git.sesse.net Git - vlc/blob - modules/codec/avcodec/audio.c

   1 /*****************************************************************************
   2  * audio.c: audio decoder using libavcodec library
   3  *****************************************************************************
   4  * Copyright (C) 1999-2003 VLC authors and VideoLAN
   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 it
  11  * under the terms of the GNU Lesser General Public License as published by
  12  * the Free Software Foundation; either version 2.1 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 Lesser General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public License
  21  * along with this program; if not, write to the Free Software Foundation,
  22  * 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 <assert.h>
  33
  34 #include <vlc_common.h>
  35 #include <vlc_aout.h>
  36 #include <vlc_codec.h>
  37 #include <vlc_avcodec.h>
  38
  39 #include <libavcodec/avcodec.h>
  40 #include <libavutil/mem.h>
  41
  42 #include <libavutil/audioconvert.h>
  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     /*
  54      * Output properties
  55      */
  56     audio_sample_format_t aout_format;
  57     date_t                end_date;
  58
  59     /* */
  60     int     i_reject_count;
  61
  62     /* */
  63     bool    b_extract;
  64     int     pi_extraction[AOUT_CHAN_MAX];
  65     int     i_previous_channels;
  66     uint64_t i_previous_layout;
  67 };
  68
  69 #define BLOCK_FLAG_PRIVATE_REALLOCATED (1 << BLOCK_FLAG_PRIVATE_SHIFT)
  70
  71 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust );
  72
  73 static void InitDecoderConfig( decoder_t *p_dec, AVCodecContext *p_context )
  74 {
  75     if( p_dec->fmt_in.i_extra > 0 )
  76     {
  77         const uint8_t * const p_src = p_dec->fmt_in.p_extra;
  78
  79         int i_offset = 0;
  80         int i_size = p_dec->fmt_in.i_extra;
  81
  82         if( p_dec->fmt_in.i_codec == VLC_CODEC_ALAC )
  83         {
  84             static const uint8_t p_pattern[] = { 0, 0, 0, 36, 'a', 'l', 'a', 'c' };
  85             /* Find alac atom XXX it is a bit ugly */
  86             for( i_offset = 0; i_offset < i_size - (int)sizeof(p_pattern); i_offset++ )
  87             {
  88                 if( !memcmp( &p_src[i_offset], p_pattern, sizeof(p_pattern) ) )
  89                     break;
  90             }
  91             i_size = __MIN( p_dec->fmt_in.i_extra - i_offset, 36 );
  92             if( i_size < 36 )
  93                 i_size = 0;
  94         }
  95
  96         if( i_size > 0 )
  97         {
  98             p_context->extradata =
  99                 av_malloc( i_size + FF_INPUT_BUFFER_PADDING_SIZE );
 100             if( p_context->extradata )
 101             {
 102                 uint8_t *p_dst = p_context->extradata;
 103
 104                 p_context->extradata_size = i_size;
 105
 106                 memcpy( &p_dst[0],            &p_src[i_offset], i_size );
 107                 memset( &p_dst[i_size], 0, FF_INPUT_BUFFER_PADDING_SIZE );
 108             }
 109         }
 110     }
 111     else
 112     {
 113         p_context->extradata_size = 0;
 114         p_context->extradata = NULL;
 115     }
 116 }
 117
 118 /**
 119  * Allocates decoded audio buffer for libavcodec to use.
 120  */
 121 #if (LIBAVCODEC_VERSION_MAJOR >= 55)
 122 typedef struct
 123 {
 124     block_t self;
 125     AVFrame *frame;
 126 } vlc_av_frame_t;
 127
 128 static void vlc_av_frame_Release(block_t *block)
 129 {
 130     vlc_av_frame_t *b = (void *)block;
 131
 132     av_frame_free(&b->frame);
 133     free(b);
 134 }
 135
 136 static block_t *vlc_av_frame_Wrap(AVFrame *frame)
 137 {
 138     for (unsigned i = 1; i < AV_NUM_DATA_POINTERS; i++)
 139         assert(frame->linesize[i] == 0); /* only packed frame supported */
 140
 141     if (av_frame_make_writable(frame)) /* TODO: read-only block_t */
 142         return NULL;
 143
 144     vlc_av_frame_t *b = malloc(sizeof (*b));
 145     if (unlikely(b == NULL))
 146         return NULL;
 147
 148     block_t *block = &b->self;
 149
 150     block_Init(block, frame->extended_data[0], frame->linesize[0]);
 151     block->i_nb_samples = frame->nb_samples;
 152     block->pf_release = vlc_av_frame_Release;
 153     b->frame = frame;
 154     return block;
 155 }
 156 #else
 157 static int GetAudioBuf( AVCodecContext *ctx, AVFrame *buf )
 158 {
 159     block_t *block;
 160     bool planar = av_sample_fmt_is_planar( ctx->sample_fmt );
 161     unsigned channels = planar ? 1 : ctx->channels;
 162     unsigned planes = planar ? ctx->channels : 1;
 163
 164     int bytes = av_samples_get_buffer_size( &buf->linesize[0], channels,
 165                                             buf->nb_samples, ctx->sample_fmt,
 166                                             16 );
 167     assert( bytes >= 0 );
 168     block = block_Alloc( bytes * planes );
 169     if( unlikely(block == NULL) )
 170         return AVERROR(ENOMEM);
 171
 172     block->i_nb_samples = buf->nb_samples;
 173     buf->opaque = block;
 174
 175     if( planes > AV_NUM_DATA_POINTERS )
 176     {
 177         uint8_t **ext = malloc( sizeof( *ext ) * planes );
 178         if( unlikely(ext == NULL) )
 179         {
 180             block_Release( block );
 181             return AVERROR(ENOMEM);
 182         }
 183         buf->extended_data = ext;
 184     }
 185     else
 186         buf->extended_data = buf->data;
 187
 188     uint8_t *buffer = block->p_buffer;
 189     for( unsigned i = 0; i < planes; i++ )
 190     {
 191         buf->linesize[i] = buf->linesize[0];
 192         buf->extended_data[i] = buffer;
 193         buffer += bytes;
 194     }
 195
 196     return 0;
 197 }
 198 #endif
 199
 200 /*****************************************************************************
 201  * InitAudioDec: initialize audio decoder
 202  *****************************************************************************
 203  * The avcodec codec will be opened, some memory allocated.
 204  *****************************************************************************/
 205 int InitAudioDec( decoder_t *p_dec, AVCodecContext *p_context,
 206                       AVCodec *p_codec, int i_codec_id, const char *psz_namecodec )
 207 {
 208     decoder_sys_t *p_sys;
 209
 210     /* Allocate the memory needed to store the decoder's structure */
 211     if( ( p_dec->p_sys = p_sys = malloc(sizeof(*p_sys)) ) == NULL )
 212     {
 213         return VLC_ENOMEM;
 214     }
 215
 216     p_codec->type = AVMEDIA_TYPE_AUDIO;
 217     p_context->codec_type = AVMEDIA_TYPE_AUDIO;
 218     p_context->codec_id = i_codec_id;
 219 #if (LIBAVCODEC_VERSION_MAJOR >= 55)
 220     p_context->refcounted_frames = true;
 221 #else
 222     p_context->get_buffer = GetAudioBuf;
 223 #endif
 224     p_sys->p_context = p_context;
 225     p_sys->p_codec = p_codec;
 226     p_sys->i_codec_id = i_codec_id;
 227     p_sys->psz_namecodec = psz_namecodec;
 228     p_sys->b_delayed_open = true;
 229
 230     // Initialize decoder extradata
 231     InitDecoderConfig( p_dec, p_context);
 232
 233     /* ***** Open the codec ***** */
 234     if( ffmpeg_OpenCodec( p_dec ) < 0 )
 235     {
 236         msg_Err( p_dec, "cannot open codec (%s)", p_sys->psz_namecodec );
 237         av_free( p_sys->p_context->extradata );
 238         free( p_sys );
 239         return VLC_EGENERIC;
 240     }
 241
 242     p_sys->i_reject_count = 0;
 243     p_sys->b_extract = false;
 244     p_sys->i_previous_channels = 0;
 245     p_sys->i_previous_layout = 0;
 246
 247     /* */
 248     p_dec->fmt_out.i_cat = AUDIO_ES;
 249     /* Try to set as much information as possible but do not trust it */
 250     SetupOutputFormat( p_dec, false );
 251
 252     date_Set( &p_sys->end_date, 0 );
 253     if( p_dec->fmt_out.audio.i_rate )
 254         date_Init( &p_sys->end_date, p_dec->fmt_out.audio.i_rate, 1 );
 255     else if( p_dec->fmt_in.audio.i_rate )
 256         date_Init( &p_sys->end_date, p_dec->fmt_in.audio.i_rate, 1 );
 257
 258     return VLC_SUCCESS;
 259 }
 260
 261 /*****************************************************************************
 262  * DecodeAudio: Called to decode one frame
 263  *****************************************************************************/
 264 block_t * DecodeAudio ( decoder_t *p_dec, block_t **pp_block )
 265 {
 266     decoder_sys_t *p_sys = p_dec->p_sys;
 267     AVCodecContext *ctx = p_sys->p_context;
 268
 269     if( !pp_block || !*pp_block )
 270         return NULL;
 271
 272     block_t *p_block = *pp_block;
 273
 274     if( !ctx->extradata_size && p_dec->fmt_in.i_extra && p_sys->b_delayed_open)
 275     {
 276         InitDecoderConfig( p_dec, ctx );
 277         if( ffmpeg_OpenCodec( p_dec ) )
 278             msg_Err( p_dec, "Cannot open decoder %s", p_sys->psz_namecodec );
 279     }
 280
 281     if( p_sys->b_delayed_open )
 282         goto end;
 283
 284     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 285     {
 286         avcodec_flush_buffers( ctx );
 287         date_Set( &p_sys->end_date, 0 );
 288
 289         if( p_sys->i_codec_id == AV_CODEC_ID_MP2 || p_sys->i_codec_id == AV_CODEC_ID_MP3 )
 290             p_sys->i_reject_count = 3;
 291
 292         goto end;
 293     }
 294
 295     /* We've just started the stream, wait for the first PTS. */
 296     if( !date_Get( &p_sys->end_date ) && p_block->i_pts <= VLC_TS_INVALID )
 297         goto end;
 298
 299     if( p_block->i_buffer <= 0 )
 300         goto end;
 301
 302     if( (p_block->i_flags & BLOCK_FLAG_PRIVATE_REALLOCATED) == 0 )
 303     {
 304         p_block = block_Realloc( p_block, 0, p_block->i_buffer + FF_INPUT_BUFFER_PADDING_SIZE );
 305         if( !p_block )
 306             return NULL;
 307         *pp_block = p_block;
 308         p_block->i_buffer -= FF_INPUT_BUFFER_PADDING_SIZE;
 309         memset( &p_block->p_buffer[p_block->i_buffer], 0, FF_INPUT_BUFFER_PADDING_SIZE );
 310
 311         p_block->i_flags |= BLOCK_FLAG_PRIVATE_REALLOCATED;
 312     }
 313
 314 #if (LIBAVCODEC_VERSION_MAJOR >= 55)
 315     AVFrame *frame = av_frame_alloc();
 316     if (unlikely(frame == NULL))
 317         goto end;
 318 #else
 319     AVFrame *frame = &(AVFrame) { };
 320 #endif
 321
 322     for( int got_frame = 0; !got_frame; )
 323     {
 324         if( p_block->i_buffer == 0 )
 325             goto end;
 326
 327         AVPacket pkt;
 328         av_init_packet( &pkt );
 329         pkt.data = p_block->p_buffer;
 330         pkt.size = p_block->i_buffer;
 331
 332         int used = avcodec_decode_audio4( ctx, frame, &got_frame, &pkt );
 333         if( used < 0 )
 334         {
 335             msg_Warn( p_dec, "cannot decode one frame (%zu bytes)",
 336                       p_block->i_buffer );
 337             goto end;
 338         }
 339
 340         assert( p_block->i_buffer >= (unsigned)used );
 341         p_block->p_buffer += used;
 342         p_block->i_buffer -= used;
 343     }
 344
 345     if( ctx->channels <= 0 || ctx->channels > 8 || ctx->sample_rate <= 0 )
 346     {
 347         msg_Warn( p_dec, "invalid audio properties channels count %d, sample rate %d",
 348                   ctx->channels, ctx->sample_rate );
 349         goto end;
 350     }
 351
 352     if( p_dec->fmt_out.audio.i_rate != (unsigned int)ctx->sample_rate )
 353         date_Init( &p_sys->end_date, ctx->sample_rate, 1 );
 354
 355     if( p_block->i_pts > VLC_TS_INVALID &&
 356         p_block->i_pts > date_Get( &p_sys->end_date ) )
 357     {
 358         date_Set( &p_sys->end_date, p_block->i_pts );
 359     }
 360
 361     if( p_block->i_buffer == 0 )
 362     {   /* Done with this buffer */
 363         block_Release( p_block );
 364         *pp_block = NULL;
 365     }
 366
 367 #if (LIBAVCODEC_VERSION_MAJOR < 55)
 368     /* NOTE WELL: Beyond this point, p_block refers to the DECODED block! */
 369     p_block = frame->opaque;
 370 #endif
 371     SetupOutputFormat( p_dec, true );
 372     if( decoder_UpdateAudioFormat( p_dec ) )
 373         goto drop;
 374
 375     /* Interleave audio if required */
 376     if( av_sample_fmt_is_planar( ctx->sample_fmt ) )
 377 #if (LIBAVCODEC_VERSION_MAJOR >= 55)
 378     {
 379         p_block = block_Alloc(frame->linesize[0] * ctx->channels);
 380         if (unlikely(p_block == NULL))
 381             goto drop;
 382
 383         const void *planes[ctx->channels];
 384         for (int i = 0; i < ctx->channels; i++)
 385             planes[i] = frame->extended_data[i];
 386
 387         aout_Interleave(p_block->p_buffer, planes, frame->nb_samples,
 388                         ctx->channels, p_dec->fmt_out.audio.i_format);
 389         p_block->i_nb_samples = frame->nb_samples;
 390         av_frame_free(&frame);
 391     }
 392     else
 393     {
 394         p_block = vlc_av_frame_Wrap(frame);
 395         if (unlikely(p_block == NULL))
 396             goto drop;
 397     }
 398 #else
 399     {
 400         block_t *p_buffer = block_Alloc( p_block->i_buffer );
 401         if( unlikely(p_buffer == NULL) )
 402             goto drop;
 403
 404         const void *planes[ctx->channels];
 405         for( int i = 0; i < ctx->channels; i++)
 406             planes[i] = frame->extended_data[i];
 407
 408         aout_Interleave( p_buffer->p_buffer, planes, frame->nb_samples,
 409                          ctx->channels, p_dec->fmt_out.audio.i_format );
 410         if( ctx->channels > AV_NUM_DATA_POINTERS )
 411             free( frame->extended_data );
 412         block_Release( p_block );
 413         p_block = p_buffer;
 414     }
 415     p_block->i_nb_samples = frame->nb_samples;
 416 #endif
 417
 418     if (p_sys->b_extract)
 419     {   /* TODO: do not drop channels... at least not here */
 420         block_t *p_buffer = block_Alloc( p_dec->fmt_out.audio.i_bytes_per_frame
 421                                          * p_block->i_nb_samples );
 422         if( unlikely(p_buffer == NULL) )
 423             goto drop;
 424         aout_ChannelExtract( p_buffer->p_buffer,
 425                              p_dec->fmt_out.audio.i_channels,
 426                              p_block->p_buffer, ctx->channels,
 427                              p_block->i_nb_samples, p_sys->pi_extraction,
 428                              p_dec->fmt_out.audio.i_bitspersample );
 429         p_buffer->i_nb_samples = p_block->i_nb_samples;
 430         block_Release( p_block );
 431         p_block = p_buffer;
 432     }
 433
 434     /* Silent unwanted samples */
 435     if( p_sys->i_reject_count > 0 )
 436     {
 437         memset( p_block->p_buffer, 0, p_block->i_buffer );
 438         p_sys->i_reject_count--;
 439     }
 440
 441     p_block->i_buffer = p_block->i_nb_samples
 442                         * p_dec->fmt_out.audio.i_bytes_per_frame;
 443     p_block->i_pts = date_Get( &p_sys->end_date );
 444     p_block->i_length = date_Increment( &p_sys->end_date,
 445                                       p_block->i_nb_samples ) - p_block->i_pts;
 446     return p_block;
 447
 448 end:
 449     *pp_block = NULL;
 450 drop:
 451     block_Release(p_block);
 452     return NULL;
 453 }
 454
 455 /*****************************************************************************
 456  *
 457  *****************************************************************************/
 458
 459 vlc_fourcc_t GetVlcAudioFormat( int fmt )
 460 {
 461     static const vlc_fourcc_t fcc[] = {
 462         [AV_SAMPLE_FMT_U8]    = VLC_CODEC_U8,
 463         [AV_SAMPLE_FMT_S16]   = VLC_CODEC_S16N,
 464         [AV_SAMPLE_FMT_S32]   = VLC_CODEC_S32N,
 465         [AV_SAMPLE_FMT_FLT]   = VLC_CODEC_FL32,
 466         [AV_SAMPLE_FMT_DBL]   = VLC_CODEC_FL64,
 467         [AV_SAMPLE_FMT_U8P]   = VLC_CODEC_U8,
 468         [AV_SAMPLE_FMT_S16P]  = VLC_CODEC_S16N,
 469         [AV_SAMPLE_FMT_S32P]  = VLC_CODEC_S32N,
 470         [AV_SAMPLE_FMT_FLTP]  = VLC_CODEC_FL32,
 471         [AV_SAMPLE_FMT_DBLP]  = VLC_CODEC_FL64,
 472     };
 473     if( (sizeof(fcc) / sizeof(fcc[0])) > (unsigned)fmt )
 474         return fcc[fmt];
 475     return VLC_CODEC_S16N;
 476 }
 477
 478 static const uint64_t pi_channels_map[][2] =
 479 {
 480     { AV_CH_FRONT_LEFT,        AOUT_CHAN_LEFT },
 481     { AV_CH_FRONT_RIGHT,       AOUT_CHAN_RIGHT },
 482     { AV_CH_FRONT_CENTER,      AOUT_CHAN_CENTER },
 483     { AV_CH_LOW_FREQUENCY,     AOUT_CHAN_LFE },
 484     { AV_CH_BACK_LEFT,         AOUT_CHAN_REARLEFT },
 485     { AV_CH_BACK_RIGHT,        AOUT_CHAN_REARRIGHT },
 486     { AV_CH_FRONT_LEFT_OF_CENTER, 0 },
 487     { AV_CH_FRONT_RIGHT_OF_CENTER, 0 },
 488     { AV_CH_BACK_CENTER,       AOUT_CHAN_REARCENTER },
 489     { AV_CH_SIDE_LEFT,         AOUT_CHAN_MIDDLELEFT },
 490     { AV_CH_SIDE_RIGHT,        AOUT_CHAN_MIDDLERIGHT },
 491     { AV_CH_TOP_CENTER,        0 },
 492     { AV_CH_TOP_FRONT_LEFT,    0 },
 493     { AV_CH_TOP_FRONT_CENTER,  0 },
 494     { AV_CH_TOP_FRONT_RIGHT,   0 },
 495     { AV_CH_TOP_BACK_LEFT,     0 },
 496     { AV_CH_TOP_BACK_CENTER,   0 },
 497     { AV_CH_TOP_BACK_RIGHT,    0 },
 498     { AV_CH_STEREO_LEFT,       0 },
 499     { AV_CH_STEREO_RIGHT,      0 },
 500 };
 501
 502 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust )
 503 {
 504     decoder_sys_t *p_sys = p_dec->p_sys;
 505
 506     p_dec->fmt_out.i_codec = GetVlcAudioFormat( p_sys->p_context->sample_fmt );
 507     p_dec->fmt_out.audio.i_format = p_dec->fmt_out.i_codec;
 508     p_dec->fmt_out.audio.i_rate = p_sys->p_context->sample_rate;
 509
 510     /* */
 511     if( p_sys->i_previous_channels == p_sys->p_context->channels &&
 512         p_sys->i_previous_layout == p_sys->p_context->channel_layout )
 513         return;
 514     if( b_trust )
 515     {
 516         p_sys->i_previous_channels = p_sys->p_context->channels;
 517         p_sys->i_previous_layout = p_sys->p_context->channel_layout;
 518     }
 519
 520     /* Specified order
 521      * FIXME should we use fmt_in.audio.i_physical_channels or not ?
 522      */
 523     const unsigned i_order_max = 8 * sizeof(p_sys->p_context->channel_layout);
 524     uint32_t pi_order_src[i_order_max];
 525     int i_channels_src = 0;
 526
 527     if( p_sys->p_context->channel_layout )
 528     {
 529         for( unsigned i = 0; i < sizeof(pi_channels_map)/sizeof(*pi_channels_map); i++ )
 530         {
 531             if( p_sys->p_context->channel_layout & pi_channels_map[i][0] )
 532                 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
 533         }
 534     }
 535     else
 536     {
 537         /* Create default order  */
 538         if( b_trust )
 539             msg_Warn( p_dec, "Physical channel configuration not set : guessing" );
 540         for( unsigned int i = 0; i < __MIN( i_order_max, (unsigned)p_sys->p_context->channels ); i++ )
 541         {
 542             if( i < sizeof(pi_channels_map)/sizeof(*pi_channels_map) )
 543                 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
 544         }
 545     }
 546     if( i_channels_src != p_sys->p_context->channels && b_trust )
 547         msg_Err( p_dec, "Channel layout not understood" );
 548
 549     uint32_t i_layout_dst;
 550     int      i_channels_dst;
 551     p_sys->b_extract = aout_CheckChannelExtraction( p_sys->pi_extraction,
 552                                                     &i_layout_dst, &i_channels_dst,
 553                                                     NULL, pi_order_src, i_channels_src );
 554     if( i_channels_dst != i_channels_src && b_trust )
 555         msg_Warn( p_dec, "%d channels are dropped", i_channels_src - i_channels_dst );
 556
 557     p_dec->fmt_out.audio.i_physical_channels =
 558     p_dec->fmt_out.audio.i_original_channels = i_layout_dst;
 559     aout_FormatPrepare( &p_dec->fmt_out.audio );
 560 }
 561