X-Git-Url: http://git.osdn.jp/view?a=blobdiff_plain;f=libhb%2Fdecavcodec.c;h=09f15ecd839360f2cfc721bf440dfa815e63e413;hb=9972486f44c586225d98967441dcd3f3fd920636;hp=f8a06992f80d946ca64de7e1c3a10ce03304b5aa;hpb=b06cc6f04d0fcd061593536a03fd6e26d99c8cf8;p=handbrake-jp%2Fhandbrake-jp-git.git diff --git a/libhb/decavcodec.c b/libhb/decavcodec.c index f8a06992..09f15ecd 100644 --- a/libhb/decavcodec.c +++ b/libhb/decavcodec.c @@ -61,6 +61,7 @@ #include "hb.h" #include "hbffmpeg.h" +#include "downmix.h" #include "libavcodec/audioconvert.h" static int decavcodecInit( hb_work_object_t *, hb_job_t * ); @@ -108,9 +109,11 @@ struct hb_work_private_s pts_heap_t pts_heap; void* buffer; struct SwsContext *sws_context; // if we have to rescale or convert color space + hb_downmix_t *downmix; + hb_sample_t *downmix_buffer; }; -static void decodeAudio( hb_work_private_t *pv, uint8_t *data, int size ); +static void decodeAudio( hb_audio_t * audio, hb_work_private_t *pv, uint8_t *data, int size ); static hb_buffer_t *link_buf_list( hb_work_private_t *pv ); @@ -200,6 +203,15 @@ static int decavcodecInit( hb_work_object_t * w, hb_job_t * job ) pv->context = avcodec_alloc_context(); hb_avcodec_open( pv->context, codec ); + if ( w->audio != NULL && + hb_need_downmix( w->audio->config.in.channel_layout, + w->audio->config.out.mixdown) ) + { + pv->downmix = hb_downmix_init(w->audio->config.in.channel_layout, + w->audio->config.out.mixdown); + hb_downmix_set_chan_map( pv->downmix, &hb_smpte_chan_map, &hb_qt_chan_map ); + } + return 0; } @@ -241,6 +253,15 @@ static void decavcodecClose( hb_work_object_t * w ) av_free( pv->buffer ); pv->buffer = NULL; } + if ( pv->downmix ) + { + hb_downmix_close( &(pv->downmix) ); + } + if ( pv->downmix_buffer ) + { + free( pv->downmix_buffer ); + pv->downmix_buffer = NULL; + } free( pv ); w->private_data = NULL; } @@ -286,9 +307,17 @@ static int decavcodecWork( hb_work_object_t * w, hb_buffer_t ** buf_in, int parser_output_buffer_len; int64_t cur = pv->pts_next; - len = av_parser_parse2( pv->parser, pv->context, - &parser_output_buffer, &parser_output_buffer_len, - in->data + pos, in->size - pos, cur, cur, AV_NOPTS_VALUE ); + if ( pv->parser != NULL ) + { + len = av_parser_parse2( pv->parser, pv->context, + &parser_output_buffer, &parser_output_buffer_len, + in->data + pos, in->size - pos, cur, cur, AV_NOPTS_VALUE ); + } + else + { + parser_output_buffer = in->data; + len = parser_output_buffer_len = in->size; + } if (parser_output_buffer_len) { // set the duration on every frame since the stream format can @@ -303,7 +332,7 @@ static int decavcodecWork( hb_work_object_t * w, hb_buffer_t ** buf_in, pv->duration = 90000. / (double)( pv->context->sample_rate * pv->context->channels ); } - decodeAudio( pv, parser_output_buffer, parser_output_buffer_len ); + decodeAudio( w->audio, pv, parser_output_buffer, parser_output_buffer_len ); } } *buf_out = link_buf_list( pv ); @@ -329,18 +358,6 @@ static int decavcodecInfo( hb_work_object_t *w, hb_work_info_t *info ) return 0; } -static const int chan2layout[] = { - HB_INPUT_CH_LAYOUT_MONO, // We should allow no audio really. - HB_INPUT_CH_LAYOUT_MONO, - HB_INPUT_CH_LAYOUT_STEREO, - HB_INPUT_CH_LAYOUT_2F1R, - HB_INPUT_CH_LAYOUT_2F2R, - HB_INPUT_CH_LAYOUT_3F2R, - HB_INPUT_CH_LAYOUT_4F2R, - HB_INPUT_CH_LAYOUT_STEREO, - HB_INPUT_CH_LAYOUT_STEREO, -}; - static int decavcodecBSInfo( hb_work_object_t *w, const hb_buffer_t *buf, hb_work_info_t *info ) { @@ -373,35 +390,56 @@ static int decavcodecBSInfo( hb_work_object_t *w, const hb_buffer_t *buf, uint8_t *buffer = av_malloc( AVCODEC_MAX_AUDIO_FRAME_SIZE ); int out_size = AVCODEC_MAX_AUDIO_FRAME_SIZE; unsigned char *pbuffer; - int pos = 0, pbuffer_size; + int pos, pbuffer_size; - while ( pos < buf->size ) + while ( buf && !ret ) { - int len = av_parser_parse2( parser, context, &pbuffer, &pbuffer_size, - buf->data + pos, buf->size - pos, - buf->start, buf->start, AV_NOPTS_VALUE ); - pos += len; - if ( pbuffer_size > 0 ) + pos = 0; + while ( pos < buf->size ) { - AVPacket avp; - av_init_packet( &avp ); - avp.data = pbuffer; - avp.size = pbuffer_size; + int len; - len = avcodec_decode_audio3( context, (int16_t*)buffer, &out_size, &avp ); - if ( len > 0 && context->sample_rate > 0 ) + if (parser != NULL ) + { + len = av_parser_parse2( parser, context, &pbuffer, + &pbuffer_size, buf->data + pos, + buf->size - pos, buf->start, + buf->start, AV_NOPTS_VALUE ); + } + else { - info->bitrate = context->bit_rate; - info->rate = context->sample_rate; - info->rate_base = 1; - info->channel_layout = chan2layout[context->channels & 7]; - ret = 1; - break; + pbuffer = buf->data; + len = pbuffer_size = buf->size; + } + pos += len; + if ( pbuffer_size > 0 ) + { + AVPacket avp; + av_init_packet( &avp ); + avp.data = pbuffer; + avp.size = pbuffer_size; + + len = avcodec_decode_audio3( context, (int16_t*)buffer, + &out_size, &avp ); + if ( len > 0 && context->sample_rate > 0 ) + { + info->bitrate = context->bit_rate; + info->rate = context->sample_rate; + info->rate_base = 1; + info->channel_layout = + hb_ff_layout_xlat(context->channel_layout, + context->channels); + ret = 1; + break; + } } } + buf = buf->next; } + av_free( buffer ); - av_parser_close( parser ); + if ( parser != NULL ) + av_parser_close( parser ); hb_avcodec_close( context ); return ret; } @@ -485,6 +523,14 @@ static int get_frame_buf( AVCodecContext *context, AVFrame *frame ) return avcodec_default_get_buffer( context, frame ); } +static int reget_frame_buf( AVCodecContext *context, AVFrame *frame ) +{ + hb_work_private_t *pv = context->opaque; + frame->pts = pv->pts; + pv->pts = -1; + return avcodec_default_reget_buffer( context, frame ); +} + static void log_chapter( hb_work_private_t *pv, int chap_num, int64_t pts ) { hb_chapter_t *c = hb_list_item( pv->job->title->list_chapter, chap_num - 1 ); @@ -515,7 +561,20 @@ static void flushDelayQueue( hb_work_private_t *pv ) } } -static int decodeFrame( hb_work_private_t *pv, uint8_t *data, int size ) +/* + * Decodes a video frame from the specified raw packet data ('data', 'size', 'sequence'). + * The output of this function is stored in 'pv->list', which contains a list + * of zero or more decoded packets. + * + * The returned packets are guaranteed to have their timestamps in the correct order, + * even if the original packets decoded by libavcodec have misordered timestamps, + * due to the use of 'packed B-frames'. + * + * Internally the set of decoded packets may be buffered in 'pv->delayq' + * until enough packets have been decoded so that the timestamps can be + * correctly rewritten, if this is necessary. + */ +static int decodeFrame( hb_work_private_t *pv, uint8_t *data, int size, int sequence ) { int got_picture, oldlevel = 0; AVFrame frame; @@ -588,6 +647,7 @@ static int decodeFrame( hb_work_private_t *pv, uint8_t *data, int size ) { buf = copy_frame( pv, &frame ); buf->start = pts; + buf->sequence = sequence; hb_list_add( pv->list, buf ); ++pv->nframes; return got_picture; @@ -635,7 +695,9 @@ static int decodeFrame( hb_work_private_t *pv, uint8_t *data, int size ) } // add the new frame to the delayq & push its timestamp on the heap - pv->delayq[slot] = copy_frame( pv, &frame ); + buf = copy_frame( pv, &frame ); + buf->sequence = sequence; + pv->delayq[slot] = buf; heap_push( &pv->pts_heap, pts ); ++pv->nframes; @@ -644,7 +706,7 @@ static int decodeFrame( hb_work_private_t *pv, uint8_t *data, int size ) return got_picture; } -static void decodeVideo( hb_work_private_t *pv, uint8_t *data, int size, +static void decodeVideo( hb_work_private_t *pv, uint8_t *data, int size, int sequence, int64_t pts, int64_t dts ) { /* @@ -664,20 +726,24 @@ static void decodeVideo( hb_work_private_t *pv, uint8_t *data, int size, if ( pout_len > 0 ) { pv->pts = pv->parser->pts; - decodeFrame( pv, pout, pout_len ); + decodeFrame( pv, pout, pout_len, sequence ); } } while ( pos < size ); /* the stuff above flushed the parser, now flush the decoder */ if ( size <= 0 ) { - while ( decodeFrame( pv, NULL, 0 ) ) + while ( decodeFrame( pv, NULL, 0, sequence ) ) { } flushDelayQueue( pv ); } } +/* + * Removes all packets from 'pv->list', links them together into + * a linked-list, and returns the first packet in the list. + */ static hb_buffer_t *link_buf_list( hb_work_private_t *pv ) { hb_buffer_t *head = hb_list_item( pv->list, 0 ); @@ -714,6 +780,7 @@ static int decavcodecvInit( hb_work_object_t * w, hb_job_t * job ) /* we have to wrap ffmpeg's get_buffer to be able to set the pts (?!) */ pv->context->opaque = pv; pv->context->get_buffer = get_frame_buf; + pv->context->reget_buffer = reget_frame_buf; return 0; } @@ -816,7 +883,7 @@ static int decavcodecvWork( hb_work_object_t * w, hb_buffer_t ** buf_in, /* if we got an empty buffer signaling end-of-stream send it downstream */ if ( in->size == 0 ) { - decodeVideo( pv, in->data, in->size, pts, dts ); + decodeVideo( pv, in->data, in->size, in->sequence, pts, dts ); hb_list_add( pv->list, in ); *buf_out = link_buf_list( pv ); return HB_WORK_DONE; @@ -853,7 +920,7 @@ static int decavcodecvWork( hb_work_object_t * w, hb_buffer_t ** buf_in, pv->new_chap = in->new_chap; pv->chap_time = pts >= 0? pts : pv->pts_next; } - decodeVideo( pv, in->data, in->size, pts, dts ); + decodeVideo( pv, in->data, in->size, in->sequence, pts, dts ); hb_buffer_close( &in ); *buf_out = link_buf_list( pv ); return HB_WORK_OK; @@ -978,8 +1045,14 @@ static void init_ffmpeg_context( hb_work_object_t *w ) // Because the time bases are so screwed up, we only take values // in the range 8fps - 64fps. AVRational tb; - if ( st->time_base.num * 64 > st->time_base.den && - st->time_base.den > st->time_base.num * 8 ) + if ( st->avg_frame_rate.den * 64 > st->avg_frame_rate.num && + st->avg_frame_rate.num > st->avg_frame_rate.den * 8 ) + { + tb.num = st->avg_frame_rate.den; + tb.den = st->avg_frame_rate.num; + } + else if ( st->time_base.num * 64 > st->time_base.den && + st->time_base.den > st->time_base.num * 8 ) { tb = st->time_base; } @@ -1001,6 +1074,7 @@ static void init_ffmpeg_context( hb_work_object_t *w ) // we have to wrap ffmpeg's get_buffer to be able to set the pts (?!) pv->context->opaque = pv; pv->context->get_buffer = get_frame_buf; + pv->context->reget_buffer = reget_frame_buf; // avi, mkv and possibly mp4 containers can contain the M$ VFW packed // b-frames abortion that messes up frame ordering and timestamps. @@ -1037,6 +1111,16 @@ static int decavcodecviInit( hb_work_object_t * w, hb_job_t * job ) pv->list = hb_list_init(); pv->pts_next = -1; pv->pts = -1; + + if ( w->audio != NULL && + hb_need_downmix( w->audio->config.in.channel_layout, + w->audio->config.out.mixdown) ) + { + pv->downmix = hb_downmix_init(w->audio->config.in.channel_layout, + w->audio->config.out.mixdown); + hb_downmix_set_chan_map( pv->downmix, &hb_smpte_chan_map, &hb_qt_chan_map ); + } + return 0; } @@ -1044,10 +1128,6 @@ static int decavcodecviWork( hb_work_object_t * w, hb_buffer_t ** buf_in, hb_buffer_t ** buf_out ) { hb_work_private_t *pv = w->private_data; - if ( ! pv->context ) - { - init_ffmpeg_context( w ); - } hb_buffer_t *in = *buf_in; *buf_in = NULL; @@ -1055,7 +1135,7 @@ static int decavcodecviWork( hb_work_object_t * w, hb_buffer_t ** buf_in, if ( in->size == 0 ) { /* flush any frames left in the decoder */ - while ( decodeFrame( pv, NULL, 0 ) ) + while ( pv->context && decodeFrame( pv, NULL, 0, in->sequence ) ) { } flushDelayQueue( pv ); @@ -1064,6 +1144,11 @@ static int decavcodecviWork( hb_work_object_t * w, hb_buffer_t ** buf_in, return HB_WORK_DONE; } + if ( ! pv->context ) + { + init_ffmpeg_context( w ); + } + int64_t pts = in->start; if( pts >= 0 ) { @@ -1081,7 +1166,7 @@ static int decavcodecviWork( hb_work_object_t * w, hb_buffer_t ** buf_in, pv->chap_time = pts >= 0? pts : pv->pts_next; } prepare_ffmpeg_buffer( in ); - decodeFrame( pv, in->data, in->size ); + decodeFrame( pv, in->data, in->size, in->sequence ); hb_buffer_close( &in ); *buf_out = link_buf_list( pv ); return HB_WORK_OK; @@ -1105,10 +1190,11 @@ static int decavcodecviInfo( hb_work_object_t *w, hb_work_info_t *info ) return 0; } -static void decodeAudio( hb_work_private_t *pv, uint8_t *data, int size ) +static void decodeAudio( hb_audio_t * audio, hb_work_private_t *pv, uint8_t *data, int size ) { AVCodecContext *context = pv->context; int pos = 0; + int loop_limit = 256; while ( pos < size ) { @@ -1125,11 +1211,20 @@ static void decodeAudio( hb_work_private_t *pv, uint8_t *data, int size ) avp.size = size - pos; int out_size = AVCODEC_MAX_AUDIO_FRAME_SIZE; + int nsamples; int len = avcodec_decode_audio3( context, buffer, &out_size, &avp ); - if ( len <= 0 ) + if ( len < 0 ) { return; } + if ( len == 0 ) + { + if ( !(loop_limit--) ) + return; + } + else + loop_limit = 256; + pos += len; if( out_size > 0 ) { @@ -1149,8 +1244,8 @@ static void decodeAudio( hb_work_private_t *pv, uint8_t *data, int size ) context->sample_fmt, 1, NULL, 0 ); // get output buffer size (in 2-byte samples) then malloc a buffer - out_size = ( out_size * 2 ) / isamp; - buffer = av_malloc( out_size ); + nsamples = out_size / isamp; + buffer = av_malloc( nsamples * 2 ); // we're doing straight sample format conversion which behaves as if // there were only one channel. @@ -1159,26 +1254,54 @@ static void decodeAudio( hb_work_private_t *pv, uint8_t *data, int size ) const int istride[6] = { isamp }; const int ostride[6] = { 2 }; - av_audio_convert( ctx, obuf, ostride, ibuf, istride, out_size >> 1 ); + av_audio_convert( ctx, obuf, ostride, ibuf, istride, nsamples ); av_audio_convert_free( ctx ); } - hb_buffer_t *buf = hb_buffer_init( 2 * out_size ); + else + { + nsamples = out_size / 2; + } - // convert from bytes to total samples - out_size >>= 1; + hb_buffer_t * buf; + + if ( pv->downmix ) + { + pv->downmix_buffer = realloc(pv->downmix_buffer, nsamples * sizeof(hb_sample_t)); + + int i; + for( i = 0; i < nsamples; ++i ) + { + pv->downmix_buffer[i] = buffer[i]; + } + + int n_ch_samples = nsamples / context->channels; + int channels = HB_AMIXDOWN_GET_DISCRETE_CHANNEL_COUNT(audio->config.out.mixdown); + + buf = hb_buffer_init( n_ch_samples * channels * sizeof(float) ); + hb_sample_t *samples = (hb_sample_t *)buf->data; + hb_downmix(pv->downmix, samples, pv->downmix_buffer, n_ch_samples); + } + else + { + buf = hb_buffer_init( nsamples * sizeof(float) ); + float *fl32 = (float *)buf->data; + int i; + for( i = 0; i < nsamples; ++i ) + { + fl32[i] = buffer[i]; + } + int n_ch_samples = nsamples / context->channels; + hb_layout_remap( &hb_smpte_chan_map, &hb_qt_chan_map, + audio->config.in.channel_layout, + fl32, n_ch_samples ); + } double pts = pv->pts_next; buf->start = pts; - pts += out_size * pv->duration; + pts += nsamples * pv->duration; buf->stop = pts; pv->pts_next = pts; - float *fl32 = (float *)buf->data; - int i; - for( i = 0; i < out_size; ++i ) - { - fl32[i] = buffer[i]; - } hb_list_add( pv->list, buf ); // if we allocated a buffer for sample format conversion, free it @@ -1231,7 +1354,7 @@ static int decavcodecaiWork( hb_work_object_t *w, hb_buffer_t **buf_in, pv->pts_next = in->start; } prepare_ffmpeg_buffer( in ); - decodeAudio( pv, in->data, in->size ); + decodeAudio( w->audio, pv, in->data, in->size ); *buf_out = link_buf_list( pv ); return HB_WORK_OK;