10 // work arounds (centos)
13 #define INT64_MAX 9223372036854775807LL
15 #define MAX_RETRY 1000
18 #include "bccmodels.h"
20 #include "fileffmpeg.h"
23 #include "indexfile.h"
26 #include "mainerror.h"
31 #define VIDEO_INBUF_SIZE 0x10000
32 #define AUDIO_INBUF_SIZE 0x10000
33 #define VIDEO_REFILL_THRESH 0
34 #define AUDIO_REFILL_THRESH 0x1000
36 Mutex FFMPEG::fflock("FFMPEG::fflock");
38 static void ff_err(int ret, const char *fmt, ...)
43 vsnprintf(msg, sizeof(msg), fmt, ap);
45 char errmsg[BCSTRLEN];
46 av_strerror(ret, errmsg, sizeof(errmsg));
47 fprintf(stderr,_("%s err: %s\n"),msg, errmsg);
53 pkt.data = 0; pkt.size = 0;
55 void FFPacket::finit()
57 av_packet_unref(&pkt);
60 FFrame::FFrame(FFStream *fst)
63 frm = av_frame_alloc();
64 init = fst->init_frame(frm);
72 void FFrame::queue(int64_t pos)
78 void FFrame::dequeue()
83 int FFAudioStream::read(float *fp, long len)
91 while( --k >= 0 ) *fp++ = *op++;
92 if( op >= lmt ) op = bfr;
97 void FFAudioStream::realloc(long nsz, int nch, long len)
100 float *np = new float[bsz];
101 inp = np + read(np, len) * nch;
106 delete [] bfr; bfr = np;
109 void FFAudioStream::realloc(long nsz, int nch)
111 if( nsz > sz || this->nch != nch ) {
112 long len = this->nch != nch ? 0 : hpos;
113 if( len > sz ) len = sz;
115 realloc(nsz, nch, len);
119 void FFAudioStream::reserve(long nsz, int nch)
121 long len = (inp - outp) / nch;
123 if( nsz > sz || this->nch != nch ) {
124 if( this->nch != nch ) len = 0;
125 realloc(nsz, nch, len);
128 if( (len*=nch) > 0 && bfr != outp )
129 memmove(bfr, outp, len*sizeof(*bfr));
134 long FFAudioStream::used()
136 long len = inp>=outp ? inp-outp : inp-bfr + lmt-outp;
139 long FFAudioStream::avail()
142 if( in1 >= lmt ) in1 = bfr;
143 long len = outp >= in1 ? outp-in1 : outp-bfr + lmt-in1;
146 void FFAudioStream::reset_history()
152 void FFAudioStream::iseek(int64_t ofs)
154 outp = inp - ofs*nch;
155 if( outp < bfr ) outp += sz*nch;
158 float *FFAudioStream::get_outp(int ofs)
165 int64_t FFAudioStream::put_inp(int ofs)
168 return (inp-outp) / nch;
171 int FFAudioStream::write(const float *fp, long len)
179 while( --k >= 0 ) *ip++ = *fp++;
180 if( ip >= lmt ) ip = bfr;
187 int FFAudioStream::zero(long len)
195 while( --k >= 0 ) *ip++ = 0;
196 if( ip >= lmt ) ip = bfr;
203 // does not advance outp
204 int FFAudioStream::read(double *dp, long len, int ch)
207 float *op = outp + ch;
208 float *lmt1 = lmt + nch-1;
210 int k = (lmt1 - op) / nch;
213 while( --k >= 0 ) { *dp++ = *op; op += nch; }
214 if( op >= lmt ) op -= sz*nch;
219 // load linear buffer, no wrapping allowed, does not advance inp
220 int FFAudioStream::write(const double *dp, long len, int ch)
223 float *ip = inp + ch;
224 while( --n >= 0 ) { *ip = *dp++; ip += nch; }
229 FFStream::FFStream(FFMPEG *ffmpeg, AVStream *st, int fidx)
231 this->ffmpeg = ffmpeg;
234 frm_lock = new Mutex("FFStream::frm_lock");
240 nudge = AV_NOPTS_VALUE;
241 seek_pos = curr_pos = 0;
243 reading = writing = 0;
249 FFStream::~FFStream()
251 if( reading > 0 || writing > 0 ) avcodec_close(st->codec);
252 if( fmt_ctx ) avformat_close_input(&fmt_ctx);
253 while( frms.first ) frms.remove(frms.first);
254 if( filter_graph ) avfilter_graph_free(&filter_graph);
255 if( frame ) av_frame_free(&frame);
256 if( fframe ) av_frame_free(&fframe);
257 bsfilter.remove_all_objects();
261 void FFStream::ff_lock(const char *cp)
263 FFMPEG::fflock.lock(cp);
266 void FFStream::ff_unlock()
268 FFMPEG::fflock.unlock();
271 void FFStream::queue(FFrame *frm)
273 frm_lock->lock("FFStream::queue");
277 ffmpeg->mux_lock->unlock();
280 void FFStream::dequeue(FFrame *frm)
282 frm_lock->lock("FFStream::dequeue");
284 frms.remove_pointer(frm);
288 int FFStream::encode_activate()
291 writing = ffmpeg->encode_activate();
295 int FFStream::decode_activate()
297 if( reading < 0 && (reading=ffmpeg->decode_activate()) > 0 ) {
298 ff_lock("FFStream::decode_activate");
300 AVDictionary *copts = 0;
301 av_dict_copy(&copts, ffmpeg->opts, 0);
303 // this should be avformat_copy_context(), but no copy avail
304 ret = avformat_open_input(&fmt_ctx, ffmpeg->fmt_ctx->filename, NULL, &copts);
306 ret = avformat_find_stream_info(fmt_ctx, 0);
307 st = fmt_ctx->streams[fidx];
311 AVCodecID codec_id = st->codec->codec_id;
312 AVCodec *decoder = avcodec_find_decoder(codec_id);
313 ret = avcodec_open2(st->codec, decoder, &copts);
317 eprintf(_("open decoder failed\n"));
320 eprintf(_("can't clone input file\n"));
321 av_dict_free(&copts);
327 int FFStream::read_packet()
329 av_packet_unref(ipkt);
330 int ret = av_read_frame(fmt_ctx, ipkt);
333 if( ret == AVERROR_EOF ) {
334 ipkt->stream_index = st->index;
337 ff_err(ret, "FFStream::read_packet: av_read_frame failed\n");
344 int FFStream::decode(AVFrame *frame)
347 int retries = MAX_RETRY;
350 while( ret >= 0 && !flushed && --retries >= 0 && !got_frame ) {
353 if( (ret=read_packet()) < 0 ) break;
355 if( ipkt->stream_index == st->index ) {
356 while( (ipkt->size > 0 || !ipkt->data) && !got_frame ) {
357 ret = decode_frame(ipkt, frame, got_frame);
358 if( ret < 0 ) need_packet = 1;
359 if( ret <= 0 || !ipkt->data ) break;
372 fprintf(stderr, "FFStream::decode: Retry limit\n");
376 fprintf(stderr, "FFStream::decode: failed\n");
381 int FFStream::load_filter(AVFrame *frame)
383 int ret = av_buffersrc_add_frame_flags(buffersrc_ctx,
384 frame, AV_BUFFERSRC_FLAG_KEEP_REF);
386 av_frame_unref(frame);
387 eprintf(_("av_buffersrc_add_frame_flags failed\n"));
392 int FFStream::read_filter(AVFrame *frame)
394 int ret = av_buffersink_get_frame(buffersink_ctx, frame);
396 if( ret == AVERROR(EAGAIN) ) return 0;
397 if( ret == AVERROR_EOF ) { st_eof(1); return -1; }
398 ff_err(ret, "FFStream::read_filter: av_buffersink_get_frame failed\n");
404 int FFStream::read_frame(AVFrame *frame)
406 if( !filter_graph || !buffersrc_ctx || !buffersink_ctx )
407 return decode(frame);
408 if( !fframe && !(fframe=av_frame_alloc()) ) {
409 fprintf(stderr, "FFStream::read_frame: av_frame_alloc failed\n");
413 while( !flushed && !(ret=read_filter(frame)) ) {
414 if( (ret=decode(fframe)) < 0 ) break;
415 if( ret > 0 && (ret=load_filter(fframe)) < 0 ) break;
420 int FFStream::write_packet(FFPacket &pkt)
423 av_packet_rescale_ts(pkt, st->codec->time_base, st->time_base);
424 pkt->stream_index = st->index;
425 return av_interleaved_write_frame(ffmpeg->fmt_ctx, pkt);
428 int FFStream::flush()
434 ret = encode_frame(pkt, 0, got_packet);
435 if( ret < 0 || !got_packet ) break;
436 ret = write_packet(pkt);
439 ff_err(ret, "FFStream::flush");
440 return ret >= 0 ? 0 : 1;
443 int FFStream::seek(int64_t no, double rate)
445 int64_t tstmp = -INT64_MAX+1;
446 // default ffmpeg native seek
448 int64_t pos = no, plmt = -1;
449 IndexMarks *index_markers = get_markers();
450 if( index_markers && index_markers->size() > 1 ) {
451 IndexMarks &marks = *index_markers;
452 int i = marks.find(pos);
453 int64_t n = i < 0 ? (i=0) : marks[i].no;
454 // if indexed seek point not too far away (<30 secs), use index
455 if( no-n < 30*rate ) {
458 if( i < marks.size() ) plmt = marks[i].pos;
463 double secs = pos / rate;
464 tstmp = secs * st->time_base.den / st->time_base.num;
465 if( nudge != AV_NOPTS_VALUE ) tstmp += nudge;
467 int ret = avformat_seek_file(fmt_ctx, st->index,
468 -INT64_MAX, tstmp, INT64_MAX, AVSEEK_FLAG_ANY);
470 avcodec_flush_buffers(st->codec);
471 ipkt.finit(); ipkt.init();
472 need_packet = 0; flushed = 0;
473 seeked = 1; st_eof(0);
474 // read up to retry packets, limited to npkts in stream, and not past pkt.pos plmt
476 if( read_packet() <= 0 ) { ret = -1; break; }
477 if( plmt >= 0 && ipkt->pos >= plmt ) break;
478 if( ipkt->stream_index != st->index ) continue;
479 if( --npkts <= 0 ) break;
480 int64_t pkt_ts = ipkt->dts != AV_NOPTS_VALUE ? ipkt->dts : ipkt->pts;
481 if( pkt_ts == AV_NOPTS_VALUE ) continue;
482 if( pkt_ts >= tstmp ) break;
486 //printf("** seek fail %ld, %ld\n", pos, tstmp);
487 seeked = need_packet = 0;
491 //printf("seeked pos = %ld, %ld\n", pos, tstmp);
492 seek_pos = curr_pos = pos;
496 FFAudioStream::FFAudioStream(FFMPEG *ffmpeg, AVStream *strm, int idx, int fidx)
497 : FFStream(ffmpeg, strm, fidx)
500 channel0 = channels = 0;
504 resample_context = 0;
513 bfr = new float[bsz];
518 FFAudioStream::~FFAudioStream()
520 if( resample_context ) swr_free(&resample_context);
525 int FFAudioStream::get_samples(float *&samples, uint8_t **data, int len)
527 samples = *(float **)data;
528 if( resample_context ) {
529 if( len > aud_bfr_sz ) {
535 aud_bfr = new float[aud_bfr_sz*channels];
537 int ret = swr_convert(resample_context,
538 (uint8_t**)&aud_bfr, aud_bfr_sz, (const uint8_t**)data, len);
540 ff_err(ret, "FFAudioStream::get_samples: swr_convert failed\n");
549 int FFAudioStream::load_history(uint8_t **data, int len)
552 len = get_samples(samples, data, len);
554 // biggest user bfr since seek + frame
555 realloc(mbsz + len + 1, channels);
561 int FFAudioStream::decode_frame(AVPacket *pkt, AVFrame *frame, int &got_frame)
563 int first_frame = seeked; seeked = 0;
564 int ret = avcodec_decode_audio4(st->codec, frame, &got_frame, pkt);
566 if( first_frame ) return 0;
567 ff_err(ret, "FFAudioStream::decode_frame: Could not read audio frame\n");
571 int64_t pkt_ts = av_frame_get_best_effort_timestamp(frame);
572 if( pkt_ts != AV_NOPTS_VALUE )
573 curr_pos = ffmpeg->to_secs(pkt_ts - nudge, st->time_base) * sample_rate + 0.5;
578 int FFAudioStream::encode_activate()
580 if( writing >= 0 ) return writing;
581 AVCodecContext *ctx = st->codec;
582 frame_sz = ctx->codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE ?
583 10000 : ctx->frame_size;
584 return FFStream::encode_activate();
587 int FFAudioStream::nb_samples()
589 AVCodecContext *ctx = st->codec;
590 return ctx->codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE ?
591 10000 : ctx->frame_size;
594 int64_t FFAudioStream::load_buffer(double ** const sp, int len)
596 reserve(len+1, st->codec->channels);
597 for( int ch=0; ch<nch; ++ch )
598 write(sp[ch], len, ch);
602 int FFAudioStream::in_history(int64_t pos)
604 if( pos > curr_pos ) return 0;
606 if( len > sz ) len = sz;
607 if( pos < curr_pos - len ) return 0;
612 int FFAudioStream::init_frame(AVFrame *frame)
614 AVCodecContext *ctx = st->codec;
615 frame->nb_samples = frame_sz;
616 frame->format = ctx->sample_fmt;
617 frame->channel_layout = ctx->channel_layout;
618 frame->sample_rate = ctx->sample_rate;
619 int ret = av_frame_get_buffer(frame, 0);
621 ff_err(ret, "FFAudioStream::init_frame: av_frame_get_buffer failed\n");
625 int FFAudioStream::load(int64_t pos, int len)
627 if( audio_seek(pos) < 0 ) return -1;
628 if( !frame && !(frame=av_frame_alloc()) ) {
629 fprintf(stderr, "FFAudioStream::load: av_frame_alloc failed\n");
632 if( mbsz < len ) mbsz = len;
633 int64_t end_pos = pos + len;
635 for( int i=0; ret>=0 && !flushed && curr_pos<end_pos && i<MAX_RETRY; ++i ) {
636 ret = read_frame(frame);
638 load_history(&frame->extended_data[0], frame->nb_samples);
639 curr_pos += frame->nb_samples;
642 if( end_pos > curr_pos ) {
643 zero(end_pos - curr_pos);
646 len = curr_pos - pos;
651 int FFAudioStream::audio_seek(int64_t pos)
653 if( decode_activate() < 0 ) return -1;
654 if( !st->codec || !st->codec->codec ) return -1;
655 if( in_history(pos) ) return 0;
656 if( pos == curr_pos ) return 0;
657 reset_history(); mbsz = 0;
658 // guarentee preload > 1sec samples
659 if( seek(pos-sample_rate, sample_rate) < 0 ) return -1;
663 int FFAudioStream::encode(double **samples, int len)
665 if( encode_activate() <= 0 ) return -1;
668 int64_t count = load_buffer(samples, len);
671 while( ret >= 0 && count >= frame_sz ) {
672 frm = new FFrame(this);
673 if( (ret=frm->initted()) < 0 ) break;
674 AVFrame *frame = *frm;
675 float *bfrp = get_outp(frame_sz);
676 ret = swr_convert(resample_context,
677 (uint8_t **)frame->extended_data, frame_sz,
678 (const uint8_t **)&bfrp, frame_sz);
680 ff_err(ret, "FFAudioStream::encode: swr_convert failed\n");
683 frm->queue(curr_pos);
685 curr_pos += frame_sz;
690 return ret >= 0 ? 0 : 1;
693 int FFAudioStream::encode_frame(AVPacket *pkt, AVFrame *frame, int &got_packet)
695 int ret = avcodec_encode_audio2(st->codec, pkt, frame, &got_packet);
697 ff_err(ret, "FFAudioStream::encode_frame: encode audio failed\n");
703 void FFAudioStream::load_markers()
705 IndexState *index_state = ffmpeg->file_base->asset->index_state;
706 if( !index_state || idx >= index_state->audio_markers.size() ) return;
707 if( index_state->marker_status == MARKERS_NOTTESTED ) return;
708 FFStream::load_markers(*index_state->audio_markers[idx], sample_rate);
711 IndexMarks *FFAudioStream::get_markers()
713 IndexState *index_state = ffmpeg->file_base->asset->index_state;
714 if( !index_state || idx >= index_state->audio_markers.size() ) return 0;
715 return index_state->audio_markers[idx];
718 FFVideoStream::FFVideoStream(FFMPEG *ffmpeg, AVStream *strm, int idx, int fidx)
719 : FFStream(ffmpeg, strm, fidx)
728 FFVideoStream::~FFVideoStream()
732 int FFVideoStream::decode_frame(AVPacket *pkt, AVFrame *frame, int &got_frame)
734 int first_frame = seeked; seeked = 0;
735 int ret = avcodec_decode_video2(st->codec, frame, &got_frame, pkt);
737 if( first_frame ) return 0;
738 ff_err(ret, "FFVideoStream::decode_frame: Could not read video frame\n");
741 else // this is right out of ffplay, looks questionable ???
745 int64_t pkt_ts = av_frame_get_best_effort_timestamp(frame);
746 if( pkt_ts != AV_NOPTS_VALUE )
747 curr_pos = ffmpeg->to_secs(pkt_ts - nudge, st->time_base) * frame_rate + 0.5;
752 int FFVideoStream::load(VFrame *vframe, int64_t pos)
754 int ret = video_seek(pos);
755 if( ret < 0 ) return -1;
756 if( !frame && !(frame=av_frame_alloc()) ) {
757 fprintf(stderr, "FFVideoStream::load: av_frame_alloc failed\n");
760 for( int i=0; ret>=0 && !flushed && curr_pos<=pos && i<MAX_RETRY; ++i ) {
761 ret = read_frame(frame);
762 if( ret > 0 ) ++curr_pos;
764 if( frame->format == AV_PIX_FMT_NONE || frame->width <= 0 || frame->height <= 0 )
767 ret = convert_cmodel(vframe, frame);
769 ret = ret > 0 ? 1 : ret < 0 ? -1 : 0;
773 int FFVideoStream::video_seek(int64_t pos)
775 if( decode_activate() < 0 ) return -1;
776 if( !st->codec || !st->codec->codec ) return -1;
777 if( pos == curr_pos-1 && !seeked ) return 0;
778 // if close enough, just read up to current
779 int gop = st->codec->gop_size;
780 if( gop < 4 ) gop = 4;
781 if( gop > 64 ) gop = 64;
782 int read_limit = curr_pos + 3*gop;
783 if( pos >= curr_pos && pos <= read_limit ) return 0;
784 // guarentee preload more than 2*gop frames
785 if( seek(pos - 3*gop, frame_rate) < 0 ) return -1;
789 int FFVideoStream::init_frame(AVFrame *picture)
791 AVCodecContext *ctx = st->codec;
792 picture->format = ctx->pix_fmt;
793 picture->width = ctx->width;
794 picture->height = ctx->height;
795 int ret = av_frame_get_buffer(picture, 32);
799 int FFVideoStream::encode(VFrame *vframe)
801 if( encode_activate() <= 0 ) return -1;
803 FFrame *picture = new FFrame(this);
804 int ret = picture->initted();
806 AVFrame *frame = *picture;
807 frame->pts = curr_pos;
808 ret = convert_pixfmt(vframe, frame);
811 picture->queue(curr_pos);
815 fprintf(stderr, "FFVideoStream::encode: encode failed\n");
818 return ret >= 0 ? 0 : 1;
821 int FFVideoStream::encode_frame(AVPacket *pkt, AVFrame *frame, int &got_packet)
823 int ret = avcodec_encode_video2(st->codec, pkt, frame, &got_packet);
825 ff_err(ret, "FFVideoStream::encode_frame: encode video failed\n");
831 AVPixelFormat FFVideoConvert::color_model_to_pix_fmt(int color_model)
833 switch( color_model ) {
834 case BC_YUV422: return AV_PIX_FMT_YUYV422;
835 case BC_RGB888: return AV_PIX_FMT_RGB24;
836 case BC_RGBA8888: return AV_PIX_FMT_RGBA;
837 case BC_BGR8888: return AV_PIX_FMT_BGR0;
838 case BC_BGR888: return AV_PIX_FMT_BGR24;
839 case BC_ARGB8888: return AV_PIX_FMT_ARGB;
840 case BC_ABGR8888: return AV_PIX_FMT_ABGR;
841 case BC_RGB8: return AV_PIX_FMT_RGB8;
842 case BC_YUV420P: return AV_PIX_FMT_YUV420P;
843 case BC_YUV422P: return AV_PIX_FMT_YUV422P;
844 case BC_YUV444P: return AV_PIX_FMT_YUV444P;
845 case BC_YUV411P: return AV_PIX_FMT_YUV411P;
846 case BC_RGB565: return AV_PIX_FMT_RGB565;
847 case BC_RGB161616: return AV_PIX_FMT_RGB48LE;
848 case BC_RGBA16161616: return AV_PIX_FMT_RGBA64LE;
852 return AV_PIX_FMT_NB;
855 int FFVideoConvert::pix_fmt_to_color_model(AVPixelFormat pix_fmt)
858 case AV_PIX_FMT_YUYV422: return BC_YUV422;
859 case AV_PIX_FMT_RGB24: return BC_RGB888;
860 case AV_PIX_FMT_RGBA: return BC_RGBA8888;
861 case AV_PIX_FMT_BGR0: return BC_BGR8888;
862 case AV_PIX_FMT_BGR24: return BC_BGR888;
863 case AV_PIX_FMT_ARGB: return BC_ARGB8888;
864 case AV_PIX_FMT_ABGR: return BC_ABGR8888;
865 case AV_PIX_FMT_RGB8: return BC_RGB8;
866 case AV_PIX_FMT_YUV420P: return BC_YUV420P;
867 case AV_PIX_FMT_YUV422P: return BC_YUV422P;
868 case AV_PIX_FMT_YUV444P: return BC_YUV444P;
869 case AV_PIX_FMT_YUV411P: return BC_YUV411P;
870 case AV_PIX_FMT_RGB565: return BC_RGB565;
871 case AV_PIX_FMT_RGB48LE: return BC_RGB161616;
872 case AV_PIX_FMT_RGBA64LE: return BC_RGBA16161616;
879 int FFVideoConvert::convert_picture_vframe(VFrame *frame, AVFrame *ip)
881 AVFrame *ipic = av_frame_alloc();
882 int ret = convert_picture_vframe(frame, ip, ipic);
883 av_frame_free(&ipic);
887 int FFVideoConvert::convert_picture_vframe(VFrame *frame, AVFrame *ip, AVFrame *ipic)
889 int cmodel = frame->get_color_model();
890 AVPixelFormat ofmt = color_model_to_pix_fmt(cmodel);
891 if( ofmt == AV_PIX_FMT_NB ) return -1;
892 int size = av_image_fill_arrays(ipic->data, ipic->linesize,
893 frame->get_data(), ofmt, frame->get_w(), frame->get_h(), 1);
894 if( size < 0 ) return -1;
896 int bpp = BC_CModels::calculate_pixelsize(cmodel);
897 int ysz = bpp * frame->get_w(), usz = ysz;
906 // override av_image_fill_arrays() for planar types
907 ipic->data[0] = frame->get_y(); ipic->linesize[0] = ysz;
908 ipic->data[1] = frame->get_u(); ipic->linesize[1] = usz;
909 ipic->data[2] = frame->get_v(); ipic->linesize[2] = usz;
912 ipic->data[0] = frame->get_data();
913 ipic->linesize[0] = frame->get_bytes_per_line();
917 AVPixelFormat pix_fmt = (AVPixelFormat)ip->format;
918 convert_ctx = sws_getCachedContext(convert_ctx, ip->width, ip->height, pix_fmt,
919 frame->get_w(), frame->get_h(), ofmt, SWS_BICUBIC, NULL, NULL, NULL);
921 fprintf(stderr, "FFVideoConvert::convert_picture_frame:"
922 " sws_getCachedContext() failed\n");
925 int ret = sws_scale(convert_ctx, ip->data, ip->linesize, 0, ip->height,
926 ipic->data, ipic->linesize);
928 ff_err(ret, "FFVideoConvert::convert_picture_frame: sws_scale() failed\n");
934 int FFVideoConvert::convert_cmodel(VFrame *frame, AVFrame *ip)
936 // try direct transfer
937 if( !convert_picture_vframe(frame, ip) ) return 1;
938 // use indirect transfer
939 AVPixelFormat pix_fmt = (AVPixelFormat)ip->format;
940 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
942 for( int i = 0; i <desc->nb_components; ++i ) {
943 int bits = desc->comp[i].depth;
944 if( bits > max_bits ) max_bits = bits;
946 // from libavcodec/pixdesc.c
947 #define pixdesc_has_alpha(pixdesc) ((pixdesc)->nb_components == 2 || \
948 (pixdesc)->nb_components == 4 || (pixdesc)->flags & AV_PIX_FMT_FLAG_PAL)
949 int icolor_model = pixdesc_has_alpha(desc) ?
950 (max_bits > 8 ? BC_RGBA16161616 : BC_RGBA8888) :
951 (max_bits > 8 ? BC_RGB161616 : BC_RGB888) ;
952 VFrame vframe(ip->width, ip->height, icolor_model);
953 if( convert_picture_vframe(&vframe, ip) ) return -1;
954 frame->transfer_from(&vframe);
958 int FFVideoConvert::transfer_cmodel(VFrame *frame, AVFrame *ifp)
960 int ret = convert_cmodel(frame, ifp);
962 const AVDictionary *src = av_frame_get_metadata(ifp);
963 AVDictionaryEntry *t = NULL;
964 BC_Hash *hp = frame->get_params();
966 while( (t=av_dict_get(src, "", t, AV_DICT_IGNORE_SUFFIX)) )
967 hp->update(t->key, t->value);
972 int FFVideoConvert::convert_vframe_picture(VFrame *frame, AVFrame *op)
974 AVFrame *opic = av_frame_alloc();
975 int ret = convert_vframe_picture(frame, op, opic);
976 av_frame_free(&opic);
980 int FFVideoConvert::convert_vframe_picture(VFrame *frame, AVFrame *op, AVFrame *opic)
982 int cmodel = frame->get_color_model();
983 AVPixelFormat ifmt = color_model_to_pix_fmt(cmodel);
984 if( ifmt == AV_PIX_FMT_NB ) return -1;
985 int size = av_image_fill_arrays(opic->data, opic->linesize,
986 frame->get_data(), ifmt, frame->get_w(), frame->get_h(), 1);
987 if( size < 0 ) return -1;
989 int bpp = BC_CModels::calculate_pixelsize(cmodel);
990 int ysz = bpp * frame->get_w(), usz = ysz;
999 // override av_image_fill_arrays() for planar types
1000 opic->data[0] = frame->get_y(); opic->linesize[0] = ysz;
1001 opic->data[1] = frame->get_u(); opic->linesize[1] = usz;
1002 opic->data[2] = frame->get_v(); opic->linesize[2] = usz;
1005 opic->data[0] = frame->get_data();
1006 opic->linesize[0] = frame->get_bytes_per_line();
1010 AVPixelFormat ofmt = (AVPixelFormat)op->format;
1011 convert_ctx = sws_getCachedContext(convert_ctx, frame->get_w(), frame->get_h(),
1012 ifmt, op->width, op->height, ofmt, SWS_BICUBIC, NULL, NULL, NULL);
1013 if( !convert_ctx ) {
1014 fprintf(stderr, "FFVideoConvert::convert_frame_picture:"
1015 " sws_getCachedContext() failed\n");
1018 int ret = sws_scale(convert_ctx, opic->data, opic->linesize, 0, frame->get_h(),
1019 op->data, op->linesize);
1021 ff_err(ret, "FFVideoConvert::convert_frame_picture: sws_scale() failed\n");
1027 int FFVideoConvert::convert_pixfmt(VFrame *frame, AVFrame *op)
1029 // try direct transfer
1030 if( !convert_vframe_picture(frame, op) ) return 1;
1031 // use indirect transfer
1032 int colormodel = frame->get_color_model();
1033 int bits = BC_CModels::calculate_pixelsize(colormodel) * 8;
1034 bits /= BC_CModels::components(colormodel);
1035 int icolor_model = BC_CModels::has_alpha(colormodel) ?
1036 (bits > 8 ? BC_RGBA16161616 : BC_RGBA8888) :
1037 (bits > 8 ? BC_RGB161616: BC_RGB888) ;
1038 VFrame vframe(frame->get_w(), frame->get_h(), icolor_model);
1039 vframe.transfer_from(frame);
1040 if( !convert_vframe_picture(&vframe, op) ) return 1;
1044 int FFVideoConvert::transfer_pixfmt(VFrame *frame, AVFrame *ofp)
1046 int ret = convert_pixfmt(frame, ofp);
1048 BC_Hash *hp = frame->get_params();
1049 AVDictionary **dict = avpriv_frame_get_metadatap(ofp);
1050 //av_dict_free(dict);
1051 for( int i=0; i<hp->size(); ++i ) {
1052 char *key = hp->get_key(i), *val = hp->get_value(i);
1053 av_dict_set(dict, key, val, 0);
1059 void FFVideoStream::load_markers()
1061 IndexState *index_state = ffmpeg->file_base->asset->index_state;
1062 if( !index_state || idx >= index_state->video_markers.size() ) return;
1063 FFStream::load_markers(*index_state->video_markers[idx], frame_rate);
1066 IndexMarks *FFVideoStream::get_markers()
1068 IndexState *index_state = ffmpeg->file_base->asset->index_state;
1069 if( !index_state || idx >= index_state->video_markers.size() ) return 0;
1070 return !index_state ? 0 : index_state->video_markers[idx];
1074 FFMPEG::FFMPEG(FileBase *file_base)
1077 this->file_base = file_base;
1078 memset(file_format,0,sizeof(file_format));
1079 mux_lock = new Condition(0,"FFMPEG::mux_lock",0);
1080 flow_lock = new Condition(1,"FFStream::flow_lock",0);
1083 decoding = encoding = 0;
1084 has_audio = has_video = 0;
1087 opt_video_filter = 0;
1088 opt_audio_filter = 0;
1089 char option_path[BCTEXTLEN];
1090 set_option_path(option_path, "%s", "ffmpeg.opts");
1091 read_options(option_path, opts);
1096 ff_lock("FFMPEG::~FFMPEG()");
1098 ffaudio.remove_all_objects();
1099 ffvideo.remove_all_objects();
1100 if( fmt_ctx ) avformat_close_input(&fmt_ctx);
1104 av_dict_free(&opts);
1105 delete [] opt_video_filter;
1106 delete [] opt_audio_filter;
1109 int FFMPEG::check_sample_rate(AVCodec *codec, int sample_rate)
1111 const int *p = codec->supported_samplerates;
1112 if( !p ) return sample_rate;
1114 if( *p == sample_rate ) return *p;
1120 static inline AVRational std_frame_rate(int i)
1122 static const int m1 = 1001*12, m2 = 1000*12;
1123 static const int freqs[] = {
1124 40*m1, 48*m1, 50*m1, 60*m1, 80*m1,120*m1, 240*m1,
1125 24*m2, 30*m2, 60*m2, 12*m2, 15*m2, 48*m2, 0,
1127 int freq = i<30*12 ? (i+1)*1001 : freqs[i-30*12];
1128 return (AVRational) { freq, 1001*12 };
1131 AVRational FFMPEG::check_frame_rate(AVCodec *codec, double frame_rate)
1133 const AVRational *p = codec->supported_framerates;
1134 AVRational rate, best_rate = (AVRational) { 0, 0 };
1135 double max_err = 1.; int i = 0;
1136 while( ((p ? (rate=*p++) : (rate=std_frame_rate(i++))), rate.num) != 0 ) {
1137 double framerate = (double) rate.num / rate.den;
1138 double err = fabs(frame_rate/framerate - 1.);
1139 if( err >= max_err ) continue;
1143 return max_err < 0.0001 ? best_rate : (AVRational) { 0, 0 };
1146 AVRational FFMPEG::to_sample_aspect_ratio(Asset *asset)
1149 double display_aspect = asset->width / (double)asset->height;
1150 double sample_aspect = asset->aspect_ratio / display_aspect;
1151 int width = 1000000, height = width * sample_aspect + 0.5;
1153 MWindow::create_aspect_ratio(w, h, width, height);
1154 return (AVRational){(int)h, (int)w};
1157 return (AVRational){1, 1};
1161 AVRational FFMPEG::to_time_base(int sample_rate)
1163 return (AVRational){1, sample_rate};
1166 void FFMPEG::set_option_path(char *path, const char *fmt, ...)
1168 char *ep = path + BCTEXTLEN-1;
1169 strncpy(path, File::get_cindat_path(), ep-path);
1170 strncat(path, "/ffmpeg/", ep-path);
1171 path += strlen(path);
1174 path += vsnprintf(path, ep-path, fmt, ap);
1179 void FFMPEG::get_option_path(char *path, const char *type, const char *spec)
1184 set_option_path(path, "%s/%s", type, spec);
1187 int FFMPEG::get_format(char *format, const char *path, const char *spec)
1189 char option_path[BCTEXTLEN], line[BCTEXTLEN], codec[BCTEXTLEN];
1190 get_option_path(option_path, path, spec);
1191 FILE *fp = fopen(option_path,"r");
1194 if( !fgets(line, sizeof(line), fp) ) ret = 1;
1196 line[sizeof(line)-1] = 0;
1197 ret = scan_option_line(line, format, codec);
1203 int FFMPEG::get_codec(char *codec, const char *path, const char *spec)
1205 char option_path[BCTEXTLEN], line[BCTEXTLEN], format[BCTEXTLEN];
1206 get_option_path(option_path, path, spec);
1207 FILE *fp = fopen(option_path,"r");
1210 if( !fgets(line, sizeof(line), fp) ) ret = 1;
1213 line[sizeof(line)-1] = 0;
1214 ret = scan_option_line(line, format, codec);
1217 char *vp = codec, *ep = vp+BCTEXTLEN-1;
1218 while( vp < ep && *vp && *vp != '|' ) ++vp;
1219 if( *vp == '|' ) --vp;
1220 while( vp > codec && (*vp==' ' || *vp=='\t') ) *vp-- = 0;
1225 int FFMPEG::get_file_format()
1228 char audio_format[BCSTRLEN], video_format[BCSTRLEN];
1229 file_format[0] = audio_format[0] = video_format[0] = 0;
1230 Asset *asset = file_base->asset;
1231 if( !ret && asset->audio_data )
1232 ret = get_format(audio_format, "audio", asset->acodec);
1233 if( !ret && asset->video_data )
1234 ret = get_format(video_format, "video", asset->vcodec);
1235 if( !ret && !audio_format[0] && !video_format[0] )
1237 if( !ret && audio_format[0] && video_format[0] &&
1238 strcmp(audio_format, video_format) ) ret = -1;
1240 strcpy(file_format, audio_format[0] ? audio_format : video_format);
1244 int FFMPEG::scan_option_line(char *cp, char *tag, char *val)
1246 while( *cp == ' ' || *cp == '\t' ) ++cp;
1248 while( *cp && *cp != ' ' && *cp != '\t' && *cp != '=' ) ++cp;
1250 if( !len || len > BCSTRLEN-1 ) return 1;
1251 while( bp < cp ) *tag++ = *bp++;
1253 while( *cp == ' ' || *cp == '\t' ) ++cp;
1254 if( *cp == '=' ) ++cp;
1255 while( *cp == ' ' || *cp == '\t' ) ++cp;
1257 while( *cp && *cp != '\n' ) ++cp;
1259 if( len > BCTEXTLEN-1 ) return 1;
1260 while( bp < cp ) *val++ = *bp++;
1265 int FFMPEG::load_defaults(const char *path, const char *type,
1266 char *codec, char *codec_options, int len)
1268 char default_file[BCTEXTLEN];
1269 FFMPEG::set_option_path(default_file, "%s/%s.dfl", path, type);
1270 FILE *fp = fopen(default_file,"r");
1272 fgets(codec, BCSTRLEN, fp);
1274 while( *cp && *cp!='\n' ) ++cp;
1276 while( len > 0 && fgets(codec_options, len, fp) ) {
1277 int n = strlen(codec_options);
1278 codec_options += n; len -= n;
1281 FFMPEG::set_option_path(default_file, "%s/%s", path, codec);
1282 return FFMPEG::load_options(default_file, codec_options, len);
1285 void FFMPEG::set_asset_format(Asset *asset, const char *text)
1287 if( asset->format != FILE_FFMPEG ) return;
1288 strcpy(asset->fformat, text);
1289 if( !asset->ff_audio_options[0] ) {
1290 asset->audio_data = !load_defaults("audio", text, asset->acodec,
1291 asset->ff_audio_options, sizeof(asset->ff_audio_options));
1293 if( !asset->ff_video_options[0] ) {
1294 asset->video_data = !load_defaults("video", text, asset->vcodec,
1295 asset->ff_video_options, sizeof(asset->ff_video_options));
1299 int FFMPEG::get_encoder(const char *options,
1300 char *format, char *codec, char *bsfilter, char *bsargs)
1302 FILE *fp = fopen(options,"r");
1304 eprintf(_("options open failed %s\n"),options);
1307 if( get_encoder(fp, format, codec, bsfilter, bsargs) )
1308 eprintf(_("format/codec not found %s\n"), options);
1313 int FFMPEG::get_encoder(FILE *fp,
1314 char *format, char *codec, char *bsfilter, char *bsargs)
1316 format[0] = codec[0] = bsfilter[0] = bsargs[0] = 0;
1317 char line[BCTEXTLEN];
1318 if( !fgets(line, sizeof(line), fp) ) return 1;
1319 line[sizeof(line)-1] = 0;
1320 if( scan_option_line(line, format, codec) ) return 1;
1322 while( *cp && *cp != '|' ) ++cp;
1323 if( !*cp ) return 0;
1324 if( scan_option_line(cp+1, bsfilter, bsargs) ) return 1;
1325 do { *cp-- = 0; } while( cp>=codec && (*cp==' ' || *cp == '\t' ) );
1329 int FFMPEG::read_options(const char *options, AVDictionary *&opts)
1331 FILE *fp = fopen(options,"r");
1333 int ret = read_options(fp, options, opts);
1338 int FFMPEG::scan_options(const char *options, AVDictionary *&opts, AVStream *st)
1340 FILE *fp = fmemopen((void *)options,strlen(options),"r");
1342 int ret = read_options(fp, options, opts);
1344 AVDictionaryEntry *tag = av_dict_get(opts, "id", NULL, 0);
1345 if( tag ) st->id = strtol(tag->value,0,0);
1349 int FFMPEG::read_options(FILE *fp, const char *options, AVDictionary *&opts)
1351 int ret = 0, no = 0;
1352 char line[BCTEXTLEN];
1353 while( !ret && fgets(line, sizeof(line), fp) ) {
1354 line[sizeof(line)-1] = 0;
1356 if( line[0] == '#' ) continue;
1357 if( line[0] == '\n' ) continue;
1358 char key[BCSTRLEN], val[BCTEXTLEN];
1359 if( scan_option_line(line, key, val) ) {
1360 eprintf(_("err reading %s: line %d\n"), options, no);
1364 if( !strcmp(key, "duration") )
1365 opt_duration = strtod(val, 0);
1366 else if( !strcmp(key, "video_filter") )
1367 opt_video_filter = cstrdup(val);
1368 else if( !strcmp(key, "audio_filter") )
1369 opt_audio_filter = cstrdup(val);
1370 else if( !strcmp(key, "loglevel") )
1373 av_dict_set(&opts, key, val, 0);
1379 int FFMPEG::load_options(const char *options, AVDictionary *&opts)
1381 char option_path[BCTEXTLEN];
1382 set_option_path(option_path, "%s", options);
1383 return read_options(option_path, opts);
1386 int FFMPEG::load_options(const char *path, char *bfr, int len)
1389 FILE *fp = fopen(path, "r");
1391 fgets(bfr, len, fp); // skip hdr
1392 len = fread(bfr, 1, len-1, fp);
1393 if( len < 0 ) len = 0;
1399 void FFMPEG::set_loglevel(const char *ap)
1401 if( !ap || !*ap ) return;
1406 { "quiet" , AV_LOG_QUIET },
1407 { "panic" , AV_LOG_PANIC },
1408 { "fatal" , AV_LOG_FATAL },
1409 { "error" , AV_LOG_ERROR },
1410 { "warning", AV_LOG_WARNING },
1411 { "info" , AV_LOG_INFO },
1412 { "verbose", AV_LOG_VERBOSE },
1413 { "debug" , AV_LOG_DEBUG },
1415 for( int i=0; i<(int)(sizeof(log_levels)/sizeof(log_levels[0])); ++i ) {
1416 if( !strcmp(log_levels[i].name, ap) ) {
1417 av_log_set_level(log_levels[i].level);
1421 av_log_set_level(atoi(ap));
1424 double FFMPEG::to_secs(int64_t time, AVRational time_base)
1426 double base_time = time == AV_NOPTS_VALUE ? 0 :
1427 av_rescale_q(time, time_base, AV_TIME_BASE_Q);
1428 return base_time / AV_TIME_BASE;
1431 int FFMPEG::info(char *text, int len)
1433 if( len <= 0 ) return 0;
1435 #define report(s...) do { int n = snprintf(cp,len,s); cp += n; len -= n; } while(0)
1437 if( ffvideo.size() > 0 )
1438 report("\n%d video stream%s\n",ffvideo.size(), ffvideo.size()!=1 ? "s" : "");
1439 for( int vidx=0; vidx<ffvideo.size(); ++vidx ) {
1440 FFVideoStream *vid = ffvideo[vidx];
1441 AVStream *st = vid->st;
1442 AVCodecContext *avctx = st->codec;
1443 report(_("vid%d (%d), id 0x%06x:\n"), vid->idx, vid->fidx, avctx->codec_id);
1444 const AVCodecDescriptor *desc = avcodec_descriptor_get(avctx->codec_id);
1445 report(" video%d %s", vidx+1, desc ? desc->name : " (unkn)");
1446 report(" %dx%d %5.2f", vid->width, vid->height, vid->frame_rate);
1447 const char *pfn = av_get_pix_fmt_name(avctx->pix_fmt);
1448 report(" pix %s\n", pfn ? pfn : "(unkn)");
1449 double secs = to_secs(st->duration, st->time_base);
1450 int64_t length = secs * vid->frame_rate + 0.5;
1451 double ofs = to_secs((vid->nudge - st->start_time), st->time_base);
1452 int64_t nudge = ofs * vid->frame_rate;
1453 int ch = nudge >= 0 ? '+' : (nudge=-nudge, '-');
1454 report(" %jd%c%jd frms %0.2f secs", length,ch,nudge, secs);
1455 int hrs = secs/3600; secs -= hrs*3600;
1456 int mins = secs/60; secs -= mins*60;
1457 report(" %d:%02d:%05.2f\n", hrs, mins, secs);
1459 if( ffaudio.size() > 0 )
1460 report("\n%d audio stream%s\n",ffaudio.size(), ffaudio.size()!=1 ? "s" : "");
1461 for( int aidx=0; aidx<ffaudio.size(); ++aidx ) {
1462 FFAudioStream *aud = ffaudio[aidx];
1463 AVStream *st = aud->st;
1464 AVCodecContext *avctx = st->codec;
1465 report(_("aud%d (%d), id 0x%06x:\n"), aud->idx, aud->fidx, avctx->codec_id);
1466 const AVCodecDescriptor *desc = avcodec_descriptor_get(avctx->codec_id);
1467 int nch = aud->channels, ch0 = aud->channel0+1;
1468 report(" audio%d-%d %s", ch0, ch0+nch-1, desc ? desc->name : " (unkn)");
1469 const char *fmt = av_get_sample_fmt_name(avctx->sample_fmt);
1470 report(" %s %d", fmt, aud->sample_rate);
1471 int sample_bits = av_get_bits_per_sample(avctx->codec_id);
1472 report(" %dbits\n", sample_bits);
1473 double secs = to_secs(st->duration, st->time_base);
1474 int64_t length = secs * aud->sample_rate + 0.5;
1475 double ofs = to_secs((aud->nudge - st->start_time), st->time_base);
1476 int64_t nudge = ofs * aud->sample_rate;
1477 int ch = nudge >= 0 ? '+' : (nudge=-nudge, '-');
1478 report(" %jd%c%jd smpl %0.2f secs", length,ch,nudge, secs);
1479 int hrs = secs/3600; secs -= hrs*3600;
1480 int mins = secs/60; secs -= mins*60;
1481 report(" %d:%02d:%05.2f\n", hrs, mins, secs);
1483 if( fmt_ctx->nb_programs > 0 )
1484 report("\n%d program%s\n",fmt_ctx->nb_programs, fmt_ctx->nb_programs!=1 ? "s" : "");
1485 for( int i=0; i<(int)fmt_ctx->nb_programs; ++i ) {
1486 report("program %d", i+1);
1487 AVProgram *pgrm = fmt_ctx->programs[i];
1488 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
1489 int idx = pgrm->stream_index[j];
1490 int vidx = ffvideo.size();
1491 while( --vidx>=0 && ffvideo[vidx]->fidx != idx );
1493 report(", vid%d", vidx);
1496 int aidx = ffaudio.size();
1497 while( --aidx>=0 && ffaudio[aidx]->fidx != idx );
1499 report(", aud%d", aidx);
1502 report(", (%d)", pgrm->stream_index[j]);
1507 AVDictionaryEntry *tag = 0;
1508 while ((tag = av_dict_get(fmt_ctx->metadata, "", tag, AV_DICT_IGNORE_SUFFIX)))
1509 report("%s=%s\n", tag->key, tag->value);
1518 int FFMPEG::init_decoder(const char *filename)
1520 ff_lock("FFMPEG::init_decoder");
1522 char file_opts[BCTEXTLEN];
1523 char *bp = strrchr(strcpy(file_opts, filename), '/');
1524 char *sp = strrchr(!bp ? file_opts : bp, '.');
1527 strcpy(sp, ".opts");
1528 fp = fopen(file_opts, "r");
1531 read_options(fp, file_opts, opts);
1535 load_options("decode.opts", opts);
1536 AVDictionary *fopts = 0;
1537 av_dict_copy(&fopts, opts, 0);
1538 int ret = avformat_open_input(&fmt_ctx, filename, NULL, &fopts);
1539 av_dict_free(&fopts);
1541 ret = avformat_find_stream_info(fmt_ctx, NULL);
1546 return !ret ? 0 : 1;
1549 int FFMPEG::open_decoder()
1552 if( stat(fmt_ctx->filename, &st) < 0 ) {
1553 eprintf(_("can't stat file: %s\n"), fmt_ctx->filename);
1557 int64_t file_bits = 8 * st.st_size;
1558 if( !fmt_ctx->bit_rate && opt_duration > 0 )
1559 fmt_ctx->bit_rate = file_bits / opt_duration;
1562 if( fmt_ctx->bit_rate > 0 ) {
1563 for( int i=0; i<(int)fmt_ctx->nb_streams; ++i ) {
1564 AVStream *st = fmt_ctx->streams[i];
1565 if( st->duration != AV_NOPTS_VALUE ) continue;
1566 if( st->time_base.num > INT64_MAX / fmt_ctx->bit_rate ) continue;
1567 st->duration = av_rescale(file_bits, st->time_base.den,
1568 fmt_ctx->bit_rate * (int64_t) st->time_base.num);
1573 printf("FFMPEG::open_decoder: some stream times estimated\n");
1575 ff_lock("FFMPEG::open_decoder");
1576 int ret = 0, bad_time = 0;
1577 for( int i=0; !ret && i<(int)fmt_ctx->nb_streams; ++i ) {
1578 AVStream *st = fmt_ctx->streams[i];
1579 if( st->duration == AV_NOPTS_VALUE ) bad_time = 1;
1580 AVCodecContext *avctx = st->codec;
1581 const AVCodecDescriptor *codec_desc = avcodec_descriptor_get(avctx->codec_id);
1582 if( !codec_desc ) continue;
1583 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1584 if( avctx->width < 1 ) continue;
1585 if( avctx->height < 1 ) continue;
1586 AVRational framerate = av_guess_frame_rate(fmt_ctx, st, 0);
1587 if( framerate.num < 1 ) continue;
1589 int vidx = ffvideo.size();
1590 FFVideoStream *vid = new FFVideoStream(this, st, vidx, i);
1591 vstrm_index.append(ffidx(vidx, 0));
1592 ffvideo.append(vid);
1593 vid->width = avctx->width;
1594 vid->height = avctx->height;
1595 vid->frame_rate = !framerate.den ? 0 : (double)framerate.num / framerate.den;
1596 double secs = to_secs(st->duration, st->time_base);
1597 vid->length = secs * vid->frame_rate;
1598 vid->aspect_ratio = (double)st->sample_aspect_ratio.num / st->sample_aspect_ratio.den;
1599 vid->nudge = st->start_time;
1601 if( opt_video_filter )
1602 ret = vid->create_filter(opt_video_filter, avctx,avctx);
1604 else if( avctx->codec_type == AVMEDIA_TYPE_AUDIO ) {
1605 if( avctx->channels < 1 ) continue;
1606 if( avctx->sample_rate < 1 ) continue;
1608 int aidx = ffaudio.size();
1609 FFAudioStream *aud = new FFAudioStream(this, st, aidx, i);
1610 ffaudio.append(aud);
1611 aud->channel0 = astrm_index.size();
1612 aud->channels = avctx->channels;
1613 for( int ch=0; ch<aud->channels; ++ch )
1614 astrm_index.append(ffidx(aidx, ch));
1615 aud->sample_rate = avctx->sample_rate;
1616 double secs = to_secs(st->duration, st->time_base);
1617 aud->length = secs * aud->sample_rate;
1618 if( avctx->sample_fmt != AV_SAMPLE_FMT_FLT ) {
1619 uint64_t layout = av_get_default_channel_layout(avctx->channels);
1620 if( !layout ) layout = ((uint64_t)1<<aud->channels) - 1;
1621 aud->resample_context = swr_alloc_set_opts(NULL,
1622 layout, AV_SAMPLE_FMT_FLT, avctx->sample_rate,
1623 layout, avctx->sample_fmt, avctx->sample_rate,
1625 swr_init(aud->resample_context);
1627 aud->nudge = st->start_time;
1629 if( opt_audio_filter )
1630 ret = aud->create_filter(opt_audio_filter, avctx,avctx);
1634 printf("FFMPEG::open_decoder: some stream have bad times\n");
1636 return ret < 0 ? -1 : 0;
1640 int FFMPEG::init_encoder(const char *filename)
1642 int fd = ::open(filename,O_WRONLY);
1643 if( fd < 0 ) fd = open(filename,O_WRONLY+O_CREAT,0666);
1645 eprintf(_("bad file path: %s\n"), filename);
1649 int ret = get_file_format();
1651 eprintf(_("bad file format: %s\n"), filename);
1655 eprintf(_("mismatch audio/video file format: %s\n"), filename);
1658 ff_lock("FFMPEG::init_encoder");
1660 avformat_alloc_output_context2(&fmt_ctx, 0, file_format, filename);
1662 eprintf(_("failed: %s\n"), filename);
1667 load_options("encode.opts", opts);
1673 int FFMPEG::open_encoder(const char *type, const char *spec)
1676 Asset *asset = file_base->asset;
1677 char *filename = asset->path;
1678 AVDictionary *sopts = 0;
1679 av_dict_copy(&sopts, opts, 0);
1680 char option_path[BCTEXTLEN];
1681 set_option_path(option_path, "%s/%s.opts", type, type);
1682 read_options(option_path, sopts);
1683 get_option_path(option_path, type, spec);
1684 char format_name[BCSTRLEN], codec_name[BCTEXTLEN];
1685 char bsfilter[BCSTRLEN], bsargs[BCTEXTLEN];
1686 if( get_encoder(option_path, format_name, codec_name, bsfilter, bsargs) ) {
1687 eprintf(_("get_encoder failed %s:%s\n"), option_path, filename);
1691 if( !strcmp(codec_name, CODEC_TAG_DVSD) ) strcpy(codec_name, "dv");
1692 else if( !strcmp(codec_name, CODEC_TAG_MJPEG) ) strcpy(codec_name, "mjpeg");
1693 else if( !strcmp(codec_name, CODEC_TAG_JPEG) ) strcpy(codec_name, "jpeg");
1696 ff_lock("FFMPEG::open_encoder");
1700 const AVCodecDescriptor *codec_desc = 0;
1701 AVCodec *codec = avcodec_find_encoder_by_name(codec_name);
1703 eprintf(_("cant find codec %s:%s\n"), codec_name, filename);
1707 codec_desc = avcodec_descriptor_get(codec->id);
1709 eprintf(_("unknown codec %s:%s\n"), codec_name, filename);
1714 st = avformat_new_stream(fmt_ctx, 0);
1716 eprintf(_("cant create stream %s:%s\n"), codec_name, filename);
1721 AVCodecContext *ctx = st->codec;
1722 switch( codec_desc->type ) {
1723 case AVMEDIA_TYPE_AUDIO: {
1725 eprintf(_("duplicate audio %s:%s\n"), codec_name, filename);
1730 if( scan_options(asset->ff_audio_options, sopts, st) ) {
1731 eprintf(_("bad audio options %s:%s\n"), codec_name, filename);
1735 if( asset->ff_audio_bitrate > 0 ) {
1736 ctx->bit_rate = asset->ff_audio_bitrate;
1738 sprintf(arg, "%d", asset->ff_audio_bitrate);
1739 av_dict_set(&sopts, "b", arg, 0);
1741 int aidx = ffaudio.size();
1742 int fidx = aidx + ffvideo.size();
1743 FFAudioStream *aud = new FFAudioStream(this, st, aidx, fidx);
1744 ffaudio.append(aud); fst = aud;
1745 aud->sample_rate = asset->sample_rate;
1746 ctx->channels = aud->channels = asset->channels;
1747 for( int ch=0; ch<aud->channels; ++ch )
1748 astrm_index.append(ffidx(aidx, ch));
1749 ctx->channel_layout = av_get_default_channel_layout(ctx->channels);
1750 ctx->sample_rate = check_sample_rate(codec, asset->sample_rate);
1751 if( !ctx->sample_rate ) {
1752 eprintf(_("check_sample_rate failed %s\n"), filename);
1756 ctx->time_base = st->time_base = (AVRational){1, aud->sample_rate};
1757 ctx->sample_fmt = codec->sample_fmts[0];
1758 uint64_t layout = av_get_default_channel_layout(ctx->channels);
1759 aud->resample_context = swr_alloc_set_opts(NULL,
1760 layout, ctx->sample_fmt, aud->sample_rate,
1761 layout, AV_SAMPLE_FMT_FLT, ctx->sample_rate,
1763 swr_init(aud->resample_context);
1766 case AVMEDIA_TYPE_VIDEO: {
1768 eprintf(_("duplicate video %s:%s\n"), codec_name, filename);
1773 if( scan_options(asset->ff_video_options, sopts, st) ) {
1774 eprintf(_("bad video options %s:%s\n"), codec_name, filename);
1778 if( asset->ff_video_bitrate > 0 ) {
1779 ctx->bit_rate = asset->ff_video_bitrate;
1781 sprintf(arg, "%d", asset->ff_video_bitrate);
1782 av_dict_set(&sopts, "b", arg, 0);
1784 else if( asset->ff_video_quality > 0 ) {
1785 ctx->global_quality = asset->ff_video_quality * FF_QP2LAMBDA;
1786 ctx->qmin = ctx->qmax = asset->ff_video_quality;
1787 ctx->mb_lmin = ctx->qmin * FF_QP2LAMBDA;
1788 ctx->mb_lmax = ctx->qmax * FF_QP2LAMBDA;
1789 ctx->flags |= CODEC_FLAG_QSCALE;
1791 av_dict_set(&sopts, "flags", "+qscale", 0);
1792 sprintf(arg, "%d", asset->ff_video_quality);
1793 av_dict_set(&sopts, "qscale", arg, 0);
1794 sprintf(arg, "%d", ctx->global_quality);
1795 av_dict_set(&sopts, "global_quality", arg, 0);
1797 int vidx = ffvideo.size();
1798 int fidx = vidx + ffaudio.size();
1799 FFVideoStream *vid = new FFVideoStream(this, st, vidx, fidx);
1800 vstrm_index.append(ffidx(vidx, 0));
1801 ffvideo.append(vid); fst = vid;
1802 vid->width = asset->width;
1803 ctx->width = (vid->width+3) & ~3;
1804 vid->height = asset->height;
1805 ctx->height = (vid->height+3) & ~3;
1806 vid->frame_rate = asset->frame_rate;
1807 ctx->sample_aspect_ratio = to_sample_aspect_ratio(asset);
1808 ctx->pix_fmt = codec->pix_fmts ? codec->pix_fmts[0] : AV_PIX_FMT_YUV420P;
1809 AVRational frame_rate = check_frame_rate(codec, vid->frame_rate);
1810 if( !frame_rate.num || !frame_rate.den ) {
1811 eprintf(_("check_frame_rate failed %s\n"), filename);
1815 ctx->time_base = (AVRational) { frame_rate.den, frame_rate.num };
1816 st->time_base = ctx->time_base;
1820 eprintf(_("not audio/video, %s:%s\n"), codec_name, filename);
1825 if( fmt_ctx->oformat->flags & AVFMT_GLOBALHEADER )
1826 st->codec->flags |= CODEC_FLAG_GLOBAL_HEADER;
1828 ret = avcodec_open2(st->codec, codec, &sopts);
1830 ff_err(ret,"FFMPEG::open_encoder");
1831 eprintf(_("open failed %s:%s\n"), codec_name, filename);
1838 if( fst && bsfilter[0] )
1839 fst->add_bsfilter(bsfilter, !bsargs[0] ? 0 : bsargs);
1845 av_dict_free(&sopts);
1849 int FFMPEG::close_encoder()
1852 if( encoding > 0 ) {
1853 av_write_trailer(fmt_ctx);
1854 if( !(fmt_ctx->flags & AVFMT_NOFILE) )
1855 avio_closep(&fmt_ctx->pb);
1861 int FFMPEG::decode_activate()
1863 if( decoding < 0 ) {
1865 for( int vidx=0; vidx<ffvideo.size(); ++vidx )
1866 ffvideo[vidx]->nudge = AV_NOPTS_VALUE;
1867 for( int aidx=0; aidx<ffaudio.size(); ++aidx )
1868 ffaudio[aidx]->nudge = AV_NOPTS_VALUE;
1869 // set nudges for each program stream set
1870 int npgrms = fmt_ctx->nb_programs;
1871 for( int i=0; i<npgrms; ++i ) {
1872 AVProgram *pgrm = fmt_ctx->programs[i];
1873 // first start time video stream
1874 int64_t vstart_time = -1, astart_time = -1;
1875 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
1876 int fidx = pgrm->stream_index[j];
1877 AVStream *st = fmt_ctx->streams[fidx];
1878 AVCodecContext *avctx = st->codec;
1879 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1880 if( st->start_time == AV_NOPTS_VALUE ) continue;
1881 if( vstart_time > st->start_time ) continue;
1882 vstart_time = st->start_time;
1885 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1886 if( st->start_time == AV_NOPTS_VALUE ) continue;
1887 if( astart_time > st->start_time ) continue;
1888 astart_time = st->start_time;
1892 // match program streams to max start_time
1893 int64_t nudge = vstart_time > astart_time ? vstart_time : astart_time;
1894 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
1895 int fidx = pgrm->stream_index[j];
1896 AVStream *st = fmt_ctx->streams[fidx];
1897 AVCodecContext *avctx = st->codec;
1898 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1899 for( int k=0; k<ffvideo.size(); ++k ) {
1900 if( ffvideo[k]->fidx != fidx ) continue;
1901 ffvideo[k]->nudge = nudge;
1905 if( avctx->codec_type == AVMEDIA_TYPE_AUDIO ) {
1906 for( int k=0; k<ffaudio.size(); ++k ) {
1907 if( ffaudio[k]->fidx != fidx ) continue;
1908 ffaudio[k]->nudge = nudge;
1914 // set nudges for any streams not yet set
1915 int64_t vstart_time = 0, astart_time = 0;
1916 int nstreams = fmt_ctx->nb_streams;
1917 for( int i=0; i<nstreams; ++i ) {
1918 AVStream *st = fmt_ctx->streams[i];
1919 AVCodecContext *avctx = st->codec;
1920 switch( avctx->codec_type ) {
1921 case AVMEDIA_TYPE_VIDEO: {
1922 if( st->start_time == AV_NOPTS_VALUE ) continue;
1923 int vidx = ffvideo.size();
1924 while( --vidx >= 0 && ffvideo[vidx]->fidx != i );
1925 if( vidx >= 0 && ffvideo[vidx]->nudge != AV_NOPTS_VALUE ) continue;
1926 if( vstart_time >= st->start_time ) continue;
1927 vstart_time = st->start_time;
1929 case AVMEDIA_TYPE_AUDIO: {
1930 if( st->start_time == AV_NOPTS_VALUE ) continue;
1931 int aidx = ffaudio.size();
1932 while( --aidx >= 0 && ffaudio[aidx]->fidx != i );
1933 if( aidx >= 0 && ffaudio[aidx]->nudge != AV_NOPTS_VALUE ) continue;
1934 if( astart_time >= st->start_time ) continue;
1935 astart_time = st->start_time;
1940 int64_t nudge = vstart_time > astart_time ? vstart_time : astart_time;
1941 for( int vidx=0; vidx<ffvideo.size(); ++vidx ) {
1942 if( ffvideo[vidx]->nudge != AV_NOPTS_VALUE ) continue;
1943 ffvideo[vidx]->nudge = nudge;
1945 for( int aidx=0; aidx<ffaudio.size(); ++aidx ) {
1946 if( ffaudio[aidx]->nudge != AV_NOPTS_VALUE ) continue;
1947 ffaudio[aidx]->nudge = nudge;
1954 int FFMPEG::encode_activate()
1957 if( encoding < 0 ) {
1959 if( !(fmt_ctx->flags & AVFMT_NOFILE) &&
1960 (ret=avio_open(&fmt_ctx->pb, fmt_ctx->filename, AVIO_FLAG_WRITE)) < 0 ) {
1961 ff_err(ret, "FFMPEG::encode_activate: err opening : %s\n",
1966 AVDictionary *fopts = 0;
1967 char option_path[BCTEXTLEN];
1968 set_option_path(option_path, "format/%s", file_format);
1969 read_options(option_path, fopts);
1970 ret = avformat_write_header(fmt_ctx, &fopts);
1971 av_dict_free(&fopts);
1973 ff_err(ret, "FFMPEG::encode_activate: write header failed %s\n",
1983 int FFMPEG::audio_seek(int stream, int64_t pos)
1985 int aidx = astrm_index[stream].st_idx;
1986 FFAudioStream *aud = ffaudio[aidx];
1987 aud->audio_seek(pos);
1991 int FFMPEG::video_seek(int stream, int64_t pos)
1993 int vidx = vstrm_index[stream].st_idx;
1994 FFVideoStream *vid = ffvideo[vidx];
1995 vid->video_seek(pos);
2000 int FFMPEG::decode(int chn, int64_t pos, double *samples, int len)
2002 if( !has_audio || chn >= astrm_index.size() ) return -1;
2003 int aidx = astrm_index[chn].st_idx;
2004 FFAudioStream *aud = ffaudio[aidx];
2005 if( aud->load(pos, len) < len ) return -1;
2006 int ch = astrm_index[chn].st_ch;
2007 int ret = aud->read(samples,len,ch);
2011 int FFMPEG::decode(int layer, int64_t pos, VFrame *vframe)
2013 if( !has_video || layer >= vstrm_index.size() ) return -1;
2014 int vidx = vstrm_index[layer].st_idx;
2015 FFVideoStream *vid = ffvideo[vidx];
2016 return vid->load(vframe, pos);
2020 int FFMPEG::encode(int stream, double **samples, int len)
2022 FFAudioStream *aud = ffaudio[stream];
2023 return aud->encode(samples, len);
2027 int FFMPEG::encode(int stream, VFrame *frame)
2029 FFVideoStream *vid = ffvideo[stream];
2030 return vid->encode(frame);
2033 void FFMPEG::start_muxer()
2041 void FFMPEG::stop_muxer()
2050 void FFMPEG::flow_off()
2053 flow_lock->lock("FFMPEG::flow_off");
2057 void FFMPEG::flow_on()
2061 flow_lock->unlock();
2064 void FFMPEG::flow_ctl()
2067 flow_lock->lock("FFMPEG::flow_ctl");
2068 flow_lock->unlock();
2072 int FFMPEG::mux_audio(FFrame *frm)
2075 FFStream *fst = frm->fst;
2076 AVCodecContext *ctx = fst->st->codec;
2077 AVFrame *frame = *frm;
2078 AVRational tick_rate = {1, ctx->sample_rate};
2079 frame->pts = av_rescale_q(frm->position, tick_rate, ctx->time_base);
2081 int ret = fst->encode_frame(pkt, frame, got_packet);
2082 if( ret >= 0 && got_packet )
2083 ret = fst->write_packet(pkt);
2085 ff_err(ret, "FFMPEG::mux_audio");
2086 return ret >= 0 ? 0 : 1;
2089 int FFMPEG::mux_video(FFrame *frm)
2092 FFStream *fst = frm->fst;
2093 AVFrame *frame = *frm;
2094 frame->pts = frm->position;
2096 int ret = fst->encode_frame(pkt, frame, got_packet);
2097 if( ret >= 0 && got_packet )
2098 ret = fst->write_packet(pkt);
2100 ff_err(ret, "FFMPEG::mux_video");
2101 return ret >= 0 ? 0 : 1;
2107 double atm = -1, vtm = -1;
2108 FFrame *afrm = 0, *vfrm = 0;
2110 for( int i=0; i<ffaudio.size(); ++i ) { // earliest audio
2111 FFStream *fst = ffaudio[i];
2112 if( fst->frm_count < 3 ) { demand = 1; flow_on(); }
2113 FFrame *frm = fst->frms.first;
2114 if( !frm ) { if( !done ) return; continue; }
2115 double tm = to_secs(frm->position, fst->st->codec->time_base);
2116 if( atm < 0 || tm < atm ) { atm = tm; afrm = frm; }
2118 for( int i=0; i<ffvideo.size(); ++i ) { // earliest video
2119 FFStream *fst = ffvideo[i];
2120 if( fst->frm_count < 2 ) { demand = 1; flow_on(); }
2121 FFrame *frm = fst->frms.first;
2122 if( !frm ) { if( !done ) return; continue; }
2123 double tm = to_secs(frm->position, fst->st->codec->time_base);
2124 if( vtm < 0 || tm < vtm ) { vtm = tm; vfrm = frm; }
2126 if( !demand ) flow_off();
2127 if( !afrm && !vfrm ) break;
2128 int v = !afrm ? -1 : !vfrm ? 1 : av_compare_ts(
2129 vfrm->position, vfrm->fst->st->codec->time_base,
2130 afrm->position, afrm->fst->st->codec->time_base);
2131 FFrame *frm = v <= 0 ? vfrm : afrm;
2132 if( frm == afrm ) mux_audio(frm);
2133 if( frm == vfrm ) mux_video(frm);
2142 mux_lock->lock("FFMPEG::run");
2146 for( int i=0; i<ffaudio.size(); ++i )
2147 ffaudio[i]->flush();
2148 for( int i=0; i<ffvideo.size(); ++i )
2149 ffvideo[i]->flush();
2153 int FFMPEG::ff_total_audio_channels()
2155 return astrm_index.size();
2158 int FFMPEG::ff_total_astreams()
2160 return ffaudio.size();
2163 int FFMPEG::ff_audio_channels(int stream)
2165 return ffaudio[stream]->channels;
2168 int FFMPEG::ff_sample_rate(int stream)
2170 return ffaudio[stream]->sample_rate;
2173 const char* FFMPEG::ff_audio_format(int stream)
2175 AVStream *st = ffaudio[stream]->st;
2176 AVCodecID id = st->codec->codec_id;
2177 const AVCodecDescriptor *desc = avcodec_descriptor_get(id);
2178 return desc ? desc->name : _("Unknown");
2181 int FFMPEG::ff_audio_pid(int stream)
2183 return ffaudio[stream]->st->id;
2186 int64_t FFMPEG::ff_audio_samples(int stream)
2188 return ffaudio[stream]->length;
2191 // find audio astream/channels with this program,
2192 // or all program audio channels (astream=-1)
2193 int FFMPEG::ff_audio_for_video(int vstream, int astream, int64_t &channel_mask)
2197 int vidx = ffvideo[vstream]->fidx;
2198 // find first program with this video stream
2199 for( int i=0; pidx<0 && i<(int)fmt_ctx->nb_programs; ++i ) {
2200 AVProgram *pgrm = fmt_ctx->programs[i];
2201 for( int j=0; pidx<0 && j<(int)pgrm->nb_stream_indexes; ++j ) {
2202 int st_idx = pgrm->stream_index[j];
2203 AVStream *st = fmt_ctx->streams[st_idx];
2204 if( st->codec->codec_type != AVMEDIA_TYPE_VIDEO ) continue;
2205 if( st_idx == vidx ) pidx = i;
2208 if( pidx < 0 ) return -1;
2210 int64_t channels = 0;
2211 AVProgram *pgrm = fmt_ctx->programs[pidx];
2212 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
2213 int aidx = pgrm->stream_index[j];
2214 AVStream *st = fmt_ctx->streams[aidx];
2215 if( st->codec->codec_type != AVMEDIA_TYPE_AUDIO ) continue;
2216 if( astream > 0 ) { --astream; continue; }
2218 for( int i=0; astrm<0 && i<ffaudio.size(); ++i )
2219 if( ffaudio[i]->fidx == aidx ) astrm = i;
2221 if( ret < 0 ) ret = astrm;
2222 int64_t mask = (1 << ffaudio[astrm]->channels) - 1;
2223 channels |= mask << ffaudio[astrm]->channel0;
2225 if( !astream ) break;
2227 channel_mask = channels;
2232 int FFMPEG::ff_total_video_layers()
2234 return vstrm_index.size();
2237 int FFMPEG::ff_total_vstreams()
2239 return ffvideo.size();
2242 int FFMPEG::ff_video_width(int stream)
2244 return ffvideo[stream]->width;
2247 int FFMPEG::ff_video_height(int stream)
2249 return ffvideo[stream]->height;
2252 int FFMPEG::ff_set_video_width(int stream, int width)
2254 int w = ffvideo[stream]->width;
2255 ffvideo[stream]->width = width;
2259 int FFMPEG::ff_set_video_height(int stream, int height)
2261 int h = ffvideo[stream]->height;
2262 ffvideo[stream]->height = height;
2266 int FFMPEG::ff_coded_width(int stream)
2268 AVStream *st = ffvideo[stream]->st;
2269 return st->codec->coded_width;
2272 int FFMPEG::ff_coded_height(int stream)
2274 AVStream *st = ffvideo[stream]->st;
2275 return st->codec->coded_height;
2278 float FFMPEG::ff_aspect_ratio(int stream)
2280 return ffvideo[stream]->aspect_ratio;
2283 const char* FFMPEG::ff_video_format(int stream)
2285 AVStream *st = ffvideo[stream]->st;
2286 AVCodecID id = st->codec->codec_id;
2287 const AVCodecDescriptor *desc = avcodec_descriptor_get(id);
2288 return desc ? desc->name : _("Unknown");
2291 double FFMPEG::ff_frame_rate(int stream)
2293 return ffvideo[stream]->frame_rate;
2296 int64_t FFMPEG::ff_video_frames(int stream)
2298 return ffvideo[stream]->length;
2301 int FFMPEG::ff_video_pid(int stream)
2303 return ffvideo[stream]->st->id;
2307 int FFMPEG::ff_cpus()
2309 return file_base->file->cpus;
2312 int FFVideoStream::create_filter(const char *filter_spec,
2313 AVCodecContext *src_ctx, AVCodecContext *sink_ctx)
2315 avfilter_register_all();
2316 const char *sp = filter_spec;
2317 char filter_name[BCSTRLEN], *np = filter_name;
2318 int i = sizeof(filter_name);
2319 while( --i>=0 && *sp!=0 && !strchr(" \t:=,",*sp) ) *np++ = *sp++;
2321 AVFilter *filter = !filter_name[0] ? 0 : avfilter_get_by_name(filter_name);
2322 if( !filter || avfilter_pad_get_type(filter->inputs,0) != AVMEDIA_TYPE_VIDEO ) {
2323 ff_err(AVERROR(EINVAL), "FFVideoStream::create_filter: %s\n", filter_spec);
2326 filter_graph = avfilter_graph_alloc();
2327 AVFilter *buffersrc = avfilter_get_by_name("buffer");
2328 AVFilter *buffersink = avfilter_get_by_name("buffersink");
2330 int ret = 0; char args[BCTEXTLEN];
2331 snprintf(args, sizeof(args),
2332 "video_size=%dx%d:pix_fmt=%d:time_base=%d/%d:pixel_aspect=%d/%d",
2333 src_ctx->width, src_ctx->height, src_ctx->pix_fmt,
2334 src_ctx->time_base.num, src_ctx->time_base.den,
2335 src_ctx->sample_aspect_ratio.num, src_ctx->sample_aspect_ratio.den);
2337 ret = avfilter_graph_create_filter(&buffersrc_ctx, buffersrc, "in",
2338 args, NULL, filter_graph);
2340 ret = avfilter_graph_create_filter(&buffersink_ctx, buffersink, "out",
2341 NULL, NULL, filter_graph);
2343 ret = av_opt_set_bin(buffersink_ctx, "pix_fmts",
2344 (uint8_t*)&sink_ctx->pix_fmt, sizeof(sink_ctx->pix_fmt),
2345 AV_OPT_SEARCH_CHILDREN);
2347 ff_err(ret, "FFVideoStream::create_filter");
2349 ret = FFStream::create_filter(filter_spec);
2350 return ret >= 0 ? 0 : -1;
2353 int FFAudioStream::create_filter(const char *filter_spec,
2354 AVCodecContext *src_ctx, AVCodecContext *sink_ctx)
2356 avfilter_register_all();
2357 const char *sp = filter_spec;
2358 char filter_name[BCSTRLEN], *np = filter_name;
2359 int i = sizeof(filter_name);
2360 while( --i>=0 && *sp!=0 && !strchr(" \t:=,",*sp) ) *np++ = *sp++;
2362 AVFilter *filter = !filter_name[0] ? 0 : avfilter_get_by_name(filter_name);
2363 if( !filter || avfilter_pad_get_type(filter->inputs,0) != AVMEDIA_TYPE_AUDIO ) {
2364 ff_err(AVERROR(EINVAL), "FFAudioStream::create_filter: %s\n", filter_spec);
2367 filter_graph = avfilter_graph_alloc();
2368 AVFilter *buffersrc = avfilter_get_by_name("abuffer");
2369 AVFilter *buffersink = avfilter_get_by_name("abuffersink");
2370 int ret = 0; char args[BCTEXTLEN];
2371 snprintf(args, sizeof(args),
2372 "time_base=%d/%d:sample_rate=%d:sample_fmt=%s:channel_layout=0x%jx",
2373 src_ctx->time_base.num, src_ctx->time_base.den, src_ctx->sample_rate,
2374 av_get_sample_fmt_name(src_ctx->sample_fmt), src_ctx->channel_layout);
2376 ret = avfilter_graph_create_filter(&buffersrc_ctx, buffersrc, "in",
2377 args, NULL, filter_graph);
2379 ret = avfilter_graph_create_filter(&buffersink_ctx, buffersink, "out",
2380 NULL, NULL, filter_graph);
2382 ret = av_opt_set_bin(buffersink_ctx, "sample_fmts",
2383 (uint8_t*)&sink_ctx->sample_fmt, sizeof(sink_ctx->sample_fmt),
2384 AV_OPT_SEARCH_CHILDREN);
2386 ret = av_opt_set_bin(buffersink_ctx, "channel_layouts",
2387 (uint8_t*)&sink_ctx->channel_layout,
2388 sizeof(sink_ctx->channel_layout), AV_OPT_SEARCH_CHILDREN);
2390 ret = av_opt_set_bin(buffersink_ctx, "sample_rates",
2391 (uint8_t*)&sink_ctx->sample_rate, sizeof(sink_ctx->sample_rate),
2392 AV_OPT_SEARCH_CHILDREN);
2394 ff_err(ret, "FFAudioStream::create_filter");
2396 ret = FFStream::create_filter(filter_spec);
2397 return ret >= 0 ? 0 : -1;
2400 int FFStream::create_filter(const char *filter_spec)
2402 /* Endpoints for the filter graph. */
2403 AVFilterInOut *outputs = avfilter_inout_alloc();
2404 outputs->name = av_strdup("in");
2405 outputs->filter_ctx = buffersrc_ctx;
2406 outputs->pad_idx = 0;
2409 AVFilterInOut *inputs = avfilter_inout_alloc();
2410 inputs->name = av_strdup("out");
2411 inputs->filter_ctx = buffersink_ctx;
2412 inputs->pad_idx = 0;
2415 int ret = !outputs->name || !inputs->name ? -1 : 0;
2417 ret = avfilter_graph_parse_ptr(filter_graph, filter_spec,
2418 &inputs, &outputs, NULL);
2420 ret = avfilter_graph_config(filter_graph, NULL);
2423 ff_err(ret, "FFStream::create_filter");
2424 avfilter_graph_free(&filter_graph);
2427 avfilter_inout_free(&inputs);
2428 avfilter_inout_free(&outputs);
2432 void FFStream::add_bsfilter(const char *bsf, const char *ap)
2434 bsfilter.append(new BSFilter(bsf,ap));
2437 int FFStream::bs_filter(AVPacket *pkt)
2439 if( !bsfilter.size() ) return 0;
2440 av_packet_split_side_data(pkt);
2443 for( int i=0; i<bsfilter.size(); ++i ) {
2444 AVPacket bspkt = *pkt;
2445 ret = av_bitstream_filter_filter(bsfilter[i]->bsfc,
2446 st->codec, bsfilter[i]->args, &bspkt.data, &bspkt.size,
2447 pkt->data, pkt->size, pkt->flags & AV_PKT_FLAG_KEY);
2448 if( ret < 0 ) break;
2449 int size = bspkt.size;
2450 uint8_t *data = bspkt.data;
2451 if( !ret && bspkt.data != pkt->data ) {
2453 data = (uint8_t *)av_malloc(size + FF_INPUT_BUFFER_PADDING_SIZE);
2454 if( !data ) { ret = AVERROR(ENOMEM); break; }
2455 memcpy(data, bspkt.data, size);
2456 memset(data+size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
2460 pkt->side_data = 0; pkt->side_data_elems = 0;
2461 av_packet_unref(pkt);
2462 ret = av_packet_from_data(&bspkt, data, size);
2463 if( ret < 0 ) break;
2468 ff_err(ret,"FFStream::bs_filter");
2472 int FFMPEG::scan(IndexState *index_state, int64_t *scan_position, int *canceled)
2475 av_init_packet(&pkt);
2476 AVFrame *frame = av_frame_alloc();
2478 fprintf(stderr,"FFMPEG::scan: ");
2479 fprintf(stderr,_("av_frame_alloc failed\n"));
2483 index_state->add_video_markers(ffvideo.size());
2484 index_state->add_audio_markers(ffaudio.size());
2486 for( int i=0; i<(int)fmt_ctx->nb_streams; ++i ) {
2487 AVDictionary *copts = 0;
2488 av_dict_copy(&copts, opts, 0);
2489 AVStream *st = fmt_ctx->streams[i];
2490 AVCodecID codec_id = st->codec->codec_id;
2491 AVCodec *decoder = avcodec_find_decoder(codec_id);
2492 if( avcodec_open2(st->codec, decoder, &copts) < 0 ) {
2493 fprintf(stderr,"FFMPEG::scan: ");
2494 fprintf(stderr,_("codec open failed\n"));
2496 av_dict_free(&copts);
2499 for( int64_t count=0; !*canceled; ++count ) {
2500 av_packet_unref(&pkt);
2501 pkt.data = 0; pkt.size = 0;
2503 int ret = av_read_frame(fmt_ctx, &pkt);
2505 if( ret == AVERROR_EOF ) break;
2506 if( ++errs > 100 ) {
2507 ff_err(ret,_("over 100 read_frame errs\n"));
2512 if( !pkt.data ) continue;
2513 int i = pkt.stream_index;
2514 if( i < 0 || i >= (int)fmt_ctx->nb_streams ) continue;
2515 AVStream *st = fmt_ctx->streams[i];
2516 AVCodecContext *avctx = st->codec;
2517 if( pkt.pos > *scan_position ) *scan_position = pkt.pos;
2519 switch( avctx->codec_type ) {
2520 case AVMEDIA_TYPE_VIDEO: {
2521 int vidx = ffvideo.size();
2522 while( --vidx>=0 && ffvideo[vidx]->fidx != i );
2523 if( vidx < 0 ) break;
2524 FFVideoStream *vid = ffvideo[vidx];
2525 int64_t tstmp = pkt.dts;
2526 if( tstmp == AV_NOPTS_VALUE ) tstmp = pkt.pts;
2527 if( tstmp != AV_NOPTS_VALUE && (pkt.flags & AV_PKT_FLAG_KEY) && pkt.pos > 0 ) {
2528 if( vid->nudge != AV_NOPTS_VALUE ) tstmp -= vid->nudge;
2529 double secs = to_secs(tstmp, st->time_base);
2530 int64_t frm = secs * vid->frame_rate + 0.5;
2531 if( frm < 0 ) frm = 0;
2532 index_state->put_video_mark(vidx, frm, pkt.pos);
2535 while( pkt.size > 0 ) {
2536 av_frame_unref(frame);
2538 int ret = vid->decode_frame(&pkt, frame, got_frame);
2539 if( ret <= 0 ) break;
2540 // if( got_frame ) {}
2546 case AVMEDIA_TYPE_AUDIO: {
2547 int aidx = ffaudio.size();
2548 while( --aidx>=0 && ffaudio[aidx]->fidx != i );
2549 if( aidx < 0 ) break;
2550 FFAudioStream *aud = ffaudio[aidx];
2551 int64_t tstmp = pkt.pts;
2552 if( tstmp == AV_NOPTS_VALUE ) tstmp = pkt.dts;
2553 if( tstmp != AV_NOPTS_VALUE && (pkt.flags & AV_PKT_FLAG_KEY) && pkt.pos > 0 ) {
2554 if( aud->nudge != AV_NOPTS_VALUE ) tstmp -= aud->nudge;
2555 double secs = to_secs(tstmp, st->time_base);
2556 int64_t sample = secs * aud->sample_rate + 0.5;
2557 if( sample < 0 ) sample = 0;
2558 index_state->put_audio_mark(aidx, sample, pkt.pos);
2560 while( pkt.size > 0 ) {
2561 int ch = aud->channel0, nch = aud->channels;
2562 int64_t pos = index_state->pos(ch);
2563 if( pos != aud->curr_pos ) {
2564 if( abs(pos-aud->curr_pos) > 1 )
2565 printf("audio%d pad %jd %jd (%jd)\n", aud->idx, pos, aud->curr_pos, pos-aud->curr_pos);
2566 index_state->pad_data(ch, nch, aud->curr_pos);
2568 av_frame_unref(frame);
2570 int ret = aud->decode_frame(&pkt, frame, got_frame);
2571 if( ret <= 0 ) break;
2572 if( got_frame && frame->channels == nch ) {
2574 int len = aud->get_samples(samples,
2575 &frame->extended_data[0], frame->nb_samples);
2576 for( int i=0; i<nch; ++i )
2577 index_state->put_data(ch+i,nch,samples+i,len);
2578 aud->curr_pos += len;
2587 av_frame_free(&frame);
2591 void FFStream::load_markers(IndexMarks &marks, double rate)
2594 int64_t sz = marks.size();
2595 int max_entries = fmt_ctx->max_index_size / sizeof(AVIndexEntry) - 1;
2596 int nb_ent = st->nb_index_entries;
2597 // some formats already have an index
2599 AVIndexEntry *ep = &st->index_entries[nb_ent-1];
2600 int64_t tstmp = ep->timestamp;
2601 if( nudge != AV_NOPTS_VALUE ) tstmp -= nudge;
2602 double secs = ffmpeg->to_secs(tstmp, st->time_base);
2603 int64_t no = secs * rate;
2604 while( in < sz && marks[in].no <= no ) ++in;
2606 int64_t len = sz - in;
2607 int64_t count = max_entries - nb_ent;
2608 if( count > len ) count = len;
2609 for( int i=0; i<count; ++i ) {
2610 int k = in + i * len / count;
2611 int64_t no = marks[k].no, pos = marks[k].pos;
2612 double secs = (double)no / rate;
2613 int64_t tstmp = secs * st->time_base.den / st->time_base.num;
2614 if( nudge != AV_NOPTS_VALUE ) tstmp += nudge;
2615 av_add_index_entry(st, pos, tstmp, 0, 0, AVINDEX_KEYFRAME);