latency-tracer: Exclude synchronization time

The goal of this tracer is to measure the processing latency between a
src and a sink. In push mode, the time was read after the chain function
have returned. As the amount of time we wait to get synched is reverse
to the amount of latency the source introduced, the result was quite
surprising.

This patch moves the latency calculation in the pre-push hook. When
there is no processing in a a pipeline (e.g. fakesrc ! fakesink), the
latency will now be 0 as it's supposed to. For pull mode, the code was
already correct. When GstBaseSink operate in pull mode, the processing
time is done durring the pull, so pull-post is the right hook. The
synchronization will happen after the pull has ended. Note that
GstBaseSink rarely operate in pull mode.

https://bugzilla.gnome.org/show_bug.cgi?id=788431

diff --git a/plugins/tracers/gstlatency.c b/plugins/tracers/gstlatency.c
index b92dabb..2d21230 100644 (file)
--- a/plugins/tracers/gstlatency.c
+++ b/plugins/tracers/gstlatency.c
@@ -126,23 +126,6 @@ send_latency_probe (GstElement * parent, GstPad * pad, guint64 ts)
 }
 
 static void
-do_push_buffer_pre (GstTracer * self, guint64 ts, GstPad * pad)
-{
-  GstElement *parent = get_real_pad_parent (pad);
-
-  send_latency_probe (parent, pad, ts);
-}
-
-static void
-do_pull_range_pre (GstTracer * self, guint64 ts, GstPad * pad)
-{
-  GstPad *peer_pad = GST_PAD_PEER (pad);
-  GstElement *parent = get_real_pad_parent (peer_pad);
-
-  send_latency_probe (parent, peer_pad, ts);
-}
-
-static void
 calculate_latency (GstElement * parent, GstPad * pad, guint64 ts)
 {
   if (parent && (!GST_IS_BIN (parent)) &&
@@ -158,12 +141,23 @@ calculate_latency (GstElement * parent, GstPad * pad, guint64 ts)
 }
 
 static void
-do_push_buffer_post (GstTracer * self, guint64 ts, GstPad * pad)
+do_push_buffer_pre (GstTracer * self, guint64 ts, GstPad * pad)
+{
+  GstPad *peer_pad = GST_PAD_PEER (pad);
+  GstElement *parent = get_real_pad_parent (pad);
+  GstElement *peer_parent = get_real_pad_parent (peer_pad);
+
+  send_latency_probe (parent, pad, ts);
+  calculate_latency (peer_parent, peer_pad, ts);
+}
+
+static void
+do_pull_range_pre (GstTracer * self, guint64 ts, GstPad * pad)
 {
   GstPad *peer_pad = GST_PAD_PEER (pad);
   GstElement *parent = get_real_pad_parent (peer_pad);
 
-  calculate_latency (parent, peer_pad, ts);
+  send_latency_probe (parent, peer_pad, ts);
 }
 
 static void
@@ -230,18 +224,24 @@ static void
 gst_latency_tracer_init (GstLatencyTracer * self)
 {
   GstTracer *tracer = GST_TRACER (self);
+
+  /* in push mode, pre/post will be called before/after the peer chain
+   * function has been called. For this reaosn, we only use -pre to avoid
+   * accounting for the processing time of the peer element (the sink) */
   gst_tracing_register_hook (tracer, "pad-push-pre",
       G_CALLBACK (do_push_buffer_pre));
   gst_tracing_register_hook (tracer, "pad-push-list-pre",
       G_CALLBACK (do_push_buffer_pre));
-  gst_tracing_register_hook (tracer, "pad-push-post",
-      G_CALLBACK (do_push_buffer_post));
-  gst_tracing_register_hook (tracer, "pad-push-list-post",
-      G_CALLBACK (do_push_buffer_post));
+
+  /* while in pull mode, pre/post will happend before and after the upstream
+   * pull_range call is made, so it already only account for the upstream
+   * processing time. As a side effect, in pull mode, we can measure the
+   * source processing latency, while in push mode, we can't */
   gst_tracing_register_hook (tracer, "pad-pull-range-pre",
       G_CALLBACK (do_pull_range_pre));
   gst_tracing_register_hook (tracer, "pad-pull-range-post",
       G_CALLBACK (do_pull_range_post));
+
   gst_tracing_register_hook (tracer, "pad-push-event-pre",
       G_CALLBACK (do_push_event_pre));
 }