I use Nvidia Systems 2022.1.1
We have a one-hop remote SSH profiling, but we use on the Login Node a self written port forwarder to reach the Compute Node.
I can start the application and get a timeline, but it is empty. No CUDA nor NVTX samples.
The Diagnostic Summary of the project shows no Injection.
When I run the project on my local system, I get Injections and a timeline with CUDA nd NVTX samples.
Should I setup the environment variables NVTX_INJECTION64_PATH or CUDA_INJECTION64_PATH on the Compute Node? If yes: to which location ?
Best regards, Stephan