include/vlc_codecs.h

   1 /*****************************************************************************
   2  * codecs.h: codec related structures needed by the demuxers and decoders
   3  *****************************************************************************
   4  * Copyright (C) 1999-2001 the VideoLAN team
   5  * $Id$
   6  *
   7  * Author: Gildas Bazin <gbazin@videolan.org>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  * GNU General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  22  *****************************************************************************/
  23
  24 #ifndef VLC_CODECS_H
  25 #define VLC_CODECS_H 1
  26
  27 #include <vlc_fourcc.h>
  28 /**
  29  * \file
  30  * This file defines codec related structures needed by the demuxers and decoders
  31  */
  32
  33 #ifdef HAVE_ATTRIBUTE_PACKED
  34 #   define ATTR_PACKED __attribute__((__packed__))
  35 #else
  36 #   error FIXME
  37 #endif
  38
  39 /* Structures exported to the demuxers and decoders */
  40
  41 #if !(defined _GUID_DEFINED || defined GUID_DEFINED)
  42 #define GUID_DEFINED
  43 typedef struct _GUID
  44 {
  45     uint32_t Data1;
  46     uint16_t Data2;
  47     uint16_t Data3;
  48     uint8_t  Data4[8];
  49 } GUID, *REFGUID, *LPGUID;
  50 #endif /* GUID_DEFINED */
  51
  52 #ifndef _WAVEFORMATEX_
  53 #define _WAVEFORMATEX_
  54 typedef struct
  55 ATTR_PACKED
  56 _WAVEFORMATEX {
  57     uint16_t   wFormatTag;
  58     uint16_t   nChannels;
  59     uint32_t   nSamplesPerSec;
  60     uint32_t   nAvgBytesPerSec;
  61     uint16_t   nBlockAlign;
  62     uint16_t   wBitsPerSample;
  63     uint16_t   cbSize;
  64 } WAVEFORMATEX, *PWAVEFORMATEX, *NPWAVEFORMATEX, *LPWAVEFORMATEX;
  65 #endif /* _WAVEFORMATEX_ */
  66
  67 #ifndef _WAVEFORMATEXTENSIBLE_
  68 #define _WAVEFORMATEXTENSIBLE_
  69 typedef struct
  70 ATTR_PACKED
  71 _WAVEFORMATEXTENSIBLE {
  72     WAVEFORMATEX Format;
  73     union {
  74         uint16_t wValidBitsPerSample;
  75         uint16_t wSamplesPerBlock;
  76         uint16_t wReserved;
  77     } Samples;
  78     uint32_t     dwChannelMask;
  79     GUID SubFormat;
  80 } WAVEFORMATEXTENSIBLE, *PWAVEFORMATEXTENSIBLE;
  81 #endif /* _WAVEFORMATEXTENSIBLE_ */
  82
  83 #ifndef _WAVEHEADER_
  84 #define _WAVEHEADER_
  85 typedef struct
  86 ATTR_PACKED
  87 _WAVEHEADER {
  88     uint32_t MainChunkID;
  89     uint32_t Length;
  90     uint32_t ChunkTypeID;
  91     uint32_t SubChunkID;
  92     uint32_t SubChunkLength;
  93     uint16_t Format;
  94     uint16_t Modus;
  95     uint32_t SampleFreq;
  96     uint32_t BytesPerSec;
  97     uint16_t BytesPerSample;
  98     uint16_t BitsPerSample;
  99     uint32_t DataChunkID;
 100     uint32_t DataLength;
 101 } WAVEHEADER;
 102 #endif /* _WAVEHEADER_ */
 103
 104 #if !defined(_BITMAPINFOHEADER_) && !defined(WIN32)
 105 #define _BITMAPINFOHEADER_
 106 typedef struct
 107 ATTR_PACKED
 108 {
 109     uint32_t   biSize;
 110     uint32_t   biWidth;
 111     uint32_t   biHeight;
 112     uint16_t   biPlanes;
 113     uint16_t   biBitCount;
 114     uint32_t   biCompression;
 115     uint32_t   biSizeImage;
 116     uint32_t   biXPelsPerMeter;
 117     uint32_t   biYPelsPerMeter;
 118     uint32_t   biClrUsed;
 119     uint32_t   biClrImportant;
 120 } BITMAPINFOHEADER, *PBITMAPINFOHEADER, *LPBITMAPINFOHEADER;
 121
 122 typedef struct
 123 ATTR_PACKED
 124 {
 125     BITMAPINFOHEADER bmiHeader;
 126     int        bmiColors[1];
 127 } BITMAPINFO, *LPBITMAPINFO;
 128 #endif
 129
 130 #ifndef _RECT32_
 131 #define _RECT32_
 132 typedef struct
 133 ATTR_PACKED
 134 {
 135     int left, top, right, bottom;
 136 } RECT32;
 137 #endif
 138
 139 #ifndef _REFERENCE_TIME_
 140 #define _REFERENCE_TIME_
 141 typedef int64_t REFERENCE_TIME;
 142 #endif
 143
 144 #ifndef _VIDEOINFOHEADER_
 145 #define _VIDEOINFOHEADER_
 146 typedef struct
 147 ATTR_PACKED
 148 {
 149     RECT32            rcSource;
 150     RECT32            rcTarget;
 151     uint32_t          dwBitRate;
 152     uint32_t          dwBitErrorRate;
 153     REFERENCE_TIME    AvgTimePerFrame;
 154     BITMAPINFOHEADER  bmiHeader;
 155 } VIDEOINFOHEADER;
 156 #endif
 157
 158 #ifndef _RGBQUAD_
 159 #define _RGBQUAD_
 160 typedef struct
 161 ATTR_PACKED
 162 {
 163     uint8_t rgbBlue;
 164     uint8_t rgbGreen;
 165     uint8_t rgbRed;
 166     uint8_t rgbReserved;
 167 } RGBQUAD1;
 168 #endif
 169
 170 #ifndef _TRUECOLORINFO_
 171 #define _TRUECOLORINFO_
 172 typedef struct
 173 ATTR_PACKED
 174 {
 175     uint32_t dwBitMasks[3];
 176     RGBQUAD1 bmiColors[256];
 177 } TRUECOLORINFO;
 178 #endif
 179
 180 #ifndef _VIDEOINFO_
 181 #define _VIDEOINFO_
 182 typedef struct
 183 ATTR_PACKED
 184 {
 185     RECT32            rcSource;
 186     RECT32            rcTarget;
 187     uint32_t          dwBitRate;
 188     uint32_t          dwBitErrorRate;
 189     REFERENCE_TIME    AvgTimePerFrame;
 190     BITMAPINFOHEADER  bmiHeader;
 191
 192     union
 193     {
 194         RGBQUAD1 bmiColors[256]; /* Colour palette */
 195         uint32_t dwBitMasks[3]; /* True colour masks */
 196         TRUECOLORINFO TrueColorInfo; /* Both of the above */
 197     };
 198
 199 } VIDEOINFO;
 200 #endif
 201
 202 /* WAVE format wFormatTag IDs */
 203 #define WAVE_FORMAT_UNKNOWN             0x0000 /* Microsoft Corporation */
 204 #define WAVE_FORMAT_PCM                 0x0001 /* Microsoft Corporation */
 205 #define WAVE_FORMAT_ADPCM               0x0002 /* Microsoft Corporation */
 206 #define WAVE_FORMAT_IEEE_FLOAT          0x0003 /* Microsoft Corporation */
 207 #define WAVE_FORMAT_ALAW                0x0006 /* Microsoft Corporation */
 208 #define WAVE_FORMAT_MULAW               0x0007 /* Microsoft Corporation */
 209 #define WAVE_FORMAT_DTS_MS              0x0008 /* Microsoft Corporation */
 210 #define WAVE_FORMAT_WMAS                0x000a /* WMA 9 Speech */
 211 #define WAVE_FORMAT_IMA_ADPCM           0x0011 /* Intel Corporation */
 212 #define WAVE_FORMAT_TRUESPEECH          0x0022 /* TrueSpeech */
 213 #define WAVE_FORMAT_GSM610              0x0031 /* Microsoft Corporation */
 214 #define WAVE_FORMAT_MSNAUDIO            0x0032 /* Microsoft Corporation */
 215 #define WAVE_FORMAT_G726                0x0045 /* ITU-T standard  */
 216 #define WAVE_FORMAT_MPEG                0x0050 /* Microsoft Corporation */
 217 #define WAVE_FORMAT_MPEGLAYER3          0x0055 /* ISO/MPEG Layer3 Format Tag */
 218 #define WAVE_FORMAT_DOLBY_AC3_SPDIF     0x0092 /* Sonic Foundry */
 219
 220 #define WAVE_FORMAT_A52                 0x2000
 221 #define WAVE_FORMAT_DTS                 0x2001
 222 #define WAVE_FORMAT_WMA1                0x0160 /* WMA version 1 */
 223 #define WAVE_FORMAT_WMA2                0x0161 /* WMA (v2) 7, 8, 9 Series */
 224 #define WAVE_FORMAT_WMAP                0x0162 /* WMA 9 Professional */
 225 #define WAVE_FORMAT_WMAL                0x0163 /* WMA 9 Lossless */
 226 #define WAVE_FORMAT_DIVIO_AAC           0x4143
 227 #define WAVE_FORMAT_AAC                 0x00FF
 228 #define WAVE_FORMAT_FFMPEG_AAC          0x706D
 229
 230 /* Need to check these */
 231 #define WAVE_FORMAT_DK3                 0x0061
 232 #define WAVE_FORMAT_DK4                 0x0062
 233
 234 /* At least FFmpeg use that ID: from libavformat/riff.c ('Vo' == 0x566f)
 235  * { CODEC_ID_VORBIS, ('V'<<8)+'o' }, //HACK/FIXME, does vorbis in WAV/AVI have an (in)official id?
 236  */
 237 #define WAVE_FORMAT_VORBIS              0x566f
 238
 239 /* It seems that these IDs are used by braindead & obsolete VorbisACM encoder
 240  * (Windows only)
 241  * A few info is available except VorbisACM source (remember, Windows only)
 242  * (available on http://svn.xiph.org), but it seems that vo3+ at least is
 243  * made of Vorbis data encapsulated in Ogg container...
 244  */
 245 #define WAVE_FORMAT_VORB_1              0x674f
 246 #define WAVE_FORMAT_VORB_2              0x6750
 247 #define WAVE_FORMAT_VORB_3              0x6751
 248 #define WAVE_FORMAT_VORB_1PLUS          0x676f
 249 #define WAVE_FORMAT_VORB_2PLUS          0x6770
 250 #define WAVE_FORMAT_VORB_3PLUS          0x6771
 251
 252 #define WAVE_FORMAT_SPEEX               0xa109 /* Speex audio */
 253
 254
 255 #if !defined(WAVE_FORMAT_EXTENSIBLE)
 256 #define WAVE_FORMAT_EXTENSIBLE          0xFFFE /* Microsoft */
 257 #endif
 258
 259 /* GUID SubFormat IDs */
 260 /* We need both b/c const variables are not compile-time constants in C, giving
 261  * us an error if we use the const GUID in an enum */
 262
 263 #ifndef _KSDATAFORMAT_SUBTYPE_PCM_
 264 #define _KSDATAFORMAT_SUBTYPE_PCM_ {0x00000001, 0x0000, 0x0010, {0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71}}
 265 static const GUID VLC_KSDATAFORMAT_SUBTYPE_PCM = {0xE923AABF, 0xCB58, 0x4471, {0xA1, 0x19, 0xFF, 0xFA, 0x01, 0xE4, 0xCE, 0x62}};
 266 #define KSDATAFORMAT_SUBTYPE_PCM VLC_KSDATAFORMAT_SUBTYPE_PCM
 267 #endif
 268
 269 #ifndef _KSDATAFORMAT_SUBTYPE_UNKNOWN_
 270 #define _KSDATAFORMAT_SUBTYPE_UNKNOWN_ {0x00000000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}}
 271 static const GUID VLC_KSDATAFORMAT_SUBTYPE_UNKNOWN = {0x00000000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}};
 272 #define KSDATAFORMAT_SUBTYPE_UNKNOWN VLC_KSDATAFORMAT_SUBTYPE_UNKNOWN
 273 #endif
 274
 275 /* Microsoft speaker definitions */
 276 #define WAVE_SPEAKER_FRONT_LEFT             0x1
 277 #define WAVE_SPEAKER_FRONT_RIGHT            0x2
 278 #define WAVE_SPEAKER_FRONT_CENTER           0x4
 279 #define WAVE_SPEAKER_LOW_FREQUENCY          0x8
 280 #define WAVE_SPEAKER_BACK_LEFT              0x10
 281 #define WAVE_SPEAKER_BACK_RIGHT             0x20
 282 #define WAVE_SPEAKER_FRONT_LEFT_OF_CENTER   0x40
 283 #define WAVE_SPEAKER_FRONT_RIGHT_OF_CENTER  0x80
 284 #define WAVE_SPEAKER_BACK_CENTER            0x100
 285 #define WAVE_SPEAKER_SIDE_LEFT              0x200
 286 #define WAVE_SPEAKER_SIDE_RIGHT             0x400
 287 #define WAVE_SPEAKER_TOP_CENTER             0x800
 288 #define WAVE_SPEAKER_TOP_FRONT_LEFT         0x1000
 289 #define WAVE_SPEAKER_TOP_FRONT_CENTER       0x2000
 290 #define WAVE_SPEAKER_TOP_FRONT_RIGHT        0x4000
 291 #define WAVE_SPEAKER_TOP_BACK_LEFT          0x8000
 292 #define WAVE_SPEAKER_TOP_BACK_CENTER        0x10000
 293 #define WAVE_SPEAKER_TOP_BACK_RIGHT         0x20000
 294 #define WAVE_SPEAKER_RESERVED               0x80000000
 295
 296 static const struct
 297 {
 298     uint16_t     i_tag;
 299     vlc_fourcc_t i_fourcc;
 300     const char  *psz_name;
 301 }
 302 wave_format_tag_to_fourcc[] =
 303 {
 304     { WAVE_FORMAT_PCM,        VLC_FOURCC( 'a', 'r', 'a', 'w' ), "Raw audio" },
 305     { WAVE_FORMAT_ADPCM,      VLC_CODEC_ADPCM_MS,               "ADPCM" },
 306     { WAVE_FORMAT_IEEE_FLOAT, VLC_FOURCC( 'a', 'f', 'l', 't' ), "IEEE Float audio" },
 307     { WAVE_FORMAT_ALAW,       VLC_CODEC_ALAW,                   "A-Law" },
 308     { WAVE_FORMAT_MULAW,      VLC_CODEC_MULAW,                  "Mu-Law" },
 309     { WAVE_FORMAT_IMA_ADPCM,  VLC_CODEC_ADPCM_IMA_WAV,          "Ima-ADPCM" },
 310     { WAVE_FORMAT_TRUESPEECH, VLC_FOURCC(0x22, 0x0, 0x0, 0x0 ), "Truespeech" },
 311     { WAVE_FORMAT_GSM610,     VLC_CODEC_GSM_MS,                 "Microsoft WAV GSM" },
 312     { WAVE_FORMAT_G726,       VLC_CODEC_ADPCM_G726,             "G.726 ADPCM" },
 313     { WAVE_FORMAT_MPEGLAYER3, VLC_CODEC_MPGA,                   "Mpeg Audio" },
 314     { WAVE_FORMAT_MPEG,       VLC_CODEC_MPGA,                   "Mpeg Audio" },
 315     { WAVE_FORMAT_A52,        VLC_CODEC_A52,                    "A/52" },
 316     { WAVE_FORMAT_WMA1,       VLC_CODEC_WMA1,                   "Window Media Audio v1" },
 317     { WAVE_FORMAT_WMA2,       VLC_CODEC_WMA2,                   "Window Media Audio v2" },
 318     { WAVE_FORMAT_WMAP,       VLC_CODEC_WMAP,                   "Window Media Audio 9 Professional" },
 319     { WAVE_FORMAT_WMAL,       VLC_CODEC_WMAL,                   "Window Media Audio 9 Lossless" },
 320     { WAVE_FORMAT_WMAS,       VLC_CODEC_WMAS,                   "Window Media Audio 9 Speech" },
 321     { WAVE_FORMAT_DK3,        VLC_FOURCC( 'm', 's', 0x00,0x61), "Duck DK3" },
 322     { WAVE_FORMAT_DK4,        VLC_FOURCC( 'm', 's', 0x00,0x62), "Duck DK4" },
 323     { WAVE_FORMAT_DTS,        VLC_CODEC_DTS,                    "DTS Coherent Acoustics" },
 324     { WAVE_FORMAT_DTS_MS,     VLC_CODEC_DTS,                    "DTS Coherent Acoustics" },
 325     { WAVE_FORMAT_DIVIO_AAC,  VLC_CODEC_MP4A,                   "MPEG-4 Audio (Divio)" },
 326     { WAVE_FORMAT_AAC,        VLC_CODEC_MP4A,                   "MPEG-4 Audio" },
 327     { WAVE_FORMAT_FFMPEG_AAC, VLC_CODEC_MP4A,                   "MPEG-4 Audio" },
 328     { WAVE_FORMAT_VORBIS,     VLC_CODEC_VORBIS,                 "Vorbis Audio" },
 329     { WAVE_FORMAT_VORB_1,     VLC_FOURCC( 'v', 'o', 'r', '1' ), "Vorbis 1 Audio" },
 330     { WAVE_FORMAT_VORB_1PLUS, VLC_FOURCC( 'v', 'o', '1', '+' ), "Vorbis 1+ Audio" },
 331     { WAVE_FORMAT_VORB_2,     VLC_FOURCC( 'v', 'o', 'r', '2' ), "Vorbis 2 Audio" },
 332     { WAVE_FORMAT_VORB_2PLUS, VLC_FOURCC( 'v', 'o', '2', '+' ), "Vorbis 2+ Audio" },
 333     { WAVE_FORMAT_VORB_3,     VLC_FOURCC( 'v', 'o', 'r', '3' ), "Vorbis 3 Audio" },
 334     { WAVE_FORMAT_VORB_3PLUS, VLC_FOURCC( 'v', 'o', '3', '+' ), "Vorbis 3+ Audio" },
 335     { WAVE_FORMAT_SPEEX,      VLC_CODEC_SPEEX,                  "Speex Audio" },
 336     { WAVE_FORMAT_UNKNOWN,    VLC_FOURCC( 'u', 'n', 'd', 'f' ), "Unknown" }
 337 };
 338
 339 static inline void wf_tag_to_fourcc( uint16_t i_tag, vlc_fourcc_t *fcc,
 340                                      const char **ppsz_name )
 341 {
 342     int i;
 343     for( i = 0; wave_format_tag_to_fourcc[i].i_tag != 0; i++ )
 344     {
 345         if( wave_format_tag_to_fourcc[i].i_tag == i_tag ) break;
 346     }
 347     if( fcc ) *fcc = wave_format_tag_to_fourcc[i].i_fourcc;
 348     if( ppsz_name ) *ppsz_name = wave_format_tag_to_fourcc[i].psz_name;
 349 }
 350
 351 static inline void fourcc_to_wf_tag( vlc_fourcc_t fcc, uint16_t *pi_tag )
 352 {
 353     int i;
 354     for( i = 0; wave_format_tag_to_fourcc[i].i_tag != 0; i++ )
 355     {
 356         if( wave_format_tag_to_fourcc[i].i_fourcc == fcc ) break;
 357     }
 358     if( pi_tag ) *pi_tag = wave_format_tag_to_fourcc[i].i_tag;
 359 }
 360
 361 /* If wFormatTag is WAVEFORMATEXTENSIBLE, we must look at the SubFormat tag
 362  * to determine the actual format.  Microsoft has stopped giving out wFormatTag
 363  * assignments in lieu of letting 3rd parties generate their own GUIDs
 364  */
 365 static const struct
 366 {
 367     GUID         guid_tag;
 368     vlc_fourcc_t i_fourcc;
 369     const char  *psz_name;
 370 }
 371 sub_format_tag_to_fourcc[] =
 372 {
 373     { _KSDATAFORMAT_SUBTYPE_PCM_, VLC_FOURCC( 'p', 'c', 'm', ' ' ), "PCM" },
 374     { _KSDATAFORMAT_SUBTYPE_UNKNOWN_, VLC_FOURCC( 'u', 'n', 'd', 'f' ), "Unknown" }
 375 };
 376
 377 /* compares two GUIDs, returns 1 if identical, 0 otherwise */
 378 static inline int guidcmp( const GUID *s1, const GUID *s2 )
 379 {
 380     return( s1->Data1 == s2->Data1 && s1->Data2 == s2->Data2 &&
 381             s1->Data3 == s2->Data3 && !memcmp( s1->Data4, s2->Data4, 8 ) );
 382 }
 383
 384 static inline void sf_tag_to_fourcc( GUID *guid_tag,
 385                                      vlc_fourcc_t *fcc, const char **ppsz_name )
 386 {
 387     int i;
 388
 389     for( i = 0; !guidcmp( &sub_format_tag_to_fourcc[i].guid_tag,
 390                           &KSDATAFORMAT_SUBTYPE_UNKNOWN ); i++ )
 391     {
 392         if( guidcmp( &sub_format_tag_to_fourcc[i].guid_tag, guid_tag ) ) break;
 393     }
 394     if( fcc ) *fcc = sub_format_tag_to_fourcc[i].i_fourcc;
 395     if( ppsz_name ) *ppsz_name = sub_format_tag_to_fourcc[i].psz_name;
 396 }
 397
 398 #endif /* "codecs.h" */