For the purpose of profiling my application I’d like to be able to identify the entry point each megakernel corresponds to.
In a test scenario I have 2 entry points and Visual Profiler displays 3 compute kernels namely Megakernel_CUDA_[0,1,2]. I would assume #0 and #1 correspond to my entry points due to the way they are scheduled (would these numbers always correspond to their entry points?), however #2 it’s barely called and takes very little execution time, would this be due to some error with my scene setup or something else?
Thank you.