_t min_value, max_value, min_shifted, max_shifted; + + switch (flags & BYTES_STORED) { + case 0: + min_shifted = (min_value = -128 >> shift) << shift; + max_shifted = (max_value = 127 >> shift) << shift; + break; + + case 1: + min_shifted = (min_value = -32768 >> shift) << shift; + max_shifted = (max_value = 32767 >> shift) << shift; + break; + + case 2: + min_shifted = (min_value = -8388608 >> shift) << shift; + max_shifted = (max_value = 8388607 >> shift) << shift; + break; + + case 3: + default: + min_shifted = (min_value = (int32_t) 0x80000000 >> shift) << shift; + max_shifted = (max_value = (int32_t) 0x7FFFFFFF >> shift) << shift; + break; + } + + if (!(flags & MONO_FLAG)) + sample_count *= 2; + + while (sample_count--) { + if (*buffer < min_value) + *buffer++ = min_shifted; + else if (*buffer > max_value) + *buffer++ = max_shifted; + else + *buffer++ <<= shift; + } + } + else if (shift) { + if (!(flags & MONO_FLAG)) + sample_count *= 2; + + while (sample_count--) + *buffer++ <<= shift; + } +} + +// This function checks the crc value(s) for an unpacked block, returning the +// number of actual crc errors detected for the block. The block must be +// completely unpacked before this test is valid. For losslessly unpacked +// blocks of float or extended integer data the extended crc is also checked. +// Note that WavPack's crc is not a CCITT approved polynomial algorithm, but +// is a much simpler method that is virtually as robust for real world data. + +int check_crc_error (WavpackContext *wpc) +{ + WavpackStream *wps = &wpc->stream; + int result = 0; + + if (wps->crc != wps->wphdr.crc) + ++result; + + return result; +} diff --git a/lib/wavpack/wavpack.h b/lib/wavpack/wavpack.h new file mode 100644 index 00000000..d6c35131 --- /dev/null +++ b/lib/wavpack/wavpack.h @@ -0,0 +1,394 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2004 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// wavpack.h + +#ifdef __cplusplus +extern "C" { +#endif + +#include + +// This header file contains all the definitions required by WavPack. + +#ifdef __BORLANDC__ +typedef unsigned long uint32_t; +typedef long int32_t; +#elif defined(_WIN32) && !defined(__MINGW32__) +#include +typedef unsigned __int64 uint64_t; +typedef unsigned __int32 uint32_t; +typedef __int64 int64_t; +typedef __int32 int32_t; +#else +#include +#endif + +typedef unsigned char uchar; + +#if !defined(__GNUC__) || defined(WIN32) +typedef unsigned short ushort; +typedef unsigned int uint; +#endif + +#include + +#define FALSE 0 +#define TRUE 1 + +////////////////////////////// WavPack Header ///////////////////////////////// + +// Note that this is the ONLY structure that is written to (or read from) +// WavPack 4.0 files, and is the preamble to every block in both the .wv +// and .wvc files. + +typedef struct { + char ckID [4]; + uint32_t ckSize; + short version; + uchar track_no, index_no; + uint32_t total_samples, block_index, block_samples, flags, crc; +} WavpackHeader; + +#define WavpackHeaderFormat "4LS2LLLLL" + +// or-values for "flags" + +#define BYTES_STORED 3 // 1-4 bytes/sample +#define MONO_FLAG 4 // not stereo +#define HYBRID_FLAG 8 // hybrid mode +#define JOINT_STEREO 0x10 // joint stereo +#define CROSS_DECORR 0x20 // no-delay cross decorrelation +#define HYBRID_SHAPE 0x40 // noise shape (hybrid mode only) +#define FLOAT_DATA 0x80 // ieee 32-bit floating point data + +#define INT32_DATA 0x100 // special extended int handling +#define HYBRID_BITRATE 0x200 // bitrate noise (hybrid mode only) +#define HYBRID_BALANCE 0x400 // balance noise (hybrid stereo mode only) + +#define INITIAL_BLOCK 0x800 // initial block of multichannel segment +#define FINAL_BLOCK 0x1000 // final block of multichannel segment + +#define SHIFT_LSB 13 +#define SHIFT_MASK (0x1fL << SHIFT_LSB) + +#define MAG_LSB 18 +#define MAG_MASK (0x1fL << MAG_LSB) + +#define SRATE_LSB 23 +#define SRATE_MASK (0xfL << SRATE_LSB) + +#define FALSE_STEREO 0x40000000 // block is stereo, but data is mono + +#define IGNORED_FLAGS 0x18000000 // reserved, but ignore if encountered +#define NEW_SHAPING 0x20000000 // use IIR filter for negative shaping +#define UNKNOWN_FLAGS 0x80000000 // also reserved, but refuse decode if + // encountered + +#define MONO_DATA (MONO_FLAG | FALSE_STEREO) + +#define MIN_STREAM_VERS 0x402 // lowest stream version we'll decode +#define MAX_STREAM_VERS 0x410 // highest stream version we'll decode + +//////////////////////////// WavPack Metadata ///////////////////////////////// + +// This is an internal representation of metadata. + +typedef struct { + int32_t byte_length; + void *data; + uchar id; +} WavpackMetadata; + +#define ID_OPTIONAL_DATA 0x20 +#define ID_ODD_SIZE 0x40 +#define ID_LARGE 0x80 + +#define ID_DUMMY 0x0 +#define ID_ENCODER_INFO 0x1 +#define ID_DECORR_TERMS 0x2 +#define ID_DECORR_WEIGHTS 0x3 +#define ID_DECORR_SAMPLES 0x4 +#define ID_ENTROPY_VARS 0x5 +#define ID_HYBRID_PROFILE 0x6 +#define ID_SHAPING_WEIGHTS 0x7 +#define ID_FLOAT_INFO 0x8 +#define ID_INT32_INFO 0x9 +#define ID_WV_BITSTREAM 0xa +#define ID_WVC_BITSTREAM 0xb +#define ID_WVX_BITSTREAM 0xc +#define ID_CHANNEL_INFO 0xd + +#define ID_RIFF_HEADER (ID_OPTIONAL_DATA | 0x1) +#define ID_RIFF_TRAILER (ID_OPTIONAL_DATA | 0x2) +#define ID_REPLAY_GAIN (ID_OPTIONAL_DATA | 0x3) +#define ID_CUESHEET (ID_OPTIONAL_DATA | 0x4) +#define ID_CONFIG_BLOCK (ID_OPTIONAL_DATA | 0x5) +#define ID_MD5_CHECKSUM (ID_OPTIONAL_DATA | 0x6) + +///////////////////////// WavPack Configuration /////////////////////////////// + +// This internal structure is used during encode to provide configuration to +// the encoding engine and during decoding to provide fle information back to +// the higher level functions. Not all fields are used in both modes. + +typedef struct { + int bits_per_sample, bytes_per_sample; + int num_channels, float_norm_exp; + uint32_t flags, sample_rate, channel_mask; +} WavpackConfig; + +#define CONFIG_BYTES_STORED 3 // 1-4 bytes/sample +#define CONFIG_MONO_FLAG 4 // not stereo +#define CONFIG_HYBRID_FLAG 8 // hybrid mode +#define CONFIG_JOINT_STEREO 0x10 // joint stereo +#define CONFIG_CROSS_DECORR 0x20 // no-delay cross decorrelation +#define CONFIG_HYBRID_SHAPE 0x40 // noise shape (hybrid mode only) +#define CONFIG_FLOAT_DATA 0x80 // ieee 32-bit floating point data + +#define CONFIG_FAST_FLAG 0x200 // fast mode +#define CONFIG_HIGH_FLAG 0x800 // high quality mode +#define CONFIG_VERY_HIGH_FLAG 0x1000 // very high +#define CONFIG_BITRATE_KBPS 0x2000 // bitrate is kbps, not bits / sample +#define CONFIG_AUTO_SHAPING 0x4000 // automatic noise shaping +#define CONFIG_SHAPE_OVERRIDE 0x8000 // shaping mode specified +#define CONFIG_JOINT_OVERRIDE 0x10000 // joint-stereo mode specified +#define CONFIG_CREATE_EXE 0x40000 // create executable +#define CONFIG_CREATE_WVC 0x80000 // create correction file +#define CONFIG_OPTIMIZE_WVC 0x100000 // maximize bybrid compression +#define CONFIG_CALC_NOISE 0x800000 // calc noise in hybrid mode +#define CONFIG_LOSSY_MODE 0x1000000 // obsolete (for information) +#define CONFIG_EXTRA_MODE 0x2000000 // extra processing mode +#define CONFIG_SKIP_WVX 0x4000000 // no wvx stream w/ floats & big ints +#define CONFIG_MD5_CHECKSUM 0x8000000 // compute & store MD5 signature +#define CONFIG_OPTIMIZE_MONO 0x80000000 // optimize for mono streams posing as stereo + +//////////////////////////////// WavPack Stream /////////////////////////////// + +// This internal structure contains everything required to handle a WavPack +// "stream", which is defined as a stereo or mono stream of audio samples. For +// multichannel audio several of these would be required. Each stream contains +// pointers to hold a complete allocated block of WavPack data, although it's +// possible to decode WavPack blocks without buffering an entire block. + +typedef int32_t (*read_stream)(void *, void *, int32_t); + +typedef struct bs { + uchar *buf, *end, *ptr; + void (*wrap)(struct bs *bs); + uint32_t file_bytes, sr; + int error, bc; + read_stream file; + void *user_data; +} Bitstream; + +#define MAX_NTERMS 16 +#define MAX_TERM 8 + +struct decorr_pass { + short term, delta, weight_A, weight_B; + int32_t samples_A [MAX_TERM], samples_B [MAX_TERM]; +}; + +struct entropy_data { + uint32_t median [3], slow_level, error_limit; +}; + +struct words_data { + uint32_t bitrate_delta [2], bitrate_acc [2]; + uint32_t pend_data, holding_one, zeros_acc; + int holding_zero, pend_count; + struct entropy_data c [2]; +}; + +typedef struct { + WavpackHeader wphdr; + Bitstream wvbits; + + struct words_data w; + + int num_terms, mute_error; + uint32_t sample_index, crc; + + uchar int32_sent_bits, int32_zeros, int32_ones, int32_dups; + uchar float_flags, float_shift, float_max_exp, float_norm_exp; + + struct decorr_pass decorr_passes [MAX_NTERMS]; + +} WavpackStream; + +// flags for float_flags: + +#define FLOAT_SHIFT_ONES 1 // bits left-shifted into float = '1' +#define FLOAT_SHIFT_SAME 2 // bits left-shifted into float are the same +#define FLOAT_SHIFT_SENT 4 // bits shifted into float are sent literally +#define FLOAT_ZEROS_SENT 8 // "zeros" are not all real zeros +#define FLOAT_NEG_ZEROS 0x10 // contains negative zeros +#define FLOAT_EXCEPTIONS 0x20 // contains exceptions (inf, nan, etc.) + +/////////////////////////////// WavPack Context /////////////////////////////// + +// This internal structure holds everything required to encode or decode WavPack +// files. It is recommended that direct access to this structure be minimized +// and the provided utilities used instead. + +typedef struct { + WavpackConfig config; + WavpackStream stream; + + uchar read_buffer [1024]; + char error_message [80]; + + read_stream infile; + void *user_data; + uint32_t total_samples, crc_errors, first_flags; + int open_flags, norm_offset, reduced_channels, lossy_blocks; + +} WavpackContext; + +//////////////////////// function prototypes and macros ////////////////////// + +#define CLEAR(destin) memset (&destin, 0, sizeof (destin)); + +// bits.c + +void bs_open_read (Bitstream *bs, uchar *buffer_start, uchar *buffer_end, read_stream file, void *user_data, uint32_t file_bytes); + +#define bs_is_open(bs) ((bs)->ptr != NULL) + +#define getbit(bs) ( \ + (((bs)->bc) ? \ + ((bs)->bc--, (bs)->sr & 1) : \ + (((++((bs)->ptr) != (bs)->end) ? (void) 0 : (bs)->wrap (bs)), (bs)->bc = 7, ((bs)->sr = *((bs)->ptr)) & 1) \ + ) ? \ + ((bs)->sr >>= 1, 1) : \ + ((bs)->sr >>= 1, 0) \ +) + +#define getbits(value, nbits, bs) { \ + while ((nbits) > (bs)->bc) { \ + if (++((bs)->ptr) == (bs)->end) (bs)->wrap (bs); \ + (bs)->sr |= (int32_t)*((bs)->ptr) << (bs)->bc; \ + (bs)->bc += 8; \ + } \ + *(value) = (bs)->sr; \ + if ((bs)->bc > 32) { \ + (bs)->bc -= (nbits); \ + (bs)->sr = *((bs)->ptr) >> (8 - (bs)->bc); \ + } \ + else { \ + (bs)->bc -= (nbits); \ + (bs)->sr >>= (nbits); \ + } \ +} + +void little_endian_to_native (void *data, char *format); +void native_to_little_endian (void *data, char *format); + +// These macros implement the weight application and update operations +// that are at the heart of the decorrelation loops. Note that when there +// are several alternative versions of the same macro (marked with PERFCOND) +// then the versions are functionally equivalent with respect to WavPack +// decoding and the user should choose the one that provides the best +// performance. This may be easier to check when NOT using the assembly +// language optimizations. + +#if 1 // PERFCOND +#define apply_weight_i(weight, sample) ((weight * sample + 512) >> 10) +#else +#define apply_weight_i(weight, sample) ((((weight * sample) >> 8) + 2) >> 2) +#endif + +#define apply_weight_f(weight, sample) (((((sample & 0xffffL) * weight) >> 9) + \ + (((sample & ~0xffffL) >> 9) * weight) + 1) >> 1) + +#if 0 // PERFCOND +#define apply_weight(weight, sample) (sample != (short) sample ? \ + apply_weight_f (weight, sample) : apply_weight_i (weight, sample)) +#else +#define apply_weight(weight, sample) ((int32_t)((weight * (int64_t) sample + 512) >> 10)) +#endif + +#if 0 // PERFCOND +#define update_weight(weight, delta, source, result) \ + if (source && result) { int32_t s = (int32_t) (source ^ result) >> 31; weight = (delta ^ s) + (weight - s); } +#elif 0 +#define update_weight(weight, delta, source, result) \ + if (source && result) weight += (((source ^ result) >> 30) | 1) * delta +#else +#define update_weight(weight, delta, source, result) \ + if (source && result) (source ^ result) < 0 ? (weight -= delta) : (weight += delta) +#endif + +#define update_weight_clip(weight, delta, source, result) \ + if (source && result && ((source ^ result) < 0 ? (weight -= delta) < -1024 : (weight += delta) > 1024)) \ + weight = weight < 0 ? -1024 : 1024 + +// unpack.c + +int unpack_init (WavpackContext *wpc); +int init_wv_bitstream (WavpackContext *wpc, WavpackMetadata *wpmd); +int read_decorr_terms (WavpackStream *wps, WavpackMetadata *wpmd); +int read_decorr_weights (WavpackStream *wps, WavpackMetadata *wpmd); +int read_decorr_samples (WavpackStream *wps, WavpackMetadata *wpmd); +int read_float_info (WavpackStream *wps, WavpackMetadata *wpmd); +int read_int32_info (WavpackStream *wps, WavpackMetadata *wpmd); +int read_channel_info (WavpackContext *wpc, WavpackMetadata *wpmd); +int read_config_info (WavpackContext *wpc, WavpackMetadata *wpmd); +int32_t unpack_samples (WavpackContext *wpc, int32_t *buffer, uint32_t sample_count); +int check_crc_error (WavpackContext *wpc); + +// metadata.c stuff + +int read_metadata_buff (WavpackContext *wpc, WavpackMetadata *wpmd); +int process_metadata (WavpackContext *wpc, WavpackMetadata *wpmd); + +// words.c stuff + +int read_entropy_vars (WavpackStream *wps, WavpackMetadata *wpmd); +int read_hybrid_profile (WavpackStream *wps, WavpackMetadata *wpmd); +int32_t get_words (int32_t *buffer, int nsamples, uint32_t flags, + struct words_data *w, Bitstream *bs); +int32_t exp2s (int log); +int restore_weight (signed char weight); + +#define WORD_EOF (1L << 31) + +// float.c + +int read_float_info (WavpackStream *wps, WavpackMetadata *wpmd); +void float_values (WavpackStream *wps, int32_t *values, int32_t num_values); + +// wputils.c + +int WavpackOpenFileInput (WavpackContext *wpc, read_stream infile, void *user_data, char *error); + +int WavpackGetMode (WavpackContext *wpc); + +#define MODE_WVC 0x1 +#define MODE_LOSSLESS 0x2 +#define MODE_HYBRID 0x4 +#define MODE_FLOAT 0x8 +#define MODE_VALID_TAG 0x10 +#define MODE_HIGH 0x20 +#define MODE_FAST 0x40 + +uint32_t WavpackUnpackSamples (WavpackContext *wpc, int32_t *buffer, uint32_t samples); +uint32_t WavpackGetNumSamples (WavpackContext *wpc); +uint32_t WavpackGetSampleIndex (WavpackContext *wpc); +int WavpackGetNumErrors (WavpackContext *wpc); +int WavpackLossyBlocks (WavpackContext *wpc); +uint32_t WavpackGetSampleRate (WavpackContext *wpc); +int WavpackGetBitsPerSample (WavpackContext *wpc); +int WavpackGetBytesPerSample (WavpackContext *wpc); +int WavpackGetNumChannels (WavpackContext *wpc); +int WavpackGetReducedChannels (WavpackContext *wpc); + +#ifdef __cplusplus +} +#endif diff --git a/lib/wavpack/words.c b/lib/wavpack/words.c new file mode 100644 index 00000000..0e5a3db7 --- /dev/null +++ b/lib/wavpack/words.c @@ -0,0 +1,560 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// words.c + +// This module provides entropy word encoding and decoding functions using +// a variation on the Rice method. This was introduced in version 3.93 +// because it allows splitting the data into a "lossy" stream and a +// "correction" stream in a very efficient manner and is therefore ideal +// for the "hybrid" mode. For 4.0, the efficiency of this method was +// significantly improved by moving away from the normal Rice restriction of +// using powers of two for the modulus divisions and now the method can be +// used for both hybrid and pure lossless encoding. + +// Samples are divided by median probabilities at 5/7 (71.43%), 10/49 (20.41%), +// and 20/343 (5.83%). Each zone has 3.5 times fewer samples than the +// previous. Using standard Rice coding on this data would result in 1.4 +// bits per sample average (not counting sign bit). However, there is a +// very simple encoding that is over 99% efficient with this data and +// results in about 1.22 bits per sample. + +#include "wavpack.h" + +#include + +//////////////////////////////// local macros ///////////////////////////////// + +#define LIMIT_ONES 16 // maximum consecutive 1s sent for "div" data + +// these control the time constant "slow_level" which is used for hybrid mode +// that controls bitrate as a function of residual level (HYBRID_BITRATE). +#define SLS 8 +#define SLO ((1 << (SLS - 1))) + +// these control the time constant of the 3 median level breakpoints +#define DIV0 128 // 5/7 of samples +#define DIV1 64 // 10/49 of samples +#define DIV2 32 // 20/343 of samples + +// this macro retrieves the specified median breakpoint (without frac; min = 1) +#define GET_MED(med) (((c->median [med]) >> 4) + 1) + +// These macros update the specified median breakpoints. Note that the median +// is incremented when the sample is higher than the median, else decremented. +// They are designed so that the median will never drop below 1 and the value +// is essentially stationary if there are 2 increments for every 5 decrements. + +#define INC_MED0() (c->median [0] += ((c->median [0] + DIV0) / DIV0) * 5) +#define DEC_MED0() (c->median [0] -= ((c->median [0] + (DIV0-2)) / DIV0) * 2) +#define INC_MED1() (c->median [1] += ((c->median [1] + DIV1) / DIV1) * 5) +#define DEC_MED1() (c->median [1] -= ((c->median [1] + (DIV1-2)) / DIV1) * 2) +#define INC_MED2() (c->median [2] += ((c->median [2] + DIV2) / DIV2) * 5) +#define DEC_MED2() (c->median [2] -= ((c->median [2] + (DIV2-2)) / DIV2) * 2) + +#define count_bits(av) ( \ + (av) < (1 << 8) ? nbits_table [av] : \ + ( \ + (av) < (1L << 16) ? nbits_table [(av) >> 8] + 8 : \ + ((av) < (1L << 24) ? nbits_table [(av) >> 16] + 16 : nbits_table [(av) >> 24] + 24) \ + ) \ +) + +///////////////////////////// local table storage //////////////////////////// + +const char nbits_table [] = { + 0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, // 0 - 15 + 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, // 16 - 31 + 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, // 32 - 47 + 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, // 48 - 63 + 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, // 64 - 79 + 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, // 80 - 95 + 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, // 96 - 111 + 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, // 112 - 127 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, // 128 - 143 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, // 144 - 159 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, // 160 - 175 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, // 176 - 191 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, // 192 - 207 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, // 208 - 223 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, // 224 - 239 + 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8 // 240 - 255 +}; + +static const uchar log2_table [] = { + 0x00, 0x01, 0x03, 0x04, 0x06, 0x07, 0x09, 0x0a, 0x0b, 0x0d, 0x0e, 0x10, 0x11, 0x12, 0x14, 0x15, + 0x16, 0x18, 0x19, 0x1a, 0x1c, 0x1d, 0x1e, 0x20, 0x21, 0x22, 0x24, 0x25, 0x26, 0x28, 0x29, 0x2a, + 0x2c, 0x2d, 0x2e, 0x2f, 0x31, 0x32, 0x33, 0x34, 0x36, 0x37, 0x38, 0x39, 0x3b, 0x3c, 0x3d, 0x3e, + 0x3f, 0x41, 0x42, 0x43, 0x44, 0x45, 0x47, 0x48, 0x49, 0x4a, 0x4b, 0x4d, 0x4e, 0x4f, 0x50, 0x51, + 0x52, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, 0x5c, 0x5d, 0x5e, 0x5f, 0x60, 0x61, 0x62, 0x63, + 0x64, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72, 0x74, 0x75, + 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, + 0x86, 0x87, 0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, + 0x96, 0x97, 0x98, 0x99, 0x9a, 0x9b, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1, 0xa2, 0xa3, 0xa4, + 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf, 0xb0, 0xb1, 0xb2, 0xb2, + 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf, 0xc0, 0xc0, + 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xcb, 0xcb, 0xcc, 0xcd, 0xce, + 0xcf, 0xd0, 0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd8, 0xd9, 0xda, 0xdb, + 0xdc, 0xdc, 0xdd, 0xde, 0xdf, 0xe0, 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe4, 0xe5, 0xe6, 0xe7, 0xe7, + 0xe8, 0xe9, 0xea, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xee, 0xef, 0xf0, 0xf1, 0xf1, 0xf2, 0xf3, 0xf4, + 0xf4, 0xf5, 0xf6, 0xf7, 0xf7, 0xf8, 0xf9, 0xf9, 0xfa, 0xfb, 0xfc, 0xfc, 0xfd, 0xfe, 0xff, 0xff +}; + +static const uchar exp2_table [] = { + 0x00, 0x01, 0x01, 0x02, 0x03, 0x03, 0x04, 0x05, 0x06, 0x06, 0x07, 0x08, 0x08, 0x09, 0x0a, 0x0b, + 0x0b, 0x0c, 0x0d, 0x0e, 0x0e, 0x0f, 0x10, 0x10, 0x11, 0x12, 0x13, 0x13, 0x14, 0x15, 0x16, 0x16, + 0x17, 0x18, 0x19, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1d, 0x1e, 0x1f, 0x20, 0x20, 0x21, 0x22, 0x23, + 0x24, 0x24, 0x25, 0x26, 0x27, 0x28, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2c, 0x2d, 0x2e, 0x2f, 0x30, + 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3a, 0x3b, 0x3c, 0x3d, + 0x3e, 0x3f, 0x40, 0x41, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x48, 0x49, 0x4a, 0x4b, + 0x4c, 0x4d, 0x4e, 0x4f, 0x50, 0x51, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, + 0x5b, 0x5c, 0x5d, 0x5e, 0x5e, 0x5f, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, + 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, + 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x87, 0x88, 0x89, 0x8a, + 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x90, 0x91, 0x92, 0x93, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9a, 0x9b, + 0x9c, 0x9d, 0x9f, 0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, + 0xaf, 0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xbc, 0xbd, 0xbe, 0xbf, 0xc0, + 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc8, 0xc9, 0xca, 0xcb, 0xcd, 0xce, 0xcf, 0xd0, 0xd2, 0xd3, 0xd4, + 0xd6, 0xd7, 0xd8, 0xd9, 0xdb, 0xdc, 0xdd, 0xde, 0xe0, 0xe1, 0xe2, 0xe4, 0xe5, 0xe6, 0xe8, 0xe9, + 0xea, 0xec, 0xed, 0xee, 0xf0, 0xf1, 0xf2, 0xf4, 0xf5, 0xf6, 0xf8, 0xf9, 0xfa, 0xfc, 0xfd, 0xff +}; + +static const char ones_count_table [] = { + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,5, + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,6, + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,5, + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,7, + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,5, + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,6, + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,5, + 0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,4,0,1,0,2,0,1,0,3,0,1,0,2,0,1,0,8 +}; + +///////////////////////////// executable code //////////////////////////////// + +void init_words (WavpackStream *wps) +{ + CLEAR (wps->w); +} + +static int mylog2 (uint32_t avalue); + +// Read the median log2 values from the specifed metadata structure, convert +// them back to 32-bit unsigned values and store them. If length is not +// exactly correct then we flag and return an error. + +int read_entropy_vars (WavpackStream *wps, WavpackMetadata *wpmd) +{ + uchar *byteptr = wpmd->data; + + if (wpmd->byte_length != ((wps->wphdr.flags & MONO_DATA) ? 6 : 12)) + return FALSE; + + wps->w.c [0].median [0] = exp2s (byteptr [0] + (byteptr [1] << 8)); + wps->w.c [0].median [1] = exp2s (byteptr [2] + (byteptr [3] << 8)); + wps->w.c [0].median [2] = exp2s (byteptr [4] + (byteptr [5] << 8)); + + if (!(wps->wphdr.flags & MONO_DATA)) { + wps->w.c [1].median [0] = exp2s (byteptr [6] + (byteptr [7] << 8)); + wps->w.c [1].median [1] = exp2s (byteptr [8] + (byteptr [9] << 8)); + wps->w.c [1].median [2] = exp2s (byteptr [10] + (byteptr [11] << 8)); + } + + return TRUE; +} + +// Read the hybrid related values from the specifed metadata structure, convert +// them back to their internal formats and store them. The extended profile +// stuff is not implemented yet, so return an error if we get more data than +// we know what to do with. + +int read_hybrid_profile (WavpackStream *wps, WavpackMetadata *wpmd) +{ + uchar *byteptr = wpmd->data; + uchar *endptr = byteptr + wpmd->byte_length; + + if (wps->wphdr.flags & HYBRID_BITRATE) { + wps->w.c [0].slow_level = exp2s (byteptr [0] + (byteptr [1] << 8)); + byteptr += 2; + + if (!(wps->wphdr.flags & MONO_DATA)) { + wps->w.c [1].slow_level = exp2s (byteptr [0] + (byteptr [1] << 8)); + byteptr += 2; + } + } + + wps->w.bitrate_acc [0] = (int32_t)(byteptr [0] + (byteptr [1] << 8)) << 16; + byteptr += 2; + + if (!(wps->wphdr.flags & MONO_DATA)) { + wps->w.bitrate_acc [1] = (int32_t)(byteptr [0] + (byteptr [1] << 8)) << 16; + byteptr += 2; + } + + if (byteptr < endptr) { + wps->w.bitrate_delta [0] = exp2s ((short)(byteptr [0] + (byteptr [1] << 8))); + byteptr += 2; + + if (!(wps->wphdr.flags & MONO_DATA)) { + wps->w.bitrate_delta [1] = exp2s ((short)(byteptr [0] + (byteptr [1] << 8))); + byteptr += 2; + } + + if (byteptr < endptr) + return FALSE; + } + else + wps->w.bitrate_delta [0] = wps->w.bitrate_delta [1] = 0; + + return TRUE; +} + +// This function is called during both encoding and decoding of hybrid data to +// update the "error_limit" variable which determines the maximum sample error +// allowed in the main bitstream. In the HYBRID_BITRATE mode (which is the only +// currently implemented) this is calculated from the slow_level values and the +// bitrate accumulators. Note that the bitrate accumulators can be changing. + +void update_error_limit (struct words_data *w, uint32_t flags) +{ + int bitrate_0 = (w->bitrate_acc [0] += w->bitrate_delta [0]) >> 16; + + if (flags & MONO_DATA) { + if (flags & HYBRID_BITRATE) { + int slow_log_0 = (w->c [0].slow_level + SLO) >> SLS; + + if (slow_log_0 - bitrate_0 > -0x100) + w->c [0].error_limit = exp2s (slow_log_0 - bitrate_0 + 0x100); + else + w->c [0].error_limit = 0; + } + else + w->c [0].error_limit = exp2s (bitrate_0); + } + else { + int bitrate_1 = (w->bitrate_acc [1] += w->bitrate_delta [1]) >> 16; + + if (flags & HYBRID_BITRATE) { + int slow_log_0 = (w->c [0].slow_level + SLO) >> SLS; + int slow_log_1 = (w->c [1].slow_level + SLO) >> SLS; + + if (flags & HYBRID_BALANCE) { + int balance = (slow_log_1 - slow_log_0 + bitrate_1 + 1) >> 1; + + if (balance > bitrate_0) { + bitrate_1 = bitrate_0 * 2; + bitrate_0 = 0; + } + else if (-balance > bitrate_0) { + bitrate_0 = bitrate_0 * 2; + bitrate_1 = 0; + } + else { + bitrate_1 = bitrate_0 + balance; + bitrate_0 = bitrate_0 - balance; + } + } + + if (slow_log_0 - bitrate_0 > -0x100) + w->c [0].error_limit = exp2s (slow_log_0 - bitrate_0 + 0x100); + else + w->c [0].error_limit = 0; + + if (slow_log_1 - bitrate_1 > -0x100) + w->c [1].error_limit = exp2s (slow_log_1 - bitrate_1 + 0x100); + else + w->c [1].error_limit = 0; + } + else { + w->c [0].error_limit = exp2s (bitrate_0); + w->c [1].error_limit = exp2s (bitrate_1); + } + } +} + +static uint32_t read_code (Bitstream *bs, uint32_t maxcode); + +// Read the next word from the bitstream "wvbits" and return the value. This +// function can be used for hybrid or lossless streams, but since an +// optimized version is available for lossless this function would normally +// be used for hybrid only. If a hybrid lossless stream is being read then +// the "correction" offset is written at the specified pointer. A return value +// of WORD_EOF indicates that the end of the bitstream was reached (all 1s) or +// some other error occurred. + +int32_t get_words (int32_t *buffer, int nsamples, uint32_t flags, + struct words_data *w, Bitstream *bs) +{ + register struct entropy_data *c = w->c; + int csamples; + + if (!(flags & MONO_DATA)) + nsamples *= 2; + + for (csamples = 0; csamples < nsamples; ++csamples) { + uint32_t ones_count, low, mid, high; + + if (!(flags & MONO_DATA)) + c = w->c + (csamples & 1); + + if (!(w->c [0].median [0] & ~1) && !w->holding_zero && !w->holding_one && !(w->c [1].median [0] & ~1)) { + uint32_t mask; + int cbits; + + if (w->zeros_acc) { + if (--w->zeros_acc) { + c->slow_level -= (c->slow_level + SLO) >> SLS; + *buffer++ = 0; + continue; + } + } + else { + for (cbits = 0; cbits < 33 && getbit (bs); ++cbits); + + if (cbits == 33) + break; + + if (cbits < 2) + w->zeros_acc = cbits; + else { + for (mask = 1, w->zeros_acc = 0; --cbits; mask <<= 1) + if (getbit (bs)) + w->zeros_acc |= mask; + + w->zeros_acc |= mask; + } + + if (w->zeros_acc) { + c->slow_level -= (c->slow_level + SLO) >> SLS; + CLEAR (w->c [0].median); + CLEAR (w->c [1].median); + *buffer++ = 0; + continue; + } + } + } + + if (w->holding_zero) + ones_count = w->holding_zero = 0; + else { + int next8; + + if (bs->bc < 8) { + if (++(bs->ptr) == bs->end) + bs->wrap (bs); + + next8 = (bs->sr |= *(bs->ptr) << bs->bc) & 0xff; + bs->bc += 8; + } + else + next8 = bs->sr & 0xff; + + if (next8 == 0xff) { + bs->bc -= 8; + bs->sr >>= 8; + + for (ones_count = 8; ones_count < (LIMIT_ONES + 1) && getbit (bs); ++ones_count); + + if (ones_count == (LIMIT_ONES + 1)) + break; + + if (ones_count == LIMIT_ONES) { + uint32_t mask; + int cbits; + + for (cbits = 0; cbits < 33 && getbit (bs); ++cbits); + + if (cbits == 33) + break; + + if (cbits < 2) + ones_count = cbits; + else { + for (mask = 1, ones_count = 0; --cbits; mask <<= 1) + if (getbit (bs)) + ones_count |= mask; + + ones_count |= mask; + } + + ones_count += LIMIT_ONES; + } + } + else { + bs->bc -= (ones_count = ones_count_table [next8]) + 1; + bs->sr >>= ones_count + 1; + } + + if (w->holding_one) { + w->holding_one = ones_count & 1; + ones_count = (ones_count >> 1) + 1; + } + else { + w->holding_one = ones_count & 1; + ones_count >>= 1; + } + + w->holding_zero = ~w->holding_one & 1; + } + + if ((flags & HYBRID_FLAG) && ((flags & MONO_DATA) || !(csamples & 1))) + update_error_limit (w, flags); + + if (ones_count == 0) { + low = 0; + high = GET_MED (0) - 1; + DEC_MED0 (); + } + else { + low = GET_MED (0); + INC_MED0 (); + + if (ones_count == 1) { + high = low + GET_MED (1) - 1; + DEC_MED1 (); + } + else { + low += GET_MED (1); + INC_MED1 (); + + if (ones_count == 2) { + high = low + GET_MED (2) - 1; + DEC_MED2 (); + } + else { + low += (ones_count - 2) * GET_MED (2); + high = low + GET_MED (2) - 1; + INC_MED2 (); + } + } + } + + mid = (high + low + 1) >> 1; + + if (!c->error_limit) + mid = read_code (bs, high - low) + low; + else while (high - low > c->error_limit) { + if (getbit (bs)) + mid = (high + (low = mid) + 1) >> 1; + else + mid = ((high = mid - 1) + low + 1) >> 1; + } + + *buffer++ = getbit (bs) ? ~mid : mid; + + if (flags & HYBRID_BITRATE) + c->slow_level = c->slow_level - ((c->slow_level + SLO) >> SLS) + mylog2 (mid); + } + + return (flags & MONO_DATA) ? csamples : (csamples / 2); +} + +// Read a single unsigned value from the specified bitstream with a value +// from 0 to maxcode. If there are exactly a power of two number of possible +// codes then this will read a fixed number of bits; otherwise it reads the +// minimum number of bits and then determines whether another bit is needed +// to define the code. + +static uint32_t read_code (Bitstream *bs, uint32_t maxcode) +{ + int bitcount = count_bits (maxcode); + uint32_t extras = (1L << bitcount) - maxcode - 1, code; + + if (!bitcount) + return 0; + + getbits (&code, bitcount - 1, bs); + code &= (1L << (bitcount - 1)) - 1; + + if (code >= extras) { + code = (code << 1) - extras; + + if (getbit (bs)) + ++code; + } + + return code; +} + +// The concept of a base 2 logarithm is used in many parts of WavPack. It is +// a way of sufficiently accurately representing 32-bit signed and unsigned +// values storing only 16 bits (actually fewer). It is also used in the hybrid +// mode for quickly comparing the relative magnitude of large values (i.e. +// division) and providing smooth exponentials using only addition. + +// These are not strict logarithms in that they become linear around zero and +// can therefore represent both zero and negative values. They have 8 bits +// of precision and in "roundtrip" conversions the total error never exceeds 1 +// part in 225 except for the cases of +/-115 and +/-195 (which error by 1). + + +// This function returns the log2 for the specified 32-bit unsigned value. +// The maximum value allowed is about 0xff800000 and returns 8447. + +static int mylog2 (uint32_t avalue) +{ + int dbits; + + if ((avalue += avalue >> 9) < (1 << 8)) { + dbits = nbits_table [avalue]; + return (dbits << 8) + log2_table [(avalue << (9 - dbits)) & 0xff]; + } + else { + if (avalue < (1L << 16)) + dbits = nbits_table [avalue >> 8] + 8; + else if (avalue < (1L << 24)) + dbits = nbits_table [avalue >> 16] + 16; + else + dbits = nbits_table [avalue >> 24] + 24; + + return (dbits << 8) + log2_table [(avalue >> (dbits - 9)) & 0xff]; + } +} + +// This function returns the log2 for the specified 32-bit signed value. +// All input values are valid and the return values are in the range of +// +/- 8192. + +int log2s (int32_t value) +{ + return (value < 0) ? -mylog2 (-value) : mylog2 (value); +} + +// This function returns the original integer represented by the supplied +// logarithm (at least within the provided accuracy). The log is signed, +// but since a full 32-bit value is returned this can be used for unsigned +// conversions as well (i.e. the input range is -8192 to +8447). + +int32_t exp2s (int log) +{ + uint32_t value; + + if (log < 0) + return -exp2s (-log); + + value = exp2_table [log & 0xff] | 0x100; + + if ((log >>= 8) <= 9) + return value >> (9 - log); + else + return value << (log - 9); +} + +// These two functions convert internal weights (which are normally +/-1024) +// to and from an 8-bit signed character version for storage in metadata. The +// weights are clipped here in the case that they are outside that range. + +int restore_weight (signed char weight) +{ + int result; + + if ((result = (int) weight << 3) > 0) + result += (result + 64) >> 7; + + return result; +} diff --git a/lib/wavpack/wputils.c b/lib/wavpack/wputils.c new file mode 100644 index 00000000..c3a23f98 --- /dev/null +++ b/lib/wavpack/wputils.c @@ -0,0 +1,350 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// wputils.c + +// This module provides a high-level interface for decoding WavPack 4.0 audio +// streams and files. WavPack data is read with a stream reading callback. No +// direct seeking is provided for, but it is possible to start decoding +// anywhere in a WavPack stream. In this case, WavPack will be able to provide +// the sample-accurate position when it synchs with the data and begins +// decoding. + +#include "wavpack.h" + +#include + +///////////////////////////// local table storage //////////////////////////// + +const uint32_t sample_rates [] = { 6000, 8000, 9600, 11025, 12000, 16000, 22050, + 24000, 32000, 44100, 48000, 64000, 88200, 96000, 192000 }; + +///////////////////////////// executable code //////////////////////////////// + +static uint32_t read_next_header (read_stream infile, void *user_data, WavpackHeader *wphdr); + +// This function reads data from the specified stream in search of a valid +// WavPack 4.0 audio block. If this fails in 1 megabyte (or an invalid or +// unsupported WavPack block is encountered) then an appropriate message is +// copied to "error" and NULL is returned, otherwise a pointer to a +// WavpackContext structure is returned (which is used to call all other +// functions in this module). This can be initiated at the beginning of a +// WavPack file, or anywhere inside a WavPack file. To determine the exact +// position within the file use WavpackGetSampleIndex(). For demonstration +// purposes this uses a single static copy of the WavpackContext structure, +// so obviously it cannot be used for more than one file at a time. Also, +// this function will not handle "correction" files, plays only the first +// two channels of multi-channel files, and is limited in resolution in some +// large integer or floating point files (but always provides at least 24 bits +// of resolution). + +int WavpackOpenFileInput (WavpackContext *wpc, read_stream infile, void *user_data, char *error) +{ + WavpackStream *wps = &wpc->stream; + uint32_t bcount; + + //CLEAR (wpc); + wpc->infile = infile; + wpc->user_data = user_data; + wpc->total_samples = (uint32_t) -1; + wpc->norm_offset = 0; + wpc->open_flags = 0; + + // open the source file for reading and store the size + + while (!wps->wphdr.block_samples) { + + bcount = read_next_header (wpc->infile, wpc->user_data, &wps->wphdr); + + if (bcount == (uint32_t) -1) { + strcpy (error, "not compatible with this version of WavPack file!"); + return 0; + } + + if (wps->wphdr.block_samples && wps->wphdr.total_samples != (uint32_t) -1) + wpc->total_samples = wps->wphdr.total_samples; + + if (!unpack_init (wpc)) { + strcpy (error, wpc->error_message [0] ? wpc->error_message : + "not compatible with this version of WavPack file!"); + + return 0; + } + } + + wpc->config.flags &= ~0xff; + wpc->config.flags |= wps->wphdr.flags & 0xff; + wpc->config.bytes_per_sample = (wps->wphdr.flags & BYTES_STORED) + 1; + wpc->config.float_norm_exp = wps->float_norm_exp; + + wpc->config.bits_per_sample = (wpc->config.bytes_per_sample * 8) - + ((wps->wphdr.flags & SHIFT_MASK) >> SHIFT_LSB); + + if (wpc->config.flags & FLOAT_DATA) { + wpc->config.bytes_per_sample = 3; + wpc->config.bits_per_sample = 24; + } + + if (!wpc->config.sample_rate) { + if (!wps || !wps->wphdr.block_samples || (wps->wphdr.flags & SRATE_MASK) == SRATE_MASK) + wpc->config.sample_rate = 44100; + else + wpc->config.sample_rate = sample_rates [(wps->wphdr.flags & SRATE_MASK) >> SRATE_LSB]; + } + + if (!wpc->config.num_channels) { + wpc->config.num_channels = (wps->wphdr.flags & MONO_FLAG) ? 1 : 2; + wpc->config.channel_mask = 0x5 - wpc->config.num_channels; + } + + if (!(wps->wphdr.flags & FINAL_BLOCK)) + wpc->reduced_channels = (wps->wphdr.flags & MONO_FLAG) ? 1 : 2; + + return 1; +} + +// This function obtains general information about an open file and returns +// a mask with the following bit values: + +// MODE_LOSSLESS: file is lossless (pure lossless only) +// MODE_HYBRID: file is hybrid mode (lossy part only) +// MODE_FLOAT: audio data is 32-bit ieee floating point (but will provided +// in 24-bit integers for convenience) +// MODE_HIGH: file was created in "high" mode (information only) +// MODE_FAST: file was created in "fast" mode (information only) + +int WavpackGetMode (WavpackContext *wpc) +{ + int mode = 0; + + if (wpc) { + if (wpc->config.flags & CONFIG_HYBRID_FLAG) + mode |= MODE_HYBRID; + else if (!(wpc->config.flags & CONFIG_LOSSY_MODE)) + mode |= MODE_LOSSLESS; + + if (wpc->lossy_blocks) + mode &= ~MODE_LOSSLESS; + + if (wpc->config.flags & CONFIG_FLOAT_DATA) + mode |= MODE_FLOAT; + + if (wpc->config.flags & CONFIG_HIGH_FLAG) + mode |= MODE_HIGH; + + if (wpc->config.flags & CONFIG_FAST_FLAG) + mode |= MODE_FAST; + } + + return mode; +} + +// Unpack the specified number of samples from the current file position. +// Note that "samples" here refers to "complete" samples, which would be +// 2 longs for stereo files. The audio data is returned right-justified in +// 32-bit longs in the endian mode native to the executing processor. So, +// if the original data was 16-bit, then the values returned would be +// +/-32k. Floating point data will be returned as 24-bit integers (and may +// also be clipped). The actual number of samples unpacked is returned, +// which should be equal to the number requested unless the end of fle is +// encountered or an error occurs. + +uint32_t WavpackUnpackSamples (WavpackContext *wpc, int32_t *buffer, uint32_t samples) +{ + WavpackStream *wps = &wpc->stream; + uint32_t bcount, samples_unpacked = 0, samples_to_unpack; + int num_channels = wpc->config.num_channels; + + while (samples) { + if (!wps->wphdr.block_samples || !(wps->wphdr.flags & INITIAL_BLOCK) || + wps->sample_index >= wps->wphdr.block_index + wps->wphdr.block_samples) { + bcount = read_next_header (wpc->infile, wpc->user_data, &wps->wphdr); + + if (bcount == (uint32_t) -1) + break; + + if (!wps->wphdr.block_samples || wps->sample_index == wps->wphdr.block_index) + if (!unpack_init (wpc)) + break; + } + + if (!wps->wphdr.block_samples || !(wps->wphdr.flags & INITIAL_BLOCK) || + wps->sample_index >= wps->wphdr.block_index + wps->wphdr.block_samples) + continue; + + if (wps->sample_index < wps->wphdr.block_index) { + samples_to_unpack = wps->wphdr.block_index - wps->sample_index; + + if (samples_to_unpack > samples) + samples_to_unpack = samples; + + wps->sample_index += samples_to_unpack; + samples_unpacked += samples_to_unpack; + samples -= samples_to_unpack; + + if (wpc->reduced_channels) + samples_to_unpack *= wpc->reduced_channels; + else + samples_to_unpack *= num_channels; + + while (samples_to_unpack--) + *buffer++ = 0; + + continue; + } + + samples_to_unpack = wps->wphdr.block_index + wps->wphdr.block_samples - wps->sample_index; + + if (samples_to_unpack > samples) + samples_to_unpack = samples; + + unpack_samples (wpc, buffer, samples_to_unpack); + + if (wpc->reduced_channels) + buffer += samples_to_unpack * wpc->reduced_channels; + else + buffer += samples_to_unpack * num_channels; + + samples_unpacked += samples_to_unpack; + samples -= samples_to_unpack; + + if (wps->sample_index == wps->wphdr.block_index + wps->wphdr.block_samples) { + if (check_crc_error (wpc)) + wpc->crc_errors++; + } + + if (wps->sample_index == wpc->total_samples) + break; + } + + return samples_unpacked; +} + +// Get total number of samples contained in the WavPack file, or -1 if unknown + +uint32_t WavpackGetNumSamples (WavpackContext *wpc) +{ + return wpc ? wpc->total_samples : (uint32_t) -1; +} + +// Get the current sample index position, or -1 if unknown + +uint32_t WavpackGetSampleIndex (WavpackContext *wpc) +{ + if (wpc) + return wpc->stream.sample_index; + + return (uint32_t) -1; +} + +// Get the number of errors encountered so far + +int WavpackGetNumErrors (WavpackContext *wpc) +{ + return wpc ? wpc->crc_errors : 0; +} + +// return TRUE if any uncorrected lossy blocks were actually written or read + +int WavpackLossyBlocks (WavpackContext *wpc) +{ + return wpc ? wpc->lossy_blocks : 0; +} + +// Returns the sample rate of the specified WavPack file + +uint32_t WavpackGetSampleRate (WavpackContext *wpc) +{ + return wpc ? wpc->config.sample_rate : 44100; +} + +// Returns the number of channels of the specified WavPack file. Note that +// this is the actual number of channels contained in the file, but this +// version can only decode the first two. + +int WavpackGetNumChannels (WavpackContext *wpc) +{ + return wpc ? wpc->config.num_channels : 2; +} + +// Returns the actual number of valid bits per sample contained in the +// original file, which may or may not be a multiple of 8. Floating data +// always has 32 bits, integers may be from 1 to 32 bits each. When this +// value is not a multiple of 8, then the "extra" bits are located in the +// LSBs of the results. That is, values are right justified when unpacked +// into longs, but are left justified in the number of bytes used by the +// original data. + +int WavpackGetBitsPerSample (WavpackContext *wpc) +{ + return wpc ? wpc->config.bits_per_sample : 16; +} + +// Returns the number of bytes used for each sample (1 to 4) in the original +// file. This is required information for the user of this module because the +// audio data is returned in the LOWER bytes of the long buffer and must be +// left-shifted 8, 16, or 24 bits if normalized longs are required. + +int WavpackGetBytesPerSample (WavpackContext *wpc) +{ + return wpc ? wpc->config.bytes_per_sample : 2; +} + +// This function will return the actual number of channels decoded from the +// file (which may or may not be less than the actual number of channels, but +// will always be 1 or 2). Normally, this will be the front left and right +// channels of a multi-channel file. + +int WavpackGetReducedChannels (WavpackContext *wpc) +{ + if (wpc) + return wpc->reduced_channels ? wpc->reduced_channels : wpc->config.num_channels; + else + return 2; +} + +// Read from current file position until a valid 32-byte WavPack 4.0 header is +// found and read into the specified pointer. The number of bytes skipped is +// returned. If no WavPack header is found within 1 meg, then a -1 is returned +// to indicate the error. No additional bytes are read past the header and it +// is returned in the processor's native endian mode. Seeking is not required. + +static uint32_t read_next_header (read_stream infile, void *user_data, WavpackHeader *wphdr) +{ + char buffer [sizeof (*wphdr)], *sp = buffer + sizeof (*wphdr), *ep = sp; + uint32_t bytes_skipped = 0; + int bleft; + + while (1) { + if (sp < ep) { + bleft = ep - sp; + memcpy (buffer, sp, bleft); + } + else + bleft = 0; + + if (infile (user_data, buffer + bleft, sizeof (*wphdr) - bleft) != (int32_t) sizeof (*wphdr) - bleft) + return -1; + + sp = buffer; + + if (*sp++ == 'w' && *sp == 'v' && *++sp == 'p' && *++sp == 'k' && + !(*++sp & 1) && sp [2] < 16 && !sp [3] && sp [5] == 4 && + sp [4] >= (MIN_STREAM_VERS & 0xff) && sp [4] <= (MAX_STREAM_VERS & 0xff)) { + memcpy (wphdr, buffer, sizeof (*wphdr)); + little_endian_to_native (wphdr, WavpackHeaderFormat); + return bytes_skipped; + } + + while (sp < ep && *sp != 'w') + sp++; + + if ((bytes_skipped += sp - buffer) > 1048576L) + return -1; + } +} diff --git a/lib/wavpack/wvfilter.c b/lib/wavpack/wvfilter.c new file mode 100644 index 00000000..f80d73dd --- /dev/null +++ b/lib/wavpack/wvfilter.c @@ -0,0 +1,200 @@ +//////////////////////////////////////////////////////////////////////////// +// **** WAVPACK **** // +// Hybrid Lossless Wavefile Compressor // +// Copyright (c) 1998 - 2006 Conifer Software. // +// All Rights Reserved. // +// Distributed under the BSD Software License (see license.txt) // +//////////////////////////////////////////////////////////////////////////// + +// wv_filter.c + +// This is the main module for the demonstration WavPack command-line +// decoder filter. It uses the tiny "hardware" version of the decoder and +// accepts WavPack files on stdin and outputs a standard MS wav file to +// stdout. Note that this involves converting the data to little-endian +// (if the executing processor is not), possibly packing the data into +// fewer bytes per sample, and generating an appropriate riff wav header. +// Note that this is NOT the copy of the RIFF header that might be stored +// in the file, and any additional RIFF information and tags are lost. +// See wputils.c for further limitations. + +#include "wavpack.h" + +#if defined(WIN32) +#include +#include +#endif + +#include + +// These structures are used to place a wav riff header at the beginning of +// the output. + +typedef struct { + char ckID [4]; + uint32_t ckSize; + char formType [4]; +} RiffChunkHeader; + +typedef struct { + char ckID [4]; + uint32_t ckSize; +} ChunkHeader; + +#define ChunkHeaderFormat "4L" + +typedef struct { + ushort FormatTag, NumChannels; + uint32_t SampleRate, BytesPerSecond; + ushort BlockAlign, BitsPerSample; +} WaveHeader; + +#define WaveHeaderFormat "SSLLSS" + +static uchar *format_samples (int bps, uchar *dst, int32_t *src, uint32_t samcnt); +static int32_t read_bytes (void *buff, int32_t bcount); +static int32_t temp_buffer [256]; + +int main () +{ + ChunkHeader FormatChunkHeader, DataChunkHeader; + RiffChunkHeader RiffChunkHeader; + WaveHeader WaveHeader; + + uint32_t total_unpacked_samples = 0, total_samples; + int num_channels, bps; + WavpackContext *wpc; + char error [80]; + +#if defined(WIN32) + setmode (fileno (stdin), O_BINARY); + setmode (fileno (stdout), O_BINARY); +#endif + + wpc = WavpackOpenFileInput (read_bytes, error); + + if (!wpc) { + fputs (error, stderr); + fputs ("\n", stderr); + return 1; + } + + num_channels = WavpackGetReducedChannels (wpc); + total_samples = WavpackGetNumSamples (wpc); + bps = WavpackGetBytesPerSample (wpc); + + strncpy (RiffChunkHeader.ckID, "RIFF", sizeof (RiffChunkHeader.ckID)); + RiffChunkHeader.ckSize = total_samples * num_channels * bps + sizeof (ChunkHeader) * 2 + sizeof (WaveHeader) + 4; + strncpy (RiffChunkHeader.formType, "WAVE", sizeof (RiffChunkHeader.formType)); + + strncpy (FormatChunkHeader.ckID, "fmt ", sizeof (FormatChunkHeader.ckID)); + FormatChunkHeader.ckSize = sizeof (WaveHeader); + + WaveHeader.FormatTag = 1; + WaveHeader.NumChannels = num_channels; + WaveHeader.SampleRate = WavpackGetSampleRate (wpc); + WaveHeader.BlockAlign = num_channels * bps; + WaveHeader.BytesPerSecond = WaveHeader.SampleRate * WaveHeader.BlockAlign; + WaveHeader.BitsPerSample = WavpackGetBitsPerSample (wpc); + + strncpy (DataChunkHeader.ckID, "data", sizeof (DataChunkHeader.ckID)); + DataChunkHeader.ckSize = total_samples * num_channels * bps; + + native_to_little_endian (&RiffChunkHeader, ChunkHeaderFormat); + native_to_little_endian (&FormatChunkHeader, ChunkHeaderFormat); + native_to_little_endian (&WaveHeader, WaveHeaderFormat); + native_to_little_endian (&DataChunkHeader, ChunkHeaderFormat); + + if (!fwrite (&RiffChunkHeader, sizeof (RiffChunkHeader), 1, stdout) || + !fwrite (&FormatChunkHeader, sizeof (FormatChunkHeader), 1, stdout) || + !fwrite (&WaveHeader, sizeof (WaveHeader), 1, stdout) || + !fwrite (&DataChunkHeader, sizeof (DataChunkHeader), 1, stdout)) { + fputs ("can't write .WAV data, disk probably full!\n", stderr); + return 1; + } + + while (1) { + uint32_t samples_unpacked; + + samples_unpacked = WavpackUnpackSamples (wpc, temp_buffer, 256 / num_channels); + total_unpacked_samples += samples_unpacked; + + if (samples_unpacked) { + format_samples (bps, (uchar *) temp_buffer, temp_buffer, samples_unpacked *= num_channels); + + if (fwrite (temp_buffer, bps, samples_unpacked, stdout) != samples_unpacked) { + fputs ("can't write .WAV data, disk probably full!\n", stderr); + return 1; + } + } + + if (!samples_unpacked) + break; + } + + fflush (stdout); + + if (WavpackGetNumSamples (wpc) != (uint32_t) -1 && + total_unpacked_samples != WavpackGetNumSamples (wpc)) { + fputs ("incorrect number of samples!\n", stderr); + return 1; + } + + if (WavpackGetNumErrors (wpc)) { + fputs ("crc errors detected!\n", stderr); + return 1; + } + + return 0; +} + +static int32_t read_bytes (void *buff, int32_t bcount) +{ + return fread (buff, 1, bcount, stdin); +} + +// Reformat samples from longs in processor's native endian mode to +// little-endian data with (possibly) less than 4 bytes / sample. + +static uchar *format_samples (int bps, uchar *dst, int32_t *src, uint32_t samcnt) +{ + int32_t temp; + + switch (bps) { + + case 1: + while (samcnt--) + *dst++ = *src++ + 128; + + break; + + case 2: + while (samcnt--) { + *dst++ = (uchar)(temp = *src++); + *dst++ = (uchar)(temp >> 8); + } + + break; + + case 3: + while (samcnt--) { + *dst++ = (uchar)(temp = *src++); + *dst++ = (uchar)(temp >> 8); + *dst++ = (uchar)(temp >> 16); + } + + break; + + case 4: + while (samcnt--) { + *dst++ = (uchar)(temp = *src++); + *dst++ = (uchar)(temp >> 8); + *dst++ = (uchar)(temp >> 16); + *dst++ = (uchar)(temp >> 24); + } + + break; + } + + return dst; +} diff --git a/src/codecs/CMakeLists.txt b/src/codecs/CMakeLists.txt index a6a48c84..1b79b863 100644 --- a/src/codecs/CMakeLists.txt +++ b/src/codecs/CMakeLists.txt @@ -4,9 +4,9 @@ idf_component_register( SRCS "dr_flac.cpp" "codec.cpp" "mad.cpp" "opus.cpp" "vorbis.cpp" - "source_buffer.cpp" "sample.cpp" "wav.cpp" "native.cpp" + "source_buffer.cpp" "sample.cpp" "wav.cpp" "native.cpp" "wavpack.cpp" INCLUDE_DIRS "include" REQUIRES "result" "libmad" "drflac" "tremor" "opusfile" "memory" "util" - "komihash") + "komihash" "wavpack") target_compile_options("${COMPONENT_LIB}" PRIVATE ${EXTRA_WARNINGS}) diff --git a/src/codecs/codec.cpp b/src/codecs/codec.cpp index af5702ff..4ddb16ad 100644 --- a/src/codecs/codec.cpp +++ b/src/codecs/codec.cpp @@ -16,6 +16,7 @@ #include "types.hpp" #include "vorbis.hpp" #include "wav.hpp" +#include "wavpack.hpp" namespace codecs { @@ -33,6 +34,8 @@ auto StreamTypeToString(StreamType t) -> std::string { return "Opus"; case StreamType::kNative: return "Native"; + case StreamType::kWavPack: + return "WavPack"; default: return ""; } @@ -52,6 +55,8 @@ auto CreateCodecForType(StreamType type) -> std::optional { return new WavDecoder(); case StreamType::kNative: return new NativeDecoder(); + case StreamType::kWavPack: + return new WavPackDecoder(); default: return {}; } diff --git a/src/codecs/include/types.hpp b/src/codecs/include/types.hpp index 2bc63b10..493a177a 100644 --- a/src/codecs/include/types.hpp +++ b/src/codecs/include/types.hpp @@ -17,6 +17,7 @@ enum class StreamType { kOpus, kWav, kNative, + kWavPack, }; auto StreamTypeToString(StreamType t) -> std::string; diff --git a/src/codecs/include/wavpack.hpp b/src/codecs/include/wavpack.hpp new file mode 100644 index 00000000..4780b6b6 --- /dev/null +++ b/src/codecs/include/wavpack.hpp @@ -0,0 +1,46 @@ +/* + * Copyright 2025 ayumi + * + * SPDX-License-Identifier: GPL-3.0-only + */ + +#pragma once + +#include +#include +#include +#include +#include +#include + +#include "wavpack.h" +#include "sample.hpp" + +#include "codec.hpp" + +namespace codecs { + +class WavPackDecoder : public ICodec { + public: + WavPackDecoder(); + ~WavPackDecoder(); + + auto OpenStream(std::shared_ptr input, uint32_t offset) + -> cpp::result override; + + auto DecodeTo(std::span destination) + -> cpp::result override; + + WavPackDecoder(const WavPackDecoder&) = delete; + WavPackDecoder& operator=(const WavPackDecoder&) = delete; + + private: + std::shared_ptr input_; + WavpackContext wavpack_; + int32_t *buf_; + uint8_t bitdepth_; + uint8_t channels_; + size_t size_; +}; + +} // namespace codecs diff --git a/src/codecs/wavpack.cpp b/src/codecs/wavpack.cpp new file mode 100644 index 00000000..fa168d32 --- /dev/null +++ b/src/codecs/wavpack.cpp @@ -0,0 +1,161 @@ +/* + * Copyright 2025 ayumi + * + * SPDX-License-Identifier: GPL-3.0-only + */ + +#include "wavpack.hpp" + +#include +#include +#include +#include + +#include "esp_heap_caps.h" +#include "codec.hpp" +#include "esp_log.h" +#include "result.hpp" +#include "sample.hpp" +#include "types.hpp" + +namespace codecs { + +[[maybe_unused]] static constexpr const char kTag[] = "wavpack"; +// kBufSize and audio::kCodecBufferLength must be equal +static constexpr const size_t kBufSize = 2048; + +static inline constexpr auto loadLe16(std::byte* data) -> uint16_t { + return *reinterpret_cast(data); +} + +static inline constexpr auto loadLe32(std::byte* data) -> uint32_t { + return *reinterpret_cast(data); +} + +static auto readProc(void* data, void* buf, long size) -> long { + IStream* stream = static_cast(data); + const int32_t res = stream->Read({ + static_cast(buf), + static_cast::size_type>(size) + }); + return res < 0 ? 0 : res; +} + +WavPackDecoder::WavPackDecoder() : input_(), buf_() { + buf_ = static_cast( + heap_caps_malloc( + kBufSize * sizeof(int32_t), + MALLOC_CAP_INTERNAL | MALLOC_CAP_CACHE_ALIGNED + )); +} + +WavPackDecoder::~WavPackDecoder() { + heap_caps_free(buf_); +} + +auto WavPackDecoder::OpenStream(std::shared_ptr input, uint32_t offset) + -> cpp::result { + char error[80]; + input_ = input; + wavpack_ = {}; + if (!WavpackOpenFileInput(&wavpack_, readProc, input_.get(), error)) { + ESP_LOGE(kTag, "WavpackOpenFileInput: %s", error); + return cpp::fail(Error::kMalformedData); + } + + channels_ = WavpackGetReducedChannels(&wavpack_); + bitdepth_ = WavpackGetBitsPerSample(&wavpack_); + size_ = kBufSize / channels_; + const std::optional total = WavpackGetNumSamples(&wavpack_) == -1 + ? std::nullopt + : std::optional( + static_cast(WavpackGetNumSamples(&wavpack_)) * channels_ + ); + const auto rate = WavpackGetSampleRate(&wavpack_); + if (offset && total && input_.get()->CanSeek()) { + const uint32_t want = offset * rate - 1; + if (total < want) { + ESP_LOGE(kTag, "seeking: offset points beyond the end of the file"); + return cpp::fail(Error::kInternalError); + } + + uint32_t target; + input_->SeekTo(0, IStream::SeekFrom::kStartOfStream); + while (true) { + std::byte header[32]; + input_->Read(header); + if (memcmp(header, "wvpk", 4) != 0) { + ESP_LOGE(kTag, "seeking: header expected, but not found"); + return cpp::fail(Error::kMalformedData); + } + const uint32_t size = loadLe32(header + 4); + const uint16_t version = loadLe16(header + 8); + if (version < 0x402 || version > 0x410) { + ESP_LOGE(kTag, "seeking: bad WavPack version (%x)", version); + return cpp::fail(Error::kMalformedData); + } + const uint32_t blockIndex = loadLe32(header + 16); + const uint32_t blockSamples = loadLe32(header + 20); + if (want >= blockIndex && want <= blockIndex + blockSamples) { + input_->SeekTo(-32, IStream::SeekFrom::kCurrentPosition); + target = want - blockIndex; + break; + } + input_->SeekTo(size - 24, IStream::SeekFrom::kCurrentPosition); + } + + wavpack_ = {}; + if (!WavpackOpenFileInput(&wavpack_, readProc, input_.get(), error)) { + ESP_LOGE(kTag, "WavpackOpenFileInput: %s", error); + return cpp::fail(Error::kMalformedData); + } + + uint32_t samples = 0; + for (size_t i = 0, n = target / size_; i < n; i++) + samples += WavpackUnpackSamples(&wavpack_, buf_, size_); + samples += WavpackUnpackSamples(&wavpack_, buf_, target % size_); + if (WavpackGetNumErrors(&wavpack_) != 0) { + ESP_LOGE(kTag, "CRC error"); + return cpp::fail(Error::kMalformedData); + } else if (samples != target || WavpackGetSampleIndex(&wavpack_) != want) { + ESP_LOGE(kTag, "seeking: seeking unsuccessful: want %lu, got %lu", + target, samples + ); + return cpp::fail(Error::kInternalError); + } + } else if (offset && (!total || !input_.get()->CanSeek())) { + ESP_LOGE(kTag, "seeking: can’t seek"); + return cpp::fail(Error::kInternalError); + } + + const auto size = input->Size(); + return OutputFormat{ + .num_channels = channels_, + .sample_rate_hz = rate, + .total_samples = total, + .bitrate_kbps = size && total + ? std::optional( + ((double)size.value() * 8.0) + / ((double)total.value() / channels_ / rate) / 1000 + ) + : std::nullopt, + }; +} + +auto WavPackDecoder::DecodeTo(std::span output) + -> cpp::result { + const auto size = std::min(size_, output.size() / channels_); + const auto samples = WavpackUnpackSamples(&wavpack_, buf_, size) * channels_; + if (WavpackGetNumErrors(&wavpack_) != 0) { + ESP_LOGE(kTag, "CRC error"); + return cpp::fail(Error::kMalformedData); + } + for (size_t i = 0; i < samples; i++) + output[i] = sample::FromSigned(buf_[i], bitdepth_); + return OutputInfo{ + .samples_written = samples, + .is_stream_finished = samples == 0, + }; +} + +} // namespace codecs diff --git a/src/tangara/audio/fatfs_stream_factory.cpp b/src/tangara/audio/fatfs_stream_factory.cpp index 94f22ae9..9089735c 100644 --- a/src/tangara/audio/fatfs_stream_factory.cpp +++ b/src/tangara/audio/fatfs_stream_factory.cpp @@ -88,6 +88,8 @@ auto FatfsStreamFactory::ContainerToStreamType(database::Container enc) return codecs::StreamType::kFlac; case database::Container::kOpus: return codecs::StreamType::kOpus; + case database::Container::kWavPack: + return codecs::StreamType::kWavPack; case database::Container::kUnsupported: default: return {}; diff --git a/src/tangara/database/tag_parser.cpp b/src/tangara/database/tag_parser.cpp index 6c95d496..0be6cb35 100644 --- a/src/tangara/database/tag_parser.cpp +++ b/src/tangara/database/tag_parser.cpp @@ -413,6 +413,9 @@ auto GenericTagParser::ReadAndParseTags(std::string_view p) case Fopus: out->encoding(Container::kOpus); break; + case Fwavpack: + out->encoding(Container::kWavPack); + break; default: out->encoding(Container::kUnsupported); } diff --git a/src/tangara/database/tag_parser.hpp b/src/tangara/database/tag_parser.hpp index 220339c0..eb0f4c7c 100644 --- a/src/tangara/database/tag_parser.hpp +++ b/src/tangara/database/tag_parser.hpp @@ -63,7 +63,8 @@ class GenericTagParser : public ITagParser { // supported audio formats here: // https://cooltech.zone/tangara/docs/music-library/ static constexpr std::string supported_exts[] = {"flac", "mp3", "ogg", - "ogx", "opus", "wav"}; + "ogx", "opus", "wav", + "wv"}; }; } // namespace database diff --git a/src/tangara/database/track.hpp b/src/tangara/database/track.hpp index c7dff425..d6039451 100644 --- a/src/tangara/database/track.hpp +++ b/src/tangara/database/track.hpp @@ -45,6 +45,7 @@ enum class Container { kOgg = 3, kFlac = 4, kOpus = 5, + kWavPack = 6, }; enum class MediaType { diff --git a/tools/cmake/common.cmake b/tools/cmake/common.cmake index f92eddb2..7afda6c1 100644 --- a/tools/cmake/common.cmake +++ b/tools/cmake/common.cmake @@ -37,6 +37,7 @@ list(APPEND EXTRA_COMPONENT_DIRS "$ENV{PROJ_PATH}/lib/result") list(APPEND EXTRA_COMPONENT_DIRS "$ENV{PROJ_PATH}/lib/speexdsp") list(APPEND EXTRA_COMPONENT_DIRS "$ENV{PROJ_PATH}/lib/tinyfsm") list(APPEND EXTRA_COMPONENT_DIRS "$ENV{PROJ_PATH}/lib/tremor") +list(APPEND EXTRA_COMPONENT_DIRS "$ENV{PROJ_PATH}/lib/wavpack") include($ENV{IDF_PATH}/tools/cmake/project.cmake)