1 /*****************************************************************************
2 * audio.c: audio decoder using ffmpeg library
3 *****************************************************************************
4 * Copyright (C) 1999-2003 the VideoLAN team
7 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
8 * Gildas Bazin <gbazin@videolan.org>
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.
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.
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 *****************************************************************************/
25 /*****************************************************************************
27 *****************************************************************************/
32 #include <vlc_common.h>
34 #include <vlc_codec.h>
35 #include <vlc_avcodec.h>
38 #ifdef HAVE_LIBAVCODEC_AVCODEC_H
39 # include <libavcodec/avcodec.h>
40 #elif defined(HAVE_FFMPEG_AVCODEC_H)
41 # include <ffmpeg/avcodec.h>
48 /*****************************************************************************
49 * decoder_sys_t : decoder descriptor
50 *****************************************************************************/
55 /* Temporary buffer for libavcodec */
62 audio_sample_format_t aout_format;
63 audio_date_t end_date;
76 int pi_extraction[AOUT_CHAN_MAX];
77 int i_previous_channels;
78 int64_t i_previous_layout;
81 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust );
83 /*****************************************************************************
84 * InitAudioDec: initialize audio decoder
85 *****************************************************************************
86 * The ffmpeg codec will be opened, some memory allocated.
87 *****************************************************************************/
88 int InitAudioDec( decoder_t *p_dec, AVCodecContext *p_context,
89 AVCodec *p_codec, int i_codec_id, const char *psz_namecodec )
93 /* Allocate the memory needed to store the decoder's structure */
94 if( ( p_dec->p_sys = p_sys = malloc(sizeof(*p_sys)) ) == NULL )
99 p_sys->p_context = p_context;
100 p_sys->p_codec = p_codec;
101 p_sys->i_codec_id = i_codec_id;
102 p_sys->psz_namecodec = psz_namecodec;
103 p_sys->b_delayed_open = false;
105 /* ***** Fill p_context with init values ***** */
106 p_sys->p_context->sample_rate = p_dec->fmt_in.audio.i_rate;
107 p_sys->p_context->channels = p_dec->fmt_in.audio.i_channels;
109 p_sys->p_context->block_align = p_dec->fmt_in.audio.i_blockalign;
110 p_sys->p_context->bit_rate = p_dec->fmt_in.i_bitrate;
111 #if LIBAVCODEC_VERSION_INT < ((52<<16)+(0<<8)+0)
112 p_sys->p_context->bits_per_sample = p_dec->fmt_in.audio.i_bitspersample;
114 p_sys->p_context->bits_per_coded_sample = p_dec->fmt_in.audio.i_bitspersample;
117 if( p_dec->fmt_in.i_extra > 0 )
119 const uint8_t * const p_src = p_dec->fmt_in.p_extra;
123 if( p_dec->fmt_in.i_codec == VLC_CODEC_FLAC )
126 i_size = p_dec->fmt_in.i_extra - 8;
128 else if( p_dec->fmt_in.i_codec == VLC_CODEC_ALAC )
130 static const uint8_t p_pattern[] = { 0, 0, 0, 36, 'a', 'l', 'a', 'c' };
131 /* Find alac atom XXX it is a bit ugly */
132 for( i_offset = 0; i_offset < p_dec->fmt_in.i_extra - sizeof(p_pattern); i_offset++ )
134 if( !memcmp( &p_src[i_offset], p_pattern, sizeof(p_pattern) ) )
137 i_size = __MIN( p_dec->fmt_in.i_extra - i_offset, 36 );
144 i_size = p_dec->fmt_in.i_extra;
149 p_sys->p_context->extradata =
150 malloc( i_size + FF_INPUT_BUFFER_PADDING_SIZE );
151 if( p_sys->p_context->extradata )
153 uint8_t *p_dst = p_sys->p_context->extradata;
155 p_sys->p_context->extradata_size = i_size;
157 memcpy( &p_dst[0], &p_src[i_offset], i_size );
158 memset( &p_dst[i_size], 0, FF_INPUT_BUFFER_PADDING_SIZE );
164 p_sys->p_context->extradata_size = 0;
165 p_sys->p_context->extradata = NULL;
168 /* ***** Open the codec ***** */
171 ret = avcodec_open( p_sys->p_context, p_sys->p_codec );
172 vlc_avcodec_unlock();
175 msg_Err( p_dec, "cannot open codec (%s)", p_sys->psz_namecodec );
176 free( p_sys->p_context->extradata );
181 msg_Dbg( p_dec, "ffmpeg codec (%s) started", p_sys->psz_namecodec );
185 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 51, 16, 0 )
186 case CODEC_ID_WAVPACK:
187 p_sys->i_output_max = 8 * sizeof(int32_t) * 131072;
190 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 51, 4, 0 )
192 p_sys->i_output_max = p_sys->p_context->channels * sizeof(int32_t) * p_sys->p_context->sample_rate * 2;
196 p_sys->i_output_max = 8 * sizeof(int32_t) * 65535;
199 p_sys->i_output_max = 0;
202 if( p_sys->i_output_max < AVCODEC_MAX_AUDIO_FRAME_SIZE )
203 p_sys->i_output_max = AVCODEC_MAX_AUDIO_FRAME_SIZE;
204 msg_Dbg( p_dec, "Using %d bytes output buffer", p_sys->i_output_max );
205 p_sys->p_output = av_malloc( p_sys->i_output_max );
207 p_sys->p_samples = NULL;
208 p_sys->i_samples = 0;
209 p_sys->i_reject_count = 0;
210 p_sys->b_extract = false;
211 p_sys->i_previous_channels = 0;
212 p_sys->i_previous_layout = 0;
214 aout_DateSet( &p_sys->end_date, 0 );
215 if( p_dec->fmt_in.audio.i_rate )
216 aout_DateInit( &p_sys->end_date, p_dec->fmt_in.audio.i_rate );
219 p_dec->fmt_out.i_cat = AUDIO_ES;
220 /* Try to set as much informations as possible but do not trust it */
221 SetupOutputFormat( p_dec, false );
226 /*****************************************************************************
227 * SplitBuffer: Needed because aout really doesn't like big audio chunk and
228 * wma produces easily > 30000 samples...
229 *****************************************************************************/
230 static aout_buffer_t *SplitBuffer( decoder_t *p_dec )
232 decoder_sys_t *p_sys = p_dec->p_sys;
233 int i_samples = __MIN( p_sys->i_samples, 4096 );
234 aout_buffer_t *p_buffer;
236 if( i_samples == 0 ) return NULL;
238 if( ( p_buffer = decoder_NewAudioBuffer( p_dec, i_samples ) ) == NULL )
241 p_buffer->start_date = aout_DateGet( &p_sys->end_date );
242 p_buffer->end_date = aout_DateIncrement( &p_sys->end_date, i_samples );
244 if( p_sys->b_extract )
245 aout_ChannelExtract( p_buffer->p_buffer, p_dec->fmt_out.audio.i_channels,
246 p_sys->p_samples, p_sys->p_context->channels, i_samples,
247 p_sys->pi_extraction, p_dec->fmt_out.audio.i_bitspersample );
249 memcpy( p_buffer->p_buffer, p_sys->p_samples, p_buffer->i_nb_bytes );
251 p_sys->p_samples += i_samples * p_sys->p_context->channels * ( p_dec->fmt_out.audio.i_bitspersample / 8 );
252 p_sys->i_samples -= i_samples;
257 /*****************************************************************************
258 * DecodeAudio: Called to decode one frame
259 *****************************************************************************/
260 aout_buffer_t * DecodeAudio ( decoder_t *p_dec, block_t **pp_block )
262 decoder_sys_t *p_sys = p_dec->p_sys;
263 int i_used, i_output;
264 aout_buffer_t *p_buffer;
267 if( !pp_block || !*pp_block ) return NULL;
271 if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
273 block_Release( p_block );
274 avcodec_flush_buffers( p_sys->p_context );
275 p_sys->i_samples = 0;
276 aout_DateSet( &p_sys->end_date, 0 );
278 if( p_sys->i_codec_id == CODEC_ID_MP2 || p_sys->i_codec_id == CODEC_ID_MP3 )
279 p_sys->i_reject_count = 3;
283 if( p_sys->i_samples > 0 )
286 p_buffer = SplitBuffer( p_dec );
287 if( !p_buffer ) block_Release( p_block );
291 if( !aout_DateGet( &p_sys->end_date ) && !p_block->i_pts )
293 /* We've just started the stream, wait for the first PTS. */
294 block_Release( p_block );
298 if( p_block->i_buffer <= 0 )
300 block_Release( p_block );
304 i_output = __MAX( p_block->i_buffer, p_sys->i_output_max );
305 if( i_output > p_sys->i_output_max )
307 /* Grow output buffer if necessary (eg. for PCM data) */
308 p_sys->p_output = av_realloc( p_sys->p_output, i_output );
311 *pp_block = p_block = block_Realloc( p_block, 0, p_block->i_buffer + FF_INPUT_BUFFER_PADDING_SIZE );
314 p_block->i_buffer -= FF_INPUT_BUFFER_PADDING_SIZE;
315 memset( &p_block->p_buffer[p_block->i_buffer], 0, FF_INPUT_BUFFER_PADDING_SIZE );
317 #if LIBAVCODEC_VERSION_INT >= ((52<<16)+(0<<8)+0)
318 i_used = avcodec_decode_audio2( p_sys->p_context,
319 (int16_t*)p_sys->p_output, &i_output,
320 p_block->p_buffer, p_block->i_buffer );
322 i_used = avcodec_decode_audio( p_sys->p_context,
323 (int16_t*)p_sys->p_output, &i_output,
324 p_block->p_buffer, p_block->i_buffer );
327 if( i_used < 0 || i_output < 0 )
330 msg_Warn( p_dec, "cannot decode one frame (%zu bytes)",
333 block_Release( p_block );
336 else if( (size_t)i_used > p_block->i_buffer )
338 i_used = p_block->i_buffer;
341 p_block->i_buffer -= i_used;
342 p_block->p_buffer += i_used;
344 if( p_sys->p_context->channels <= 0 || p_sys->p_context->channels > 8 ||
345 p_sys->p_context->sample_rate <= 0 )
347 msg_Warn( p_dec, "invalid audio properties channels count %d, sample rate %d",
348 p_sys->p_context->channels, p_sys->p_context->sample_rate );
349 block_Release( p_block );
353 if( p_dec->fmt_out.audio.i_rate != (unsigned int)p_sys->p_context->sample_rate )
355 aout_DateInit( &p_sys->end_date, p_sys->p_context->sample_rate );
356 aout_DateSet( &p_sys->end_date, p_block->i_pts );
359 /* **** Set audio output parameters **** */
360 SetupOutputFormat( p_dec, true );
362 if( p_block->i_pts != 0 &&
363 p_block->i_pts != aout_DateGet( &p_sys->end_date ) )
365 aout_DateSet( &p_sys->end_date, p_block->i_pts );
369 /* **** Now we can output these samples **** */
370 p_sys->i_samples = i_output / (p_dec->fmt_out.audio.i_bitspersample / 8) / p_sys->p_context->channels;
371 p_sys->p_samples = p_sys->p_output;
373 /* Silent unwanted samples */
374 if( p_sys->i_reject_count > 0 )
376 memset( p_sys->p_output, 0, i_output );
377 p_sys->i_reject_count--;
380 p_buffer = SplitBuffer( p_dec );
381 if( !p_buffer ) block_Release( p_block );
385 /*****************************************************************************
386 * EndAudioDec: audio decoder destruction
387 *****************************************************************************/
388 void EndAudioDec( decoder_t *p_dec )
390 decoder_sys_t *p_sys = p_dec->p_sys;
392 av_free( p_sys->p_output );
395 /*****************************************************************************
397 *****************************************************************************/
398 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 52, 2, 0 )
399 # define LIBAVCODEC_AUDIO_LAYOUT
401 # warning "Audio channel layout is unsupported by your avcodec version."
404 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
405 static const uint64_t pi_channels_map[][2] =
407 { CH_FRONT_LEFT, AOUT_CHAN_LEFT },
408 { CH_FRONT_RIGHT, AOUT_CHAN_RIGHT },
409 { CH_FRONT_CENTER, AOUT_CHAN_CENTER },
410 { CH_LOW_FREQUENCY, AOUT_CHAN_LFE },
411 { CH_BACK_LEFT, AOUT_CHAN_REARLEFT },
412 { CH_BACK_RIGHT, AOUT_CHAN_REARRIGHT },
413 { CH_FRONT_LEFT_OF_CENTER, 0 },
414 { CH_FRONT_RIGHT_OF_CENTER, 0 },
415 { CH_BACK_CENTER, AOUT_CHAN_REARCENTER },
416 { CH_SIDE_LEFT, AOUT_CHAN_MIDDLELEFT },
417 { CH_SIDE_RIGHT, AOUT_CHAN_MIDDLERIGHT },
418 { CH_TOP_CENTER, 0 },
419 { CH_TOP_FRONT_LEFT, 0 },
420 { CH_TOP_FRONT_CENTER, 0 },
421 { CH_TOP_FRONT_RIGHT, 0 },
422 { CH_TOP_BACK_LEFT, 0 },
423 { CH_TOP_BACK_CENTER, 0 },
424 { CH_TOP_BACK_RIGHT, 0 },
425 { CH_STEREO_LEFT, 0 },
426 { CH_STEREO_RIGHT, 0 },
429 static const uint64_t pi_channels_map[][2] =
431 { 0, AOUT_CHAN_LEFT },
432 { 0, AOUT_CHAN_RIGHT },
433 { 0, AOUT_CHAN_CENTER },
434 { 0, AOUT_CHAN_LFE },
435 { 0, AOUT_CHAN_REARLEFT },
436 { 0, AOUT_CHAN_REARRIGHT },
439 { 0, AOUT_CHAN_REARCENTER },
440 { 0, AOUT_CHAN_MIDDLELEFT },
441 { 0, AOUT_CHAN_MIDDLERIGHT },
454 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust )
456 decoder_sys_t *p_sys = p_dec->p_sys;
458 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 51, 65, 0 )
459 switch( p_sys->p_context->sample_fmt )
462 p_dec->fmt_out.i_codec = VLC_CODEC_U8;
463 p_dec->fmt_out.audio.i_bitspersample = 8;
466 p_dec->fmt_out.i_codec = VLC_CODEC_S32N;
467 p_dec->fmt_out.audio.i_bitspersample = 32;
470 p_dec->fmt_out.i_codec = VLC_CODEC_FL32;
471 p_dec->fmt_out.audio.i_bitspersample = 32;
474 p_dec->fmt_out.i_codec = VLC_CODEC_FL64;
475 p_dec->fmt_out.audio.i_bitspersample = 64;
480 p_dec->fmt_out.i_codec = VLC_CODEC_S16N;
481 p_dec->fmt_out.audio.i_bitspersample = 16;
485 p_dec->fmt_out.i_codec = VLC_CODEC_S16N;
486 p_dec->fmt_out.audio.i_bitspersample = 16;
488 p_dec->fmt_out.audio.i_rate = p_sys->p_context->sample_rate;
491 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
492 if( p_sys->i_previous_channels == p_sys->p_context->channels &&
493 p_sys->i_previous_layout == p_sys->p_context->channel_layout )
496 if( p_sys->i_previous_channels == p_sys->p_context->channels )
501 p_sys->i_previous_channels = p_sys->p_context->channels;
502 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
503 p_sys->i_previous_layout = p_sys->p_context->channel_layout;
508 * FIXME should we use fmt_in.audio.i_physical_channels or not ?
510 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
511 const unsigned i_order_max = 8 * sizeof(p_sys->p_context->channel_layout);
513 const unsigned i_order_max = 64;
515 uint32_t pi_order_src[i_order_max];
516 int i_channels_src = 0;
518 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
519 if( p_sys->p_context->channel_layout )
521 for( unsigned i = 0; i < sizeof(pi_channels_map)/sizeof(*pi_channels_map); i++ )
523 if( p_sys->p_context->channel_layout & pi_channels_map[i][0] )
524 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
530 /* Create default order */
532 msg_Warn( p_dec, "Physical channel configuration not set : guessing" );
533 for( unsigned int i = 0; i < __MIN( i_order_max, (unsigned)p_sys->p_context->channels ); i++ )
535 if( i < sizeof(pi_channels_map)/sizeof(*pi_channels_map) )
536 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
539 if( i_channels_src != p_sys->p_context->channels && b_trust )
540 msg_Err( p_dec, "Channel layout not understood" );
542 uint32_t i_layout_dst;
544 p_sys->b_extract = aout_CheckChannelExtraction( p_sys->pi_extraction,
545 &i_layout_dst, &i_channels_dst,
546 NULL, pi_order_src, i_channels_src );
547 if( i_channels_dst != i_channels_src && b_trust )
548 msg_Warn( p_dec, "%d channels are dropped", i_channels_src - i_channels_dst );
550 p_dec->fmt_out.audio.i_physical_channels =
551 p_dec->fmt_out.audio.i_original_channels = i_layout_dst;
552 p_dec->fmt_out.audio.i_channels = i_channels_dst;