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

   1 /*****************************************************************************
   2  * faad.c: AAC decoder using libfaad2
   3  *****************************************************************************
   4  * Copyright (C) 2001, 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  * NOTA BENE: this module requires the linking against a library which is
  27  * known to require licensing under the GNU General Public License version 2
  28  * (or later). Therefore, the result of compiling this module will normally
  29  * be subject to the terms of that later license.
  30  *****************************************************************************/
  31
  32
  33 #ifdef HAVE_CONFIG_H
  34 # include "config.h"
  35 #endif
  36
  37 #include <vlc_common.h>
  38 #include <vlc_plugin.h>
  39 #include <vlc_input.h>
  40 #include <vlc_codec.h>
  41 #include <vlc_cpu.h>
  42
  43 #include <faad.h>
  44
  45 /*****************************************************************************
  46  * Module descriptor
  47  *****************************************************************************/
  48 static int  Open( vlc_object_t * );
  49 static void Close( vlc_object_t * );
  50
  51 vlc_module_begin ()
  52     set_description( N_("AAC audio decoder (using libfaad2)") )
  53     set_capability( "decoder", 100 )
  54     set_category( CAT_INPUT )
  55     set_subcategory( SUBCAT_INPUT_ACODEC )
  56     set_callbacks( Open, Close )
  57 vlc_module_end ()
  58
  59 /****************************************************************************
  60  * Local prototypes
  61  ****************************************************************************/
  62 static block_t *DecodeBlock( decoder_t *, block_t ** );
  63 static void DoReordering( uint32_t *, uint32_t *, int, int, uint32_t * );
  64
  65 #define MAX_CHANNEL_POSITIONS 9
  66
  67 struct decoder_sys_t
  68 {
  69     /* faad handler */
  70     faacDecHandle *hfaad;
  71
  72     /* samples */
  73     date_t date;
  74
  75     /* temporary buffer */
  76     uint8_t *p_buffer;
  77     int     i_buffer;
  78     size_t  i_buffer_size;
  79
  80     /* Channel positions of the current stream (for re-ordering) */
  81     uint32_t pi_channel_positions[MAX_CHANNEL_POSITIONS];
  82
  83     bool b_sbr, b_ps;
  84 };
  85
  86 static const uint32_t pi_channels_in[MAX_CHANNEL_POSITIONS] =
  87     { FRONT_CHANNEL_CENTER, FRONT_CHANNEL_LEFT, FRONT_CHANNEL_RIGHT,
  88       SIDE_CHANNEL_LEFT, SIDE_CHANNEL_RIGHT,
  89       BACK_CHANNEL_LEFT, BACK_CHANNEL_RIGHT,
  90       BACK_CHANNEL_CENTER, LFE_CHANNEL };
  91 static const uint32_t pi_channels_out[MAX_CHANNEL_POSITIONS] =
  92     { AOUT_CHAN_CENTER, AOUT_CHAN_LEFT, AOUT_CHAN_RIGHT,
  93       AOUT_CHAN_MIDDLELEFT, AOUT_CHAN_MIDDLERIGHT,
  94       AOUT_CHAN_REARLEFT, AOUT_CHAN_REARRIGHT,
  95       AOUT_CHAN_REARCENTER, AOUT_CHAN_LFE };
  96 static const uint32_t pi_channels_ordered[MAX_CHANNEL_POSITIONS] =
  97     { AOUT_CHAN_LEFT, AOUT_CHAN_RIGHT,
  98       AOUT_CHAN_MIDDLELEFT, AOUT_CHAN_MIDDLERIGHT,
  99       AOUT_CHAN_REARLEFT, AOUT_CHAN_REARRIGHT,
 100       AOUT_CHAN_CENTER, AOUT_CHAN_REARCENTER, AOUT_CHAN_LFE
 101     };
 102 static const uint32_t pi_channels_guessed[MAX_CHANNEL_POSITIONS] =
 103     { 0, AOUT_CHAN_CENTER, AOUT_CHAN_LEFT | AOUT_CHAN_RIGHT,
 104       AOUT_CHAN_LEFT | AOUT_CHAN_RIGHT | AOUT_CHAN_LFE,
 105       AOUT_CHAN_LEFT | AOUT_CHAN_RIGHT | AOUT_CHAN_REARLEFT
 106           | AOUT_CHAN_REARRIGHT,
 107       AOUT_CHAN_LEFT | AOUT_CHAN_RIGHT | AOUT_CHAN_REARLEFT
 108           | AOUT_CHAN_REARRIGHT | AOUT_CHAN_CENTER,
 109       AOUT_CHAN_LEFT | AOUT_CHAN_RIGHT | AOUT_CHAN_REARLEFT
 110           | AOUT_CHAN_REARRIGHT | AOUT_CHAN_CENTER | AOUT_CHAN_LFE,
 111       AOUT_CHAN_LEFT | AOUT_CHAN_RIGHT | AOUT_CHAN_MIDDLELEFT
 112           | AOUT_CHAN_MIDDLERIGHT | AOUT_CHAN_REARLEFT | AOUT_CHAN_REARRIGHT
 113           | AOUT_CHAN_CENTER,
 114       AOUT_CHAN_LEFT | AOUT_CHAN_RIGHT | AOUT_CHAN_MIDDLELEFT
 115           | AOUT_CHAN_MIDDLERIGHT | AOUT_CHAN_REARLEFT | AOUT_CHAN_REARRIGHT
 116           | AOUT_CHAN_CENTER | AOUT_CHAN_LFE
 117     };
 118
 119 /*****************************************************************************
 120  * OpenDecoder: probe the decoder and return score
 121  *****************************************************************************/
 122 static int Open( vlc_object_t *p_this )
 123 {
 124     decoder_t *p_dec = (decoder_t*)p_this;
 125     decoder_sys_t *p_sys;
 126     faacDecConfiguration *cfg;
 127
 128     if( p_dec->fmt_in.i_codec != VLC_CODEC_MP4A )
 129     {
 130         return VLC_EGENERIC;
 131     }
 132
 133     /* Allocate the memory needed to store the decoder's structure */
 134     if( ( p_dec->p_sys = p_sys = malloc( sizeof(*p_sys) ) ) == NULL )
 135         return VLC_ENOMEM;
 136
 137     /* Open a faad context */
 138     if( ( p_sys->hfaad = faacDecOpen() ) == NULL )
 139     {
 140         msg_Err( p_dec, "cannot initialize faad" );
 141         free( p_sys );
 142         return VLC_EGENERIC;
 143     }
 144
 145     /* Misc init */
 146     date_Set( &p_sys->date, 0 );
 147     p_dec->fmt_out.i_cat = AUDIO_ES;
 148
 149     p_dec->fmt_out.i_codec = HAVE_FPU ? VLC_CODEC_FL32 : VLC_CODEC_S16N;
 150
 151     p_dec->fmt_out.audio.i_physical_channels =
 152         p_dec->fmt_out.audio.i_original_channels = 0;
 153
 154     if( p_dec->fmt_in.i_extra > 0 )
 155     {
 156         /* We have a decoder config so init the handle */
 157         unsigned long i_rate;
 158         unsigned char i_channels;
 159
 160         if( faacDecInit2( p_sys->hfaad, p_dec->fmt_in.p_extra,
 161                           p_dec->fmt_in.i_extra,
 162                           &i_rate, &i_channels ) < 0 )
 163         {
 164             msg_Err( p_dec, "Failed to initialize faad using extra data" );
 165             faacDecClose( p_sys->hfaad );
 166             free( p_sys );
 167             return VLC_EGENERIC;
 168         }
 169
 170         p_dec->fmt_out.audio.i_rate = i_rate;
 171         p_dec->fmt_out.audio.i_channels = i_channels;
 172         p_dec->fmt_out.audio.i_physical_channels
 173             = p_dec->fmt_out.audio.i_original_channels
 174             = pi_channels_guessed[i_channels];
 175         date_Init( &p_sys->date, i_rate, 1 );
 176     }
 177     else
 178     {
 179         /* Will be initalised from first frame */
 180         p_dec->fmt_out.audio.i_rate = 0;
 181         p_dec->fmt_out.audio.i_channels = 0;
 182     }
 183
 184     /* Set the faad config */
 185     cfg = faacDecGetCurrentConfiguration( p_sys->hfaad );
 186     cfg->outputFormat = HAVE_FPU ? FAAD_FMT_FLOAT : FAAD_FMT_16BIT;
 187     faacDecSetConfiguration( p_sys->hfaad, cfg );
 188
 189     /* buffer */
 190     p_sys->i_buffer = p_sys->i_buffer_size = 0;
 191     p_sys->p_buffer = NULL;
 192
 193     /* Faad2 can't deal with truncated data (eg. from MPEG TS) */
 194     p_dec->b_need_packetized = true;
 195
 196     p_sys->b_sbr = p_sys->b_ps = false;
 197
 198     p_dec->pf_decode_audio = DecodeBlock;
 199     return VLC_SUCCESS;
 200 }
 201
 202 /*****************************************************************************
 203  * DecodeBlock:
 204  *****************************************************************************/
 205 static block_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 206 {
 207     decoder_sys_t *p_sys = p_dec->p_sys;
 208     block_t *p_block;
 209
 210     if( !pp_block || !*pp_block ) return NULL;
 211
 212     p_block = *pp_block;
 213
 214     if( p_block->i_flags&(BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 215     {
 216         block_Release( p_block );
 217         return NULL;
 218     }
 219
 220     /* Remove ADTS header if we have decoder specific config */
 221     if( p_dec->fmt_in.i_extra && p_block->i_buffer > 7 )
 222     {
 223         if( p_block->p_buffer[0] == 0xff &&
 224             ( p_block->p_buffer[1] & 0xf0 ) == 0xf0 ) /* syncword */
 225         {   /* ADTS header present */
 226             size_t i_header_size; /* 7 bytes (+ 2 bytes for crc) */
 227             i_header_size = 7 + ( ( p_block->p_buffer[1] & 0x01 ) ? 0 : 2 );
 228             /* FIXME: multiple blocks per frame */
 229             if( p_block->i_buffer > i_header_size )
 230             {
 231                 p_block->p_buffer += i_header_size;
 232                 p_block->i_buffer -= i_header_size;
 233             }
 234         }
 235     }
 236
 237     /* Append the block to the temporary buffer */
 238     if( p_sys->i_buffer_size < p_sys->i_buffer + p_block->i_buffer )
 239     {
 240         size_t  i_buffer_size = p_sys->i_buffer + p_block->i_buffer;
 241         uint8_t *p_buffer     = realloc( p_sys->p_buffer, i_buffer_size );
 242         if( p_buffer )
 243         {
 244             p_sys->i_buffer_size = i_buffer_size;
 245             p_sys->p_buffer      = p_buffer;
 246         }
 247         else
 248         {
 249             p_block->i_buffer = 0;
 250         }
 251     }
 252
 253     if( p_block->i_buffer > 0 )
 254     {
 255         memcpy( &p_sys->p_buffer[p_sys->i_buffer],
 256                      p_block->p_buffer, p_block->i_buffer );
 257         p_sys->i_buffer += p_block->i_buffer;
 258         p_block->i_buffer = 0;
 259     }
 260
 261     if( p_dec->fmt_out.audio.i_rate == 0 && p_dec->fmt_in.i_extra > 0 )
 262     {
 263         /* We have a decoder config so init the handle */
 264         unsigned long i_rate;
 265         unsigned char i_channels;
 266
 267         if( faacDecInit2( p_sys->hfaad, p_dec->fmt_in.p_extra,
 268                           p_dec->fmt_in.i_extra,
 269                           &i_rate, &i_channels ) >= 0 )
 270         {
 271             p_dec->fmt_out.audio.i_rate = i_rate;
 272             p_dec->fmt_out.audio.i_channels = i_channels;
 273             p_dec->fmt_out.audio.i_physical_channels
 274                 = p_dec->fmt_out.audio.i_original_channels
 275                 = pi_channels_guessed[i_channels];
 276
 277             date_Init( &p_sys->date, i_rate, 1 );
 278         }
 279     }
 280
 281     if( p_dec->fmt_out.audio.i_rate == 0 && p_sys->i_buffer )
 282     {
 283         unsigned long i_rate;
 284         unsigned char i_channels;
 285
 286         /* Init faad with the first frame */
 287         if( faacDecInit( p_sys->hfaad,
 288                          p_sys->p_buffer, p_sys->i_buffer,
 289                          &i_rate, &i_channels ) < 0 )
 290         {
 291             block_Release( p_block );
 292             return NULL;
 293         }
 294
 295         p_dec->fmt_out.audio.i_rate = i_rate;
 296         p_dec->fmt_out.audio.i_channels = i_channels;
 297         p_dec->fmt_out.audio.i_physical_channels
 298             = p_dec->fmt_out.audio.i_original_channels
 299             = pi_channels_guessed[i_channels];
 300         date_Init( &p_sys->date, i_rate, 1 );
 301     }
 302
 303     if( p_block->i_pts > VLC_TS_INVALID && p_block->i_pts != date_Get( &p_sys->date ) )
 304     {
 305         date_Set( &p_sys->date, p_block->i_pts );
 306     }
 307     else if( !date_Get( &p_sys->date ) )
 308     {
 309         /* We've just started the stream, wait for the first PTS. */
 310         block_Release( p_block );
 311         p_sys->i_buffer = 0;
 312         return NULL;
 313     }
 314
 315     /* Decode all data */
 316     if( p_sys->i_buffer > 1)
 317     {
 318         void *samples;
 319         faacDecFrameInfo frame;
 320         block_t *p_out;
 321
 322         samples = faacDecDecode( p_sys->hfaad, &frame,
 323                                  p_sys->p_buffer, p_sys->i_buffer );
 324
 325         if( frame.error > 0 )
 326         {
 327             msg_Warn( p_dec, "%s", faacDecGetErrorMessage( frame.error ) );
 328
 329             if( frame.error == 21 || frame.error == 12 )
 330             {
 331                 /*
 332                  * Once an "Unexpected channel configuration change"
 333                  * or a "Invalid number of channels" error
 334                  * occurs, it will occurs afterwards, and we got no sound.
 335                  * Reinitialization of the decoder is required.
 336                  */
 337                 unsigned long i_rate;
 338                 unsigned char i_channels;
 339                 faacDecHandle *hfaad;
 340                 faacDecConfiguration *cfg,*oldcfg;
 341
 342                 oldcfg = faacDecGetCurrentConfiguration( p_sys->hfaad );
 343                 hfaad = faacDecOpen();
 344                 cfg = faacDecGetCurrentConfiguration( hfaad );
 345                 if( oldcfg->defSampleRate )
 346                     cfg->defSampleRate = oldcfg->defSampleRate;
 347                 cfg->defObjectType = oldcfg->defObjectType;
 348                 cfg->outputFormat = oldcfg->outputFormat;
 349                 faacDecSetConfiguration( hfaad, cfg );
 350
 351                 if( faacDecInit( hfaad, p_sys->p_buffer, p_sys->i_buffer,
 352                                 &i_rate,&i_channels ) < 0 )
 353                 {
 354                     /* reinitialization failed */
 355                     faacDecClose( hfaad );
 356                     faacDecSetConfiguration( p_sys->hfaad, oldcfg );
 357                 }
 358                 else
 359                 {
 360                     faacDecClose( p_sys->hfaad );
 361                     p_sys->hfaad = hfaad;
 362                     p_dec->fmt_out.audio.i_rate = i_rate;
 363                     p_dec->fmt_out.audio.i_channels = i_channels;
 364                     p_dec->fmt_out.audio.i_physical_channels
 365                         = p_dec->fmt_out.audio.i_original_channels
 366                         = pi_channels_guessed[i_channels];
 367                     date_Init( &p_sys->date, i_rate, 1 );
 368                 }
 369             }
 370
 371             /* Flush the buffer */
 372             p_sys->i_buffer = 0;
 373             block_Release( p_block );
 374             return NULL;
 375         }
 376
 377         if( frame.channels <= 0 || frame.channels > 8 || frame.channels == 7 )
 378         {
 379             msg_Warn( p_dec, "invalid channels count: %i", frame.channels );
 380
 381             /* Flush the buffer */
 382             p_sys->i_buffer -= frame.bytesconsumed;
 383             if( p_sys->i_buffer > 0 )
 384             {
 385                 memmove( p_sys->p_buffer,&p_sys->p_buffer[frame.bytesconsumed],
 386                          p_sys->i_buffer );
 387             }
 388             block_Release( p_block );
 389             return NULL;
 390         }
 391
 392         if( frame.samples <= 0 )
 393         {
 394             msg_Warn( p_dec, "decoded zero sample" );
 395
 396             /* Flush the buffer */
 397             p_sys->i_buffer -= frame.bytesconsumed;
 398             if( p_sys->i_buffer > 1 )
 399             {
 400                 memmove( p_sys->p_buffer,&p_sys->p_buffer[frame.bytesconsumed],
 401                          p_sys->i_buffer );
 402             }
 403             else
 404             {
 405                 /* Drop byte of padding */
 406                 p_sys->i_buffer = 0;
 407             }
 408             block_Release( p_block );
 409             return NULL;
 410         }
 411
 412         /* We decoded a valid frame */
 413         if( p_dec->fmt_out.audio.i_rate != frame.samplerate )
 414         {
 415             date_Init( &p_sys->date, frame.samplerate, 1 );
 416             date_Set( &p_sys->date, p_block->i_pts );
 417         }
 418         p_block->i_pts = VLC_TS_INVALID;  /* PTS is valid only once */
 419
 420         p_dec->fmt_out.audio.i_rate = frame.samplerate;
 421         p_dec->fmt_out.audio.i_channels = frame.channels;
 422
 423         /* Adjust stream info when dealing with SBR/PS */
 424         bool b_sbr = (frame.sbr == 1) || (frame.sbr == 2);
 425         if( p_sys->b_sbr != b_sbr || p_sys->b_ps != frame.ps )
 426         {
 427             const char *psz_ext = (b_sbr && frame.ps) ? "SBR+PS" :
 428                                     b_sbr ? "SBR" : "PS";
 429
 430             msg_Dbg( p_dec, "AAC %s (channels: %u, samplerate: %lu)",
 431                     psz_ext, frame.channels, frame.samplerate );
 432
 433             if( !p_dec->p_description )
 434                 p_dec->p_description = vlc_meta_New();
 435             if( p_dec->p_description )
 436                 vlc_meta_AddExtra( p_dec->p_description, _("AAC extension"), psz_ext );
 437
 438             p_sys->b_sbr = b_sbr;
 439             p_sys->b_ps = frame.ps;
 440         }
 441
 442         /* Convert frame.channel_position to our own channel values */
 443         p_dec->fmt_out.audio.i_physical_channels = 0;
 444         const uint32_t nbChannels = frame.channels;
 445         unsigned j;
 446         for( unsigned i = 0; i < nbChannels; i++ )
 447         {
 448             /* Find the channel code */
 449             for( j = 0; j < MAX_CHANNEL_POSITIONS; j++ )
 450             {
 451                 if( frame.channel_position[i] == pi_channels_in[j] )
 452                     break;
 453             }
 454             if( j >= MAX_CHANNEL_POSITIONS )
 455             {
 456                 msg_Warn( p_dec, "unknown channel ordering" );
 457                 /* Invent something */
 458                 j = i;
 459             }
 460             /* */
 461             p_sys->pi_channel_positions[i] = pi_channels_out[j];
 462             if( p_dec->fmt_out.audio.i_physical_channels & pi_channels_out[j] )
 463                 frame.channels--; /* We loose a duplicated channel */
 464             else
 465                 p_dec->fmt_out.audio.i_physical_channels |= pi_channels_out[j];
 466         }
 467         if ( nbChannels != frame.channels )
 468         {
 469             p_dec->fmt_out.audio.i_physical_channels
 470                 = p_dec->fmt_out.audio.i_original_channels
 471                 = pi_channels_guessed[nbChannels];
 472         }
 473         else
 474         {
 475             p_dec->fmt_out.audio.i_original_channels =
 476                 p_dec->fmt_out.audio.i_physical_channels;
 477         }
 478         p_dec->fmt_out.audio.i_channels = nbChannels;
 479         p_out = decoder_NewAudioBuffer( p_dec, frame.samples / nbChannels );
 480         if( p_out == NULL )
 481         {
 482             p_sys->i_buffer = 0;
 483             block_Release( p_block );
 484             return NULL;
 485         }
 486
 487         p_out->i_pts = date_Get( &p_sys->date );
 488         p_out->i_length = date_Increment( &p_sys->date,
 489                                           frame.samples / nbChannels )
 490                           - p_out->i_pts;
 491
 492         DoReordering( (uint32_t *)p_out->p_buffer, samples,
 493                       frame.samples / nbChannels, nbChannels,
 494                       p_sys->pi_channel_positions );
 495
 496         p_sys->i_buffer -= frame.bytesconsumed;
 497         if( p_sys->i_buffer > 0 )
 498         {
 499             memmove( p_sys->p_buffer, &p_sys->p_buffer[frame.bytesconsumed],
 500                      p_sys->i_buffer );
 501         }
 502
 503         return p_out;
 504     }
 505     else
 506     {
 507         /* Drop byte of padding */
 508         p_sys->i_buffer = 0;
 509     }
 510
 511     block_Release( p_block );
 512     return NULL;
 513 }
 514
 515 /*****************************************************************************
 516  * Close:
 517  *****************************************************************************/
 518 static void Close( vlc_object_t *p_this )
 519 {
 520     decoder_t *p_dec = (decoder_t *)p_this;
 521     decoder_sys_t *p_sys = p_dec->p_sys;
 522
 523     faacDecClose( p_sys->hfaad );
 524     free( p_sys->p_buffer );
 525     free( p_sys );
 526 }
 527
 528 /*****************************************************************************
 529  * DoReordering: do some channel re-ordering (the ac3 channel order is
 530  *   different from the aac one).
 531  *****************************************************************************/
 532 static void DoReordering( uint32_t *p_out, uint32_t *p_in, int i_samples,
 533                           int i_nb_channels, uint32_t *pi_chan_positions )
 534 {
 535     int pi_chan_table[MAX_CHANNEL_POSITIONS] = {0};
 536     int i, j, k;
 537
 538     /* Find the channels mapping */
 539     for( i = 0, j = 0; i < MAX_CHANNEL_POSITIONS; i++ )
 540     {
 541         for( k = 0; k < i_nb_channels; k++ )
 542         {
 543             if( pi_channels_ordered[i] == pi_chan_positions[k] )
 544             {
 545                 pi_chan_table[k] = j++;
 546                 break;
 547             }
 548         }
 549     }
 550
 551     /* Do the actual reordering */
 552     if( HAVE_FPU )
 553         for( i = 0; i < i_samples; i++ )
 554             for( j = 0; j < i_nb_channels; j++ )
 555                 p_out[i * i_nb_channels + pi_chan_table[j]] =
 556                     p_in[i * i_nb_channels + j];
 557     else
 558         for( i = 0; i < i_samples; i++ )
 559             for( j = 0; j < i_nb_channels; j++ )
 560                 ((uint16_t *)p_out)[i * i_nb_channels + pi_chan_table[j]] =
 561                     ((uint16_t *)p_in)[i * i_nb_channels + j];
 562 }
 563