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 /**
  28  * \file
  29  * This file defines codec related structures needed by the demuxers and decoders
  30  */
  31
  32 #ifdef HAVE_ATTRIBUTE_PACKED
  33 #   define ATTR_PACKED __attribute__((__packed__))
  34 #else
  35 #   error FIXME
  36 #endif
  37
  38 /* Structures exported to the demuxers and decoders */
  39
  40 #if !(defined _GUID_DEFINED || defined GUID_DEFINED)
  41 #define GUID_DEFINED
  42 typedef struct _GUID
  43 {
  44     uint32_t Data1;
  45     uint16_t Data2;
  46     uint16_t Data3;
  47     uint8_t  Data4[8];
  48 } GUID, *REFGUID, *LPGUID;
  49 #endif /* GUID_DEFINED */
  50
  51 #ifndef _WAVEFORMATEX_
  52 #define _WAVEFORMATEX_
  53 typedef struct
  54 ATTR_PACKED
  55 _WAVEFORMATEX {
  56     uint16_t   wFormatTag;
  57     uint16_t   nChannels;
  58     uint32_t   nSamplesPerSec;
  59     uint32_t   nAvgBytesPerSec;
  60     uint16_t   nBlockAlign;
  61     uint16_t   wBitsPerSample;
  62     uint16_t   cbSize;
  63 } WAVEFORMATEX, *PWAVEFORMATEX, *NPWAVEFORMATEX, *LPWAVEFORMATEX;
  64 #endif /* _WAVEFORMATEX_ */
  65
  66 #ifndef _WAVEFORMATEXTENSIBLE_
  67 #define _WAVEFORMATEXTENSIBLE_
  68 typedef struct
  69 ATTR_PACKED
  70 _WAVEFORMATEXTENSIBLE {
  71     WAVEFORMATEX Format;
  72     union {
  73         uint16_t wValidBitsPerSample;
  74         uint16_t wSamplesPerBlock;
  75         uint16_t wReserved;
  76     } Samples;
  77     uint32_t     dwChannelMask;
  78     GUID SubFormat;
  79 } WAVEFORMATEXTENSIBLE, *PWAVEFORMATEXTENSIBLE;
  80 #endif /* _WAVEFORMATEXTENSIBLE_ */
  81
  82 #ifndef _WAVEHEADER_
  83 #define _WAVEHEADER_
  84 typedef struct
  85 ATTR_PACKED
  86 _WAVEHEADER {
  87     uint32_t MainChunkID;
  88     uint32_t Length;
  89     uint32_t ChunkTypeID;
  90     uint32_t SubChunkID;
  91     uint32_t SubChunkLength;
  92     uint16_t Format;
  93     uint16_t Modus;
  94     uint32_t SampleFreq;
  95     uint32_t BytesPerSec;
  96     uint16_t BytesPerSample;
  97     uint16_t BitsPerSample;
  98     uint32_t DataChunkID;
  99     uint32_t DataLength;
 100 } WAVEHEADER;
 101 #endif /* _WAVEHEADER_ */
 102
 103 #if !defined(_BITMAPINFOHEADER_) && !defined(WIN32)
 104 #define _BITMAPINFOHEADER_
 105 typedef struct
 106 ATTR_PACKED
 107 {
 108     uint32_t   biSize;
 109     uint32_t   biWidth;
 110     uint32_t   biHeight;
 111     uint16_t   biPlanes;
 112     uint16_t   biBitCount;
 113     uint32_t   biCompression;
 114     uint32_t   biSizeImage;
 115     uint32_t   biXPelsPerMeter;
 116     uint32_t   biYPelsPerMeter;
 117     uint32_t   biClrUsed;
 118     uint32_t   biClrImportant;
 119 } BITMAPINFOHEADER, *PBITMAPINFOHEADER, *LPBITMAPINFOHEADER;
 120
 121 typedef struct
 122 ATTR_PACKED
 123 {
 124     BITMAPINFOHEADER bmiHeader;
 125     int        bmiColors[1];
 126 } BITMAPINFO, *LPBITMAPINFO;
 127 #endif
 128
 129 #ifndef _RECT32_
 130 #define _RECT32_
 131 typedef struct
 132 ATTR_PACKED
 133 {
 134     int left, top, right, bottom;
 135 } RECT32;
 136 #endif
 137
 138 #ifndef _REFERENCE_TIME_
 139 #define _REFERENCE_TIME_
 140 typedef int64_t REFERENCE_TIME;
 141 #endif
 142
 143 #ifndef _VIDEOINFOHEADER_
 144 #define _VIDEOINFOHEADER_
 145 typedef struct
 146 ATTR_PACKED
 147 {
 148     RECT32            rcSource;
 149     RECT32            rcTarget;
 150     uint32_t          dwBitRate;
 151     uint32_t          dwBitErrorRate;
 152     REFERENCE_TIME    AvgTimePerFrame;
 153     BITMAPINFOHEADER  bmiHeader;
 154 } VIDEOINFOHEADER;
 155 #endif
 156
 157 #ifndef _RGBQUAD_
 158 #define _RGBQUAD_
 159 typedef struct
 160 ATTR_PACKED
 161 {
 162     uint8_t rgbBlue;
 163     uint8_t rgbGreen;
 164     uint8_t rgbRed;
 165     uint8_t rgbReserved;
 166 } RGBQUAD1;
 167 #endif
 168
 169 #ifndef _TRUECOLORINFO_
 170 #define _TRUECOLORINFO_
 171 typedef struct
 172 ATTR_PACKED
 173 {
 174     uint32_t dwBitMasks[3];
 175     RGBQUAD1 bmiColors[256];
 176 } TRUECOLORINFO;
 177 #endif
 178
 179 #ifndef _VIDEOINFO_
 180 #define _VIDEOINFO_
 181 typedef struct
 182 ATTR_PACKED
 183 {
 184     RECT32            rcSource;
 185     RECT32            rcTarget;
 186     uint32_t          dwBitRate;
 187     uint32_t          dwBitErrorRate;
 188     REFERENCE_TIME    AvgTimePerFrame;
 189     BITMAPINFOHEADER  bmiHeader;
 190
 191     union
 192     {
 193         RGBQUAD1 bmiColors[256]; /* Colour palette */
 194         uint32_t dwBitMasks[3]; /* True colour masks */
 195         TRUECOLORINFO TrueColorInfo; /* Both of the above */
 196     };
 197
 198 } VIDEOINFO;
 199 #endif
 200
 201 /* WAVE format wFormatTag IDs */
 202 #define WAVE_FORMAT_UNKNOWN             0x0000 /* Microsoft Corporation */
 203 #define WAVE_FORMAT_PCM                 0x0001 /* Microsoft Corporation */
 204 #define WAVE_FORMAT_ADPCM               0x0002 /* Microsoft Corporation */
 205 #define WAVE_FORMAT_IEEE_FLOAT          0x0003 /* Microsoft Corporation */
 206 #define WAVE_FORMAT_ALAW                0x0006 /* Microsoft Corporation */
 207 #define WAVE_FORMAT_MULAW               0x0007 /* Microsoft Corporation */
 208 #define WAVE_FORMAT_DTS_MS              0x0008 /* Microsoft Corporation */
 209 #define WAVE_FORMAT_WMAS                0x000a /* WMA 9 Speech */
 210 #define WAVE_FORMAT_IMA_ADPCM           0x0011 /* Intel Corporation */
 211 #define WAVE_FORMAT_GSM610              0x0031 /* Microsoft Corporation */
 212 #define WAVE_FORMAT_MSNAUDIO            0x0032 /* Microsoft Corporation */
 213 #define WAVE_FORMAT_G726                0x0045 /* ITU-T standard  */
 214 #define WAVE_FORMAT_MPEG                0x0050 /* Microsoft Corporation */
 215 #define WAVE_FORMAT_MPEGLAYER3          0x0055 /* ISO/MPEG Layer3 Format Tag */
 216 #define WAVE_FORMAT_DOLBY_AC3_SPDIF     0x0092 /* Sonic Foundry */
 217
 218 #define WAVE_FORMAT_A52                 0x2000
 219 #define WAVE_FORMAT_DTS                 0x2001
 220 #define WAVE_FORMAT_WMA1                0x0160 /* WMA version 1 */
 221 #define WAVE_FORMAT_WMA2                0x0161 /* WMA (v2) 7, 8, 9 Series */
 222 #define WAVE_FORMAT_WMAP                0x0162 /* WMA 9 Professional */
 223 #define WAVE_FORMAT_WMAL                0x0163 /* WMA 9 Lossless */
 224 #define WAVE_FORMAT_DIVIO_AAC           0x4143
 225 #define WAVE_FORMAT_AAC                 0x00FF
 226 #define WAVE_FORMAT_FFMPEG_AAC          0x706D
 227
 228 /* Need to check these */
 229 #define WAVE_FORMAT_DK3                 0x0061
 230 #define WAVE_FORMAT_DK4                 0x0062
 231
 232 /* At least FFmpeg use that ID: from libavformat/riff.c ('Vo' == 0x566f)
 233  * { CODEC_ID_VORBIS, ('V'<<8)+'o' }, //HACK/FIXME, does vorbis in WAV/AVI have an (in)official id?
 234  */
 235 #define WAVE_FORMAT_VORBIS              0x566f
 236
 237 /* It seems that these IDs are used by braindead & obsolete VorbisACM encoder
 238  * (Windows only)
 239  * A few info is available except VorbisACM source (remember, Windows only)
 240  * (available on http://svn.xiph.org), but it seems that vo3+ at least is
 241  * made of Vorbis data encapsulated in Ogg container...
 242  */
 243 #define WAVE_FORMAT_VORB_1              0x674f
 244 #define WAVE_FORMAT_VORB_2              0x6750
 245 #define WAVE_FORMAT_VORB_3              0x6751
 246 #define WAVE_FORMAT_VORB_1PLUS          0x676f
 247 #define WAVE_FORMAT_VORB_2PLUS          0x6770
 248 #define WAVE_FORMAT_VORB_3PLUS          0x6771
 249
 250 #define WAVE_FORMAT_SPEEX               0xa109 /* Speex audio */
 251
 252
 253 #if !defined(WAVE_FORMAT_EXTENSIBLE)
 254 #define WAVE_FORMAT_EXTENSIBLE          0xFFFE /* Microsoft */
 255 #endif
 256
 257 /* GUID SubFormat IDs */
 258 /* We need both b/c const variables are not compile-time constants in C, giving
 259  * us an error if we use the const GUID in an enum */
 260
 261 #ifndef _KSDATAFORMAT_SUBTYPE_PCM_
 262 #define _KSDATAFORMAT_SUBTYPE_PCM_ {0x00000001, 0x0000, 0x0010, {0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71}}
 263 static const GUID VLC_KSDATAFORMAT_SUBTYPE_PCM = {0xE923AABF, 0xCB58, 0x4471, {0xA1, 0x19, 0xFF, 0xFA, 0x01, 0xE4, 0xCE, 0x62}};
 264 #define KSDATAFORMAT_SUBTYPE_PCM VLC_KSDATAFORMAT_SUBTYPE_PCM
 265 #endif
 266
 267 #ifndef _KSDATAFORMAT_SUBTYPE_UNKNOWN_
 268 #define _KSDATAFORMAT_SUBTYPE_UNKNOWN_ {0x00000000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}}
 269 static const GUID VLC_KSDATAFORMAT_SUBTYPE_UNKNOWN = {0x00000000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}};
 270 #define KSDATAFORMAT_SUBTYPE_UNKNOWN VLC_KSDATAFORMAT_SUBTYPE_UNKNOWN
 271 #endif
 272
 273 /* Microsoft speaker definitions */
 274 #define WAVE_SPEAKER_FRONT_LEFT             0x1
 275 #define WAVE_SPEAKER_FRONT_RIGHT            0x2
 276 #define WAVE_SPEAKER_FRONT_CENTER           0x4
 277 #define WAVE_SPEAKER_LOW_FREQUENCY          0x8
 278 #define WAVE_SPEAKER_BACK_LEFT              0x10
 279 #define WAVE_SPEAKER_BACK_RIGHT             0x20
 280 #define WAVE_SPEAKER_FRONT_LEFT_OF_CENTER   0x40
 281 #define WAVE_SPEAKER_FRONT_RIGHT_OF_CENTER  0x80
 282 #define WAVE_SPEAKER_BACK_CENTER            0x100
 283 #define WAVE_SPEAKER_SIDE_LEFT              0x200
 284 #define WAVE_SPEAKER_SIDE_RIGHT             0x400
 285 #define WAVE_SPEAKER_TOP_CENTER             0x800
 286 #define WAVE_SPEAKER_TOP_FRONT_LEFT         0x1000
 287 #define WAVE_SPEAKER_TOP_FRONT_CENTER       0x2000
 288 #define WAVE_SPEAKER_TOP_FRONT_RIGHT        0x4000
 289 #define WAVE_SPEAKER_TOP_BACK_LEFT          0x8000
 290 #define WAVE_SPEAKER_TOP_BACK_CENTER        0x10000
 291 #define WAVE_SPEAKER_TOP_BACK_RIGHT         0x20000
 292 #define WAVE_SPEAKER_RESERVED               0x80000000
 293
 294 static const struct
 295 {
 296     uint16_t     i_tag;
 297     vlc_fourcc_t i_fourcc;
 298     const char  *psz_name;
 299 }
 300 wave_format_tag_to_fourcc[] =
 301 {
 302     { WAVE_FORMAT_PCM,        VLC_FOURCC( 'a', 'r', 'a', 'w' ), "Raw audio" },
 303     { WAVE_FORMAT_ADPCM,      VLC_FOURCC( 'm', 's', 0x00,0x02), "ADPCM" },
 304     { WAVE_FORMAT_IEEE_FLOAT, VLC_FOURCC( 'a', 'f', 'l', 't' ), "IEEE Float audio" },
 305     { WAVE_FORMAT_ALAW,       VLC_FOURCC( 'a', 'l', 'a', 'w' ), "A-Law" },
 306     { WAVE_FORMAT_MULAW,      VLC_FOURCC( 'm', 'l', 'a', 'w' ), "Mu-Law" },
 307     { WAVE_FORMAT_IMA_ADPCM,  VLC_FOURCC( 'm', 's', 0x00,0x11), "Ima-ADPCM" },
 308     { WAVE_FORMAT_GSM610,     VLC_FOURCC( 'g', 's', 'm', ' ' ), "GSM 610" },
 309     { WAVE_FORMAT_G726,       VLC_FOURCC( 'g', '7', '2', '6' ), "G.726 ADPCM" },
 310     { WAVE_FORMAT_MPEGLAYER3, VLC_FOURCC( 'm', 'p', 'g', 'a' ), "Mpeg Audio" },
 311     { WAVE_FORMAT_MPEG,       VLC_FOURCC( 'm', 'p', 'g', 'a' ), "Mpeg Audio" },
 312     { WAVE_FORMAT_A52,        VLC_FOURCC( 'a', '5', '2', ' ' ), "A/52" },
 313     { WAVE_FORMAT_WMA1,       VLC_FOURCC( 'w', 'm', 'a', '1' ), "Window Media Audio v1" },
 314     { WAVE_FORMAT_WMA2,       VLC_FOURCC( 'w', 'm', 'a', '2' ), "Window Media Audio v2" },
 315     { WAVE_FORMAT_WMA2,       VLC_FOURCC( 'w', 'm', 'a', ' ' ), "Window Media Audio v2" },
 316     { WAVE_FORMAT_WMAP,       VLC_FOURCC( 'w', 'm', 'a', 'p' ), "Window Media Audio 9 Professional" },
 317     { WAVE_FORMAT_WMAL,       VLC_FOURCC( 'w', 'm', 'a', 'l' ), "Window Media Audio 9 Lossless" },
 318     { WAVE_FORMAT_WMAS,       VLC_FOURCC( 'w', 'm', 'a', 's' ), "Window Media Audio 9 Speech" },
 319     { WAVE_FORMAT_DK3,        VLC_FOURCC( 'm', 's', 0x00,0x61), "Duck DK3" },
 320     { WAVE_FORMAT_DK4,        VLC_FOURCC( 'm', 's', 0x00,0x62), "Duck DK4" },
 321     { WAVE_FORMAT_DTS,        VLC_FOURCC( 'd', 't', 's', ' ' ), "DTS Coherent Acoustics" },
 322     { WAVE_FORMAT_DTS_MS,     VLC_FOURCC( 'd', 't', 's', ' ' ), "DTS Coherent Acoustics" },
 323     { WAVE_FORMAT_DIVIO_AAC,  VLC_FOURCC( 'm', 'p', '4', 'a' ), "MPEG-4 Audio (Divio)" },
 324     { WAVE_FORMAT_AAC,        VLC_FOURCC( 'm', 'p', '4', 'a' ), "MPEG-4 Audio" },
 325     { WAVE_FORMAT_FFMPEG_AAC, VLC_FOURCC( 'm', 'p', '4', 'a' ), "MPEG-4 Audio" },
 326     { WAVE_FORMAT_VORBIS,     VLC_FOURCC( 'v', 'o', 'r', 'b' ), "Vorbis Audio" },
 327     { WAVE_FORMAT_VORB_1,     VLC_FOURCC( 'v', 'o', 'r', '1' ), "Vorbis 1 Audio" },
 328     { WAVE_FORMAT_VORB_1PLUS, VLC_FOURCC( 'v', 'o', '1', '+' ), "Vorbis 1+ Audio" },
 329     { WAVE_FORMAT_VORB_2,     VLC_FOURCC( 'v', 'o', 'r', '2' ), "Vorbis 2 Audio" },
 330     { WAVE_FORMAT_VORB_2PLUS, VLC_FOURCC( 'v', 'o', '2', '+' ), "Vorbis 2+ Audio" },
 331     { WAVE_FORMAT_VORB_3,     VLC_FOURCC( 'v', 'o', 'r', '3' ), "Vorbis 3 Audio" },
 332     { WAVE_FORMAT_VORB_3PLUS, VLC_FOURCC( 'v', 'o', '3', '+' ), "Vorbis 3+ Audio" },
 333     { WAVE_FORMAT_SPEEX,      VLC_FOURCC( 's', 'p', 'x', ' ' ), "Speex Audio" },
 334     { WAVE_FORMAT_UNKNOWN,    VLC_FOURCC( 'u', 'n', 'd', 'f' ), "Unknown" }
 335 };
 336
 337 static inline void wf_tag_to_fourcc( uint16_t i_tag, vlc_fourcc_t *fcc,
 338                                      const char **ppsz_name )
 339 {
 340     int i;
 341     for( i = 0; wave_format_tag_to_fourcc[i].i_tag != 0; i++ )
 342     {
 343         if( wave_format_tag_to_fourcc[i].i_tag == i_tag ) break;
 344     }
 345     if( fcc ) *fcc = wave_format_tag_to_fourcc[i].i_fourcc;
 346     if( ppsz_name ) *ppsz_name = wave_format_tag_to_fourcc[i].psz_name;
 347 }
 348
 349 static inline void fourcc_to_wf_tag( vlc_fourcc_t fcc, uint16_t *pi_tag )
 350 {
 351     int i;
 352     for( i = 0; wave_format_tag_to_fourcc[i].i_tag != 0; i++ )
 353     {
 354         if( wave_format_tag_to_fourcc[i].i_fourcc == fcc ) break;
 355     }
 356     if( pi_tag ) *pi_tag = wave_format_tag_to_fourcc[i].i_tag;
 357 }
 358
 359 /* If wFormatTag is WAVEFORMATEXTENSIBLE, we must look at the SubFormat tag
 360  * to determine the actual format.  Microsoft has stopped giving out wFormatTag
 361  * assignments in lieu of letting 3rd parties generate their own GUIDs
 362  */
 363 static const struct
 364 {
 365     GUID         guid_tag;
 366     vlc_fourcc_t i_fourcc;
 367     const char  *psz_name;
 368 }
 369 sub_format_tag_to_fourcc[] =
 370 {
 371     { _KSDATAFORMAT_SUBTYPE_PCM_, VLC_FOURCC( 'p', 'c', 'm', ' ' ), "PCM" },
 372     { _KSDATAFORMAT_SUBTYPE_UNKNOWN_, VLC_FOURCC( 'u', 'n', 'd', 'f' ), "Unknown" }
 373 };
 374
 375 /* compares two GUIDs, returns 1 if identical, 0 otherwise */
 376 static inline int guidcmp( const GUID *s1, const GUID *s2 )
 377 {
 378     return( s1->Data1 == s2->Data1 && s1->Data2 == s2->Data2 &&
 379             s1->Data3 == s2->Data3 && !memcmp( s1->Data4, s2->Data4, 8 ) );
 380 }
 381
 382 static inline void sf_tag_to_fourcc( GUID *guid_tag,
 383                                      vlc_fourcc_t *fcc, const char **ppsz_name )
 384 {
 385     int i;
 386
 387     for( i = 0; !guidcmp( &sub_format_tag_to_fourcc[i].guid_tag,
 388                           &KSDATAFORMAT_SUBTYPE_UNKNOWN ); i++ )
 389     {
 390         if( guidcmp( &sub_format_tag_to_fourcc[i].guid_tag, guid_tag ) ) break;
 391     }
 392     if( fcc ) *fcc = sub_format_tag_to_fourcc[i].i_fourcc;
 393     if( ppsz_name ) *ppsz_name = sub_format_tag_to_fourcc[i].psz_name;
 394 }
 395
 396 /**
 397  * Structure to hold information concerning subtitles.
 398  * Used between demuxers and decoders of subtitles.
 399  */
 400 typedef struct es_sys_t
 401 {
 402     char               *psz_header; /* for 'ssa ' and 'subt' */
 403
 404     /* for spudec */
 405     unsigned int        i_orig_height;
 406     unsigned int        i_orig_width;
 407     unsigned int        i_origin_x;
 408     unsigned int        i_origin_y;
 409     unsigned int        i_scale_h;
 410     unsigned int        i_scale_v;
 411     unsigned int        i_alpha;
 412     bool          b_smooth;
 413     mtime_t             i_fade_in;
 414     mtime_t             i_fade_out;
 415     unsigned int        i_align;
 416     mtime_t             i_time_offset;
 417     bool          b_forced_subs;
 418     unsigned int        palette[16];
 419     unsigned int        colors[4];
 420
 421 } subtitle_data_t;
 422
 423 #endif /* "codecs.h" */