I would like to know if compute preemption of the GPU can be controlled by the user or the OS (L4T) on the Jetson TX2 platform.
This presentation explains the basics but does not mention how to use it in practice: http://on-demand.gputechconf.com/gtc/2017/presentation/s7105-venugopala-madumbu-adas-ad-challenges-gpu-scheduling-and-synchronization.pdf
As well, this paper states that there is a runlist construction algorithm sample in the L4T sources: https://algo.ing.unimo.it/people/marko/papers/RTSS18.pdf
Unfortunately I was not able to find it.
I am thankful for any help.