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;
76 int pi_extraction[AOUT_CHAN_MAX];
77 int i_previous_channels;
78 int64_t i_previous_layout;
81 #define BLOCK_FLAG_PRIVATE_REALLOCATED (1 << BLOCK_FLAG_PRIVATE_SHIFT)
83 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust );
85 /*****************************************************************************
86 * InitAudioDec: initialize audio decoder
87 *****************************************************************************
88 * The ffmpeg codec will be opened, some memory allocated.
89 *****************************************************************************/
90 int InitAudioDec( decoder_t *p_dec, AVCodecContext *p_context,
91 AVCodec *p_codec, int i_codec_id, const char *psz_namecodec )
95 /* Allocate the memory needed to store the decoder's structure */
96 if( ( p_dec->p_sys = p_sys = malloc(sizeof(*p_sys)) ) == NULL )
101 p_codec->type = CODEC_TYPE_AUDIO;
102 p_context->codec_type = CODEC_TYPE_AUDIO;
103 p_context->codec_id = i_codec_id;
104 p_sys->p_context = p_context;
105 p_sys->p_codec = p_codec;
106 p_sys->i_codec_id = i_codec_id;
107 p_sys->psz_namecodec = psz_namecodec;
108 p_sys->b_delayed_open = false;
110 /* ***** Fill p_context with init values ***** */
111 p_sys->p_context->sample_rate = p_dec->fmt_in.audio.i_rate;
112 p_sys->p_context->channels = p_dec->fmt_in.audio.i_channels;
114 p_sys->p_context->block_align = p_dec->fmt_in.audio.i_blockalign;
115 p_sys->p_context->bit_rate = p_dec->fmt_in.i_bitrate;
116 #if LIBAVCODEC_VERSION_INT < AV_VERSION_INT( 52, 0, 0 )
117 p_sys->p_context->bits_per_sample = p_dec->fmt_in.audio.i_bitspersample;
119 p_sys->p_context->bits_per_coded_sample = p_dec->fmt_in.audio.i_bitspersample;
122 if( p_dec->fmt_in.i_extra > 0 )
124 const uint8_t * const p_src = p_dec->fmt_in.p_extra;
128 if( p_dec->fmt_in.i_codec == VLC_CODEC_FLAC )
131 i_size = p_dec->fmt_in.i_extra - 8;
133 else if( p_dec->fmt_in.i_codec == VLC_CODEC_ALAC )
135 static const uint8_t p_pattern[] = { 0, 0, 0, 36, 'a', 'l', 'a', 'c' };
136 /* Find alac atom XXX it is a bit ugly */
137 for( i_offset = 0; i_offset < p_dec->fmt_in.i_extra - sizeof(p_pattern); i_offset++ )
139 if( !memcmp( &p_src[i_offset], p_pattern, sizeof(p_pattern) ) )
142 i_size = __MIN( p_dec->fmt_in.i_extra - i_offset, 36 );
149 i_size = p_dec->fmt_in.i_extra;
154 p_sys->p_context->extradata =
155 malloc( i_size + FF_INPUT_BUFFER_PADDING_SIZE );
156 if( p_sys->p_context->extradata )
158 uint8_t *p_dst = p_sys->p_context->extradata;
160 p_sys->p_context->extradata_size = i_size;
162 memcpy( &p_dst[0], &p_src[i_offset], i_size );
163 memset( &p_dst[i_size], 0, FF_INPUT_BUFFER_PADDING_SIZE );
169 p_sys->p_context->extradata_size = 0;
170 p_sys->p_context->extradata = NULL;
173 /* ***** Open the codec ***** */
176 ret = avcodec_open( p_sys->p_context, p_sys->p_codec );
177 vlc_avcodec_unlock();
180 msg_Err( p_dec, "cannot open codec (%s)", p_sys->psz_namecodec );
181 free( p_sys->p_context->extradata );
186 msg_Dbg( p_dec, "ffmpeg codec (%s) started", p_sys->psz_namecodec );
190 case CODEC_ID_WAVPACK:
191 p_sys->i_output_max = 8 * sizeof(int32_t) * 131072;
194 p_sys->i_output_max = p_sys->p_context->channels * sizeof(int32_t) * p_sys->p_context->sample_rate * 2;
197 p_sys->i_output_max = 8 * sizeof(int32_t) * 65535;
199 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 52, 35, 0 )
200 case CODEC_ID_WMAPRO:
201 p_sys->i_output_max = 8 * sizeof(float) * 6144; /* (1 << 12) * 3/2 */
205 p_sys->i_output_max = 0;
208 if( p_sys->i_output_max < AVCODEC_MAX_AUDIO_FRAME_SIZE )
209 p_sys->i_output_max = AVCODEC_MAX_AUDIO_FRAME_SIZE;
210 msg_Dbg( p_dec, "Using %d bytes output buffer", p_sys->i_output_max );
211 p_sys->p_output = av_malloc( p_sys->i_output_max );
213 p_sys->p_samples = NULL;
214 p_sys->i_samples = 0;
215 p_sys->i_reject_count = 0;
216 p_sys->b_extract = false;
217 p_sys->i_previous_channels = 0;
218 p_sys->i_previous_layout = 0;
220 date_Set( &p_sys->end_date, 0 );
221 if( p_dec->fmt_in.audio.i_rate )
222 date_Init( &p_sys->end_date, p_dec->fmt_in.audio.i_rate, 1 );
225 p_dec->fmt_out.i_cat = AUDIO_ES;
226 /* Try to set as much informations as possible but do not trust it */
227 SetupOutputFormat( p_dec, false );
232 /*****************************************************************************
233 * SplitBuffer: Needed because aout really doesn't like big audio chunk and
234 * wma produces easily > 30000 samples...
235 *****************************************************************************/
236 static aout_buffer_t *SplitBuffer( decoder_t *p_dec )
238 decoder_sys_t *p_sys = p_dec->p_sys;
239 int i_samples = __MIN( p_sys->i_samples, 4096 );
240 aout_buffer_t *p_buffer;
242 if( i_samples == 0 ) return NULL;
244 if( ( p_buffer = decoder_NewAudioBuffer( p_dec, i_samples ) ) == NULL )
247 p_buffer->i_pts = date_Get( &p_sys->end_date );
248 p_buffer->i_length = date_Increment( &p_sys->end_date, i_samples )
251 if( p_sys->b_extract )
252 aout_ChannelExtract( p_buffer->p_buffer, p_dec->fmt_out.audio.i_channels,
253 p_sys->p_samples, p_sys->p_context->channels, i_samples,
254 p_sys->pi_extraction, p_dec->fmt_out.audio.i_bitspersample );
256 memcpy( p_buffer->p_buffer, p_sys->p_samples, p_buffer->i_buffer );
258 p_sys->p_samples += i_samples * p_sys->p_context->channels * ( p_dec->fmt_out.audio.i_bitspersample / 8 );
259 p_sys->i_samples -= i_samples;
264 /*****************************************************************************
265 * DecodeAudio: Called to decode one frame
266 *****************************************************************************/
267 aout_buffer_t * DecodeAudio ( decoder_t *p_dec, block_t **pp_block )
269 decoder_sys_t *p_sys = p_dec->p_sys;
270 int i_used, i_output;
271 aout_buffer_t *p_buffer;
274 if( !pp_block || !*pp_block ) return NULL;
278 if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
280 block_Release( p_block );
281 avcodec_flush_buffers( p_sys->p_context );
282 p_sys->i_samples = 0;
283 date_Set( &p_sys->end_date, 0 );
285 if( p_sys->i_codec_id == CODEC_ID_MP2 || p_sys->i_codec_id == CODEC_ID_MP3 )
286 p_sys->i_reject_count = 3;
290 if( p_sys->i_samples > 0 )
293 p_buffer = SplitBuffer( p_dec );
294 if( !p_buffer ) block_Release( p_block );
298 if( !date_Get( &p_sys->end_date ) && !p_block->i_pts )
300 /* We've just started the stream, wait for the first PTS. */
301 block_Release( p_block );
305 if( p_block->i_buffer <= 0 )
307 block_Release( p_block );
311 if( (p_block->i_flags & BLOCK_FLAG_PRIVATE_REALLOCATED) == 0 )
313 *pp_block = p_block = block_Realloc( p_block, 0, p_block->i_buffer + FF_INPUT_BUFFER_PADDING_SIZE );
316 p_block->i_buffer -= FF_INPUT_BUFFER_PADDING_SIZE;
317 memset( &p_block->p_buffer[p_block->i_buffer], 0, FF_INPUT_BUFFER_PADDING_SIZE );
319 p_block->i_flags |= BLOCK_FLAG_PRIVATE_REALLOCATED;
324 i_output = __MAX( p_block->i_buffer, p_sys->i_output_max );
325 if( i_output > p_sys->i_output_max )
327 /* Grow output buffer if necessary (eg. for PCM data) */
328 p_sys->p_output = av_realloc( p_sys->p_output, i_output );
331 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 52, 0, 0 )
332 i_used = avcodec_decode_audio2( p_sys->p_context,
333 (int16_t*)p_sys->p_output, &i_output,
334 p_block->p_buffer, p_block->i_buffer );
336 i_used = avcodec_decode_audio( p_sys->p_context,
337 (int16_t*)p_sys->p_output, &i_output,
338 p_block->p_buffer, p_block->i_buffer );
341 if( i_used < 0 || i_output < 0 )
344 msg_Warn( p_dec, "cannot decode one frame (%zu bytes)",
347 block_Release( p_block );
350 else if( (size_t)i_used > p_block->i_buffer )
352 i_used = p_block->i_buffer;
355 p_block->i_buffer -= i_used;
356 p_block->p_buffer += i_used;
358 } while( p_block->i_buffer > 0 && i_output <= 0 );
360 if( p_sys->p_context->channels <= 0 || p_sys->p_context->channels > 8 ||
361 p_sys->p_context->sample_rate <= 0 )
363 msg_Warn( p_dec, "invalid audio properties channels count %d, sample rate %d",
364 p_sys->p_context->channels, p_sys->p_context->sample_rate );
365 block_Release( p_block );
369 if( p_dec->fmt_out.audio.i_rate != (unsigned int)p_sys->p_context->sample_rate )
371 date_Init( &p_sys->end_date, p_sys->p_context->sample_rate, 1 );
372 date_Set( &p_sys->end_date, p_block->i_pts );
375 /* **** Set audio output parameters **** */
376 SetupOutputFormat( p_dec, true );
378 if( p_block->i_pts != 0 &&
379 p_block->i_pts != date_Get( &p_sys->end_date ) )
381 date_Set( &p_sys->end_date, p_block->i_pts );
385 /* **** Now we can output these samples **** */
386 p_sys->i_samples = i_output / (p_dec->fmt_out.audio.i_bitspersample / 8) / p_sys->p_context->channels;
387 p_sys->p_samples = p_sys->p_output;
389 /* Silent unwanted samples */
390 if( p_sys->i_reject_count > 0 )
392 memset( p_sys->p_output, 0, i_output );
393 p_sys->i_reject_count--;
396 p_buffer = SplitBuffer( p_dec );
397 if( !p_buffer ) block_Release( p_block );
401 /*****************************************************************************
402 * EndAudioDec: audio decoder destruction
403 *****************************************************************************/
404 void EndAudioDec( decoder_t *p_dec )
406 decoder_sys_t *p_sys = p_dec->p_sys;
408 av_free( p_sys->p_output );
411 /*****************************************************************************
413 *****************************************************************************/
414 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 52, 2, 0 )
415 # define LIBAVCODEC_AUDIO_LAYOUT
417 # warning "Audio channel layout is unsupported by your avcodec version."
420 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
421 static const uint64_t pi_channels_map[][2] =
423 { CH_FRONT_LEFT, AOUT_CHAN_LEFT },
424 { CH_FRONT_RIGHT, AOUT_CHAN_RIGHT },
425 { CH_FRONT_CENTER, AOUT_CHAN_CENTER },
426 { CH_LOW_FREQUENCY, AOUT_CHAN_LFE },
427 { CH_BACK_LEFT, AOUT_CHAN_REARLEFT },
428 { CH_BACK_RIGHT, AOUT_CHAN_REARRIGHT },
429 { CH_FRONT_LEFT_OF_CENTER, 0 },
430 { CH_FRONT_RIGHT_OF_CENTER, 0 },
431 { CH_BACK_CENTER, AOUT_CHAN_REARCENTER },
432 { CH_SIDE_LEFT, AOUT_CHAN_MIDDLELEFT },
433 { CH_SIDE_RIGHT, AOUT_CHAN_MIDDLERIGHT },
434 { CH_TOP_CENTER, 0 },
435 { CH_TOP_FRONT_LEFT, 0 },
436 { CH_TOP_FRONT_CENTER, 0 },
437 { CH_TOP_FRONT_RIGHT, 0 },
438 { CH_TOP_BACK_LEFT, 0 },
439 { CH_TOP_BACK_CENTER, 0 },
440 { CH_TOP_BACK_RIGHT, 0 },
441 { CH_STEREO_LEFT, 0 },
442 { CH_STEREO_RIGHT, 0 },
445 static const uint64_t pi_channels_map[][2] =
447 { 0, AOUT_CHAN_LEFT },
448 { 0, AOUT_CHAN_RIGHT },
449 { 0, AOUT_CHAN_CENTER },
450 { 0, AOUT_CHAN_LFE },
451 { 0, AOUT_CHAN_REARLEFT },
452 { 0, AOUT_CHAN_REARRIGHT },
455 { 0, AOUT_CHAN_REARCENTER },
456 { 0, AOUT_CHAN_MIDDLELEFT },
457 { 0, AOUT_CHAN_MIDDLERIGHT },
470 static void SetupOutputFormat( decoder_t *p_dec, bool b_trust )
472 decoder_sys_t *p_sys = p_dec->p_sys;
474 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT( 51, 65, 0 )
475 switch( p_sys->p_context->sample_fmt )
478 p_dec->fmt_out.i_codec = VLC_CODEC_U8;
479 p_dec->fmt_out.audio.i_bitspersample = 8;
482 p_dec->fmt_out.i_codec = VLC_CODEC_S32N;
483 p_dec->fmt_out.audio.i_bitspersample = 32;
486 p_dec->fmt_out.i_codec = VLC_CODEC_FL32;
487 p_dec->fmt_out.audio.i_bitspersample = 32;
490 p_dec->fmt_out.i_codec = VLC_CODEC_FL64;
491 p_dec->fmt_out.audio.i_bitspersample = 64;
496 p_dec->fmt_out.i_codec = VLC_CODEC_S16N;
497 p_dec->fmt_out.audio.i_bitspersample = 16;
501 p_dec->fmt_out.i_codec = VLC_CODEC_S16N;
502 p_dec->fmt_out.audio.i_bitspersample = 16;
504 p_dec->fmt_out.audio.i_rate = p_sys->p_context->sample_rate;
507 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
508 if( p_sys->i_previous_channels == p_sys->p_context->channels &&
509 p_sys->i_previous_layout == p_sys->p_context->channel_layout )
512 if( p_sys->i_previous_channels == p_sys->p_context->channels )
517 p_sys->i_previous_channels = p_sys->p_context->channels;
518 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
519 p_sys->i_previous_layout = p_sys->p_context->channel_layout;
524 * FIXME should we use fmt_in.audio.i_physical_channels or not ?
526 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
527 const unsigned i_order_max = 8 * sizeof(p_sys->p_context->channel_layout);
529 const unsigned i_order_max = 64;
531 uint32_t pi_order_src[i_order_max];
532 int i_channels_src = 0;
534 #if defined(LIBAVCODEC_AUDIO_LAYOUT)
535 if( p_sys->p_context->channel_layout )
537 for( unsigned i = 0; i < sizeof(pi_channels_map)/sizeof(*pi_channels_map); i++ )
539 if( p_sys->p_context->channel_layout & pi_channels_map[i][0] )
540 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
546 /* Create default order */
548 msg_Warn( p_dec, "Physical channel configuration not set : guessing" );
549 for( unsigned int i = 0; i < __MIN( i_order_max, (unsigned)p_sys->p_context->channels ); i++ )
551 if( i < sizeof(pi_channels_map)/sizeof(*pi_channels_map) )
552 pi_order_src[i_channels_src++] = pi_channels_map[i][1];
555 if( i_channels_src != p_sys->p_context->channels && b_trust )
556 msg_Err( p_dec, "Channel layout not understood" );
558 uint32_t i_layout_dst;
560 p_sys->b_extract = aout_CheckChannelExtraction( p_sys->pi_extraction,
561 &i_layout_dst, &i_channels_dst,
562 NULL, pi_order_src, i_channels_src );
563 if( i_channels_dst != i_channels_src && b_trust )
564 msg_Warn( p_dec, "%d channels are dropped", i_channels_src - i_channels_dst );
566 p_dec->fmt_out.audio.i_physical_channels =
567 p_dec->fmt_out.audio.i_original_channels = i_layout_dst;
568 p_dec->fmt_out.audio.i_channels = i_channels_dst;