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 <neaacdec.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     NeAACDecHandle *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     NeAACDecConfiguration *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 = NeAACDecOpen() ) == 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( NeAACDecInit2( 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             NeAACDecClose( 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 = NeAACDecGetCurrentConfiguration( p_sys->hfaad );
 186     if( p_dec->fmt_in.audio.i_rate )
 187         cfg->defSampleRate = p_dec->fmt_in.audio.i_rate;
 188     cfg->outputFormat = HAVE_FPU ? FAAD_FMT_FLOAT : FAAD_FMT_16BIT;
 189     NeAACDecSetConfiguration( p_sys->hfaad, cfg );
 190
 191     /* buffer */
 192     p_sys->i_buffer = p_sys->i_buffer_size = 0;
 193     p_sys->p_buffer = NULL;
 194
 195     p_sys->b_sbr = p_sys->b_ps = false;
 196
 197     p_dec->pf_decode_audio = DecodeBlock;
 198     return VLC_SUCCESS;
 199 }
 200
 201 /*****************************************************************************
 202  * DecodeBlock:
 203  *****************************************************************************/
 204 static block_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 205 {
 206     decoder_sys_t *p_sys = p_dec->p_sys;
 207     block_t *p_block;
 208
 209     if( !pp_block || !*pp_block ) return NULL;
 210
 211     p_block = *pp_block;
 212
 213     if( p_block->i_flags&(BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 214     {
 215         block_Release( p_block );
 216         return NULL;
 217     }
 218
 219     /* Remove ADTS header if we have decoder specific config */
 220     if( p_dec->fmt_in.i_extra && p_block->i_buffer > 7 )
 221     {
 222         if( p_block->p_buffer[0] == 0xff &&
 223             ( p_block->p_buffer[1] & 0xf0 ) == 0xf0 ) /* syncword */
 224         {   /* ADTS header present */
 225             size_t i_header_size; /* 7 bytes (+ 2 bytes for crc) */
 226             i_header_size = 7 + ( ( p_block->p_buffer[1] & 0x01 ) ? 0 : 2 );
 227             /* FIXME: multiple blocks per frame */
 228             if( p_block->i_buffer > i_header_size )
 229             {
 230                 p_block->p_buffer += i_header_size;
 231                 p_block->i_buffer -= i_header_size;
 232             }
 233         }
 234     }
 235
 236     /* Append the block to the temporary buffer */
 237     if( p_sys->i_buffer_size < p_sys->i_buffer + p_block->i_buffer )
 238     {
 239         size_t  i_buffer_size = p_sys->i_buffer + p_block->i_buffer;
 240         uint8_t *p_buffer     = realloc( p_sys->p_buffer, i_buffer_size );
 241         if( p_buffer )
 242         {
 243             p_sys->i_buffer_size = i_buffer_size;
 244             p_sys->p_buffer      = p_buffer;
 245         }
 246         else
 247         {
 248             p_block->i_buffer = 0;
 249         }
 250     }
 251
 252     if( p_block->i_buffer > 0 )
 253     {
 254         memcpy( &p_sys->p_buffer[p_sys->i_buffer],
 255                      p_block->p_buffer, p_block->i_buffer );
 256         p_sys->i_buffer += p_block->i_buffer;
 257         p_block->i_buffer = 0;
 258     }
 259
 260     if( p_dec->fmt_out.audio.i_rate == 0 && p_dec->fmt_in.i_extra > 0 )
 261     {
 262         /* We have a decoder config so init the handle */
 263         unsigned long i_rate;
 264         unsigned char i_channels;
 265
 266         if( NeAACDecInit2( p_sys->hfaad, p_dec->fmt_in.p_extra,
 267                            p_dec->fmt_in.i_extra,
 268                            &i_rate, &i_channels ) >= 0 )
 269         {
 270             p_dec->fmt_out.audio.i_rate = i_rate;
 271             p_dec->fmt_out.audio.i_channels = i_channels;
 272             p_dec->fmt_out.audio.i_physical_channels
 273                 = p_dec->fmt_out.audio.i_original_channels
 274                 = pi_channels_guessed[i_channels];
 275
 276             date_Init( &p_sys->date, i_rate, 1 );
 277         }
 278     }
 279
 280     if( p_dec->fmt_out.audio.i_rate == 0 && p_sys->i_buffer )
 281     {
 282         unsigned long i_rate;
 283         unsigned char i_channels;
 284
 285         /* Init faad with the first frame */
 286         if( NeAACDecInit( p_sys->hfaad,
 287                           p_sys->p_buffer, p_sys->i_buffer,
 288                           &i_rate, &i_channels ) < 0 )
 289         {
 290             block_Release( p_block );
 291             return NULL;
 292         }
 293
 294         p_dec->fmt_out.audio.i_rate = i_rate;
 295         p_dec->fmt_out.audio.i_channels = i_channels;
 296         p_dec->fmt_out.audio.i_physical_channels
 297             = p_dec->fmt_out.audio.i_original_channels
 298             = pi_channels_guessed[i_channels];
 299         date_Init( &p_sys->date, i_rate, 1 );
 300     }
 301
 302     if( p_block->i_pts > VLC_TS_INVALID && p_block->i_pts != date_Get( &p_sys->date ) )
 303     {
 304         date_Set( &p_sys->date, p_block->i_pts );
 305     }
 306     else if( !date_Get( &p_sys->date ) )
 307     {
 308         /* We've just started the stream, wait for the first PTS. */
 309         block_Release( p_block );
 310         p_sys->i_buffer = 0;
 311         return NULL;
 312     }
 313
 314     /* Decode all data */
 315     if( p_sys->i_buffer > 1)
 316     {
 317         void *samples;
 318         NeAACDecFrameInfo frame;
 319         block_t *p_out;
 320
 321         samples = NeAACDecDecode( p_sys->hfaad, &frame,
 322                                   p_sys->p_buffer, p_sys->i_buffer );
 323
 324         if( frame.error > 0 )
 325         {
 326             msg_Warn( p_dec, "%s", NeAACDecGetErrorMessage( frame.error ) );
 327
 328             if( frame.error == 21 || frame.error == 12 )
 329             {
 330                 /*
 331                  * Once an "Unexpected channel configuration change"
 332                  * or a "Invalid number of channels" error
 333                  * occurs, it will occurs afterwards, and we got no sound.
 334                  * Reinitialization of the decoder is required.
 335                  */
 336                 unsigned long i_rate;
 337                 unsigned char i_channels;
 338                 NeAACDecHandle *hfaad;
 339                 NeAACDecConfiguration *cfg,*oldcfg;
 340
 341                 oldcfg = NeAACDecGetCurrentConfiguration( p_sys->hfaad );
 342                 hfaad = NeAACDecOpen();
 343                 cfg = NeAACDecGetCurrentConfiguration( hfaad );
 344                 if( oldcfg->defSampleRate )
 345                     cfg->defSampleRate = oldcfg->defSampleRate;
 346                 cfg->defObjectType = oldcfg->defObjectType;
 347                 cfg->outputFormat = oldcfg->outputFormat;
 348                 NeAACDecSetConfiguration( hfaad, cfg );
 349
 350                 if( NeAACDecInit( hfaad, p_sys->p_buffer, p_sys->i_buffer,
 351                                   &i_rate,&i_channels ) < 0 )
 352                 {
 353                     /* reinitialization failed */
 354                     NeAACDecClose( hfaad );
 355                     NeAACDecSetConfiguration( p_sys->hfaad, oldcfg );
 356                 }
 357                 else
 358                 {
 359                     NeAACDecClose( p_sys->hfaad );
 360                     p_sys->hfaad = hfaad;
 361                     p_dec->fmt_out.audio.i_rate = i_rate;
 362                     p_dec->fmt_out.audio.i_channels = i_channels;
 363                     p_dec->fmt_out.audio.i_physical_channels
 364                         = p_dec->fmt_out.audio.i_original_channels
 365                         = pi_channels_guessed[i_channels];
 366                     date_Init( &p_sys->date, i_rate, 1 );
 367                 }
 368             }
 369
 370             /* Flush the buffer */
 371             p_sys->i_buffer = 0;
 372             block_Release( p_block );
 373             return NULL;
 374         }
 375
 376         if( frame.channels <= 0 || frame.channels > 8 || frame.channels == 7 )
 377         {
 378             msg_Warn( p_dec, "invalid channels count: %i", frame.channels );
 379
 380             /* Flush the buffer */
 381             p_sys->i_buffer -= frame.bytesconsumed;
 382             if( p_sys->i_buffer > 0 )
 383             {
 384                 memmove( p_sys->p_buffer,&p_sys->p_buffer[frame.bytesconsumed],
 385                          p_sys->i_buffer );
 386             }
 387             block_Release( p_block );
 388             return NULL;
 389         }
 390
 391         if( frame.samples <= 0 )
 392         {
 393             msg_Warn( p_dec, "decoded zero sample" );
 394
 395             /* Flush the buffer */
 396             p_sys->i_buffer -= frame.bytesconsumed;
 397             if( p_sys->i_buffer > 1 )
 398             {
 399                 memmove( p_sys->p_buffer,&p_sys->p_buffer[frame.bytesconsumed],
 400                          p_sys->i_buffer );
 401             }
 402             else
 403             {
 404                 /* Drop byte of padding */
 405                 p_sys->i_buffer = 0;
 406             }
 407             block_Release( p_block );
 408             return NULL;
 409         }
 410
 411         /* We decoded a valid frame */
 412         if( p_dec->fmt_out.audio.i_rate != frame.samplerate )
 413         {
 414             date_Init( &p_sys->date, frame.samplerate, 1 );
 415             date_Set( &p_sys->date, p_block->i_pts );
 416         }
 417         p_block->i_pts = VLC_TS_INVALID;  /* PTS is valid only once */
 418
 419         p_dec->fmt_out.audio.i_rate = frame.samplerate;
 420         p_dec->fmt_out.audio.i_channels = frame.channels;
 421
 422         /* Adjust stream info when dealing with SBR/PS */
 423         bool b_sbr = (frame.sbr == 1) || (frame.sbr == 2);
 424         if( p_sys->b_sbr != b_sbr || p_sys->b_ps != frame.ps )
 425         {
 426             const char *psz_ext = (b_sbr && frame.ps) ? "SBR+PS" :
 427                                     b_sbr ? "SBR" : "PS";
 428
 429             msg_Dbg( p_dec, "AAC %s (channels: %u, samplerate: %lu)",
 430                     psz_ext, frame.channels, frame.samplerate );
 431
 432             if( !p_dec->p_description )
 433                 p_dec->p_description = vlc_meta_New();
 434             if( p_dec->p_description )
 435                 vlc_meta_AddExtra( p_dec->p_description, _("AAC extension"), psz_ext );
 436
 437             p_sys->b_sbr = b_sbr;
 438             p_sys->b_ps = frame.ps;
 439         }
 440
 441         /* Convert frame.channel_position to our own channel values */
 442         p_dec->fmt_out.audio.i_physical_channels = 0;
 443         const uint32_t nbChannels = frame.channels;
 444         unsigned j;
 445         for( unsigned i = 0; i < nbChannels; i++ )
 446         {
 447             /* Find the channel code */
 448             for( j = 0; j < MAX_CHANNEL_POSITIONS; j++ )
 449             {
 450                 if( frame.channel_position[i] == pi_channels_in[j] )
 451                     break;
 452             }
 453             if( j >= MAX_CHANNEL_POSITIONS )
 454             {
 455                 msg_Warn( p_dec, "unknown channel ordering" );
 456                 /* Invent something */
 457                 j = i;
 458             }
 459             /* */
 460             p_sys->pi_channel_positions[i] = pi_channels_out[j];
 461             if( p_dec->fmt_out.audio.i_physical_channels & pi_channels_out[j] )
 462                 frame.channels--; /* We loose a duplicated channel */
 463             else
 464                 p_dec->fmt_out.audio.i_physical_channels |= pi_channels_out[j];
 465         }
 466         if ( nbChannels != frame.channels )
 467         {
 468             p_dec->fmt_out.audio.i_physical_channels
 469                 = p_dec->fmt_out.audio.i_original_channels
 470                 = pi_channels_guessed[nbChannels];
 471         }
 472         else
 473         {
 474             p_dec->fmt_out.audio.i_original_channels =
 475                 p_dec->fmt_out.audio.i_physical_channels;
 476         }
 477         p_dec->fmt_out.audio.i_channels = nbChannels;
 478         p_out = decoder_NewAudioBuffer( p_dec, frame.samples / nbChannels );
 479         if( p_out == NULL )
 480         {
 481             p_sys->i_buffer = 0;
 482             block_Release( p_block );
 483             return NULL;
 484         }
 485
 486         p_out->i_pts = date_Get( &p_sys->date );
 487         p_out->i_length = date_Increment( &p_sys->date,
 488                                           frame.samples / nbChannels )
 489                           - p_out->i_pts;
 490
 491         DoReordering( (uint32_t *)p_out->p_buffer, samples,
 492                       frame.samples / nbChannels, nbChannels,
 493                       p_sys->pi_channel_positions );
 494
 495         p_sys->i_buffer -= frame.bytesconsumed;
 496         if( p_sys->i_buffer > 0 )
 497         {
 498             memmove( p_sys->p_buffer, &p_sys->p_buffer[frame.bytesconsumed],
 499                      p_sys->i_buffer );
 500         }
 501
 502         return p_out;
 503     }
 504     else
 505     {
 506         /* Drop byte of padding */
 507         p_sys->i_buffer = 0;
 508     }
 509
 510     block_Release( p_block );
 511     return NULL;
 512 }
 513
 514 /*****************************************************************************
 515  * Close:
 516  *****************************************************************************/
 517 static void Close( vlc_object_t *p_this )
 518 {
 519     decoder_t *p_dec = (decoder_t *)p_this;
 520     decoder_sys_t *p_sys = p_dec->p_sys;
 521
 522     NeAACDecClose( p_sys->hfaad );
 523     free( p_sys->p_buffer );
 524     free( p_sys );
 525 }
 526
 527 /*****************************************************************************
 528  * DoReordering: do some channel re-ordering (the ac3 channel order is
 529  *   different from the aac one).
 530  *****************************************************************************/
 531 static void DoReordering( uint32_t *p_out, uint32_t *p_in, int i_samples,
 532                           int i_nb_channels, uint32_t *pi_chan_positions )
 533 {
 534     int pi_chan_table[MAX_CHANNEL_POSITIONS] = {0};
 535     int i, j, k;
 536
 537     /* Find the channels mapping */
 538     for( i = 0, j = 0; i < MAX_CHANNEL_POSITIONS; i++ )
 539     {
 540         for( k = 0; k < i_nb_channels; k++ )
 541         {
 542             if( pi_channels_ordered[i] == pi_chan_positions[k] )
 543             {
 544                 pi_chan_table[k] = j++;
 545                 break;
 546             }
 547         }
 548     }
 549
 550     /* Do the actual reordering */
 551     if( HAVE_FPU )
 552         for( i = 0; i < i_samples; i++ )
 553             for( j = 0; j < i_nb_channels; j++ )
 554                 p_out[i * i_nb_channels + pi_chan_table[j]] =
 555                     p_in[i * i_nb_channels + j];
 556     else
 557         for( i = 0; i < i_samples; i++ )
 558             for( j = 0; j < i_nb_channels; j++ )
 559                 ((uint16_t *)p_out)[i * i_nb_channels + pi_chan_table[j]] =
 560                     ((uint16_t *)p_in)[i * i_nb_channels + j];
 561 }
 562