mirror of
https://gitlab.freedesktop.org/gstreamer/gstreamer.git
synced 2024-12-23 16:50:47 +00:00
h265decoder: Report latency
Similar to that of h264decoder Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2726>
This commit is contained in:
parent
94e1ddad43
commit
5386a426f3
1 changed files with 72 additions and 0 deletions
|
@ -346,6 +346,77 @@ gst_h265_decoder_drain_output_queue (GstH265Decoder * self, guint num,
|
|||
}
|
||||
}
|
||||
|
||||
static void
|
||||
gst_h265_decoder_set_latency (GstH265Decoder * self, const GstH265SPS * sps,
|
||||
gint max_dpb_size)
|
||||
{
|
||||
GstH265DecoderPrivate *priv = self->priv;
|
||||
GstCaps *caps;
|
||||
GstClockTime min, max;
|
||||
GstStructure *structure;
|
||||
gint fps_d = 1, fps_n = 0;
|
||||
guint frames_delay;
|
||||
|
||||
caps = gst_pad_get_current_caps (GST_VIDEO_DECODER_SRC_PAD (self));
|
||||
if (!caps)
|
||||
return;
|
||||
|
||||
structure = gst_caps_get_structure (caps, 0);
|
||||
if (gst_structure_get_fraction (structure, "framerate", &fps_n, &fps_d)) {
|
||||
if (fps_n == 0) {
|
||||
/* variable framerate: see if we have a max-framerate */
|
||||
gst_structure_get_fraction (structure, "max-framerate", &fps_n, &fps_d);
|
||||
}
|
||||
}
|
||||
gst_caps_unref (caps);
|
||||
|
||||
/* if no fps or variable, then 25/1 */
|
||||
if (fps_n == 0) {
|
||||
fps_n = 25;
|
||||
fps_d = 1;
|
||||
}
|
||||
|
||||
/* Minimum possible latency could be calculated based on C.5.2.3
|
||||
* 1) # of pictures (marked as "needed for output") in DPB > sps_max_num_reorder_pics
|
||||
* - We will assume all pictures in DPB are marked as "needed for output"
|
||||
* 2) sps_max_latency_increase_plus1 != 0 and
|
||||
* PicLatencyCount >= SpsMaxLatencyPictures
|
||||
* - SpsMaxLatencyPictures is equal to
|
||||
* "sps_max_num_reorder_pics + sps_max_latency_increase_plus1 - 1"
|
||||
* and PicLatencyCount of each picture in DPB is increased by 1 per
|
||||
* decoding loop. Note that PicLatencyCount of the currently decoded
|
||||
* picture is zero. So, in case that all pictures in DPB are marked as
|
||||
* "needed for output", Only condition 1) will have an effect
|
||||
* regardless of sps_max_latency_increase_plus1.
|
||||
*
|
||||
* For example, assume sps_max_num_reorder_pics is 2 and
|
||||
* sps_max_latency_increase_plus1 is 1, then SpsMaxLatencyPictures is 2.
|
||||
* For a picture in DPB to have PicLatencyCount >= SpsMaxLatencyPictures,
|
||||
* there must be at least 3 pictures including current picture in DPB
|
||||
* (current picture's PicLatencyCount is zero).
|
||||
* This is already covered by the condition 1). So, this condition 2)
|
||||
* will have effect only when there are pictures marked as
|
||||
* "not needed for output" in DPB.
|
||||
*
|
||||
* Thus, we can take sps_max_num_reorder_pics as a min latency value
|
||||
*/
|
||||
frames_delay = sps->max_num_reorder_pics[sps->max_sub_layers_minus1];
|
||||
|
||||
/* Consider output delay wanted by subclass */
|
||||
frames_delay += priv->preferred_output_delay;
|
||||
|
||||
min = gst_util_uint64_scale_int (frames_delay * GST_SECOND, fps_d, fps_n);
|
||||
max = gst_util_uint64_scale_int ((max_dpb_size + priv->preferred_output_delay)
|
||||
* GST_SECOND, fps_d, fps_n);
|
||||
|
||||
GST_DEBUG_OBJECT (self,
|
||||
"latency min %" GST_TIME_FORMAT " max %" GST_TIME_FORMAT
|
||||
" min-frames-delay %d", GST_TIME_ARGS (min), GST_TIME_ARGS (max),
|
||||
frames_delay);
|
||||
|
||||
gst_video_decoder_set_latency (GST_VIDEO_DECODER (self), min, max);
|
||||
}
|
||||
|
||||
static GstFlowReturn
|
||||
gst_h265_decoder_process_sps (GstH265Decoder * self, GstH265SPS * sps)
|
||||
{
|
||||
|
@ -435,6 +506,7 @@ gst_h265_decoder_process_sps (GstH265Decoder * self, GstH265SPS * sps)
|
|||
priv->interlaced_source_flag = interlaced_source_flag;
|
||||
|
||||
gst_h265_dpb_set_max_num_pics (priv->dpb, max_dpb_size);
|
||||
gst_h265_decoder_set_latency (self, sps, max_dpb_size);
|
||||
}
|
||||
|
||||
if (sps->max_latency_increase_plus1[sps->max_sub_layers_minus1]) {
|
||||
|
|
Loading…
Reference in a new issue