10 // work arounds (centos)
13 #define INT64_MAX 9223372036854775807LL
17 #include "bccmodels.h"
19 #include "fileffmpeg.h"
22 #include "indexfile.h"
25 #include "mainerror.h"
30 #define VIDEO_INBUF_SIZE 0x10000
31 #define AUDIO_INBUF_SIZE 0x10000
32 #define VIDEO_REFILL_THRESH 0
33 #define AUDIO_REFILL_THRESH 0x1000
35 Mutex FFMPEG::fflock("FFMPEG::fflock");
37 static void ff_err(int ret, const char *fmt, ...)
42 vsnprintf(msg, sizeof(msg), fmt, ap);
44 char errmsg[BCSTRLEN];
45 av_strerror(ret, errmsg, sizeof(errmsg));
46 fprintf(stderr,_("%s err: %s\n"),msg, errmsg);
62 pkt.data = 0; pkt.size = 0;
65 FFrame::FFrame(FFStream *fst)
68 frm = av_frame_alloc();
69 init = fst->init_frame(frm);
77 void FFrame::queue(int64_t pos)
83 void FFrame::dequeue()
88 int FFAudioStream::read(float *fp, long len)
96 while( --k >= 0 ) *fp++ = *op++;
97 if( op >= lmt ) op = bfr;
102 void FFAudioStream::realloc(long nsz, int nch, long len)
104 long bsz = nsz * nch;
105 float *np = new float[bsz];
106 inp = np + read(np, len) * nch;
111 delete [] bfr; bfr = np;
114 void FFAudioStream::realloc(long nsz, int nch)
116 if( nsz > sz || this->nch != nch ) {
117 long len = this->nch != nch ? 0 : hpos;
118 if( len > sz ) len = sz;
120 realloc(nsz, nch, len);
124 void FFAudioStream::reserve(long nsz, int nch)
126 long len = (inp - outp) / nch;
128 if( nsz > sz || this->nch != nch ) {
129 if( this->nch != nch ) len = 0;
130 realloc(nsz, nch, len);
133 if( (len*=nch) > 0 && bfr != outp )
134 memmove(bfr, outp, len*sizeof(*bfr));
139 long FFAudioStream::used()
141 long len = inp>=outp ? inp-outp : inp-bfr + lmt-outp;
144 long FFAudioStream::avail()
147 if( in1 >= lmt ) in1 = bfr;
148 long len = outp >= in1 ? outp-in1 : outp-bfr + lmt-in1;
151 void FFAudioStream::reset_history()
157 void FFAudioStream::iseek(int64_t ofs)
159 outp = inp - ofs*nch;
160 if( outp < bfr ) outp += sz*nch;
163 float *FFAudioStream::get_outp(int ofs)
170 int64_t FFAudioStream::put_inp(int ofs)
173 return (inp-outp) / nch;
176 int FFAudioStream::write(const float *fp, long len)
184 while( --k >= 0 ) *ip++ = *fp++;
185 if( ip >= lmt ) ip = bfr;
192 int FFAudioStream::zero(long len)
200 while( --k >= 0 ) *ip++ = 0;
201 if( ip >= lmt ) ip = bfr;
208 // does not advance outp
209 int FFAudioStream::read(double *dp, long len, int ch)
212 float *op = outp + ch;
213 float *lmt1 = lmt + nch-1;
215 int k = (lmt1 - op) / nch;
218 while( --k >= 0 ) { *dp++ = *op; op += nch; }
219 if( op >= lmt ) op -= sz*nch;
224 // load linear buffer, no wrapping allowed, does not advance inp
225 int FFAudioStream::write(const double *dp, long len, int ch)
228 float *ip = inp + ch;
229 while( --n >= 0 ) { *ip = *dp++; ip += nch; }
234 FFStream::FFStream(FFMPEG *ffmpeg, AVStream *st, int fidx)
236 this->ffmpeg = ffmpeg;
239 frm_lock = new Mutex("FFStream::frm_lock");
245 nudge = AV_NOPTS_VALUE;
246 seek_pos = curr_pos = 0;
249 reading = writing = 0;
255 FFStream::~FFStream()
257 if( reading > 0 || writing > 0 ) avcodec_close(st->codec);
258 if( fmt_ctx ) avformat_close_input(&fmt_ctx);
259 while( frms.first ) frms.remove(frms.first);
260 if( filter_graph ) avfilter_graph_free(&filter_graph);
261 if( frame ) av_frame_free(&frame);
262 if( fframe ) av_frame_free(&fframe);
263 bsfilter.remove_all_objects();
267 void FFStream::ff_lock(const char *cp)
269 FFMPEG::fflock.lock(cp);
272 void FFStream::ff_unlock()
274 FFMPEG::fflock.unlock();
277 void FFStream::queue(FFrame *frm)
279 frm_lock->lock("FFStream::queue");
283 ffmpeg->mux_lock->unlock();
286 void FFStream::dequeue(FFrame *frm)
288 frm_lock->lock("FFStream::dequeue");
290 frms.remove_pointer(frm);
294 int FFStream::encode_activate()
297 writing = ffmpeg->encode_activate();
301 int FFStream::decode_activate()
303 if( reading < 0 && (reading=ffmpeg->decode_activate()) > 0 ) {
304 ff_lock("FFStream::decode_activate");
306 AVDictionary *copts = 0;
307 av_dict_copy(&copts, ffmpeg->opts, 0);
309 // this should be avformat_copy_context(), but no copy avail
310 ret = avformat_open_input(&fmt_ctx, ffmpeg->fmt_ctx->filename, NULL, &copts);
312 ret = avformat_find_stream_info(fmt_ctx, 0);
313 st = fmt_ctx->streams[fidx];
317 AVCodecID codec_id = st->codec->codec_id;
318 AVCodec *decoder = avcodec_find_decoder(codec_id);
319 ret = avcodec_open2(st->codec, decoder, &copts);
323 eprintf("FFStream::decode_activate: open decoder failed\n");
326 eprintf("FFStream::decode_activate: can't clone input file\n");
327 av_dict_free(&copts);
333 int FFStream::read_packet()
335 av_packet_unref(ipkt);
336 int ret = av_read_frame(fmt_ctx, ipkt);
339 if( ret == AVERROR_EOF ) {
340 ipkt->stream_index = st->index;
343 ff_err(ret, "FFStream::read_packet: av_read_frame failed\n");
350 int FFStream::decode(AVFrame *frame)
356 while( ret >= 0 && !flushed && --retries >= 0 && !got_frame ) {
359 if( (ret=read_packet()) < 0 ) break;
361 if( ipkt->stream_index == st->index ) {
362 while( (ipkt->size > 0 || !ipkt->data) && !got_frame ) {
363 ret = decode_frame(ipkt, frame, got_frame);
364 if( ret <= 0 || !ipkt->data ) break;
377 fprintf(stderr, "FFStream::decode: Retry limit\n");
381 fprintf(stderr, "FFStream::decode: failed\n");
386 int FFStream::load_filter(AVFrame *frame)
388 int ret = av_buffersrc_add_frame_flags(buffersrc_ctx,
389 frame, AV_BUFFERSRC_FLAG_KEEP_REF);
391 av_frame_unref(frame);
392 eprintf("FFStream::load_filter: av_buffersrc_add_frame_flags failed\n");
397 int FFStream::read_filter(AVFrame *frame)
399 int ret = av_buffersink_get_frame(buffersink_ctx, frame);
401 if( ret == AVERROR(EAGAIN) ) return 0;
402 if( ret == AVERROR_EOF ) { st_eof(1); return -1; }
403 ff_err(ret, "FFStream::read_filter: av_buffersink_get_frame failed\n");
409 int FFStream::read_frame(AVFrame *frame)
411 if( !filter_graph || !buffersrc_ctx || !buffersink_ctx )
412 return decode(frame);
413 if( !fframe && !(fframe=av_frame_alloc()) ) {
414 fprintf(stderr, "FFStream::read_frame: av_frame_alloc failed\n");
418 while( !flushed && !(ret=read_filter(frame)) ) {
419 if( (ret=decode(fframe)) < 0 ) break;
420 if( ret > 0 && (ret=load_filter(fframe)) < 0 ) break;
425 int FFStream::write_packet(FFPacket &pkt)
428 av_packet_rescale_ts(pkt, st->codec->time_base, st->time_base);
429 pkt->stream_index = st->index;
430 return av_interleaved_write_frame(ffmpeg->fmt_ctx, pkt);
433 int FFStream::flush()
439 ret = encode_frame(pkt, 0, got_packet);
440 if( ret < 0 || !got_packet ) break;
441 ret = write_packet(pkt);
444 ff_err(ret, "FFStream::flush");
445 return ret >= 0 ? 0 : 1;
448 int FFStream::seek(int64_t no, double rate)
451 // default ffmpeg native seek
453 int64_t pos = no, plmt = -1;
454 if( index_markers && index_markers->size() > 1 ) {
455 IndexMarks &marks = *index_markers;
456 int i = marks.find(pos);
457 int64_t n = i < 0 ? (i=0) : marks[i].no;
458 // if indexed seek point not too far away (<30 secs), use index
459 if( no-n < 30*rate ) {
466 double secs = pos / rate;
467 int64_t tstmp = secs * st->time_base.den / st->time_base.num;
468 if( nudge != AV_NOPTS_VALUE ) tstmp += nudge;
469 if( avformat_seek_file(fmt_ctx, st->index,
470 -INT64_MAX, tstmp, INT64_MAX, AVSEEK_FLAG_ANY) < 0 ) return -1;
471 avcodec_flush_buffers(st->codec);
472 need_packet = 0; flushed = 0;
473 seeked = 1; st_eof(0);
474 int64_t pkt_ts = AV_NOPTS_VALUE;
475 int ret = 1, retry = 1000;
477 // read up to retry packets, limited to npkts in stream, and not past pkt.pos plmt
478 while( ret > 0 && npkts > 0 && --retry >= 0 ) {
479 if( (ret=read_packet()) <= 0 ) break;
480 if( ipkt->stream_index != st->index ) continue;
481 if( (pkt_ts=ipkt->dts) == AV_NOPTS_VALUE ) pkt_ts = ipkt->pts;
482 if( pkt_ts != AV_NOPTS_VALUE && pkt_ts >= tstmp ) break;
483 if( plmt >= 0 && ipkt->pos >= plmt ) break;
487 if( ret <= 0 || retry < 0 ) return -1;
488 seek_pos = curr_pos = pos;
489 return npkts > 0 ? 1 : 0;
492 FFAudioStream::FFAudioStream(FFMPEG *ffmpeg, AVStream *strm, int idx, int fidx)
493 : FFStream(ffmpeg, strm, fidx)
496 channel0 = channels = 0;
500 resample_context = 0;
509 bfr = new float[bsz];
514 FFAudioStream::~FFAudioStream()
516 if( resample_context ) swr_free(&resample_context);
521 int FFAudioStream::get_samples(float *&samples, uint8_t **data, int len)
523 samples = *(float **)data;
524 if( resample_context ) {
525 if( len > aud_bfr_sz ) {
531 aud_bfr = new float[aud_bfr_sz*channels];
533 int ret = swr_convert(resample_context,
534 (uint8_t**)&aud_bfr, aud_bfr_sz, (const uint8_t**)data, len);
536 ff_err(ret, "FFAudioStream::get_samples: swr_convert failed\n");
545 int FFAudioStream::load_history(uint8_t **data, int len)
548 len = get_samples(samples, data, len);
550 // biggest user bfr since seek + frame
551 realloc(mbsz + len + 1, channels);
557 int FFAudioStream::decode_frame(AVPacket *pkt, AVFrame *frame, int &got_frame)
559 int first_frame = seeked; seeked = 0;
560 int ret = avcodec_decode_audio4(st->codec, frame, &got_frame, pkt);
562 if( first_frame ) return 0;
563 ff_err(ret, "FFAudioStream::decode_frame: Could not read audio frame\n");
567 int64_t pkt_ts = av_frame_get_best_effort_timestamp(frame);
568 if( pkt_ts != AV_NOPTS_VALUE )
569 curr_pos = ffmpeg->to_secs(pkt_ts - nudge, st->time_base) * sample_rate + 0.5;
574 int FFAudioStream::encode_activate()
576 if( writing >= 0 ) return writing;
577 AVCodecContext *ctx = st->codec;
578 frame_sz = ctx->codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE ?
579 10000 : ctx->frame_size;
580 return FFStream::encode_activate();
583 int FFAudioStream::nb_samples()
585 AVCodecContext *ctx = st->codec;
586 return ctx->codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE ?
587 10000 : ctx->frame_size;
590 int64_t FFAudioStream::load_buffer(double ** const sp, int len)
592 reserve(len+1, st->codec->channels);
593 for( int ch=0; ch<nch; ++ch )
594 write(sp[ch], len, ch);
598 int FFAudioStream::in_history(int64_t pos)
600 if( pos > curr_pos ) return 0;
602 if( len > sz ) len = sz;
603 if( pos < curr_pos - len ) return 0;
608 int FFAudioStream::init_frame(AVFrame *frame)
610 AVCodecContext *ctx = st->codec;
611 frame->nb_samples = frame_sz;
612 frame->format = ctx->sample_fmt;
613 frame->channel_layout = ctx->channel_layout;
614 frame->sample_rate = ctx->sample_rate;
615 int ret = av_frame_get_buffer(frame, 0);
617 ff_err(ret, "FFAudioStream::init_frame: av_frame_get_buffer failed\n");
621 int FFAudioStream::load(int64_t pos, int len)
623 if( audio_seek(pos) < 0 ) return -1;
624 if( !frame && !(frame=av_frame_alloc()) ) {
625 fprintf(stderr, "FFAudioStream::load: av_frame_alloc failed\n");
628 if( mbsz < len ) mbsz = len;
629 int64_t end_pos = pos + len;
631 for( int i=0; ret>=0 && !flushed && curr_pos<end_pos && i<1000; ++i ) {
632 ret = read_frame(frame);
634 load_history(&frame->extended_data[0], frame->nb_samples);
635 curr_pos += frame->nb_samples;
638 if( end_pos > curr_pos ) {
639 zero(end_pos - curr_pos);
642 len = curr_pos - pos;
647 int FFAudioStream::audio_seek(int64_t pos)
649 if( decode_activate() < 0 ) return -1;
650 if( !st->codec || !st->codec->codec ) return -1;
651 if( in_history(pos) ) return 0;
652 if( pos == curr_pos ) return 0;
653 reset_history(); mbsz = 0;
654 // guarentee preload > 1sec samples
655 if( seek(pos-sample_rate, sample_rate) < 0 ) return -1;
659 int FFAudioStream::encode(double **samples, int len)
661 if( encode_activate() <= 0 ) return -1;
664 int64_t count = load_buffer(samples, len);
667 while( ret >= 0 && count >= frame_sz ) {
668 frm = new FFrame(this);
669 if( (ret=frm->initted()) < 0 ) break;
670 AVFrame *frame = *frm;
671 float *bfrp = get_outp(frame_sz);
672 ret = swr_convert(resample_context,
673 (uint8_t **)frame->extended_data, frame_sz,
674 (const uint8_t **)&bfrp, frame_sz);
676 ff_err(ret, "FFAudioStream::encode: swr_convert failed\n");
679 frm->queue(curr_pos);
681 curr_pos += frame_sz;
686 return ret >= 0 ? 0 : 1;
689 int FFAudioStream::encode_frame(AVPacket *pkt, AVFrame *frame, int &got_packet)
691 int ret = avcodec_encode_audio2(st->codec, pkt, frame, &got_packet);
693 ff_err(ret, "FFAudioStream::encode_frame: encode audio failed\n");
699 void FFAudioStream::load_markers()
701 IndexState *index_state = ffmpeg->file_base->asset->index_state;
702 if( !index_state || idx >= index_state->audio_markers.size() ) return;
703 FFStream::load_markers(*index_state->audio_markers[idx], sample_rate);
706 FFVideoStream::FFVideoStream(FFMPEG *ffmpeg, AVStream *strm, int idx, int fidx)
707 : FFStream(ffmpeg, strm, fidx)
716 FFVideoStream::~FFVideoStream()
720 int FFVideoStream::decode_frame(AVPacket *pkt, AVFrame *frame, int &got_frame)
722 int first_frame = seeked; seeked = 0;
723 int ret = avcodec_decode_video2(st->codec, frame, &got_frame, pkt);
725 if( first_frame ) return 0;
726 ff_err(ret, "FFVideoStream::decode_frame: Could not read video frame\n");
730 int64_t pkt_ts = av_frame_get_best_effort_timestamp(frame);
731 if( pkt_ts != AV_NOPTS_VALUE )
732 curr_pos = ffmpeg->to_secs(pkt_ts - nudge, st->time_base) * frame_rate + 0.5;
737 int FFVideoStream::load(VFrame *vframe, int64_t pos)
739 int ret = video_seek(pos);
740 if( ret < 0 ) return -1;
741 if( !frame && !(frame=av_frame_alloc()) ) {
742 fprintf(stderr, "FFVideoStream::load: av_frame_alloc failed\n");
745 for( int i=0; ret>=0 && !flushed && curr_pos<=pos && i<1000; ++i ) {
746 ret = read_frame(frame);
747 if( ret > 0 ) ++curr_pos;
750 AVCodecContext *ctx = st->codec;
751 ret = convert_cmodel(vframe, (AVPicture *)frame,
752 ctx->pix_fmt, ctx->width, ctx->height);
754 ret = ret > 0 ? 1 : ret < 0 ? -1 : 0;
758 int FFVideoStream::video_seek(int64_t pos)
760 if( decode_activate() < 0 ) return -1;
761 if( !st->codec || !st->codec->codec ) return -1;
762 if( pos == curr_pos-1 && !seeked ) return 0;
763 // if close enough, just read up to current
764 int gop = st->codec->gop_size;
765 if( gop < 4 ) gop = 4;
766 if( gop > 64 ) gop = 64;
767 int read_limit = curr_pos + 3*gop;
768 if( pos >= curr_pos && pos <= read_limit ) return 0;
769 // guarentee preload more than 2*gop frames
770 if( seek(pos - 3*gop, frame_rate) < 0 ) return -1;
774 int FFVideoStream::init_frame(AVFrame *picture)
776 AVCodecContext *ctx = st->codec;
777 picture->format = ctx->pix_fmt;
778 picture->width = ctx->width;
779 picture->height = ctx->height;
780 int ret = av_frame_get_buffer(picture, 32);
784 int FFVideoStream::encode(VFrame *vframe)
786 if( encode_activate() <= 0 ) return -1;
788 FFrame *picture = new FFrame(this);
789 int ret = picture->initted();
791 AVFrame *frame = *picture;
792 frame->pts = curr_pos;
793 AVCodecContext *ctx = st->codec;
794 ret = convert_pixfmt(vframe, (AVPicture*)frame,
795 ctx->pix_fmt, ctx->width, ctx->height);
798 picture->queue(curr_pos);
802 fprintf(stderr, "FFVideoStream::encode: encode failed\n");
805 return ret >= 0 ? 0 : 1;
808 int FFVideoStream::encode_frame(AVPacket *pkt, AVFrame *frame, int &got_packet)
810 int ret = avcodec_encode_video2(st->codec, pkt, frame, &got_packet);
812 ff_err(ret, "FFVideoStream::encode_frame: encode video failed\n");
818 PixelFormat FFVideoConvert::color_model_to_pix_fmt(int color_model)
820 switch( color_model ) {
821 case BC_YUV422: return AV_PIX_FMT_YUYV422;
822 case BC_RGB888: return AV_PIX_FMT_RGB24;
823 case BC_RGBA8888: return AV_PIX_FMT_RGBA;
824 case BC_BGR8888: return AV_PIX_FMT_BGR0;
825 case BC_BGR888: return AV_PIX_FMT_BGR24;
826 case BC_YUV420P: return AV_PIX_FMT_YUV420P;
827 case BC_YUV422P: return AV_PIX_FMT_YUV422P;
828 case BC_YUV444P: return AV_PIX_FMT_YUV444P;
829 case BC_YUV411P: return AV_PIX_FMT_YUV411P;
830 case BC_RGB565: return AV_PIX_FMT_RGB565;
831 case BC_RGB161616: return AV_PIX_FMT_RGB48LE;
832 case BC_RGBA16161616: return AV_PIX_FMT_RGBA64LE;
836 return AV_PIX_FMT_NB;
839 int FFVideoConvert::pix_fmt_to_color_model(PixelFormat pix_fmt)
842 case AV_PIX_FMT_YUYV422: return BC_YUV422;
843 case AV_PIX_FMT_RGB24: return BC_RGB888;
844 case AV_PIX_FMT_RGBA: return BC_RGBA8888;
845 case AV_PIX_FMT_BGR0: return BC_BGR8888;
846 case AV_PIX_FMT_BGR24: return BC_BGR888;
847 case AV_PIX_FMT_YUV420P: return BC_YUV420P;
848 case AV_PIX_FMT_YUV422P: return BC_YUV422P;
849 case AV_PIX_FMT_YUV444P: return BC_YUV444P;
850 case AV_PIX_FMT_YUV411P: return BC_YUV411P;
851 case AV_PIX_FMT_RGB565: return BC_RGB565;
852 case AV_PIX_FMT_RGB48LE: return BC_RGB161616;
853 case AV_PIX_FMT_RGBA64LE: return BC_RGBA16161616;
857 return BC_TRANSPARENCY;
860 int FFVideoConvert::convert_picture_vframe(VFrame *frame,
861 AVPicture *ip, PixelFormat ifmt, int iw, int ih)
864 int cmodel = frame->get_color_model();
865 PixelFormat ofmt = color_model_to_pix_fmt(cmodel);
866 if( ofmt == AV_PIX_FMT_NB ) return -1;
867 int size = avpicture_fill(&opic, frame->get_data(), ofmt,
868 frame->get_w(), frame->get_h());
869 if( size < 0 ) return -1;
871 // transfer line sizes must match also
872 int planar = BC_CModels::is_planar(cmodel);
873 int packed_width = !planar ? frame->get_bytes_per_line() :
874 BC_CModels::calculate_pixelsize(cmodel) * frame->get_w();
875 if( packed_width != opic.linesize[0] ) return -1;
878 // override avpicture_fill() for planar types
879 opic.data[0] = frame->get_y();
880 opic.data[1] = frame->get_u();
881 opic.data[2] = frame->get_v();
884 convert_ctx = sws_getCachedContext(convert_ctx, iw, ih, ifmt,
885 frame->get_w(), frame->get_h(), ofmt, SWS_BICUBIC, NULL, NULL, NULL);
887 fprintf(stderr, "FFVideoConvert::convert_picture_frame:"
888 " sws_getCachedContext() failed\n");
891 int ret = sws_scale(convert_ctx, ip->data, ip->linesize, 0, ih,
892 opic.data, opic.linesize);
894 ff_err(ret, "FFVideoConvert::convert_picture_frame: sws_scale() failed\n");
900 int FFVideoConvert::convert_cmodel(VFrame *frame,
901 AVPicture *ip, PixelFormat ifmt, int iw, int ih)
903 // try direct transfer
904 if( !convert_picture_vframe(frame, ip, ifmt, iw, ih) ) return 1;
905 // use indirect transfer
906 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(ifmt);
908 for( int i = 0; i <desc->nb_components; ++i ) {
909 int bits = desc->comp[i].depth_minus1 + 1;
910 if( bits > max_bits ) max_bits = bits;
912 // from libavcodec/pixdesc.c
913 #define pixdesc_has_alpha(pixdesc) ((pixdesc)->nb_components == 2 || \
914 (pixdesc)->nb_components == 4 || (pixdesc)->flags & AV_PIX_FMT_FLAG_PAL)
915 int icolor_model = pixdesc_has_alpha(desc) ?
916 (max_bits > 8 ? BC_RGBA16161616 : BC_RGBA8888) :
917 (max_bits > 8 ? BC_RGB161616 : BC_RGB888) ;
918 VFrame vframe(iw, ih, icolor_model);
919 if( convert_picture_vframe(&vframe, ip, ifmt, iw, ih) ) return -1;
920 frame->transfer_from(&vframe);
924 int FFVideoConvert::transfer_cmodel(VFrame *frame,
925 AVFrame *ifp, PixelFormat ifmt, int iw, int ih)
927 int ret = convert_cmodel(frame, (AVPicture *)ifp, ifmt, iw, ih);
929 const AVDictionary *src = av_frame_get_metadata(ifp);
930 AVDictionaryEntry *t = NULL;
931 BC_Hash *hp = frame->get_params();
933 while( (t=av_dict_get(src, "", t, AV_DICT_IGNORE_SUFFIX)) )
934 hp->update(t->key, t->value);
939 int FFVideoConvert::convert_vframe_picture(VFrame *frame,
940 AVPicture *op, PixelFormat ofmt, int ow, int oh)
943 int cmodel = frame->get_color_model();
944 PixelFormat ifmt = color_model_to_pix_fmt(cmodel);
945 if( ifmt == AV_PIX_FMT_NB ) return -1;
946 int size = avpicture_fill(&opic, frame->get_data(), ifmt,
947 frame->get_w(), frame->get_h());
948 if( size < 0 ) return -1;
950 // transfer line sizes must match also
951 int planar = BC_CModels::is_planar(cmodel);
952 int packed_width = !planar ? frame->get_bytes_per_line() :
953 BC_CModels::calculate_pixelsize(cmodel) * frame->get_w();
954 if( packed_width != opic.linesize[0] ) return -1;
957 // override avpicture_fill() for planar types
958 opic.data[0] = frame->get_y();
959 opic.data[1] = frame->get_u();
960 opic.data[2] = frame->get_v();
963 convert_ctx = sws_getCachedContext(convert_ctx, frame->get_w(), frame->get_h(), ifmt,
964 ow, oh, ofmt, SWS_BICUBIC, NULL, NULL, NULL);
966 fprintf(stderr, "FFVideoConvert::convert_frame_picture:"
967 " sws_getCachedContext() failed\n");
970 int ret = sws_scale(convert_ctx, opic.data, opic.linesize, 0, frame->get_h(),
971 op->data, op->linesize);
973 ff_err(ret, "FFVideoConvert::convert_frame_picture: sws_scale() failed\n");
979 int FFVideoConvert::convert_pixfmt(VFrame *frame,
980 AVPicture *op, PixelFormat ofmt, int ow, int oh)
982 // try direct transfer
983 if( !convert_vframe_picture(frame, op, ofmt, ow, oh) ) return 1;
984 // use indirect transfer
985 int colormodel = frame->get_color_model();
986 int bits = BC_CModels::calculate_pixelsize(colormodel) * 8;
987 bits /= BC_CModels::components(colormodel);
988 int icolor_model = BC_CModels::has_alpha(colormodel) ?
989 (bits > 8 ? BC_RGBA16161616 : BC_RGBA8888) :
990 (bits > 8 ? BC_RGB161616: BC_RGB888) ;
991 VFrame vframe(frame->get_w(), frame->get_h(), icolor_model);
992 vframe.transfer_from(frame);
993 if( !convert_vframe_picture(&vframe, op, ofmt, ow, oh) ) return 1;
997 int FFVideoConvert::transfer_pixfmt(VFrame *frame,
998 AVFrame *ofp, PixelFormat ofmt, int ow, int oh)
1000 int ret = convert_pixfmt(frame, (AVPicture *)ofp, ofmt, ow, oh);
1002 BC_Hash *hp = frame->get_params();
1003 AVDictionary **dict = avpriv_frame_get_metadatap(ofp);
1004 //av_dict_free(dict);
1005 for( int i=0; i<hp->size(); ++i ) {
1006 char *key = hp->get_key(i), *val = hp->get_value(i);
1007 av_dict_set(dict, key, val, 0);
1013 void FFVideoStream::load_markers()
1015 IndexState *index_state = ffmpeg->file_base->asset->index_state;
1016 if( idx >= index_state->video_markers.size() ) return;
1017 FFStream::load_markers(*index_state->video_markers[idx], frame_rate);
1021 FFMPEG::FFMPEG(FileBase *file_base)
1024 this->file_base = file_base;
1025 memset(file_format,0,sizeof(file_format));
1026 mux_lock = new Condition(0,"FFMPEG::mux_lock",0);
1027 flow_lock = new Condition(1,"FFStream::flow_lock",0);
1030 decoding = encoding = 0;
1031 has_audio = has_video = 0;
1034 opt_video_filter = 0;
1035 opt_audio_filter = 0;
1036 char option_path[BCTEXTLEN];
1037 set_option_path(option_path, "%s", "ffmpeg.opts");
1038 read_options(option_path, opts);
1043 ff_lock("FFMPEG::~FFMPEG()");
1045 ffaudio.remove_all_objects();
1046 ffvideo.remove_all_objects();
1047 if( fmt_ctx ) avformat_close_input(&fmt_ctx);
1051 av_dict_free(&opts);
1052 delete [] opt_video_filter;
1053 delete [] opt_audio_filter;
1056 int FFMPEG::check_sample_rate(AVCodec *codec, int sample_rate)
1058 const int *p = codec->supported_samplerates;
1059 if( !p ) return sample_rate;
1061 if( *p == sample_rate ) return *p;
1067 static inline AVRational std_frame_rate(int i)
1069 static const int m1 = 1001*12, m2 = 1000*12;
1070 static const int freqs[] = {
1071 40*m1, 48*m1, 50*m1, 60*m1, 80*m1,120*m1, 240*m1,
1072 24*m2, 30*m2, 60*m2, 12*m2, 15*m2, 48*m2, 0,
1074 int freq = i<30*12 ? (i+1)*1001 : freqs[i-30*12];
1075 return (AVRational) { freq, 1001*12 };
1078 AVRational FFMPEG::check_frame_rate(AVCodec *codec, double frame_rate)
1080 const AVRational *p = codec->supported_framerates;
1081 AVRational rate, best_rate = (AVRational) { 0, 0 };
1082 double max_err = 1.; int i = 0;
1083 while( ((p ? (rate=*p++) : (rate=std_frame_rate(i++))), rate.num) != 0 ) {
1084 double framerate = (double) rate.num / rate.den;
1085 double err = fabs(frame_rate/framerate - 1.);
1086 if( err >= max_err ) continue;
1090 return max_err < 0.0001 ? best_rate : (AVRational) { 0, 0 };
1093 AVRational FFMPEG::to_sample_aspect_ratio(Asset *asset)
1096 double display_aspect = asset->width / (double)asset->height;
1097 double sample_aspect = asset->aspect_ratio / display_aspect;
1098 int width = 1000000, height = width * sample_aspect + 0.5;
1100 MWindow::create_aspect_ratio(w, h, width, height);
1101 return (AVRational){(int)h, (int)w};
1104 return (AVRational){1, 1};
1108 AVRational FFMPEG::to_time_base(int sample_rate)
1110 return (AVRational){1, sample_rate};
1113 void FFMPEG::set_option_path(char *path, const char *fmt, ...)
1116 strcat(path, "/ffmpeg/");
1117 path += strlen(path);
1120 vsprintf(path, fmt, ap);
1124 void FFMPEG::get_option_path(char *path, const char *type, const char *spec)
1129 set_option_path(path, "%s/%s", type, spec);
1132 int FFMPEG::get_format(char *format, const char *path, char *spec)
1134 char option_path[BCTEXTLEN], line[BCTEXTLEN], codec[BCTEXTLEN];
1135 get_option_path(option_path, path, spec);
1136 FILE *fp = fopen(option_path,"r");
1139 if( !fgets(line, sizeof(line), fp) ) ret = 1;
1141 line[sizeof(line)-1] = 0;
1142 ret = scan_option_line(line, format, codec);
1148 int FFMPEG::get_file_format()
1151 char audio_format[BCSTRLEN], video_format[BCSTRLEN];
1152 file_format[0] = audio_format[0] = video_format[0] = 0;
1153 Asset *asset = file_base->asset;
1154 if( !ret && asset->audio_data )
1155 ret = get_format(audio_format, "audio", asset->acodec);
1156 if( !ret && asset->video_data )
1157 ret = get_format(video_format, "video", asset->vcodec);
1158 if( !ret && !audio_format[0] && !video_format[0] )
1160 if( !ret && audio_format[0] && video_format[0] &&
1161 strcmp(audio_format, video_format) ) ret = -1;
1163 strcpy(file_format, audio_format[0] ? audio_format : video_format);
1167 int FFMPEG::scan_option_line(char *cp, char *tag, char *val)
1169 while( *cp == ' ' || *cp == '\t' ) ++cp;
1171 while( *cp && *cp != ' ' && *cp != '\t' && *cp != '=' ) ++cp;
1173 if( !len || len > BCSTRLEN-1 ) return 1;
1174 while( bp < cp ) *tag++ = *bp++;
1176 while( *cp == ' ' || *cp == '\t' ) ++cp;
1177 if( *cp == '=' ) ++cp;
1178 while( *cp == ' ' || *cp == '\t' ) ++cp;
1180 while( *cp && *cp != '\n' ) ++cp;
1182 if( len > BCTEXTLEN-1 ) return 1;
1183 while( bp < cp ) *val++ = *bp++;
1188 int FFMPEG::get_encoder(const char *options,
1189 char *format, char *codec, char *bsfilter, char *bsargs)
1191 FILE *fp = fopen(options,"r");
1193 eprintf("FFMPEG::get_encoder: options open failed %s\n",options);
1196 if( get_encoder(fp, format, codec, bsfilter, bsargs) )
1197 eprintf(_("FFMPEG::get_encoder:"
1198 " err: format/codec not found %s\n"), options);
1203 int FFMPEG::get_encoder(FILE *fp,
1204 char *format, char *codec, char *bsfilter, char *bsargs)
1206 format[0] = codec[0] = bsfilter[0] = bsargs[0] = 0;
1207 char line[BCTEXTLEN];
1208 if( !fgets(line, sizeof(line), fp) ) return 1;
1209 line[sizeof(line)-1] = 0;
1210 if( scan_option_line(line, format, codec) ) return 1;
1212 while( *cp && *cp != '|' ) ++cp;
1213 if( !*cp ) return 0;
1214 if( scan_option_line(cp+1, bsfilter, bsargs) ) return 1;
1215 do { *cp-- = 0; } while( cp>=codec && (*cp==' ' || *cp == '\t' ) );
1219 int FFMPEG::read_options(const char *options, AVDictionary *&opts)
1221 FILE *fp = fopen(options,"r");
1223 int ret = read_options(fp, options, opts);
1228 int FFMPEG::scan_options(const char *options, AVDictionary *&opts, AVStream *st)
1230 FILE *fp = fmemopen((void *)options,strlen(options),"r");
1232 int ret = read_options(fp, options, opts);
1234 AVDictionaryEntry *tag = av_dict_get(opts, "id", NULL, 0);
1235 if( tag ) st->id = strtol(tag->value,0,0);
1239 int FFMPEG::read_options(FILE *fp, const char *options, AVDictionary *&opts)
1241 int ret = 0, no = 0;
1242 char line[BCTEXTLEN];
1243 while( !ret && fgets(line, sizeof(line), fp) ) {
1244 line[sizeof(line)-1] = 0;
1246 if( line[0] == '#' ) continue;
1247 if( line[0] == '\n' ) continue;
1248 char key[BCSTRLEN], val[BCTEXTLEN];
1249 if( scan_option_line(line, key, val) ) {
1250 eprintf(_("FFMPEG::read_options:"
1251 " err reading %s: line %d\n"), options, no);
1255 if( !strcmp(key, "duration") )
1256 opt_duration = strtod(val, 0);
1257 if( !strcmp(key, "video_filter") )
1258 opt_video_filter = cstrdup(val);
1259 if( !strcmp(key, "audio_filter") )
1260 opt_audio_filter = cstrdup(val);
1261 else if( !strcmp(key, "loglevel") )
1264 av_dict_set(&opts, key, val, 0);
1270 int FFMPEG::load_options(const char *options, AVDictionary *&opts)
1272 char option_path[BCTEXTLEN];
1273 set_option_path(option_path, "%s", options);
1274 return read_options(option_path, opts);
1277 int FFMPEG::load_options(const char *path, char *bfr, int len)
1280 FILE *fp = fopen(path, "r");
1282 fgets(bfr, len, fp); // skip hdr
1283 len = fread(bfr, 1, len-1, fp);
1284 if( len < 0 ) len = 0;
1290 void FFMPEG::set_loglevel(const char *ap)
1292 if( !ap || !*ap ) return;
1297 { "quiet" , AV_LOG_QUIET },
1298 { "panic" , AV_LOG_PANIC },
1299 { "fatal" , AV_LOG_FATAL },
1300 { "error" , AV_LOG_ERROR },
1301 { "warning", AV_LOG_WARNING },
1302 { "info" , AV_LOG_INFO },
1303 { "verbose", AV_LOG_VERBOSE },
1304 { "debug" , AV_LOG_DEBUG },
1306 for( int i=0; i<(int)(sizeof(log_levels)/sizeof(log_levels[0])); ++i ) {
1307 if( !strcmp(log_levels[i].name, ap) ) {
1308 av_log_set_level(log_levels[i].level);
1312 av_log_set_level(atoi(ap));
1315 double FFMPEG::to_secs(int64_t time, AVRational time_base)
1317 double base_time = time == AV_NOPTS_VALUE ? 0 :
1318 av_rescale_q(time, time_base, AV_TIME_BASE_Q);
1319 return base_time / AV_TIME_BASE;
1322 int FFMPEG::info(char *text, int len)
1324 if( len <= 0 ) return 0;
1326 #define report(s...) do { int n = snprintf(cp,len,s); cp += n; len -= n; } while(0)
1328 if( ffvideo.size() > 0 )
1329 report("\n%d video stream%s\n",ffvideo.size(), ffvideo.size()!=1 ? "s" : "");
1330 for( int vidx=0; vidx<ffvideo.size(); ++vidx ) {
1331 FFVideoStream *vid = ffvideo[vidx];
1332 AVStream *st = vid->st;
1333 AVCodecContext *avctx = st->codec;
1334 report(_("vid%d (%d), id 0x%06x:\n"), vid->idx, vid->fidx, avctx->codec_id);
1335 const AVCodecDescriptor *desc = avcodec_descriptor_get(avctx->codec_id);
1336 report(" video%d %s", vidx+1, desc ? desc->name : " (unkn)");
1337 report(" %dx%d %5.2f", vid->width, vid->height, vid->frame_rate);
1338 const char *pfn = av_get_pix_fmt_name(avctx->pix_fmt);
1339 report(" pix %s\n", pfn ? pfn : "(unkn)");
1340 double secs = to_secs(st->duration, st->time_base);
1341 int64_t length = secs * vid->frame_rate + 0.5;
1342 double ofs = to_secs((vid->nudge - st->start_time), st->time_base);
1343 int64_t nudge = ofs * vid->frame_rate;
1344 int ch = nudge >= 0 ? '+' : (nudge=-nudge, '-');
1345 report(" %jd%c%jd frms %0.2f secs", length,ch,nudge, secs);
1346 int hrs = secs/3600; secs -= hrs*3600;
1347 int mins = secs/60; secs -= mins*60;
1348 report(" %d:%02d:%05.2f\n", hrs, mins, secs);
1350 if( ffaudio.size() > 0 )
1351 report("\n%d audio stream%s\n",ffaudio.size(), ffaudio.size()!=1 ? "s" : "");
1352 for( int aidx=0; aidx<ffaudio.size(); ++aidx ) {
1353 FFAudioStream *aud = ffaudio[aidx];
1354 AVStream *st = aud->st;
1355 AVCodecContext *avctx = st->codec;
1356 report(_("aud%d (%d), id 0x%06x:\n"), aud->idx, aud->fidx, avctx->codec_id);
1357 const AVCodecDescriptor *desc = avcodec_descriptor_get(avctx->codec_id);
1358 int nch = aud->channels, ch0 = aud->channel0+1;
1359 report(" audio%d-%d %s", ch0, ch0+nch-1, desc ? desc->name : " (unkn)");
1360 const char *fmt = av_get_sample_fmt_name(avctx->sample_fmt);
1361 report(" %s %d", fmt, aud->sample_rate);
1362 int sample_bits = av_get_bits_per_sample(avctx->codec_id);
1363 report(" %dbits\n", sample_bits);
1364 double secs = to_secs(st->duration, st->time_base);
1365 int64_t length = secs * aud->sample_rate + 0.5;
1366 double ofs = to_secs((aud->nudge - st->start_time), st->time_base);
1367 int64_t nudge = ofs * aud->sample_rate;
1368 int ch = nudge >= 0 ? '+' : (nudge=-nudge, '-');
1369 report(" %jd%c%jd smpl %0.2f secs", length,ch,nudge, secs);
1370 int hrs = secs/3600; secs -= hrs*3600;
1371 int mins = secs/60; secs -= mins*60;
1372 report(" %d:%02d:%05.2f\n", hrs, mins, secs);
1374 if( fmt_ctx->nb_programs > 0 )
1375 report("\n%d program%s\n",fmt_ctx->nb_programs, fmt_ctx->nb_programs!=1 ? "s" : "");
1376 for( int i=0; i<(int)fmt_ctx->nb_programs; ++i ) {
1377 report("program %d", i+1);
1378 AVProgram *pgrm = fmt_ctx->programs[i];
1379 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
1380 int idx = pgrm->stream_index[j];
1381 int vidx = ffvideo.size();
1382 while( --vidx>=0 && ffvideo[vidx]->fidx != idx );
1384 report(", vid%d", vidx);
1387 int aidx = ffaudio.size();
1388 while( --aidx>=0 && ffaudio[aidx]->fidx != idx );
1390 report(", aud%d", aidx);
1393 report(", (%d)", pgrm->stream_index[j]);
1398 AVDictionaryEntry *tag = 0;
1399 while ((tag = av_dict_get(fmt_ctx->metadata, "", tag, AV_DICT_IGNORE_SUFFIX)))
1400 report("%s=%s\n", tag->key, tag->value);
1409 int FFMPEG::init_decoder(const char *filename)
1411 ff_lock("FFMPEG::init_decoder");
1413 char file_opts[BCTEXTLEN];
1414 char *bp = strrchr(strcpy(file_opts, filename), '/');
1415 char *sp = strrchr(!bp ? file_opts : bp, '.');
1418 strcpy(sp, ".opts");
1419 fp = fopen(file_opts, "r");
1422 read_options(fp, file_opts, opts);
1426 load_options("decode.opts", opts);
1427 AVDictionary *fopts = 0;
1428 av_dict_copy(&fopts, opts, 0);
1429 int ret = avformat_open_input(&fmt_ctx, filename, NULL, &fopts);
1430 av_dict_free(&fopts);
1432 ret = avformat_find_stream_info(fmt_ctx, NULL);
1437 return !ret ? 0 : 1;
1440 int FFMPEG::open_decoder()
1443 if( stat(fmt_ctx->filename, &st) < 0 ) {
1444 eprintf("FFMPEG::open_decoder: can't stat file: %s\n",
1449 int64_t file_bits = 8 * st.st_size;
1450 if( !fmt_ctx->bit_rate && opt_duration > 0 )
1451 fmt_ctx->bit_rate = file_bits / opt_duration;
1454 if( fmt_ctx->bit_rate > 0 ) {
1455 for( int i=0; i<(int)fmt_ctx->nb_streams; ++i ) {
1456 AVStream *st = fmt_ctx->streams[i];
1457 if( st->duration != AV_NOPTS_VALUE ) continue;
1458 if( st->time_base.num > INT64_MAX / fmt_ctx->bit_rate ) continue;
1459 st->duration = av_rescale(file_bits, st->time_base.den,
1460 fmt_ctx->bit_rate * (int64_t) st->time_base.num);
1465 printf("FFMPEG::open_decoder: some stream times estimated\n");
1467 ff_lock("FFMPEG::open_decoder");
1469 for( int i=0; i<(int)fmt_ctx->nb_streams; ++i ) {
1470 AVStream *st = fmt_ctx->streams[i];
1471 if( st->duration == AV_NOPTS_VALUE ) bad_time = 1;
1472 AVCodecContext *avctx = st->codec;
1473 const AVCodecDescriptor *codec_desc = avcodec_descriptor_get(avctx->codec_id);
1474 if( !codec_desc ) continue;
1475 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1476 if( avctx->width < 1 ) continue;
1477 if( avctx->height < 1 ) continue;
1478 AVRational framerate = av_guess_frame_rate(fmt_ctx, st, 0);
1479 if( framerate.num < 1 ) continue;
1481 int vidx = ffvideo.size();
1482 FFVideoStream *vid = new FFVideoStream(this, st, vidx, i);
1483 vstrm_index.append(ffidx(vidx, 0));
1484 ffvideo.append(vid);
1485 vid->width = avctx->width;
1486 vid->height = avctx->height;
1487 vid->frame_rate = !framerate.den ? 0 : (double)framerate.num / framerate.den;
1488 double secs = to_secs(st->duration, st->time_base);
1489 vid->length = secs * vid->frame_rate;
1490 vid->aspect_ratio = (double)st->sample_aspect_ratio.num / st->sample_aspect_ratio.den;
1491 vid->nudge = st->start_time;
1493 if( opt_video_filter )
1494 vid->create_filter(opt_video_filter, avctx,avctx);
1496 else if( avctx->codec_type == AVMEDIA_TYPE_AUDIO ) {
1497 if( avctx->channels < 1 ) continue;
1498 if( avctx->sample_rate < 1 ) continue;
1500 int aidx = ffaudio.size();
1501 FFAudioStream *aud = new FFAudioStream(this, st, aidx, i);
1502 ffaudio.append(aud);
1503 aud->channel0 = astrm_index.size();
1504 aud->channels = avctx->channels;
1505 for( int ch=0; ch<aud->channels; ++ch )
1506 astrm_index.append(ffidx(aidx, ch));
1507 aud->sample_rate = avctx->sample_rate;
1508 double secs = to_secs(st->duration, st->time_base);
1509 aud->length = secs * aud->sample_rate;
1510 if( avctx->sample_fmt != AV_SAMPLE_FMT_FLT ) {
1511 uint64_t layout = av_get_default_channel_layout(avctx->channels);
1512 if( !layout ) layout = ((uint64_t)1<<aud->channels) - 1;
1513 aud->resample_context = swr_alloc_set_opts(NULL,
1514 layout, AV_SAMPLE_FMT_FLT, avctx->sample_rate,
1515 layout, avctx->sample_fmt, avctx->sample_rate,
1517 swr_init(aud->resample_context);
1519 aud->nudge = st->start_time;
1521 if( opt_audio_filter )
1522 aud->create_filter(opt_audio_filter, avctx,avctx);
1526 printf("FFMPEG::open_decoder: some stream have bad times\n");
1532 int FFMPEG::init_encoder(const char *filename)
1534 int fd = ::open(filename,O_WRONLY);
1535 if( fd < 0 ) fd = open(filename,O_WRONLY+O_CREAT,0666);
1537 eprintf("FFMPEG::init_encoder: bad file path: %s\n", filename);
1541 int ret = get_file_format();
1543 eprintf("FFMPEG::init_encoder: bad file format: %s\n", filename);
1547 eprintf("FFMPEG::init_encoder: mismatch audio/video file format: %s\n", filename);
1550 ff_lock("FFMPEG::init_encoder");
1552 avformat_alloc_output_context2(&fmt_ctx, 0, file_format, filename);
1554 eprintf("FFMPEG::init_encoder: failed: %s\n", filename);
1559 load_options("encode.opts", opts);
1566 int FFMPEG::open_encoder(const char *type, const char *spec)
1569 Asset *asset = file_base->asset;
1570 char *filename = asset->path;
1571 AVDictionary *sopts = 0;
1572 av_dict_copy(&sopts, opts, 0);
1573 char option_path[BCTEXTLEN];
1574 set_option_path(option_path, "%s/%s.opts", type, type);
1575 read_options(option_path, sopts);
1576 get_option_path(option_path, type, spec);
1577 char format_name[BCSTRLEN], codec_name[BCTEXTLEN];
1578 char bsfilter[BCSTRLEN], bsargs[BCTEXTLEN];
1579 if( get_encoder(option_path, format_name, codec_name, bsfilter, bsargs) ) {
1580 eprintf("FFMPEG::open_encoder: get_encoder failed %s:%s\n",
1581 option_path, filename);
1585 if( !strcmp(codec_name, CODEC_TAG_DVSD) ) strcpy(codec_name, "dv");
1586 else if( !strcmp(codec_name, CODEC_TAG_MJPEG) ) strcpy(codec_name, "mjpeg");
1587 else if( !strcmp(codec_name, CODEC_TAG_JPEG) ) strcpy(codec_name, "jpeg");
1590 ff_lock("FFMPEG::open_encoder");
1594 const AVCodecDescriptor *codec_desc = 0;
1595 AVCodec *codec = avcodec_find_encoder_by_name(codec_name);
1597 eprintf("FFMPEG::open_encoder: cant find codec %s:%s\n",
1598 codec_name, filename);
1602 codec_desc = avcodec_descriptor_get(codec->id);
1604 eprintf("FFMPEG::open_encoder: unknown codec %s:%s\n",
1605 codec_name, filename);
1610 st = avformat_new_stream(fmt_ctx, 0);
1612 eprintf("FFMPEG::open_encoder: cant create stream %s:%s\n",
1613 codec_name, filename);
1618 AVCodecContext *ctx = st->codec;
1619 switch( codec_desc->type ) {
1620 case AVMEDIA_TYPE_AUDIO: {
1622 eprintf("FFMPEG::open_encoder: duplicate audio %s:%s\n",
1623 codec_name, filename);
1628 if( scan_options(asset->ff_audio_options, sopts, st) ) {
1629 eprintf("FFMPEG::open_encoder: bad audio options %s:%s\n",
1630 codec_name, filename);
1634 if( asset->ff_audio_bitrate > 0 ) {
1635 ctx->bit_rate = asset->ff_audio_bitrate;
1637 sprintf(arg, "%d", asset->ff_audio_bitrate);
1638 av_dict_set(&sopts, "b", arg, 0);
1640 int aidx = ffaudio.size();
1641 int fidx = aidx + ffvideo.size();
1642 FFAudioStream *aud = new FFAudioStream(this, st, aidx, fidx);
1643 ffaudio.append(aud); fst = aud;
1644 aud->sample_rate = asset->sample_rate;
1645 ctx->channels = aud->channels = asset->channels;
1646 for( int ch=0; ch<aud->channels; ++ch )
1647 astrm_index.append(ffidx(aidx, ch));
1648 ctx->channel_layout = av_get_default_channel_layout(ctx->channels);
1649 ctx->sample_rate = check_sample_rate(codec, asset->sample_rate);
1650 if( !ctx->sample_rate ) {
1651 eprintf("FFMPEG::open_audio_encode:"
1652 " check_sample_rate failed %s\n", filename);
1656 ctx->time_base = st->time_base = (AVRational){1, aud->sample_rate};
1657 ctx->sample_fmt = codec->sample_fmts[0];
1658 uint64_t layout = av_get_default_channel_layout(ctx->channels);
1659 aud->resample_context = swr_alloc_set_opts(NULL,
1660 layout, ctx->sample_fmt, aud->sample_rate,
1661 layout, AV_SAMPLE_FMT_FLT, ctx->sample_rate,
1663 swr_init(aud->resample_context);
1666 case AVMEDIA_TYPE_VIDEO: {
1668 eprintf("FFMPEG::open_encoder: duplicate video %s:%s\n",
1669 codec_name, filename);
1674 if( scan_options(asset->ff_video_options, sopts, st) ) {
1675 eprintf("FFMPEG::open_encoder: bad video options %s:%s\n",
1676 codec_name, filename);
1680 if( asset->ff_video_bitrate > 0 ) {
1681 ctx->bit_rate = asset->ff_video_bitrate;
1683 sprintf(arg, "%d", asset->ff_video_bitrate);
1684 av_dict_set(&sopts, "b", arg, 0);
1686 else if( asset->ff_video_quality > 0 ) {
1687 ctx->global_quality = asset->ff_video_quality * FF_QP2LAMBDA;
1688 ctx->qmin = ctx->qmax = asset->ff_video_quality;
1689 ctx->mb_lmin = ctx->qmin * FF_QP2LAMBDA;
1690 ctx->mb_lmax = ctx->qmax * FF_QP2LAMBDA;
1691 ctx->flags |= CODEC_FLAG_QSCALE;
1693 av_dict_set(&sopts, "flags", "+qscale", 0);
1694 sprintf(arg, "%d", asset->ff_video_quality);
1695 av_dict_set(&sopts, "qscale", arg, 0);
1696 sprintf(arg, "%d", ctx->global_quality);
1697 av_dict_set(&sopts, "global_quality", arg, 0);
1699 int vidx = ffvideo.size();
1700 int fidx = vidx + ffaudio.size();
1701 FFVideoStream *vid = new FFVideoStream(this, st, vidx, fidx);
1702 vstrm_index.append(ffidx(vidx, 0));
1703 ffvideo.append(vid); fst = vid;
1704 vid->width = asset->width;
1705 ctx->width = (vid->width+3) & ~3;
1706 vid->height = asset->height;
1707 ctx->height = (vid->height+3) & ~3;
1708 vid->frame_rate = asset->frame_rate;
1709 ctx->sample_aspect_ratio = to_sample_aspect_ratio(asset);
1710 ctx->pix_fmt = codec->pix_fmts ? codec->pix_fmts[0] : AV_PIX_FMT_YUV420P;
1711 AVRational frame_rate = check_frame_rate(codec, vid->frame_rate);
1712 if( !frame_rate.num || !frame_rate.den ) {
1713 eprintf("FFMPEG::open_audio_encode:"
1714 " check_frame_rate failed %s\n", filename);
1718 ctx->time_base = (AVRational) { frame_rate.den, frame_rate.num };
1719 st->time_base = ctx->time_base;
1723 eprintf("FFMPEG::open_encoder: not audio/video, %s:%s\n",
1724 codec_name, filename);
1729 if( fmt_ctx->oformat->flags & AVFMT_GLOBALHEADER )
1730 st->codec->flags |= CODEC_FLAG_GLOBAL_HEADER;
1732 ret = avcodec_open2(st->codec, codec, &sopts);
1734 ff_err(ret,"FFMPEG::open_encoder");
1735 eprintf("FFMPEG::open_encoder: open failed %s:%s\n",
1736 codec_name, filename);
1743 if( fst && bsfilter[0] )
1744 fst->add_bsfilter(bsfilter, !bsargs[0] ? 0 : bsargs);
1748 av_dict_free(&sopts);
1752 int FFMPEG::close_encoder()
1755 if( encoding > 0 ) {
1756 av_write_trailer(fmt_ctx);
1757 if( !(fmt_ctx->flags & AVFMT_NOFILE) )
1758 avio_closep(&fmt_ctx->pb);
1764 int FFMPEG::decode_activate()
1766 if( decoding < 0 ) {
1768 for( int vidx=0; vidx<ffvideo.size(); ++vidx )
1769 ffvideo[vidx]->nudge = AV_NOPTS_VALUE;
1770 for( int aidx=0; aidx<ffaudio.size(); ++aidx )
1771 ffaudio[aidx]->nudge = AV_NOPTS_VALUE;
1772 // set nudges for each program stream set
1773 int npgrms = fmt_ctx->nb_programs;
1774 for( int i=0; i<npgrms; ++i ) {
1775 AVProgram *pgrm = fmt_ctx->programs[i];
1776 // first start time video stream
1777 int64_t vstart_time = -1, astart_time = -1;
1778 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
1779 int fidx = pgrm->stream_index[j];
1780 AVStream *st = fmt_ctx->streams[fidx];
1781 AVCodecContext *avctx = st->codec;
1782 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1783 if( st->start_time == AV_NOPTS_VALUE ) continue;
1784 if( vstart_time > st->start_time ) continue;
1785 vstart_time = st->start_time;
1788 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1789 if( st->start_time == AV_NOPTS_VALUE ) continue;
1790 if( astart_time > st->start_time ) continue;
1791 astart_time = st->start_time;
1795 // match program streams to max start_time
1796 int64_t nudge = vstart_time > astart_time ? vstart_time : astart_time;
1797 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
1798 int fidx = pgrm->stream_index[j];
1799 AVStream *st = fmt_ctx->streams[fidx];
1800 AVCodecContext *avctx = st->codec;
1801 if( avctx->codec_type == AVMEDIA_TYPE_VIDEO ) {
1802 for( int k=0; k<ffvideo.size(); ++k ) {
1803 if( ffvideo[k]->fidx != fidx ) continue;
1804 ffvideo[k]->nudge = nudge;
1808 if( avctx->codec_type == AVMEDIA_TYPE_AUDIO ) {
1809 for( int k=0; k<ffaudio.size(); ++k ) {
1810 if( ffaudio[k]->fidx != fidx ) continue;
1811 ffaudio[k]->nudge = nudge;
1817 // set nudges for any streams not yet set
1818 int64_t vstart_time = 0, astart_time = 0;
1819 int nstreams = fmt_ctx->nb_streams;
1820 for( int i=0; i<nstreams; ++i ) {
1821 AVStream *st = fmt_ctx->streams[i];
1822 AVCodecContext *avctx = st->codec;
1823 switch( avctx->codec_type ) {
1824 case AVMEDIA_TYPE_VIDEO: {
1825 if( st->start_time == AV_NOPTS_VALUE ) continue;
1826 int vidx = ffvideo.size();
1827 while( --vidx >= 0 && ffvideo[vidx]->fidx != i );
1828 if( vidx >= 0 && ffvideo[vidx]->nudge != AV_NOPTS_VALUE ) continue;
1829 if( vstart_time >= st->start_time ) continue;
1830 vstart_time = st->start_time;
1832 case AVMEDIA_TYPE_AUDIO: {
1833 if( st->start_time == AV_NOPTS_VALUE ) continue;
1834 int aidx = ffaudio.size();
1835 while( --aidx >= 0 && ffaudio[aidx]->fidx != i );
1836 if( aidx >= 0 && ffaudio[aidx]->nudge != AV_NOPTS_VALUE ) continue;
1837 if( astart_time >= st->start_time ) continue;
1838 astart_time = st->start_time;
1843 int64_t nudge = vstart_time > astart_time ? vstart_time : astart_time;
1844 for( int vidx=0; vidx<ffvideo.size(); ++vidx ) {
1845 if( ffvideo[vidx]->nudge != AV_NOPTS_VALUE ) continue;
1846 ffvideo[vidx]->nudge = nudge;
1848 for( int aidx=0; aidx<ffaudio.size(); ++aidx ) {
1849 if( ffaudio[aidx]->nudge != AV_NOPTS_VALUE ) continue;
1850 ffaudio[aidx]->nudge = nudge;
1857 int FFMPEG::encode_activate()
1860 if( encoding < 0 ) {
1862 if( !(fmt_ctx->flags & AVFMT_NOFILE) &&
1863 (ret=avio_open(&fmt_ctx->pb, fmt_ctx->filename, AVIO_FLAG_WRITE)) < 0 ) {
1864 ff_err(ret, "FFMPEG::encode_activate: err opening : %s\n",
1869 AVDictionary *fopts = 0;
1870 char option_path[BCTEXTLEN];
1871 set_option_path(option_path, "format/%s", file_format);
1872 read_options(option_path, fopts);
1873 ret = avformat_write_header(fmt_ctx, &fopts);
1874 av_dict_free(&fopts);
1876 ff_err(ret, "FFMPEG::encode_activate: write header failed %s\n",
1886 int FFMPEG::audio_seek(int stream, int64_t pos)
1888 int aidx = astrm_index[stream].st_idx;
1889 FFAudioStream *aud = ffaudio[aidx];
1890 aud->audio_seek(pos);
1894 int FFMPEG::video_seek(int stream, int64_t pos)
1896 int vidx = vstrm_index[stream].st_idx;
1897 FFVideoStream *vid = ffvideo[vidx];
1898 vid->video_seek(pos);
1903 int FFMPEG::decode(int chn, int64_t pos, double *samples, int len)
1905 if( !has_audio || chn >= astrm_index.size() ) return -1;
1906 int aidx = astrm_index[chn].st_idx;
1907 FFAudioStream *aud = ffaudio[aidx];
1908 if( aud->load(pos, len) < len ) return -1;
1909 int ch = astrm_index[chn].st_ch;
1910 int ret = aud->read(samples,len,ch);
1914 int FFMPEG::decode(int layer, int64_t pos, VFrame *vframe)
1916 if( !has_video || layer >= vstrm_index.size() ) return -1;
1917 int vidx = vstrm_index[layer].st_idx;
1918 FFVideoStream *vid = ffvideo[vidx];
1919 return vid->load(vframe, pos);
1923 int FFMPEG::encode(int stream, double **samples, int len)
1925 FFAudioStream *aud = ffaudio[stream];
1926 return aud->encode(samples, len);
1930 int FFMPEG::encode(int stream, VFrame *frame)
1932 FFVideoStream *vid = ffvideo[stream];
1933 return vid->encode(frame);
1936 void FFMPEG::start_muxer()
1944 void FFMPEG::stop_muxer()
1953 void FFMPEG::flow_off()
1956 flow_lock->lock("FFMPEG::flow_off");
1960 void FFMPEG::flow_on()
1964 flow_lock->unlock();
1967 void FFMPEG::flow_ctl()
1970 flow_lock->lock("FFMPEG::flow_ctl");
1971 flow_lock->unlock();
1975 int FFMPEG::mux_audio(FFrame *frm)
1978 FFStream *fst = frm->fst;
1979 AVCodecContext *ctx = fst->st->codec;
1980 AVFrame *frame = *frm;
1981 AVRational tick_rate = {1, ctx->sample_rate};
1982 frame->pts = av_rescale_q(frm->position, tick_rate, ctx->time_base);
1984 int ret = fst->encode_frame(pkt, frame, got_packet);
1985 if( ret >= 0 && got_packet )
1986 ret = fst->write_packet(pkt);
1988 ff_err(ret, "FFMPEG::mux_audio");
1989 return ret >= 0 ? 0 : 1;
1992 int FFMPEG::mux_video(FFrame *frm)
1995 FFStream *fst = frm->fst;
1996 AVFrame *frame = *frm;
1997 frame->pts = frm->position;
1998 int ret = 1, got_packet = 0;
1999 if( fmt_ctx->oformat->flags & AVFMT_RAWPICTURE ) {
2000 /* a hack to avoid data copy with some raw video muxers */
2001 pkt->flags |= AV_PKT_FLAG_KEY;
2002 pkt->stream_index = fst->st->index;
2003 AVPicture *picture = (AVPicture *)frame;
2004 pkt->data = (uint8_t *)picture;
2005 pkt->size = sizeof(AVPicture);
2006 pkt->pts = pkt->dts = frame->pts;
2010 ret = fst->encode_frame(pkt, frame, got_packet);
2011 if( ret >= 0 && got_packet )
2012 ret = fst->write_packet(pkt);
2014 ff_err(ret, "FFMPEG::mux_video");
2015 return ret >= 0 ? 0 : 1;
2021 double atm = -1, vtm = -1;
2022 FFrame *afrm = 0, *vfrm = 0;
2024 for( int i=0; i<ffaudio.size(); ++i ) { // earliest audio
2025 FFStream *fst = ffaudio[i];
2026 if( fst->frm_count < 3 ) { demand = 1; flow_on(); }
2027 FFrame *frm = fst->frms.first;
2028 if( !frm ) { if( !done ) return; continue; }
2029 double tm = to_secs(frm->position, fst->st->codec->time_base);
2030 if( atm < 0 || tm < atm ) { atm = tm; afrm = frm; }
2032 for( int i=0; i<ffvideo.size(); ++i ) { // earliest video
2033 FFStream *fst = ffvideo[i];
2034 if( fst->frm_count < 2 ) { demand = 1; flow_on(); }
2035 FFrame *frm = fst->frms.first;
2036 if( !frm ) { if( !done ) return; continue; }
2037 double tm = to_secs(frm->position, fst->st->codec->time_base);
2038 if( vtm < 0 || tm < vtm ) { vtm = tm; vfrm = frm; }
2040 if( !demand ) flow_off();
2041 if( !afrm && !vfrm ) break;
2042 int v = !afrm ? -1 : !vfrm ? 1 : av_compare_ts(
2043 vfrm->position, vfrm->fst->st->codec->time_base,
2044 afrm->position, afrm->fst->st->codec->time_base);
2045 FFrame *frm = v <= 0 ? vfrm : afrm;
2046 if( frm == afrm ) mux_audio(frm);
2047 if( frm == vfrm ) mux_video(frm);
2056 mux_lock->lock("FFMPEG::run");
2060 for( int i=0; i<ffaudio.size(); ++i )
2061 ffaudio[i]->flush();
2062 for( int i=0; i<ffvideo.size(); ++i )
2063 ffvideo[i]->flush();
2067 int FFMPEG::ff_total_audio_channels()
2069 return astrm_index.size();
2072 int FFMPEG::ff_total_astreams()
2074 return ffaudio.size();
2077 int FFMPEG::ff_audio_channels(int stream)
2079 return ffaudio[stream]->channels;
2082 int FFMPEG::ff_sample_rate(int stream)
2084 return ffaudio[stream]->sample_rate;
2087 const char* FFMPEG::ff_audio_format(int stream)
2089 AVStream *st = ffaudio[stream]->st;
2090 AVCodecID id = st->codec->codec_id;
2091 const AVCodecDescriptor *desc = avcodec_descriptor_get(id);
2092 return desc ? desc->name : "Unknown";
2095 int FFMPEG::ff_audio_pid(int stream)
2097 return ffaudio[stream]->st->id;
2100 int64_t FFMPEG::ff_audio_samples(int stream)
2102 return ffaudio[stream]->length;
2105 // find audio astream/channels with this program,
2106 // or all program audio channels (astream=-1)
2107 int FFMPEG::ff_audio_for_video(int vstream, int astream, int64_t &channel_mask)
2111 int vidx = ffvideo[vstream]->fidx;
2112 // find first program with this video stream
2113 for( int i=0; pidx<0 && i<(int)fmt_ctx->nb_programs; ++i ) {
2114 AVProgram *pgrm = fmt_ctx->programs[i];
2115 for( int j=0; pidx<0 && j<(int)pgrm->nb_stream_indexes; ++j ) {
2116 int st_idx = pgrm->stream_index[j];
2117 AVStream *st = fmt_ctx->streams[st_idx];
2118 if( st->codec->codec_type != AVMEDIA_TYPE_VIDEO ) continue;
2119 if( st_idx == vidx ) pidx = i;
2122 if( pidx < 0 ) return -1;
2124 int64_t channels = 0;
2125 AVProgram *pgrm = fmt_ctx->programs[pidx];
2126 for( int j=0; j<(int)pgrm->nb_stream_indexes; ++j ) {
2127 int aidx = pgrm->stream_index[j];
2128 AVStream *st = fmt_ctx->streams[aidx];
2129 if( st->codec->codec_type != AVMEDIA_TYPE_AUDIO ) continue;
2130 if( astream > 0 ) { --astream; continue; }
2132 for( int i=0; astrm<0 && i<ffaudio.size(); ++i )
2133 if( ffaudio[i]->fidx == aidx ) astrm = i;
2135 if( ret < 0 ) ret = astrm;
2136 int64_t mask = (1 << ffaudio[astrm]->channels) - 1;
2137 channels |= mask << ffaudio[astrm]->channel0;
2139 if( !astream ) break;
2141 channel_mask = channels;
2146 int FFMPEG::ff_total_video_layers()
2148 return vstrm_index.size();
2151 int FFMPEG::ff_total_vstreams()
2153 return ffvideo.size();
2156 int FFMPEG::ff_video_width(int stream)
2158 return ffvideo[stream]->width;
2161 int FFMPEG::ff_video_height(int stream)
2163 return ffvideo[stream]->height;
2166 int FFMPEG::ff_set_video_width(int stream, int width)
2168 int w = ffvideo[stream]->width;
2169 ffvideo[stream]->width = width;
2173 int FFMPEG::ff_set_video_height(int stream, int height)
2175 int h = ffvideo[stream]->height;
2176 ffvideo[stream]->height = height;
2180 int FFMPEG::ff_coded_width(int stream)
2182 AVStream *st = ffvideo[stream]->st;
2183 return st->codec->coded_width;
2186 int FFMPEG::ff_coded_height(int stream)
2188 AVStream *st = ffvideo[stream]->st;
2189 return st->codec->coded_height;
2192 float FFMPEG::ff_aspect_ratio(int stream)
2194 return ffvideo[stream]->aspect_ratio;
2197 const char* FFMPEG::ff_video_format(int stream)
2199 AVStream *st = ffvideo[stream]->st;
2200 AVCodecID id = st->codec->codec_id;
2201 const AVCodecDescriptor *desc = avcodec_descriptor_get(id);
2202 return desc ? desc->name : "Unknown";
2205 double FFMPEG::ff_frame_rate(int stream)
2207 return ffvideo[stream]->frame_rate;
2210 int64_t FFMPEG::ff_video_frames(int stream)
2212 return ffvideo[stream]->length;
2215 int FFMPEG::ff_video_pid(int stream)
2217 return ffvideo[stream]->st->id;
2221 int FFMPEG::ff_cpus()
2223 return file_base->file->cpus;
2226 int FFVideoStream::create_filter(const char *filter_spec,
2227 AVCodecContext *src_ctx, AVCodecContext *sink_ctx)
2229 avfilter_register_all();
2230 AVFilter *filter = avfilter_get_by_name(filter_spec);
2231 if( !filter || filter->inputs->type != AVMEDIA_TYPE_VIDEO ) {
2232 ff_err(AVERROR(EINVAL), "FFVideoStream::create_filter: %s\n", filter_spec);
2235 filter_graph = avfilter_graph_alloc();
2236 AVFilter *buffersrc = avfilter_get_by_name("buffer");
2237 AVFilter *buffersink = avfilter_get_by_name("buffersink");
2239 int ret = 0; char args[BCTEXTLEN];
2240 snprintf(args, sizeof(args),
2241 "video_size=%dx%d:pix_fmt=%d:time_base=%d/%d:pixel_aspect=%d/%d",
2242 src_ctx->width, src_ctx->height, src_ctx->pix_fmt,
2243 src_ctx->time_base.num, src_ctx->time_base.den,
2244 src_ctx->sample_aspect_ratio.num, src_ctx->sample_aspect_ratio.den);
2246 ret = avfilter_graph_create_filter(&buffersrc_ctx, buffersrc, "in",
2247 args, NULL, filter_graph);
2249 ret = avfilter_graph_create_filter(&buffersink_ctx, buffersink, "out",
2250 NULL, NULL, filter_graph);
2252 ret = av_opt_set_bin(buffersink_ctx, "pix_fmts",
2253 (uint8_t*)&sink_ctx->pix_fmt, sizeof(sink_ctx->pix_fmt),
2254 AV_OPT_SEARCH_CHILDREN);
2256 ff_err(ret, "FFVideoStream::create_filter");
2258 ret = FFStream::create_filter(filter_spec);
2259 return ret >= 0 ? 0 : 1;
2262 int FFAudioStream::create_filter(const char *filter_spec,
2263 AVCodecContext *src_ctx, AVCodecContext *sink_ctx)
2265 avfilter_register_all();
2266 AVFilter *filter = avfilter_get_by_name(filter_spec);
2267 if( !filter || filter->inputs->type != AVMEDIA_TYPE_AUDIO ) {
2268 ff_err(AVERROR(EINVAL), "FFAudioStream::create_filter: %s\n", filter_spec);
2271 filter_graph = avfilter_graph_alloc();
2272 AVFilter *buffersrc = avfilter_get_by_name("abuffer");
2273 AVFilter *buffersink = avfilter_get_by_name("abuffersink");
2274 int ret = 0; char args[BCTEXTLEN];
2275 snprintf(args, sizeof(args),
2276 "time_base=%d/%d:sample_rate=%d:sample_fmt=%s:channel_layout=0x%jx",
2277 src_ctx->time_base.num, src_ctx->time_base.den, src_ctx->sample_rate,
2278 av_get_sample_fmt_name(src_ctx->sample_fmt), src_ctx->channel_layout);
2280 ret = avfilter_graph_create_filter(&buffersrc_ctx, buffersrc, "in",
2281 args, NULL, filter_graph);
2283 ret = avfilter_graph_create_filter(&buffersink_ctx, buffersink, "out",
2284 NULL, NULL, filter_graph);
2286 ret = av_opt_set_bin(buffersink_ctx, "sample_fmts",
2287 (uint8_t*)&sink_ctx->sample_fmt, sizeof(sink_ctx->sample_fmt),
2288 AV_OPT_SEARCH_CHILDREN);
2290 ret = av_opt_set_bin(buffersink_ctx, "channel_layouts",
2291 (uint8_t*)&sink_ctx->channel_layout,
2292 sizeof(sink_ctx->channel_layout), AV_OPT_SEARCH_CHILDREN);
2294 ret = av_opt_set_bin(buffersink_ctx, "sample_rates",
2295 (uint8_t*)&sink_ctx->sample_rate, sizeof(sink_ctx->sample_rate),
2296 AV_OPT_SEARCH_CHILDREN);
2298 ff_err(ret, "FFAudioStream::create_filter");
2300 ret = FFStream::create_filter(filter_spec);
2301 return ret >= 0 ? 0 : 1;
2304 int FFStream::create_filter(const char *filter_spec)
2306 /* Endpoints for the filter graph. */
2307 AVFilterInOut *outputs = avfilter_inout_alloc();
2308 outputs->name = av_strdup("in");
2309 outputs->filter_ctx = buffersrc_ctx;
2310 outputs->pad_idx = 0;
2313 AVFilterInOut *inputs = avfilter_inout_alloc();
2314 inputs->name = av_strdup("out");
2315 inputs->filter_ctx = buffersink_ctx;
2316 inputs->pad_idx = 0;
2319 int ret = !outputs->name || !inputs->name ? -1 : 0;
2321 ret = avfilter_graph_parse_ptr(filter_graph, filter_spec,
2322 &inputs, &outputs, NULL);
2324 ret = avfilter_graph_config(filter_graph, NULL);
2327 ff_err(ret, "FFStream::create_filter");
2328 avfilter_inout_free(&inputs);
2329 avfilter_inout_free(&outputs);
2333 void FFStream::add_bsfilter(const char *bsf, const char *ap)
2335 bsfilter.append(new BSFilter(bsf,ap));
2338 int FFStream::bs_filter(AVPacket *pkt)
2340 if( !bsfilter.size() ) return 0;
2341 av_packet_split_side_data(pkt);
2344 for( int i=0; i<bsfilter.size(); ++i ) {
2345 AVPacket bspkt = *pkt;
2346 ret = av_bitstream_filter_filter(bsfilter[i]->bsfc,
2347 st->codec, bsfilter[i]->args, &bspkt.data, &bspkt.size,
2348 pkt->data, pkt->size, pkt->flags & AV_PKT_FLAG_KEY);
2349 if( ret < 0 ) break;
2350 int size = bspkt.size;
2351 uint8_t *data = bspkt.data;
2352 if( !ret && bspkt.data != pkt->data ) {
2354 data = (uint8_t *)av_malloc(size + FF_INPUT_BUFFER_PADDING_SIZE);
2355 if( !data ) { ret = AVERROR(ENOMEM); break; }
2356 memcpy(data, bspkt.data, size);
2357 memset(data+size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
2361 pkt->side_data = 0; pkt->side_data_elems = 0;
2362 av_free_packet(pkt);
2363 ret = av_packet_from_data(&bspkt, data, size);
2364 if( ret < 0 ) break;
2369 ff_err(ret,"FFStream::bs_filter");
2373 int FFMPEG::scan(IndexState *index_state, int64_t *scan_position, int *canceled)
2376 av_init_packet(&pkt);
2377 AVFrame *frame = av_frame_alloc();
2379 fprintf(stderr, "FFMPEG::scan: av_frame_alloc failed\n");
2383 index_state->add_video_markers(ffvideo.size());
2384 index_state->add_audio_markers(ffaudio.size());
2386 for( int i=0; i<(int)fmt_ctx->nb_streams; ++i ) {
2387 AVDictionary *copts = 0;
2388 av_dict_copy(&copts, opts, 0);
2389 AVStream *st = fmt_ctx->streams[i];
2390 AVCodecID codec_id = st->codec->codec_id;
2391 AVCodec *decoder = avcodec_find_decoder(codec_id);
2392 if( avcodec_open2(st->codec, decoder, &copts) < 0 )
2393 fprintf(stderr, "FFMPEG::scan: codec open failed\n");
2394 av_dict_free(&copts);
2397 for( int64_t count=0; !*canceled; ++count ) {
2398 av_packet_unref(&pkt);
2399 pkt.data = 0; pkt.size = 0;
2401 int ret = av_read_frame(fmt_ctx, &pkt);
2403 if( ret == AVERROR_EOF ) break;
2404 if( ++errs > 100 ) {
2405 ff_err(ret, "over 100 read_frame errs\n");
2410 if( !pkt.data ) continue;
2411 int i = pkt.stream_index;
2412 if( i < 0 || i >= (int)fmt_ctx->nb_streams ) continue;
2413 AVStream *st = fmt_ctx->streams[i];
2414 AVCodecContext *avctx = st->codec;
2415 if( pkt.pos > *scan_position ) *scan_position = pkt.pos;
2417 switch( avctx->codec_type ) {
2418 case AVMEDIA_TYPE_VIDEO: {
2419 int vidx = ffvideo.size();
2420 while( --vidx>=0 && ffvideo[vidx]->fidx != i );
2421 if( vidx < 0 ) break;
2422 FFVideoStream *vid = ffvideo[vidx];
2423 int64_t tstmp = pkt.dts;
2424 if( tstmp == AV_NOPTS_VALUE ) tstmp = pkt.pts;
2425 if( tstmp != AV_NOPTS_VALUE && (pkt.flags & AV_PKT_FLAG_KEY) && pkt.pos > 0 ) {
2426 if( vid->nudge != AV_NOPTS_VALUE ) tstmp -= vid->nudge;
2427 double secs = to_secs(tstmp, st->time_base);
2428 int64_t frm = secs * vid->frame_rate + 0.5;
2429 if( frm < 0 ) frm = 0;
2430 index_state->put_video_mark(vidx, frm, pkt.pos);
2433 while( pkt.size > 0 ) {
2434 av_frame_unref(frame);
2436 int ret = vid->decode_frame(&pkt, frame, got_frame);
2437 if( ret <= 0 ) break;
2438 // if( got_frame ) {}
2444 case AVMEDIA_TYPE_AUDIO: {
2445 int aidx = ffaudio.size();
2446 while( --aidx>=0 && ffaudio[aidx]->fidx != i );
2447 if( aidx < 0 ) break;
2448 FFAudioStream *aud = ffaudio[aidx];
2449 int64_t tstmp = pkt.pts;
2450 if( tstmp == AV_NOPTS_VALUE ) tstmp = pkt.dts;
2451 if( tstmp != AV_NOPTS_VALUE && (pkt.flags & AV_PKT_FLAG_KEY) && pkt.pos > 0 ) {
2452 if( aud->nudge != AV_NOPTS_VALUE ) tstmp -= aud->nudge;
2453 double secs = to_secs(tstmp, st->time_base);
2454 int64_t sample = secs * aud->sample_rate + 0.5;
2455 if( sample < 0 ) sample = 0;
2456 index_state->put_audio_mark(aidx, sample, pkt.pos);
2458 while( pkt.size > 0 ) {
2459 int ch = aud->channel0, nch = aud->channels;
2460 int64_t pos = index_state->pos(ch);
2461 if( pos != aud->curr_pos ) {
2462 if( abs(pos-aud->curr_pos) > 1 )
2463 printf("audio%d pad %ld %ld (%ld)\n", aud->idx, pos, aud->curr_pos, pos-aud->curr_pos);
2464 index_state->pad_data(ch, nch, aud->curr_pos);
2466 av_frame_unref(frame);
2468 int ret = aud->decode_frame(&pkt, frame, got_frame);
2469 if( ret <= 0 ) break;
2472 int len = aud->get_samples(samples,
2473 &frame->extended_data[0], frame->nb_samples);
2474 for( int i=0; i<nch; ++i )
2475 index_state->put_data(ch+i,nch,samples+i,len);
2476 aud->curr_pos += len;
2485 av_frame_free(&frame);
2489 void FFStream::load_markers(IndexMarks &marks, double rate)
2491 index_markers = &marks;
2493 int64_t sz = marks.size();
2494 int max_entries = fmt_ctx->max_index_size / sizeof(AVIndexEntry) - 1;
2495 int nb_ent = st->nb_index_entries;
2496 // some formats already have an index
2498 AVIndexEntry *ep = &st->index_entries[nb_ent-1];
2499 int64_t tstmp = ep->timestamp;
2500 if( nudge != AV_NOPTS_VALUE ) tstmp -= nudge;
2501 double secs = ffmpeg->to_secs(tstmp, st->time_base);
2502 int64_t no = secs * rate;
2503 while( in < sz && marks[in].no <= no ) ++in;
2505 int64_t len = sz - in;
2506 int64_t count = max_entries - nb_ent;
2507 if( count > len ) count = len;
2508 for( int i=0; i<count; ++i ) {
2509 int k = in + i * len / count;
2510 int64_t no = marks[k].no, pos = marks[k].pos;
2511 double secs = (double)no / rate;
2512 int64_t tstmp = secs * st->time_base.den / st->time_base.num;
2513 if( nudge != AV_NOPTS_VALUE ) tstmp += nudge;
2514 av_add_index_entry(st, pos, tstmp, 0, 0, AVINDEX_KEYFRAME);