static void* EncoderThread( void *obj )
{
sout_stream_sys_t *p_sys = (sout_stream_sys_t*)obj;
- sout_stream_id_t *id = p_sys->id_video;
+ sout_stream_id_sys_t *id = p_sys->id_video;
picture_t *p_pic = NULL;
int canc = vlc_savecancel ();
block_t *p_block = NULL;
return NULL;
}
-int transcode_video_new( sout_stream_t *p_stream, sout_stream_id_t *id )
+int transcode_video_new( sout_stream_t *p_stream, sout_stream_id_sys_t *id )
{
sout_stream_sys_t *p_sys = p_stream->p_sys;
? id->p_encoder->fmt_out.video.i_visible_height
: id->p_decoder->fmt_in.video.i_visible_height
? id->p_decoder->fmt_in.video.i_visible_height : id->p_encoder->fmt_in.video.i_height;
- id->p_encoder->fmt_in.video.i_frame_rate = id->p_decoder->fmt_out.video.i_frame_rate;
- id->p_encoder->fmt_in.video.i_frame_rate_base = id->p_decoder->fmt_out.video.i_frame_rate_base;
id->p_encoder->i_threads = p_sys->i_threads;
id->p_encoder->p_cfg = p_sys->p_video_cfg;
}
static void transcode_video_filter_init( sout_stream_t *p_stream,
- sout_stream_id_t *id )
+ sout_stream_id_sys_t *id )
{
es_format_t *p_fmt_out = &id->p_decoder->fmt_out;
id->p_encoder->fmt_in.video.i_chroma = id->p_encoder->fmt_in.i_codec;
p_stream->p_sys );
filter_chain_Reset( id->p_uf_chain, p_fmt_out,
&id->p_encoder->fmt_in );
+ if( p_fmt_out->video.i_chroma != id->p_encoder->fmt_in.video.i_chroma )
+ {
+ filter_chain_AppendFilter( id->p_uf_chain,
+ NULL, NULL,
+ p_fmt_out,
+ &id->p_encoder->fmt_in );
+ }
filter_chain_AppendFromString( id->p_uf_chain, p_stream->p_sys->psz_vf2 );
p_fmt_out = filter_chain_GetFmtOut( id->p_uf_chain );
es_format_Copy( &id->p_encoder->fmt_in, p_fmt_out );
}
/* Take care of the scaling and chroma conversions. */
-static void conversion_video_filter_append( sout_stream_id_t *id )
+static void conversion_video_filter_append( sout_stream_id_sys_t *id )
{
const es_format_t *p_fmt_out = &id->p_decoder->fmt_out;
if( id->p_f_chain )
p_fmt_out = filter_chain_GetFmtOut( id->p_f_chain );
+
if( id->p_uf_chain )
p_fmt_out = filter_chain_GetFmtOut( id->p_uf_chain );
( p_fmt_out->video.i_width != id->p_encoder->fmt_in.video.i_width ) ||
( p_fmt_out->video.i_height != id->p_encoder->fmt_in.video.i_height ) )
{
- filter_chain_AppendFilter( id->p_f_chain,
+ filter_chain_AppendFilter( id->p_uf_chain ? id->p_uf_chain : id->p_f_chain,
NULL, NULL,
p_fmt_out,
&id->p_encoder->fmt_in );
}
static void transcode_video_encoder_init( sout_stream_t *p_stream,
- sout_stream_id_t *id )
+ sout_stream_id_sys_t *id )
{
sout_stream_sys_t *p_sys = p_stream->p_sys;
}
}
+ id->p_encoder->fmt_in.video.orientation =
+ id->p_encoder->fmt_out.video.orientation =
+ id->p_decoder->fmt_in.video.orientation;
+
id->p_encoder->fmt_in.video.i_frame_rate =
id->p_encoder->fmt_out.video.i_frame_rate;
id->p_encoder->fmt_in.video.i_frame_rate_base =
id->p_encoder->fmt_in.video.i_frame_rate,
id->p_encoder->fmt_in.video.i_frame_rate_base );
- id->i_output_frame_interval = id->p_encoder->fmt_out.video.i_frame_rate_base * CLOCK_FREQ / id->p_encoder->fmt_out.video.i_frame_rate;
- id->i_input_frame_interval = id->p_decoder->fmt_out.video.i_frame_rate_base * CLOCK_FREQ / id->p_decoder->fmt_out.video.i_frame_rate;
- msg_Info( p_stream, "input interval %d (base %d) output interval %d (base %d)", id->i_input_frame_interval, id->p_decoder->fmt_out.video.i_frame_rate_base,
+ id->i_input_frame_interval = id->p_decoder->fmt_out.video.i_frame_rate_base * CLOCK_FREQ / id->p_decoder->fmt_out.video.i_frame_rate;
+ msg_Info( p_stream, "input interval %d (base %d)",
+ id->i_input_frame_interval, id->p_decoder->fmt_out.video.i_frame_rate_base );
+
+ id->i_output_frame_interval = id->p_encoder->fmt_in.video.i_frame_rate_base * CLOCK_FREQ / id->p_encoder->fmt_in.video.i_frame_rate;
+ msg_Info( p_stream, "output interval %d (base %d)",
id->i_output_frame_interval, id->p_encoder->fmt_in.video.i_frame_rate_base );
- date_Init( &id->interpolated_pts,
+ date_Init( &id->next_input_pts,
id->p_decoder->fmt_out.video.i_frame_rate,
1 );
+
date_Init( &id->next_output_pts,
id->p_encoder->fmt_in.video.i_frame_rate,
1 );
}
static int transcode_video_encoder_open( sout_stream_t *p_stream,
- sout_stream_id_t *id )
+ sout_stream_id_sys_t *id )
{
sout_stream_sys_t *p_sys = p_stream->p_sys;
}
void transcode_video_close( sout_stream_t *p_stream,
- sout_stream_id_t *id )
+ sout_stream_id_sys_t *id )
{
if( p_stream->p_sys->i_threads >= 1 )
{
filter_chain_Delete( id->p_uf_chain );
}
-static void OutputFrame( sout_stream_sys_t *p_sys, picture_t *p_pic, sout_stream_t *p_stream, sout_stream_id_t *id, block_t **out )
+static void OutputFrame( sout_stream_t *p_stream, picture_t *p_pic, sout_stream_id_sys_t *id, block_t **out )
{
-
+ sout_stream_sys_t *p_sys = p_stream->p_sys;
picture_t *p_pic2 = NULL;
+ const mtime_t original_date = p_pic->date;
bool b_need_duplicate=false;
- /* If input pts + input_frame_interval is lower than next_output_pts - output_frame_interval
+ /* If input pts is lower than next_output_pts - output_frame_interval
* Then the future input frame should fit better and we can drop this one
*
- * Duplication need is checked in OutputFrame */
- if( ( p_pic->date + (mtime_t)id->i_input_frame_interval ) <
- ( date_Get( &id->next_output_pts ) ) )
+ * We check it here also because we can have case that video filters outputs multiple
+ * pictures but we don't need to use them all, for example yadif2x and outputting to some
+ * different fps value
+ */
+ if( ( original_date ) <
+ ( date_Get( &id->next_output_pts ) - (mtime_t)id->i_output_frame_interval ) )
{
#if 0
msg_Dbg( p_stream, "dropping frame (%"PRId64" + %"PRId64" vs %"PRId64")",
subpicture_Delete( p_subpic );
}
}
+
+ /* set output pts*/
+ p_pic->date = date_Get( &id->next_output_pts );
/*This pts is handled, increase clock to next one*/
date_Increment( &id->next_output_pts, id->p_encoder->fmt_in.video.i_frame_rate_base );
/* we need to duplicate while next_output_pts + output_frame_interval < input_pts (next input pts)*/
b_need_duplicate = ( date_Get( &id->next_output_pts ) + id->i_output_frame_interval ) <
- ( date_Get( &id->interpolated_pts ) );
+ ( original_date );
if( p_sys->i_threads )
{
#endif
date_Increment( &id->next_output_pts, id->p_encoder->fmt_in.video.i_frame_rate_base );
b_need_duplicate = ( date_Get( &id->next_output_pts ) + id->i_output_frame_interval ) <
- ( date_Get( &id->interpolated_pts ) );
+ ( original_date );
}
if( p_sys->i_threads && p_pic2 )
picture_Release( p_pic );
}
-int transcode_video_process( sout_stream_t *p_stream, sout_stream_id_t *id,
+int transcode_video_process( sout_stream_t *p_stream, sout_stream_id_sys_t *id,
block_t *in, block_t **out )
{
sout_stream_sys_t *p_sys = p_stream->p_sys;
while( (p_pic = id->p_decoder->pf_decode_video( id->p_decoder, &in )) )
{
- if( p_stream->p_sout->i_out_pace_nocontrol && p_sys->b_hurry_up )
- {
- mtime_t current_date = mdate();
- if( unlikely( (current_date - 50000) > p_pic->date ) )
- {
- msg_Dbg( p_stream, "late picture skipped (%"PRId64")",
- current_date - 50000 - p_pic->date );
- picture_Release( p_pic );
- continue;
- }
- }
-
if( unlikely (
id->p_encoder->p_module &&
- !video_format_IsSimilar( &p_sys->fmt_input_video, &id->p_decoder->fmt_out.video )
+ !video_format_IsSimilar( &id->fmt_input_video, &id->p_decoder->fmt_out.video )
)
)
{
msg_Info( p_stream, "aspect-ratio changed, reiniting. %i -> %i : %i -> %i.",
- p_sys->fmt_input_video.i_sar_num, id->p_decoder->fmt_out.video.i_sar_num,
- p_sys->fmt_input_video.i_sar_den, id->p_decoder->fmt_out.video.i_sar_den
+ id->fmt_input_video.i_sar_num, id->p_decoder->fmt_out.video.i_sar_num,
+ id->fmt_input_video.i_sar_den, id->p_decoder->fmt_out.video.i_sar_den
);
/* Close filters */
if( id->p_f_chain )
transcode_video_filter_init( p_stream, id );
transcode_video_encoder_init( p_stream, id );
conversion_video_filter_append( id );
- memcpy( &p_sys->fmt_input_video, &id->p_decoder->fmt_out.video, sizeof(video_format_t));
+ memcpy( &id->fmt_input_video, &id->p_decoder->fmt_out.video, sizeof(video_format_t));
}
transcode_video_filter_init( p_stream, id );
transcode_video_encoder_init( p_stream, id );
conversion_video_filter_append( id );
- memcpy( &p_sys->fmt_input_video, &id->p_decoder->fmt_out.video, sizeof(video_format_t));
+ memcpy( &id->fmt_input_video, &id->p_decoder->fmt_out.video, sizeof(video_format_t));
if( transcode_video_encoder_open( p_stream, id ) != VLC_SUCCESS )
{
id->b_transcode = false;
return VLC_EGENERIC;
}
+ date_Set( &id->next_output_pts, p_pic->date );
+ date_Set( &id->next_input_pts, p_pic->date );
}
/*Input lipsync and drop check */
if( p_sys->b_master_sync )
{
- /* How much audio has drifted */
- mtime_t i_master_drift = p_sys->i_master_drift;
-
- /* This is the pts input should have now with constant frame rate */
- mtime_t i_pts = date_Get( &id->interpolated_pts );
-
- /* How much video pts is ahead of calculated pts */
- mtime_t i_video_drift = p_pic->date - i_pts;
-
- /* Check that we are having lipsync with input here */
- if( unlikely ( ( (i_video_drift - i_master_drift ) > MASTER_SYNC_MAX_DRIFT
- || (i_video_drift + i_master_drift ) < -MASTER_SYNC_MAX_DRIFT ) ) )
- {
- msg_Warn( p_stream,
- "video drift too big, resetting sync %"PRId64" to %"PRId64,
- (i_video_drift + i_master_drift),
- p_pic->date
- );
- date_Set( &id->interpolated_pts, p_pic->date );
- date_Set( &id->next_output_pts, p_pic->date );
- i_pts = date_Get( &id->interpolated_pts );
- }
-
- /* Set the pts of the frame being encoded */
- p_pic->date = i_pts;
-
- /* now take next input pts, pts dates are only enabled if p_module is set*/
- date_Increment( &id->interpolated_pts, id->p_decoder->fmt_out.video.i_frame_rate_base );
-
-
- /* If input pts + input_frame_interval is lower than next_output_pts - output_frame_interval
+ /* If input pts lower than next_output_pts - output_frame_interval
* Then the future input frame should fit better and we can drop this one
*
+ * We check this here as we don't need to run video filter at all for pictures
+ * we are going to drop anyway
+ *
* Duplication need is checked in OutputFrame */
- if( ( p_pic->date + (mtime_t)id->i_input_frame_interval ) <
- ( date_Get( &id->next_output_pts ) ) )
+ if( ( p_pic->date ) <
+ ( date_Get( &id->next_output_pts ) - (mtime_t)id->i_output_frame_interval ) )
{
#if 0
msg_Dbg( p_stream, "dropping frame (%"PRId64" + %"PRId64" vs %"PRId64")",
msg_Dbg( p_stream, "not dropping frame");
#endif
- /* input calculated pts isn't necessary what pts output should be, so use output pts*/
- p_pic->date = date_Get( &id->next_output_pts );
-
-
}
+ /* Check input drift regardless, if it's more than 100ms from our approximation, we most likely have lost pictures
+ * and are in danger to become out of sync, so better reset timestamps then */
+ if( likely( p_pic->date != VLC_TS_INVALID ) )
+ {
+ mtime_t input_drift = p_pic->date - date_Get( &id->next_input_pts );
+ if( unlikely( (input_drift > (CLOCK_FREQ/10)) ||
+ (input_drift < -(CLOCK_FREQ/10))
+ ) )
+ {
+ msg_Warn( p_stream, "Reseting video sync" );
+ date_Set( &id->next_output_pts, p_pic->date );
+ date_Set( &id->next_input_pts, p_pic->date );
+ }
+ }
+ date_Increment( &id->next_input_pts, id->p_decoder->fmt_out.video.i_frame_rate_base );
/* Run the filter and output chains; first with the picture,
* and then with NULL as many times as we need until they
if( !p_user_filtered_pic )
break;
- OutputFrame( p_sys, p_user_filtered_pic, p_stream, id, out );
+ OutputFrame( p_stream, p_user_filtered_pic, id, out );
p_filtered_pic = NULL;
}
}
bool transcode_video_add( sout_stream_t *p_stream, es_format_t *p_fmt,
- sout_stream_id_t *id )
+ sout_stream_id_sys_t *id )
{
sout_stream_sys_t *p_sys = p_stream->p_sys;
* all the characteristics of the decoded stream yet */
id->b_transcode = true;
- if( p_sys->f_fps > 0 )
+ if( p_sys->fps_num )
{
- id->p_encoder->fmt_out.video.i_frame_rate = (p_sys->f_fps * ENC_FRAMERATE_BASE);
- id->p_encoder->fmt_out.video.i_frame_rate_base = ENC_FRAMERATE_BASE;
+ id->p_encoder->fmt_out.video.i_frame_rate = (p_sys->fps_num );
+ id->p_encoder->fmt_out.video.i_frame_rate_base = (p_sys->fps_den ? p_sys->fps_den : 1);
}
return true;