GSt Tracer Latency vs RidgeRun GSt Shark Inter-Latency

[connor.mcguirk]

I have been trying to figure out the difference between these two tracing options so that I can better understand my DeepStream processing pipeline. I’m using Python (deepstream 6.2, cuda 11.8, ds python bindings etc.).

GStreamer seems to offer the following method to capture element latency

GST_DEBUG="GST_TRACER:7" "GST_TRACERS=latency(flags=element+pipeline)"

While Ridge Run GSt Shark (which seems popular) offers the following method

GST_DEBUG="GST_TRACER:7" GST_TRACERS="interlatency"

I was wondering if anyone knows the difference (if any) between these two tracers.

[Fiona.Chen]

For GST_TRACER, please refer to Tracing (gstreamer.freedesktop.org)

For Gst Shark, please consult @michael_gruner

[connor.mcguirk]

Thanks, yes if Michael could provide some clarity on the differences that would be great!

[michael_gruner]

You should use the GStreamer one (latency). Some years ago, the one in GstShark (interlatency) used to provide more information that latency, but they eventually caught up and absorbed these features.