How to efficiently replace vaapih265enc with omxh265enc?

[Nicolas Dufresne]

Le lundi 18 novembre 2019 à 06:20 -0600, jeyp4 a écrit :
> Thanks for replying.
> 
> <http://gstreamer-devel.966125.n4.nabble.com/file/t379085/vaapih265enc.jpg> 
> 
> Actually, above pipeline works fine, even without 'h265parse'. I am sending
> buffer data in ROS msg. What I am doing at receiver side is, modifying pts,
> dts with current receiver pipeline running time. And also I have to start
> receiver before transmitter. In this way I observe almost no delay in
> decoding ~ 2ms. I am happy with this pipeline.
> 
> Just for info: If I add h265parse in above receiver pipeline, it works then
> also. But now decoding time is ~40ms. Decoding time: Difference between when
> receiver received buffer and when vaapidecode dropped its output. I
> confirmed this latency presence by measuring glass to glass latency also.

Just a note that we are trying to get some patches that address this in
h264/h265parse. Aligned input should not yield latency in the parsers,
but right now it does.

> 
> As I said, I am happy with above pipeline without h265parse because it is
> working fine.
> 
> I am trying to replace vaapih265enc with omxh265enc !

All you need is to find which field in the caps is needed omxh265dec,
and try and figure-out what to add in omxh265enc, you have to modify
the code. Careful, NVidia have added quite an important layer of hacks
on top of these elements, so you have to start from their code.

> 
> 
> 
> --
> Sent from: http://gstreamer-devel.966125.n4.nabble.com/
> _______________________________________________
> gstreamer-devel mailing list
> gstreamer-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel