Hi,
I have ran nsight compute and could see a lot of statics on each kernel under launch section, i do not see overall kernel utilization table or chart.I’m interested in overall Bandwidth utilization, during my application run?
Is there any way i can get those info, or any recommended tool ?
I am not fully sure what you are currently seeing and what is missing from your description. Are you checking in the UI or in the CLI output? Assuming that you are using Nsight Compute 2019.5 or higher, try collecting with --set full on the CLI and inspect the report in the UI. Alternatively, enable the “full” section set in the Sections/Rules tool window in the UI when doing the Interactive Profiling activity from there to collect the full set of pre-configured sections.
https://docs.nvidia.com/nsight-compute/NsightComputeCli/index.html#command-line-options-profile
https://docs.nvidia.com/nsight-compute/NsightCompute/index.html#tool-window-sections-info
MPI statics is it available with ncu(nisghtcompute) or any other tools?
Nsight Compute does not capture any MPI details, since it focuses on individual kernels. I think you can check Nsight Systems to get more details on your MPI usage.
https://docs.nvidia.com/nsight-systems/tracing/index.html#mpi-trace