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     int64_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 static int GetAudioBuf( struct AVCodecContext *, AVFrame * );
  73
  74 static void InitDecoderConfig( decoder_t *p_dec, AVCodecContext *p_context )
  75 {
  76     if( p_dec->fmt_in.i_extra > 0 )
  77     {
  78         const uint8_t * const p_src = p_dec->fmt_in.p_extra;
  79
  80         int i_offset = 0;
  81         int i_size = p_dec->fmt_in.i_extra;
  82
  83         if( p_dec->fmt_in.i_codec == VLC_CODEC_ALAC )
  84         {
  85             static const uint8_t p_pattern[] = { 0, 0, 0, 36, 'a', 'l', 'a', 'c' };
  86             /* Find alac atom XXX it is a bit ugly */
  87             for( i_offset = 0; i_offset < p_dec->fmt_in.i_extra - sizeof(p_pattern); i_offset++ )
  88             {
  89                 if( !memcmp( &p_src[i_offset], p_pattern, sizeof(p_pattern) ) )
  90                     break;
  91             }
  92             i_size = __MIN( p_dec->fmt_in.i_extra - i_offset, 36 );
  93             if( i_size < 36 )
  94                 i_size = 0;
  95         }
  96
  97         if( i_size > 0 )
  98         {
  99             p_context->extradata =
 100                 malloc( i_size + FF_INPUT_BUFFER_PADDING_SIZE );
 101             if( p_context->extradata )
 102             {
 103                 uint8_t *p_dst = p_context->extradata;
 104
 105                 p_context->extradata_size = i_size;
 106
 107                 memcpy( &p_dst[0],            &p_src[i_offset], i_size );
 108                 memset( &p_dst[i_size], 0, FF_INPUT_BUFFER_PADDING_SIZE );
 109             }
 110         }
 111     }
 112     else
 113     {
 114         p_context->extradata_size = 0;
 115         p_context->extradata = NULL;
 116     }
 117 }
 118
 119 /*****************************************************************************
 120  * InitAudioDec: initialize audio decoder
 121  *****************************************************************************
 122  * The avcodec codec will be opened, some memory allocated.
 123  *****************************************************************************/
 124 int InitAudioDec( decoder_t *p_dec, AVCodecContext *p_context,
 125                       AVCodec *p_codec, int i_codec_id, const char *psz_namecodec )
 126 {
 127     decoder_sys_t *p_sys;
 128
 129     /* Allocate the memory needed to store the decoder's structure */
 130     if( ( p_dec->p_sys = p_sys = malloc(sizeof(*p_sys)) ) == NULL )
 131     {
 132         return VLC_ENOMEM;
 133     }
 134
 135     p_codec->type = AVMEDIA_TYPE_AUDIO;
 136     p_context->codec_type = AVMEDIA_TYPE_AUDIO;
 137     p_context->codec_id = i_codec_id;
 138     p_context->get_buffer = GetAudioBuf;
 139     p_sys->p_context = p_context;
 140     p_sys->p_codec = p_codec;
 141     p_sys->i_codec_id = i_codec_id;
 142     p_sys->psz_namecodec = psz_namecodec;
 143     p_sys->b_delayed_open = true;
 144
 145     // Initialize decoder extradata
 146     InitDecoderConfig( p_dec, p_context);
 147
 148     /* ***** Open the codec ***** */
 149     if( ffmpeg_OpenCodec( p_dec ) < 0 )
 150     {
 151         msg_Err( p_dec, "cannot open codec (%s)", p_sys->psz_namecodec );
 152         free( p_sys->p_context->extradata );
 153         free( p_sys );
 154         return VLC_EGENERIC;
 155     }
 156
 157     p_sys->i_reject_count = 0;
 158     p_sys->b_extract = false;
 159     p_sys->i_previous_channels = 0;
 160     p_sys->i_previous_layout = 0;
 161
 162     /* */
 163     p_dec->fmt_out.i_cat = AUDIO_ES;
 164     /* Try to set as much information as possible but do not trust it */
 165     SetupOutputFormat( p_dec, false );
 166
 167     date_Set( &p_sys->end_date, 0 );
 168     if( p_dec->fmt_out.audio.i_rate )
 169         date_Init( &p_sys->end_date, p_dec->fmt_out.audio.i_rate, 1 );
 170     else if( p_dec->fmt_in.audio.i_rate )
 171         date_Init( &p_sys->end_date, p_dec->fmt_in.audio.i_rate, 1 );
 172
 173     return VLC_SUCCESS;
 174 }
 175
 176 /**
 177  * Allocates decoded audio buffer for libavcodec to use.
 178  */
 179 static int GetAudioBuf( AVCodecContext *ctx, AVFrame *buf )
 180 {
 181     block_t *block;
 182     bool planar = av_sample_fmt_is_planar( ctx->sample_fmt );
 183     unsigned channels = planar ? 1 : ctx->channels;
 184     unsigned planes = planar ? ctx->channels : 1;
 185
 186     int bytes = av_samples_get_buffer_size( &buf->linesize[0], channels,
 187                                             buf->nb_samples, ctx->sample_fmt,
 188                                             16 );
 189     assert( bytes >= 0 );
 190     block = block_Alloc( bytes * planes );
 191     if( unlikely(block == NULL) )
 192         return AVERROR(ENOMEM);
 193
 194     block->i_nb_samples = buf->nb_samples;
 195     buf->opaque = block;
 196
 197     if( planes > AV_NUM_DATA_POINTERS )
 198     {
 199         uint8_t **ext = malloc( sizeof( *ext ) * planes );
 200         if( unlikely(ext == NULL) )
 201         {
 202             block_Release( block );
 203             return AVERROR(ENOMEM);
 204         }
 205         buf->extended_data = ext;
 206     }
 207     else
 208         buf->extended_data = buf->data;
 209
 210     uint8_t *buffer = block->p_buffer;
 211     for( unsigned i = 0; i < planes; i++ )
 212     {
 213         buf->linesize[i] = buf->linesize[0];
 214         buf->extended_data[i] = buffer;
 215         buffer += bytes;
 216     }
 217
 218     return 0;
 219 }
 220
 221 /*****************************************************************************
 222  * DecodeAudio: Called to decode one frame
 223  *****************************************************************************/
 224 block_t * DecodeAudio ( decoder_t *p_dec, block_t **pp_block )
 225 {
 226     decoder_sys_t *p_sys = p_dec->p_sys;
 227     AVCodecContext *ctx = p_sys->p_context;
 228
 229     if( !pp_block || !*pp_block )
 230         return NULL;
 231
 232     block_t *p_block = *pp_block;
 233
 234     if( !ctx->extradata_size && p_dec->fmt_in.i_extra && p_sys->b_delayed_open)
 235     {
 236         InitDecoderConfig( p_dec, ctx );
 237         if( ffmpeg_OpenCodec( p_dec ) )
 238             msg_Err( p_dec, "Cannot open decoder %s", p_sys->psz_namecodec );
 239     }
 240
 241     if( p_sys->b_delayed_open )
 242         goto end;
 243
 244     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 245     {
 246         avcodec_flush_buffers( ctx );
 247         date_Set( &p_sys->end_date, 0 );
 248
 249         if( p_sys->i_codec_id == AV_CODEC_ID_MP2 || p_sys->i_codec_id == AV_CODEC_ID_MP3 )
 250             p_sys->i_reject_count = 3;
 251
 252         goto end;
 253     }
 254
 255     /* We've just started the stream, wait for the first PTS. */
 256     if( !date_Get( &p_sys->end_date ) && p_block->i_pts <= VLC_TS_INVALID )
 257         goto end;
 258
 259     if( p_block->i_buffer <= 0 )
 260         goto end;
 261
 262     if( (p_block->i_flags & BLOCK_FLAG_PRIVATE_REALLOCATED) == 0 )
 263     {
 264         p_block = block_Realloc( p_block, 0, p_block->i_buffer + FF_INPUT_BUFFER_PADDING_SIZE );
 265         if( !p_block )
 266             return NULL;
 267         p_block->i_buffer -= FF_INPUT_BUFFER_PADDING_SIZE;
 268         memset( &p_block->p_buffer[p_block->i_buffer], 0, FF_INPUT_BUFFER_PADDING_SIZE );
 269
 270         p_block->i_flags |= BLOCK_FLAG_PRIVATE_REALLOCATED;
 271     }
 272
 273     AVFrame frame;
 274     memset( &frame, 0, sizeof( frame ) );
 275
 276     for( int got_frame = 0; !got_frame; )
 277     {
 278         if( p_block->i_buffer == 0 )
 279             goto end;
 280
 281         AVPacket pkt;
 282         av_init_packet( &pkt );
 283         pkt.data = p_block->p_buffer;
 284         pkt.size = p_block->i_buffer;
 285
 286         int used = avcodec_decode_audio4( ctx, &frame, &got_frame, &pkt );
 287         if( used < 0 )
 288         {
 289             msg_Warn( p_dec, "cannot decode one frame (%zu bytes)",
 290                       p_block->i_buffer );
 291             goto end;
 292         }
 293
 294         assert( p_block->i_buffer >= (unsigned)used );
 295         p_block->p_buffer += used;
 296         p_block->i_buffer -= used;
 297     }
 298
 299     if( ctx->channels <= 0 || ctx->channels > 8 || ctx->sample_rate <= 0 )
 300     {
 301         msg_Warn( p_dec, "invalid audio properties channels count %d, sample rate %d",
 302                   ctx->channels, ctx->sample_rate );
 303         goto end;
 304     }
 305
 306     if( p_dec->fmt_out.audio.i_rate != (unsigned int)ctx->sample_rate )
 307         date_Init( &p_sys->end_date, ctx->sample_rate, 1 );
 308
 309     if( p_block->i_pts > VLC_TS_INVALID &&
 310         p_block->i_pts > date_Get( &p_sys->end_date ) )
 311     {
 312         date_Set( &p_sys->end_date, p_block->i_pts );
 313     }
 314
 315     if( p_block->i_buffer == 0 )
 316     {   /* Done with this buffer */
 317         block_Release( p_block );
 318         *pp_block = NULL;
 319     }
 320
 321     /* NOTE WELL: Beyond this point, p_block now refers to the DECODED block */
 322     p_block = frame.opaque;
 323     SetupOutputFormat( p_dec, true );
 324
 325     /* Silent unwanted samples */
 326     if( p_sys->i_reject_count > 0 )
 327     {
 328         memset( p_block->p_buffer, 0, p_block->i_buffer );
 329         p_sys->i_reject_count--;
 330     }
 331
 332     block_t *p_buffer = decoder_NewAudioBuffer( p_dec, p_block->i_nb_samples );
 333     if (!p_buffer)
 334         return NULL;
 335     assert( p_block->i_nb_samples >= (unsigned)frame.nb_samples );
 336     assert( p_block->i_nb_samples == p_buffer->i_nb_samples );
 337     p_block->i_buffer = p_buffer->i_buffer; /* drop buffer padding */
 338
 339     /* Interleave audio if required */
 340     if( av_sample_fmt_is_planar( ctx->sample_fmt ) )
 341     {
 342         aout_Interleave( p_buffer->p_buffer, p_block->p_buffer,
 343                          p_block->i_nb_samples, ctx->channels,
 344                          p_dec->fmt_out.audio.i_format );
 345         if( ctx->channels > AV_NUM_DATA_POINTERS )
 346             free( frame.extended_data );
 347         block_Release( p_block );
 348         p_block = p_buffer;
 349     }
 350     else /* FIXME: improve decoder_NewAudioBuffer(), avoid useless buffer... */
 351         block_Release( p_buffer );
 352
 353     if (p_sys->b_extract)
 354     {   /* TODO: do not drop channels... at least not here */
 355         p_buffer = block_Alloc( p_dec->fmt_out.audio.i_bytes_per_frame
 356                                 * frame.nb_samples );
 357         if( unlikely(p_buffer == NULL) )
 358         {
 359             block_Release( p_block );
 360             return NULL;
 361         }
 362         aout_ChannelExtract( p_buffer->p_buffer,
 363                              p_dec->fmt_out.audio.i_channels,
 364                              p_block->p_buffer, ctx->channels,
 365                              frame.nb_samples, p_sys->pi_extraction,
 366                              p_dec->fmt_out.audio.i_bitspersample );
 367         block_Release( p_block );
 368         p_block = p_buffer;
 369     }
 370
 371     p_block->i_nb_samples = frame.nb_samples;
 372     p_block->i_buffer = frame.nb_samples
 373                         * p_dec->fmt_out.audio.i_bytes_per_frame;
 374     p_block->i_pts = date_Get( &p_sys->end_date );
 375     p_block->i_length = date_Increment( &p_sys->end_date, frame.nb_samples )
 376                         - p_block->i_pts;
 377     return p_block;
 378
 379 end:
 380     block_Release(p_block);
 381     *pp_block = NULL;
 382     return NULL;
 383 }
 384
 385 /*****************************************************************************
 386  *
 387  *****************************************************************************/
 388
 389 vlc_fourcc_t GetVlcAudioFormat( int fmt )
 390 {
 391     static const vlc_fourcc_t fcc[] = {
 392         [AV_SAMPLE_FMT_U8]    = VLC_CODEC_U8,
 393         [AV_SAMPLE_FMT_S16]   = VLC_CODEC_S16N,
 394         [AV_SAMPLE_FMT_S32]   = VLC_CODEC_S32N,
 395         [AV_SAMPLE_FMT_FLT]   = VLC_CODEC_FL32,
 396         [AV_SAMPLE_FMT_DBL]   = VLC_CODEC_FL64,
 397         [AV_SAMPLE_FMT_U8P]   = VLC_CODEC_U8,
 398         [AV_SAMPLE_FMT_S16P]  = VLC_CODEC_S16N,
 399         [AV_SAMPLE_FMT_S32P]  = VLC_CODEC_S32N,
 400         [AV_SAMPLE_FMT_FLTP]  = VLC_CODEC_FL32,
 401         [AV_SAMPLE_FMT_DBLP]  = VLC_CODEC_FL64,
 402     };
 403     if( (sizeof(fcc) / sizeof(fcc[0])) > (unsigned)fmt )
 404         return fcc[fmt];
 405     return VLC_CODEC_S16N;
 406 }
 407
 408 static const uint64_t pi_channels_map[][2] =
 409 {
 410     { AV_CH_FRONT_LEFT,        AOUT_CHAN_LEFT },
 411     { AV_CH_FRONT_RIGHT,       AOUT_CHAN_RIGHT },
 412     { AV_CH_FRONT_CENTER,      AOUT_CHAN_CENTER },
 413     { AV_CH_LOW_FREQUENCY,     AOUT_CHAN_LFE },
 414     { AV_CH_BACK_LEFT,         AOUT_CHAN_REARLEFT },
 415     { AV_CH_BACK_RIGHT,        AOUT_CHAN_REARRIGHT },
 416     { AV_CH_FRONT_LEFT_OF_CENTER, 0 },
 417     { AV_CH_FRONT_RIGHT_OF_CENTER, 0 },
 418     { AV_CH_BACK_CENTER,       AOUT_CHAN_REARCENTER },
 419     { AV_CH_SIDE_LEFT,         AOUT_CHAN_MIDDLELEFT },
 420     { AV_CH_SIDE_RIGHT,        AOUT_CHAN_MIDDLERIGHT },
 421     { AV_CH_TOP_CENTER,        0 },
 422     { AV_CH_TOP_FRONT_LEFT,    0 },
 423     { AV_CH_TOP_FRONT_CENTER,  0 },
 424     { AV_CH_TOP_FRONT_RIGHT,   0 },
 425     { AV_CH_TOP_BACK_LEFT,     0 },
 426     { AV_CH_TOP_BACK_CENTER,   0 },
 427     { AV_CH_TOP_BACK_RIGHT,    0 },
 428     { AV_CH_STEREO_LEFT,       0 },
 429     { AV_CH_STEREO_RIGHT,      0 },
 430 };
 431
 432 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust )
 433 {
 434     decoder_sys_t *p_sys = p_dec->p_sys;
 435
 436     p_dec->fmt_out.i_codec = GetVlcAudioFormat( p_sys->p_context->sample_fmt );
 437     p_dec->fmt_out.audio.i_format = p_dec->fmt_out.i_codec;
 438     p_dec->fmt_out.audio.i_rate = p_sys->p_context->sample_rate;
 439
 440     /* */
 441     if( p_sys->i_previous_channels == p_sys->p_context->channels &&
 442         p_sys->i_previous_layout == p_sys->p_context->channel_layout )
 443         return;
 444     if( b_trust )
 445     {
 446         p_sys->i_previous_channels = p_sys->p_context->channels;
 447         p_sys->i_previous_layout = p_sys->p_context->channel_layout;
 448     }
 449
 450     /* Specified order
 451      * FIXME should we use fmt_in.audio.i_physical_channels or not ?
 452      */
 453     const unsigned i_order_max = 8 * sizeof(p_sys->p_context->channel_layout);
 454     uint32_t pi_order_src[i_order_max];
 455     int i_channels_src = 0;
 456
 457     if( p_sys->p_context->channel_layout )
 458     {
 459         for( unsigned i = 0; i < sizeof(pi_channels_map)/sizeof(*pi_channels_map); i++ )
 460         {
 461             if( p_sys->p_context->channel_layout & pi_channels_map[i][0] )
 462                 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
 463         }
 464     }
 465     else
 466     {
 467         /* Create default order  */
 468         if( b_trust )
 469             msg_Warn( p_dec, "Physical channel configuration not set : guessing" );
 470         for( unsigned int i = 0; i < __MIN( i_order_max, (unsigned)p_sys->p_context->channels ); i++ )
 471         {
 472             if( i < sizeof(pi_channels_map)/sizeof(*pi_channels_map) )
 473                 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
 474         }
 475     }
 476     if( i_channels_src != p_sys->p_context->channels && b_trust )
 477         msg_Err( p_dec, "Channel layout not understood" );
 478
 479     uint32_t i_layout_dst;
 480     int      i_channels_dst;
 481     p_sys->b_extract = aout_CheckChannelExtraction( p_sys->pi_extraction,
 482                                                     &i_layout_dst, &i_channels_dst,
 483                                                     NULL, pi_order_src, i_channels_src );
 484     if( i_channels_dst != i_channels_src && b_trust )
 485         msg_Warn( p_dec, "%d channels are dropped", i_channels_src - i_channels_dst );
 486
 487     p_dec->fmt_out.audio.i_physical_channels =
 488     p_dec->fmt_out.audio.i_original_channels = i_layout_dst;
 489     aout_FormatPrepare( &p_dec->fmt_out.audio );
 490 }
 491