cudaProfilerInitialize() is now deprecated (as of CUDA 11), but I cannot see an indication of what to use in its place? Any help much appreciated.
Which profiler do you use?
Hey, thanks for swift reply. We are using nvprof on a remote embedded target running in docker
I think you can specify everything you need in the profiler command line invocation. Apart from that, you can use cudaProfilerStart
and cudaProfilerStop
without a previous cudaProfilerInitialize
.