modules/codec/avcodec/audio.c

   1 /*****************************************************************************
   2  * audio.c: audio decoder using ffmpeg library
   3  *****************************************************************************
   4  * Copyright (C) 1999-2003 the VideoLAN team
   5  * $Id$
   6  *
   7  * Authors: Laurent Aimar <fenrir@via.ecp.fr>
   8  *          Gildas Bazin <gbazin@videolan.org>
   9  *
  10  * This program is free software; you can redistribute it and/or modify
  11  * it under the terms of the GNU General Public License as published by
  12  * the Free Software Foundation; either version 2 of the License, or
  13  * (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  18  * GNU General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public License
  21  * along with this program; if not, write to the Free Software
  22  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  23  *****************************************************************************/
  24
  25 /*****************************************************************************
  26  * Preamble
  27  *****************************************************************************/
  28 #ifdef HAVE_CONFIG_H
  29 # include "config.h"
  30 #endif
  31
  32 #include <vlc_common.h>
  33 #include <vlc_aout.h>
  34 #include <vlc_codec.h>
  35 #include <vlc_avcodec.h>
  36
  37 /* ffmpeg header */
  38 #ifdef HAVE_LIBAVCODEC_AVCODEC_H
  39 #   include <libavcodec/avcodec.h>
  40 #else
  41 #   include <avcodec.h>
  42 #endif
  43
  44 #include "avcodec.h"
  45
  46 /*****************************************************************************
  47  * decoder_sys_t : decoder descriptor
  48  *****************************************************************************/
  49 struct decoder_sys_t
  50 {
  51     AVCODEC_COMMON_MEMBERS
  52
  53     /* Temporary buffer for libavcodec */
  54     int     i_output_max;
  55     uint8_t *p_output;
  56
  57     /*
  58      * Output properties
  59      */
  60     audio_sample_format_t aout_format;
  61     date_t                end_date;
  62
  63     /*
  64      *
  65      */
  66     uint8_t *p_samples;
  67     int     i_samples;
  68
  69     /* */
  70     int     i_reject_count;
  71
  72     /* */
  73     bool    b_extract;
  74     int     pi_extraction[AOUT_CHAN_MAX];
  75     int     i_previous_channels;
  76     int64_t i_previous_layout;
  77 };
  78
  79 #define BLOCK_FLAG_PRIVATE_REALLOCATED (1 << BLOCK_FLAG_PRIVATE_SHIFT)
  80
  81 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust );
  82
  83 static void InitDecoderConfig( decoder_t *p_dec, AVCodecContext *p_context )
  84 {
  85     if( p_dec->fmt_in.i_extra > 0 )
  86     {
  87         const uint8_t * const p_src = p_dec->fmt_in.p_extra;
  88         int i_offset;
  89         int i_size;
  90
  91         if( p_dec->fmt_in.i_codec == VLC_CODEC_FLAC )
  92         {
  93             i_offset = 8;
  94             i_size = p_dec->fmt_in.i_extra - 8;
  95         }
  96         else if( p_dec->fmt_in.i_codec == VLC_CODEC_ALAC )
  97         {
  98             static const uint8_t p_pattern[] = { 0, 0, 0, 36, 'a', 'l', 'a', 'c' };
  99             /* Find alac atom XXX it is a bit ugly */
 100             for( i_offset = 0; i_offset < p_dec->fmt_in.i_extra - sizeof(p_pattern); i_offset++ )
 101             {
 102                 if( !memcmp( &p_src[i_offset], p_pattern, sizeof(p_pattern) ) )
 103                     break;
 104             }
 105             i_size = __MIN( p_dec->fmt_in.i_extra - i_offset, 36 );
 106             if( i_size < 36 )
 107                 i_size = 0;
 108         }
 109         else
 110         {
 111             i_offset = 0;
 112             i_size = p_dec->fmt_in.i_extra;
 113         }
 114
 115         if( i_size > 0 )
 116         {
 117             p_context->extradata =
 118                 malloc( i_size + FF_INPUT_BUFFER_PADDING_SIZE );
 119             if( p_context->extradata )
 120             {
 121                 uint8_t *p_dst = p_context->extradata;
 122
 123                 p_context->extradata_size = i_size;
 124
 125                 memcpy( &p_dst[0],            &p_src[i_offset], i_size );
 126                 memset( &p_dst[i_size], 0, FF_INPUT_BUFFER_PADDING_SIZE );
 127             }
 128         }
 129     }
 130     else
 131     {
 132         p_context->extradata_size = 0;
 133         p_context->extradata = NULL;
 134     }
 135 }
 136
 137 /*****************************************************************************
 138  * InitAudioDec: initialize audio decoder
 139  *****************************************************************************
 140  * The ffmpeg codec will be opened, some memory allocated.
 141  *****************************************************************************/
 142 int InitAudioDec( decoder_t *p_dec, AVCodecContext *p_context,
 143                       AVCodec *p_codec, int i_codec_id, const char *psz_namecodec )
 144 {
 145     decoder_sys_t *p_sys;
 146
 147     /* Allocate the memory needed to store the decoder's structure */
 148     if( ( p_dec->p_sys = p_sys = malloc(sizeof(*p_sys)) ) == NULL )
 149     {
 150         return VLC_ENOMEM;
 151     }
 152
 153     p_codec->type = AVMEDIA_TYPE_AUDIO;
 154     p_context->codec_type = AVMEDIA_TYPE_AUDIO;
 155     p_context->codec_id = i_codec_id;
 156     p_sys->p_context = p_context;
 157     p_sys->p_codec = p_codec;
 158     p_sys->i_codec_id = i_codec_id;
 159     p_sys->psz_namecodec = psz_namecodec;
 160     p_sys->b_delayed_open = true;
 161
 162     // Initialize decoder extradata
 163     InitDecoderConfig( p_dec, p_context);
 164
 165     /* ***** Open the codec ***** */
 166     if( ffmpeg_OpenCodec( p_dec ) < 0 )
 167     {
 168         msg_Err( p_dec, "cannot open codec (%s)", p_sys->psz_namecodec );
 169         free( p_sys->p_context->extradata );
 170         free( p_sys );
 171         return VLC_EGENERIC;
 172     }
 173
 174     switch( i_codec_id )
 175     {
 176     case CODEC_ID_WAVPACK:
 177         p_sys->i_output_max = 8 * sizeof(int32_t) * 131072;
 178         break;
 179     case CODEC_ID_TTA:
 180         p_sys->i_output_max = p_sys->p_context->channels * sizeof(int32_t) * p_sys->p_context->sample_rate * 2;
 181         break;
 182     case CODEC_ID_FLAC:
 183         p_sys->i_output_max = 8 * sizeof(int32_t) * 65535;
 184         break;
 185 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 52, 35, 0 )
 186     case CODEC_ID_WMAPRO:
 187         p_sys->i_output_max = 8 * sizeof(float) * 6144; /* (1 << 12) * 3/2 */
 188         break;
 189 #endif
 190     default:
 191         p_sys->i_output_max = 0;
 192         break;
 193     }
 194     if( p_sys->i_output_max < AVCODEC_MAX_AUDIO_FRAME_SIZE )
 195         p_sys->i_output_max = AVCODEC_MAX_AUDIO_FRAME_SIZE;
 196     msg_Dbg( p_dec, "Using %d bytes output buffer", p_sys->i_output_max );
 197     p_sys->p_output = av_malloc( p_sys->i_output_max );
 198
 199     p_sys->p_samples = NULL;
 200     p_sys->i_samples = 0;
 201     p_sys->i_reject_count = 0;
 202     p_sys->b_extract = false;
 203     p_sys->i_previous_channels = 0;
 204     p_sys->i_previous_layout = 0;
 205
 206     /* */
 207     p_dec->fmt_out.i_cat = AUDIO_ES;
 208     /* Try to set as much information as possible but do not trust it */
 209     SetupOutputFormat( p_dec, false );
 210
 211     date_Set( &p_sys->end_date, 0 );
 212     if( p_dec->fmt_out.audio.i_rate )
 213         date_Init( &p_sys->end_date, p_dec->fmt_out.audio.i_rate, 1 );
 214     else if( p_dec->fmt_in.audio.i_rate )
 215         date_Init( &p_sys->end_date, p_dec->fmt_in.audio.i_rate, 1 );
 216
 217     return VLC_SUCCESS;
 218 }
 219
 220 /*****************************************************************************
 221  * SplitBuffer: Needed because aout really doesn't like big audio chunk and
 222  * wma produces easily > 30000 samples...
 223  *****************************************************************************/
 224 static aout_buffer_t *SplitBuffer( decoder_t *p_dec )
 225 {
 226     decoder_sys_t *p_sys = p_dec->p_sys;
 227     int i_samples = __MIN( p_sys->i_samples, 4096 );
 228     aout_buffer_t *p_buffer;
 229
 230     if( i_samples == 0 ) return NULL;
 231
 232     if( ( p_buffer = decoder_NewAudioBuffer( p_dec, i_samples ) ) == NULL )
 233         return NULL;
 234
 235     p_buffer->i_pts = date_Get( &p_sys->end_date );
 236     p_buffer->i_length = date_Increment( &p_sys->end_date, i_samples )
 237                          - p_buffer->i_pts;
 238
 239     if( p_sys->b_extract )
 240         aout_ChannelExtract( p_buffer->p_buffer, p_dec->fmt_out.audio.i_channels,
 241                              p_sys->p_samples, p_sys->p_context->channels, i_samples,
 242                              p_sys->pi_extraction, p_dec->fmt_out.audio.i_bitspersample );
 243     else
 244         memcpy( p_buffer->p_buffer, p_sys->p_samples, p_buffer->i_buffer );
 245
 246     p_sys->p_samples += i_samples * p_sys->p_context->channels * ( p_dec->fmt_out.audio.i_bitspersample / 8 );
 247     p_sys->i_samples -= i_samples;
 248
 249     return p_buffer;
 250 }
 251
 252 /*****************************************************************************
 253  * DecodeAudio: Called to decode one frame
 254  *****************************************************************************/
 255 aout_buffer_t * DecodeAudio ( decoder_t *p_dec, block_t **pp_block )
 256 {
 257     decoder_sys_t *p_sys = p_dec->p_sys;
 258     int i_used, i_output;
 259     aout_buffer_t *p_buffer;
 260     block_t *p_block;
 261     AVPacket pkt;
 262
 263     if( !pp_block || !*pp_block ) return NULL;
 264
 265     p_block = *pp_block;
 266
 267     if( !p_sys->p_context->extradata_size && p_dec->fmt_in.i_extra &&
 268         p_sys->b_delayed_open)
 269     {
 270         InitDecoderConfig( p_dec, p_sys->p_context);
 271         if( ffmpeg_OpenCodec( p_dec ) )
 272             msg_Err( p_dec, "Cannot open decoder %s", p_sys->psz_namecodec );
 273     }
 274     if( p_sys->b_delayed_open )
 275     {
 276         block_Release( p_block );
 277         return NULL;
 278     }
 279
 280     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 281     {
 282         block_Release( p_block );
 283         avcodec_flush_buffers( p_sys->p_context );
 284         p_sys->i_samples = 0;
 285         date_Set( &p_sys->end_date, 0 );
 286
 287         if( p_sys->i_codec_id == CODEC_ID_MP2 || p_sys->i_codec_id == CODEC_ID_MP3 )
 288             p_sys->i_reject_count = 3;
 289         return NULL;
 290     }
 291
 292     if( p_sys->i_samples > 0 )
 293     {
 294         /* More data */
 295         p_buffer = SplitBuffer( p_dec );
 296         if( !p_buffer ) block_Release( p_block );
 297         return p_buffer;
 298     }
 299
 300     if( !date_Get( &p_sys->end_date ) && !p_block->i_pts )
 301     {
 302         /* We've just started the stream, wait for the first PTS. */
 303         block_Release( p_block );
 304         return NULL;
 305     }
 306
 307     if( p_block->i_buffer <= 0 )
 308     {
 309         block_Release( p_block );
 310         return NULL;
 311     }
 312
 313     if( (p_block->i_flags & BLOCK_FLAG_PRIVATE_REALLOCATED) == 0 )
 314     {
 315         *pp_block = p_block = block_Realloc( p_block, 0, p_block->i_buffer + FF_INPUT_BUFFER_PADDING_SIZE );
 316         if( !p_block )
 317             return NULL;
 318         p_block->i_buffer -= FF_INPUT_BUFFER_PADDING_SIZE;
 319         memset( &p_block->p_buffer[p_block->i_buffer], 0, FF_INPUT_BUFFER_PADDING_SIZE );
 320
 321         p_block->i_flags |= BLOCK_FLAG_PRIVATE_REALLOCATED;
 322     }
 323
 324     do
 325     {
 326         i_output = __MAX( p_block->i_buffer, p_sys->i_output_max );
 327         if( i_output > p_sys->i_output_max )
 328         {
 329             /* Grow output buffer if necessary (eg. for PCM data) */
 330             p_sys->p_output = av_realloc( p_sys->p_output, i_output );
 331         }
 332
 333         av_init_packet( &pkt );
 334         pkt.data = p_block->p_buffer;
 335         pkt.size = p_block->i_buffer;
 336         i_used = avcodec_decode_audio3( p_sys->p_context,
 337                                        (int16_t*)p_sys->p_output, &i_output,
 338                                        &pkt );
 339
 340         if( i_used < 0 || i_output < 0 )
 341         {
 342             if( i_used < 0 )
 343                 msg_Warn( p_dec, "cannot decode one frame (%zu bytes)",
 344                           p_block->i_buffer );
 345
 346             block_Release( p_block );
 347             return NULL;
 348         }
 349         else if( (size_t)i_used > p_block->i_buffer )
 350         {
 351             i_used = p_block->i_buffer;
 352         }
 353
 354         p_block->i_buffer -= i_used;
 355         p_block->p_buffer += i_used;
 356
 357     } while( p_block->i_buffer > 0 && i_output <= 0 );
 358
 359     if( p_sys->p_context->channels <= 0 || p_sys->p_context->channels > 8 ||
 360         p_sys->p_context->sample_rate <= 0 )
 361     {
 362         msg_Warn( p_dec, "invalid audio properties channels count %d, sample rate %d",
 363                   p_sys->p_context->channels, p_sys->p_context->sample_rate );
 364         block_Release( p_block );
 365         return NULL;
 366     }
 367
 368     if( p_dec->fmt_out.audio.i_rate != (unsigned int)p_sys->p_context->sample_rate )
 369     {
 370         date_Init( &p_sys->end_date, p_sys->p_context->sample_rate, 1 );
 371         date_Set( &p_sys->end_date, p_block->i_pts );
 372     }
 373
 374     /* **** Set audio output parameters **** */
 375     SetupOutputFormat( p_dec, true );
 376
 377     if( p_block->i_pts != 0 &&
 378         p_block->i_pts != date_Get( &p_sys->end_date ) )
 379     {
 380         date_Set( &p_sys->end_date, p_block->i_pts );
 381     }
 382     p_block->i_pts = 0;
 383
 384     /* **** Now we can output these samples **** */
 385     p_sys->i_samples = i_output / (p_dec->fmt_out.audio.i_bitspersample / 8) / p_sys->p_context->channels;
 386     p_sys->p_samples = p_sys->p_output;
 387
 388     /* Silent unwanted samples */
 389     if( p_sys->i_reject_count > 0 )
 390     {
 391         memset( p_sys->p_output, 0, i_output );
 392         p_sys->i_reject_count--;
 393     }
 394
 395     p_buffer = SplitBuffer( p_dec );
 396     if( !p_buffer ) block_Release( p_block );
 397     return p_buffer;
 398 }
 399
 400 /*****************************************************************************
 401  * EndAudioDec: audio decoder destruction
 402  *****************************************************************************/
 403 void EndAudioDec( decoder_t *p_dec )
 404 {
 405     decoder_sys_t *p_sys = p_dec->p_sys;
 406
 407     av_free( p_sys->p_output );
 408 }
 409
 410 /*****************************************************************************
 411  *
 412  *****************************************************************************/
 413
 414 void GetVlcAudioFormat( vlc_fourcc_t *pi_codec, unsigned *pi_bits, int i_sample_fmt )
 415 {
 416     switch( i_sample_fmt )
 417     {
 418     case SAMPLE_FMT_U8:
 419         *pi_codec = VLC_CODEC_U8;
 420         *pi_bits = 8;
 421         break;
 422     case SAMPLE_FMT_S32:
 423         *pi_codec = VLC_CODEC_S32N;
 424         *pi_bits = 32;
 425         break;
 426     case SAMPLE_FMT_FLT:
 427         *pi_codec = VLC_CODEC_FL32;
 428         *pi_bits = 32;
 429         break;
 430     case SAMPLE_FMT_DBL:
 431         *pi_codec = VLC_CODEC_FL64;
 432         *pi_bits = 64;
 433         break;
 434
 435     case SAMPLE_FMT_S16:
 436     default:
 437         *pi_codec = VLC_CODEC_S16N;
 438         *pi_bits = 16;
 439         break;
 440     }
 441 }
 442
 443 static const uint64_t pi_channels_map[][2] =
 444 {
 445     { CH_FRONT_LEFT,        AOUT_CHAN_LEFT },
 446     { CH_FRONT_RIGHT,       AOUT_CHAN_RIGHT },
 447     { CH_FRONT_CENTER,      AOUT_CHAN_CENTER },
 448     { CH_LOW_FREQUENCY,     AOUT_CHAN_LFE },
 449     { CH_BACK_LEFT,         AOUT_CHAN_REARLEFT },
 450     { CH_BACK_RIGHT,        AOUT_CHAN_REARRIGHT },
 451     { CH_FRONT_LEFT_OF_CENTER, 0 },
 452     { CH_FRONT_RIGHT_OF_CENTER, 0 },
 453     { CH_BACK_CENTER,       AOUT_CHAN_REARCENTER },
 454     { CH_SIDE_LEFT,         AOUT_CHAN_MIDDLELEFT },
 455     { CH_SIDE_RIGHT,        AOUT_CHAN_MIDDLERIGHT },
 456     { CH_TOP_CENTER,        0 },
 457     { CH_TOP_FRONT_LEFT,    0 },
 458     { CH_TOP_FRONT_CENTER,  0 },
 459     { CH_TOP_FRONT_RIGHT,   0 },
 460     { CH_TOP_BACK_LEFT,     0 },
 461     { CH_TOP_BACK_CENTER,   0 },
 462     { CH_TOP_BACK_RIGHT,    0 },
 463     { CH_STEREO_LEFT,       0 },
 464     { CH_STEREO_RIGHT,      0 },
 465 };
 466
 467 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust )
 468 {
 469     decoder_sys_t *p_sys = p_dec->p_sys;
 470
 471     GetVlcAudioFormat( &p_dec->fmt_out.i_codec,
 472                        &p_dec->fmt_out.audio.i_bitspersample,
 473                        p_sys->p_context->sample_fmt );
 474     p_dec->fmt_out.audio.i_rate = p_sys->p_context->sample_rate;
 475
 476     /* */
 477     if( p_sys->i_previous_channels == p_sys->p_context->channels &&
 478         p_sys->i_previous_layout == p_sys->p_context->channel_layout )
 479         return;
 480     if( b_trust )
 481     {
 482         p_sys->i_previous_channels = p_sys->p_context->channels;
 483         p_sys->i_previous_layout = p_sys->p_context->channel_layout;
 484     }
 485
 486     /* Specified order
 487      * FIXME should we use fmt_in.audio.i_physical_channels or not ?
 488      */
 489     const unsigned i_order_max = 8 * sizeof(p_sys->p_context->channel_layout);
 490     uint32_t pi_order_src[i_order_max];
 491     int i_channels_src = 0;
 492
 493     if( p_sys->p_context->channel_layout )
 494     {
 495         for( unsigned i = 0; i < sizeof(pi_channels_map)/sizeof(*pi_channels_map); i++ )
 496         {
 497             if( p_sys->p_context->channel_layout & pi_channels_map[i][0] )
 498                 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
 499         }
 500     }
 501     else
 502     {
 503         /* Create default order  */
 504         if( b_trust )
 505             msg_Warn( p_dec, "Physical channel configuration not set : guessing" );
 506         for( unsigned int i = 0; i < __MIN( i_order_max, (unsigned)p_sys->p_context->channels ); i++ )
 507         {
 508             if( i < sizeof(pi_channels_map)/sizeof(*pi_channels_map) )
 509                 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
 510         }
 511     }
 512     if( i_channels_src != p_sys->p_context->channels && b_trust )
 513         msg_Err( p_dec, "Channel layout not understood" );
 514
 515     uint32_t i_layout_dst;
 516     int      i_channels_dst;
 517     p_sys->b_extract = aout_CheckChannelExtraction( p_sys->pi_extraction,
 518                                                     &i_layout_dst, &i_channels_dst,
 519                                                     NULL, pi_order_src, i_channels_src );
 520     if( i_channels_dst != i_channels_src && b_trust )
 521         msg_Warn( p_dec, "%d channels are dropped", i_channels_src - i_channels_dst );
 522
 523     p_dec->fmt_out.audio.i_physical_channels =
 524     p_dec->fmt_out.audio.i_original_channels = i_layout_dst;
 525     p_dec->fmt_out.audio.i_channels = i_channels_dst;
 526 }
 527