Hi there, I’m relatively new to CUDA so apologies for any newb references,
I’ve written a simple CUDA app that at the moment does very little aside from write out to device memory in a location that is dependant on it’s blockIdx.
Running this application via the CUDA profiler included with the toolkit, it doesn’t seem to pick up on any coalecsed / non-coalesced writes and reads that I’d expect it to, do I need to integrate it with the actual code in some way before it is able to do this?
thanks in advance