Hello,
I want to time the difference between each kernel call i.e. time cost between one kernel call and the next…
What is the best way to do that?
Thank you,
cudaCreateEvent
cudaRecordEvent
Those guys will timestamp stuff that happens on the GPU. You can record one at the beginning of each kernel. Or you can use a different mechanism on the x86/x64 code to timestamp just before each call. Or just after each call.