But when I run the program through Nsight Systems, I get a warning that says: “No NVTX events collected. Does the process use NVTX?” Instead, all the threads are in a blocked state and I cannot tell if there is any useful information. Is there something wrong with my test?
I just tested Nsight Systems on the nvToolsExt samples, but I’m getting the same issue that no NVTX samples are being collected. I tried collecting CUDA and Vulkan from other samples, both of which seem to be fine. My hardware is an RTX 3070 with an AMD Ryzen 9 5900HS.
Hi Jebbly, your test is perfectly fine. I can’t repro the problem – here is what I get when I compile your code and run it in Nsight Systems with “NVTX” enabled:
I tried capturing both from the GUI and the command line, and I get the NVTX events. Does it work for you when you disable all trace options except for NVTX?
Oh, also, Nsight Systems 2021.3 is out now. I don’t really expect that to change anything in this case, but it couldn’t hurt to grab the newest version, since that’s what I’m testing with.