summaryrefslogtreecommitdiffstats
path: root/libhb/sync.c
blob: d8b31ed9340f79f03d08f172ee4849f02369bb19 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
/* $Id: sync.c,v 1.38 2005/04/14 21:57:58 titer Exp $

   This file is part of the HandBrake source code.
   Homepage: <http://handbrake.m0k.org/>.
   It may be used under the terms of the GNU General Public License. */

#include "hb.h"

#include "samplerate.h"
#include "ffmpeg/avcodec.h"

#ifdef INT64_MIN
#undef INT64_MIN /* Because it isn't defined correctly in Zeta */
#endif
#define INT64_MIN (-9223372036854775807LL-1)

#define AC3_SAMPLES_PER_FRAME 1536

typedef struct
{
    hb_audio_t * audio;

    int64_t      next_start;    /* start time of next output frame */
    int64_t      next_pts;      /* start time of next input frame */
    int64_t      start_silence; /* if we're inserting silence, the time we started */
    int64_t      first_drop;    /* PTS of first 'went backwards' frame dropped */
    int          drop_count;    /* count of 'time went backwards' drops */
    int          inserting_silence;

    /* Raw */
    SRC_STATE  * state;
    SRC_DATA     data;

    /* AC-3 */
    int          ac3_size;
    uint8_t    * ac3_buf;

} hb_sync_audio_t;

struct hb_work_private_s
{
    hb_job_t * job;
    int        done;

    /* Video */
    hb_subtitle_t * subtitle;
    int64_t pts_offset;
    int64_t next_start;         /* start time of next output frame */
    int64_t next_pts;           /* start time of next input frame */
    int64_t first_drop;         /* PTS of first 'went backwards' frame dropped */
    int drop_count;             /* count of 'time went backwards' drops */
    int video_sequence;
    int count_frames;
    int count_frames_max;
    int chap_mark;              /* to propagate chapter mark across a drop */
    hb_buffer_t * cur; /* The next picture to process */

    /* Audio */
    hb_sync_audio_t sync_audio[8];

    /* Statistics */
    uint64_t st_counts[4];
    uint64_t st_dates[4];
    uint64_t st_first;
};

/***********************************************************************
 * Local prototypes
 **********************************************************************/
static void InitAudio( hb_work_object_t * w, int i );
static int  SyncVideo( hb_work_object_t * w );
static void SyncAudio( hb_work_object_t * w, int i );
static int  NeedSilence( hb_work_object_t * w, hb_audio_t *, int i );
static void InsertSilence( hb_work_object_t * w, int i, int64_t d );
static void UpdateState( hb_work_object_t * w );

/***********************************************************************
 * hb_work_sync_init
 ***********************************************************************
 * Initialize the work object
 **********************************************************************/
int syncInit( hb_work_object_t * w, hb_job_t * job )
{
    hb_title_t       * title = job->title;
    hb_chapter_t     * chapter;
    int                i;
    uint64_t           duration;
    hb_work_private_t * pv;

    pv = calloc( 1, sizeof( hb_work_private_t ) );
    w->private_data = pv;

    pv->job            = job;
    pv->pts_offset     = INT64_MIN;
    pv->count_frames   = 0;

    /* Calculate how many video frames we are expecting */
    duration = 0;
    for( i = job->chapter_start; i <= job->chapter_end; i++ )
    {
        chapter   = hb_list_item( title->list_chapter, i - 1 );
        duration += chapter->duration;
    }
    duration += 90000;
        /* 1 second safety so we're sure we won't miss anything */
    pv->count_frames_max = duration * job->vrate / job->vrate_base / 90000;

    hb_log( "sync: expecting %d video frames", pv->count_frames_max );

    /* Initialize libsamplerate for every audio track we have */
    for( i = 0; i < hb_list_count( title->list_audio ); i++ )
    {
        InitAudio( w, i );
    }

    /* Get subtitle info, if any */
    pv->subtitle = hb_list_item( title->list_subtitle, 0 );

    pv->video_sequence = 0;

    return 0;
}

/***********************************************************************
 * Close
 ***********************************************************************
 *
 **********************************************************************/
void syncClose( hb_work_object_t * w )
{
    hb_work_private_t * pv = w->private_data;
    hb_job_t          * job   = pv->job;
    hb_title_t        * title = job->title;
    hb_audio_t        * audio = NULL;

    int i;

    if( pv->cur ) hb_buffer_close( &pv->cur );

    for( i = 0; i < hb_list_count( title->list_audio ); i++ )
    {
        if ( pv->sync_audio[i].start_silence )
        {
            hb_log( "sync: added %d ms of silence to audio %d",
                    (int)((pv->sync_audio[i].next_pts -
                              pv->sync_audio[i].start_silence) / 90), i );
        }

        audio = hb_list_item( title->list_audio, i );
        if( audio->config.out.codec == HB_ACODEC_AC3 )
        {
            free( pv->sync_audio[i].ac3_buf );
        }
        else
        {
            src_delete( pv->sync_audio[i].state );
        }
    }

    free( pv );
    w->private_data = NULL;
}

/***********************************************************************
 * Work
 ***********************************************************************
 * The root routine of this work abject
 *
 * The way this works is that we are syncing the audio to the PTS of
 * the last video that we processed. That's why we skip the audio sync
 * if we haven't got a valid PTS from the video yet.
 *
 **********************************************************************/
int syncWork( hb_work_object_t * w, hb_buffer_t ** unused1,
              hb_buffer_t ** unused2 )
{
    hb_work_private_t * pv = w->private_data;
    int i;

    /* If we ever got a video frame, handle audio now */
    if( pv->pts_offset != INT64_MIN )
    {
        for( i = 0; i < hb_list_count( pv->job->title->list_audio ); i++ )
        {
            SyncAudio( w, i );
        }
    }

    /* Handle video */
    return SyncVideo( w );
}

hb_work_object_t hb_sync =
{
    WORK_SYNC,
    "Synchronization",
    syncInit,
    syncWork,
    syncClose
};

static void InitAudio( hb_work_object_t * w, int i )
{
    hb_work_private_t * pv = w->private_data;
    hb_job_t        * job   = pv->job;
    hb_title_t      * title = job->title;
    hb_sync_audio_t * sync;

    sync        = &pv->sync_audio[i];
    sync->audio = hb_list_item( title->list_audio, i );

    if( sync->audio->config.out.codec == HB_ACODEC_AC3 )
    {
        /* Have a silent AC-3 frame ready in case we have to fill a
           gap */
        AVCodec        * codec;
        AVCodecContext * c;
        short          * zeros;

        codec = avcodec_find_encoder( CODEC_ID_AC3 );
        c     = avcodec_alloc_context();

        c->bit_rate    = sync->audio->config.in.bitrate;
        c->sample_rate = sync->audio->config.in.samplerate;
        c->channels    = HB_INPUT_CH_LAYOUT_GET_DISCRETE_COUNT( sync->audio->config.in.channel_layout );

        if( avcodec_open( c, codec ) < 0 )
        {
            hb_log( "sync: avcodec_open failed" );
            return;
        }

        zeros          = calloc( AC3_SAMPLES_PER_FRAME *
                                 sizeof( short ) * c->channels, 1 );
        sync->ac3_size = sync->audio->config.in.bitrate * AC3_SAMPLES_PER_FRAME /
                             sync->audio->config.in.samplerate / 8;
        sync->ac3_buf  = malloc( sync->ac3_size );

        if( avcodec_encode_audio( c, sync->ac3_buf, sync->ac3_size,
                                  zeros ) != sync->ac3_size )
        {
            hb_log( "sync: avcodec_encode_audio failed" );
        }

        free( zeros );
        avcodec_close( c );
        av_free( c );
    }
    else
    {
        /* Initialize libsamplerate */
        int error;
        sync->state             = src_new( SRC_LINEAR, HB_AMIXDOWN_GET_DISCRETE_CHANNEL_COUNT(sync->audio->config.out.mixdown), &error );
        sync->data.end_of_input = 0;
    }
}

/***********************************************************************
 * SyncVideo
 ***********************************************************************
 *
 **********************************************************************/
static int SyncVideo( hb_work_object_t * w )
{
    hb_work_private_t * pv = w->private_data;
    hb_buffer_t * cur, * next, * sub = NULL;
    hb_job_t * job = pv->job;

    if( pv->done )
    {
        return HB_WORK_DONE;
    }

    if( hb_thread_has_exited( job->reader ) &&
        !hb_fifo_size( job->fifo_mpeg2 ) &&
        !hb_fifo_size( job->fifo_raw ) )
    {
        /* All video data has been processed already, we won't get
           more */
        hb_log( "sync: got %d frames, %d expected",
                pv->count_frames, pv->count_frames_max );
        pv->done = 1;

        hb_buffer_t * buf_tmp;

       // Drop an empty buffer into our output to ensure that things
       // get flushed all the way out.
        buf_tmp = hb_buffer_init(0); // Empty end buffer
        hb_fifo_push( job->fifo_sync, buf_tmp );

        return HB_WORK_DONE;
    }

    if( !pv->cur && !( pv->cur = hb_fifo_get( job->fifo_raw ) ) )
    {
        /* We haven't even got a frame yet */
        return HB_WORK_OK;
    }
    cur = pv->cur;

    /* At this point we have a frame to process. Let's check
        1) if we will be able to push into the fifo ahead
        2) if the next frame is there already, since we need it to
           compute the duration of the current frame*/
    while( !hb_fifo_is_full( job->fifo_sync ) &&
           ( next = hb_fifo_see( job->fifo_raw ) ) )
    {
        hb_buffer_t * buf_tmp;

        if( pv->pts_offset == INT64_MIN )
        {
            /* This is our first frame */
            pv->pts_offset = 0;
            if ( cur->start != 0 )
            {
                /*
                 * The first pts from a dvd should always be zero but
                 * can be non-zero with a transport or program stream since
                 * we're not guaranteed to start on an IDR frame. If we get
                 * a non-zero initial PTS extend its duration so it behaves
                 * as if it started at zero so that our audio timing will
                 * be in sync.
                 */
                hb_log( "sync: first pts is %lld", cur->start );
                cur->start = 0;
            }
        }

        /*
         * since the first frame is always 0 and the upstream reader code
         * is taking care of adjusting for pts discontinuities, we just have
         * to deal with the next frame's start being in the past. This can
         * happen when the PTS is adjusted after data loss but video frame
         * reordering causes some frames with the old clock to appear after
         * the clock change. This creates frames that overlap in time which
         * looks to us like time going backward. The downstream muxing code
         * can deal with overlaps of up to a frame time but anything larger
         * we handle by dropping frames here.
         */
        if ( (int64_t)( next->start - pv->next_pts ) <= 0 )
        {
            if ( pv->first_drop == 0 )
            {
                pv->first_drop = next->start;
            }
            ++pv->drop_count;
            buf_tmp = hb_fifo_get( job->fifo_raw );
            if ( buf_tmp->new_chap )
            {
                // don't drop a chapter mark when we drop the buffer
                pv->chap_mark = buf_tmp->new_chap;
            }
            hb_buffer_close( &buf_tmp );
            continue;
        }
        if ( pv->first_drop )
        {
            hb_log( "sync: video time didn't advance - dropped %d frames "
                    "(delta %d ms, current %lld, next %lld)",
                    pv->drop_count, (int)( pv->next_pts - pv->first_drop ) / 90,
                    pv->next_pts, pv->first_drop );
            pv->first_drop = 0;
            pv->drop_count = 0;
        }

        /*
         * Track the video sequence number localy so that we can sync the audio
         * to it using the sequence number as well as the PTS.
         */
        pv->video_sequence = cur->sequence;

        /* Look for a subtitle for this frame */
        if( pv->subtitle )
        {
            hb_buffer_t * sub2;
            while( ( sub = hb_fifo_see( pv->subtitle->fifo_raw ) ) )
            {
                /* If two subtitles overlap, make the first one stop
                   when the second one starts */
                sub2 = hb_fifo_see2( pv->subtitle->fifo_raw );
                if( sub2 && sub->stop > sub2->start )
                    sub->stop = sub2->start;

                // hb_log("0x%x: video seq: %lld  subtitle sequence: %lld",
                //       sub, cur->sequence, sub->sequence);

                if( sub->sequence > cur->sequence )
                {
                    /*
                     * The video is behind where we are, so wait until
                     * it catches up to the same reader point on the
                     * DVD. Then our PTS should be in the same region
                     * as the video.
                     */
                    sub = NULL;
                    break;
                }

                if( sub->stop > cur->start ) {
                    /*
                     * The stop time is in the future, so fall through
                     * and we'll deal with it in the next block of
                     * code.
                     */
                    break;
                }

                /*
                 * The subtitle is older than this picture, trash it
                 */
                sub = hb_fifo_get( pv->subtitle->fifo_raw );
                hb_buffer_close( &sub );
            }

            /*
             * There is a valid subtitle, is it time to display it?
             */
            if( sub )
            {
                if( sub->stop > sub->start)
                {
                    /*
                     * Normal subtitle which ends after it starts, check to
                     * see that the current video is between the start and end.
                     */
                    if( cur->start > sub->start &&
                        cur->start < sub->stop )
                    {
                        /*
                         * We should be playing this, so leave the
                         * subtitle in place.
                         *
                         * fall through to display
                         */
                        if( ( sub->stop - sub->start ) < ( 3 * 90000 ) )
                        {
                            /*
                             * Subtitle is on for less than three seconds, extend
                             * the time that it is displayed to make it easier
                             * to read. Make it 3 seconds or until the next
                             * subtitle is displayed.
                             *
                             * This is in response to Indochine which only
                             * displays subs for 1 second - too fast to read.
                             */
                            sub->stop = sub->start + ( 3 * 90000 );

                            sub2 = hb_fifo_see2( pv->subtitle->fifo_raw );

                            if( sub2 && sub->stop > sub2->start )
                            {
                                sub->stop = sub2->start;
                            }
                        }
                    }
                    else
                    {
                        /*
                         * Defer until the play point is within the subtitle
                         */
                        sub = NULL;
                    }
                }
                else
                {
                    /*
                     * The end of the subtitle is less than the start, this is a
                     * sign of a PTS discontinuity.
                     */
                    if( sub->start > cur->start )
                    {
                        /*
                         * we haven't reached the start time yet, or
                         * we have jumped backwards after having
                         * already started this subtitle.
                         */
                        if( cur->start < sub->stop )
                        {
                            /*
                             * We have jumped backwards and so should
                             * continue displaying this subtitle.
                             *
                             * fall through to display.
                             */
                        }
                        else
                        {
                            /*
                             * Defer until the play point is within the subtitle
                             */
                            sub = NULL;
                        }
                    } else {
                        /*
                         * Play this subtitle as the start is greater than our
                         * video point.
                         *
                         * fall through to display/
                         */
                    }
                }
            }
        }

        /*
         * Adjust the pts of the current frame so that it's contiguous
         * with the previous frame. The start time of the current frame
         * has to be the end time of the previous frame and the stop
         * time has to be the start of the next frame.  We don't
         * make any adjustments to the source timestamps other than removing
         * the clock offsets (which also removes pts discontinuities).
         * This means we automatically encode at the source's frame rate.
         * MP2 uses an implicit duration (frames end when the next frame
         * starts) but more advanced containers like MP4 use an explicit
         * duration. Since we're looking ahead one frame we set the
         * explicit stop time from the start time of the next frame.
         */
        buf_tmp = cur;
        pv->cur = cur = hb_fifo_get( job->fifo_raw );
        pv->next_pts = next->start;
        int64_t duration = next->start - buf_tmp->start;
        if ( duration <= 0 )
        {
            hb_log( "sync: invalid video duration %lld, start %lld, next %lld",
                    duration, buf_tmp->start, next->start );
        }
        buf_tmp->start = pv->next_start;
        pv->next_start += duration;
        buf_tmp->stop = pv->next_start;
        if ( pv->chap_mark )
        {
            // we have a pending chapter mark from a recent drop - put it on this
            // buffer (this may make it one frame late but we can't do any better).
            buf_tmp->new_chap = pv->chap_mark;
            pv->chap_mark = 0;
        }

        /* If we have a subtitle for this picture, copy it */
        /* FIXME: we should avoid this memcpy */
        if( sub )
        {
            buf_tmp->sub         = hb_buffer_init( sub->size );
            buf_tmp->sub->x      = sub->x;
            buf_tmp->sub->y      = sub->y;
            buf_tmp->sub->width  = sub->width;
            buf_tmp->sub->height = sub->height;
            memcpy( buf_tmp->sub->data, sub->data, sub->size );
        }

        /* Push the frame to the renderer */
        hb_fifo_push( job->fifo_sync, buf_tmp );

        /* Update UI */
        UpdateState( w );

        /* Make sure we won't get more frames then expected */
        if( pv->count_frames >= pv->count_frames_max * 2)
        {
            hb_log( "sync: got too many frames (%d), exiting early", pv->count_frames );
            pv->done = 1;

           // Drop an empty buffer into our output to ensure that things
           // get flushed all the way out.
           buf_tmp = hb_buffer_init(0); // Empty end buffer
           hb_fifo_push( job->fifo_sync, buf_tmp );

            break;
        }
    }

    return HB_WORK_OK;
}

static void OutputAudioFrame( hb_job_t *job, hb_audio_t *audio, hb_buffer_t *buf,
                              hb_sync_audio_t *sync, hb_fifo_t *fifo, int i )
{
    int64_t start = sync->next_start;
    int64_t duration = buf->stop - buf->start;

    sync->next_pts += duration;

    if( audio->config.in.samplerate == audio->config.out.samplerate ||
        audio->config.out.codec == HB_ACODEC_AC3 ||
        audio->config.out.codec == HB_ACODEC_DCA )
    {
        /*
         * If we don't have to do sample rate conversion or this audio is 
         * pass-thru just send the input buffer downstream after adjusting
         * its timestamps to make the output stream continuous.
         */
    }
    else
    {
        /* Not pass-thru - do sample rate conversion */
        int count_in, count_out;
        hb_buffer_t * buf_raw = buf;
        int channel_count = HB_AMIXDOWN_GET_DISCRETE_CHANNEL_COUNT(audio->config.out.mixdown) *
                            sizeof( float );

        count_in  = buf_raw->size / channel_count;
        /*
         * When using stupid rates like 44.1 there will always be some
         * truncation error. E.g., a 1536 sample AC3 frame will turn into a
         * 1536*44.1/48.0 = 1411.2 sample frame. If we just truncate the .2
         * the error will build up over time and eventually the audio will
         * substantially lag the video. libsamplerate will keep track of the
         * fractional sample & give it to us when appropriate if we give it
         * an extra sample of space in the output buffer.
         */
        count_out = ( duration * audio->config.out.samplerate ) / 90000 + 1;

        sync->data.input_frames = count_in;
        sync->data.output_frames = count_out;
        sync->data.src_ratio = (double)audio->config.out.samplerate /
                               (double)audio->config.in.samplerate;

        buf = hb_buffer_init( count_out * channel_count );
        sync->data.data_in  = (float *) buf_raw->data;
        sync->data.data_out = (float *) buf->data;
        if( src_process( sync->state, &sync->data ) )
        {
            /* XXX If this happens, we're screwed */
            hb_log( "sync: audio %d src_process failed", i );
        }
        hb_buffer_close( &buf_raw );

        buf->size = sync->data.output_frames_gen * channel_count;
        duration = ( sync->data.output_frames_gen * 90000 ) /
                   audio->config.out.samplerate;
    }
    buf->frametype = HB_FRAME_AUDIO;
    buf->start = start;
    buf->stop  = start + duration;
    sync->next_start = start + duration;
    while( hb_fifo_is_full( fifo ) )
    {
        hb_snooze( 50 );
        if ( job->done && hb_fifo_is_full( fifo ) )
        {
            /* don't block here if the job's finished */
            hb_buffer_close( &buf );
            return;
        }
    }
    hb_fifo_push( fifo, buf );
}

/***********************************************************************
 * SyncAudio
 ***********************************************************************
 *
 **********************************************************************/
static void SyncAudio( hb_work_object_t * w, int i )
{
    hb_work_private_t * pv = w->private_data;
    hb_job_t        * job = pv->job;
    hb_sync_audio_t * sync = &pv->sync_audio[i];
    hb_audio_t      * audio = sync->audio;
    hb_buffer_t     * buf;
    hb_fifo_t       * fifo;

    if( audio->config.out.codec == HB_ACODEC_AC3 )
    {
        fifo = audio->priv.fifo_out;
    }
    else
    {
        fifo = audio->priv.fifo_sync;
    }

    while( !hb_fifo_is_full( fifo ) && ( buf = hb_fifo_see( audio->priv.fifo_raw ) ) )
    {
        if ( (int64_t)( buf->start - sync->next_pts ) < 0 )
        {
            /*
             * audio time went backwards by more than a frame time (this can
             * happen when we reset the PTS because of lost data).
             * Discard data that's in the past.
             */
            if ( sync->first_drop == 0 )
            {
                sync->first_drop = buf->start;
            }
            ++sync->drop_count;
            buf = hb_fifo_get( audio->priv.fifo_raw );
            hb_buffer_close( &buf );
            continue;
        }
        if ( sync->first_drop )
        {
            hb_log( "sync: audio %d time went backwards %d ms, dropped %d frames "
                    "(next %lld, current %lld)", i,
                    (int)( sync->next_pts - sync->first_drop ) / 90,
                    sync->drop_count, sync->first_drop, sync->next_pts );
            sync->first_drop = 0;
            sync->drop_count = 0;
        }

        if ( sync->inserting_silence && (int64_t)(buf->start - sync->next_pts) > 0 )
        {
            /*
             * if we're within one frame time of the amount of silence
             * we need, insert just what we need otherwise insert a frame time.
             */
            int64_t framedur = buf->stop - buf->start;
            if ( buf->start - sync->next_pts <= framedur )
            {
                InsertSilence( w, i, buf->start - sync->next_pts );
                sync->inserting_silence = 0;
            }
            else
            {
                InsertSilence( w, i, framedur );
            }
            continue;
        }
        if ( buf->start - sync->next_pts >= (90 * 70) )
        {
            /*
             * there's a gap of at least 70ms between the last
             * frame we processed & the next. Fill it with silence.
             */
            if ( ! sync->inserting_silence )
            {
                hb_log( "sync: adding %d ms of silence to audio %d"
                        "  start %lld, next %lld",
                        (int)((buf->start - sync->next_pts) / 90),
                        i, buf->start, sync->next_pts );
                sync->inserting_silence = 1;
            }
            InsertSilence( w, i, buf->start - sync->next_pts );
            continue;
        }

        /*
         * When we get here we've taken care of all the dups and gaps in the
         * audio stream and are ready to inject the next input frame into
         * the output stream.
         */
        buf = hb_fifo_get( audio->priv.fifo_raw );
        OutputAudioFrame( job, audio, buf, sync, fifo, i );
    }

    if( NeedSilence( w, audio, i ) )
    {
        InsertSilence( w, i, (90000 * AC3_SAMPLES_PER_FRAME) /
                             sync->audio->config.in.samplerate );
    }
}

static int NeedSilence( hb_work_object_t * w, hb_audio_t * audio, int i )
{
    hb_work_private_t * pv = w->private_data;
    hb_job_t * job = pv->job;
    hb_sync_audio_t * sync = &pv->sync_audio[i];

    if( hb_fifo_size( audio->priv.fifo_in ) ||
        hb_fifo_size( audio->priv.fifo_raw ) ||
        hb_fifo_size( audio->priv.fifo_sync ) ||
        hb_fifo_size( audio->priv.fifo_out ) )
    {
        /* We have some audio, we are fine */
        return 0;
    }

    /* No audio left in fifos */

    if( hb_thread_has_exited( job->reader ) )
    {
        /* We might miss some audio to complete encoding and muxing
           the video track */
        if ( sync->start_silence == 0 )
        {
            hb_log("sync: reader has exited, adding silence to audio %d", i);
            sync->start_silence = sync->next_pts;
        }
        return 1;
    }
    return 0;
}

static void InsertSilence( hb_work_object_t * w, int i, int64_t duration )
{
    hb_work_private_t * pv = w->private_data;
    hb_job_t        *job = pv->job;
    hb_sync_audio_t *sync = &pv->sync_audio[i];
    hb_buffer_t     *buf;
    hb_fifo_t       *fifo;

    // to keep pass-thru and regular audio in sync we generate silence in
    // AC3 frame-sized units. If the silence duration isn't an integer multiple
    // of the AC3 frame duration we will truncate or round up depending on
    // which minimizes the timing error.
    const int frame_dur = ( 90000 * AC3_SAMPLES_PER_FRAME ) /
                          sync->audio->config.in.samplerate;
    int frame_count = ( duration + (frame_dur >> 1) ) / frame_dur;

    while ( --frame_count >= 0 )
    {
        if( sync->audio->config.out.codec == HB_ACODEC_AC3 )
        {
            buf        = hb_buffer_init( sync->ac3_size );
            buf->start = sync->next_pts;
            buf->stop  = buf->start + frame_dur;
            memcpy( buf->data, sync->ac3_buf, buf->size );
            fifo = sync->audio->priv.fifo_out;
        }
        else
        {
            buf = hb_buffer_init( AC3_SAMPLES_PER_FRAME * sizeof( float ) *
                                     HB_AMIXDOWN_GET_DISCRETE_CHANNEL_COUNT(
                                         sync->audio->config.out.mixdown) );
            buf->start = sync->next_pts;
            buf->stop  = buf->start + frame_dur;
            memset( buf->data, 0, buf->size );
            fifo = sync->audio->priv.fifo_sync;
        }
        OutputAudioFrame( job, sync->audio, buf, sync, fifo, i );
    }
}

static void UpdateState( hb_work_object_t * w )
{
    hb_work_private_t * pv = w->private_data;
    hb_state_t state;

    if( !pv->count_frames )
    {
        pv->st_first = hb_get_date();
    }
    pv->count_frames++;

    if( hb_get_date() > pv->st_dates[3] + 1000 )
    {
        memmove( &pv->st_dates[0], &pv->st_dates[1],
                 3 * sizeof( uint64_t ) );
        memmove( &pv->st_counts[0], &pv->st_counts[1],
                 3 * sizeof( uint64_t ) );
        pv->st_dates[3]  = hb_get_date();
        pv->st_counts[3] = pv->count_frames;
    }

#define p state.param.working
    state.state = HB_STATE_WORKING;
    p.progress  = (float) pv->count_frames / (float) pv->count_frames_max;
    if( p.progress > 1.0 )
    {
        p.progress = 1.0;
    }
    p.rate_cur   = 1000.0 *
        (float) ( pv->st_counts[3] - pv->st_counts[0] ) /
        (float) ( pv->st_dates[3] - pv->st_dates[0] );
    if( hb_get_date() > pv->st_first + 4000 )
    {
        int eta;
        p.rate_avg = 1000.0 * (float) pv->st_counts[3] /
            (float) ( pv->st_dates[3] - pv->st_first );
        eta = (float) ( pv->count_frames_max - pv->st_counts[3] ) /
            p.rate_avg;
        p.hours   = eta / 3600;
        p.minutes = ( eta % 3600 ) / 60;
        p.seconds = eta % 60;
    }
    else
    {
        p.rate_avg = 0.0;
        p.hours    = -1;
        p.minutes  = -1;
        p.seconds  = -1;
    }
#undef p

    hb_set_state( pv->job->h, &state );
}