Seeing n/a for metrics

I have run the following command on a 2080Ti device

sudo /usr/local/cuda-10.1/NsightCompute-2019.3/nv-nsight-cu-cli \
--quiet --metrics sm​_​_warps​_active.avg.pct​_of​_peak​_sustained​_active,smsp​_​_inst​_executed.avg.per​_cycle​_active \
-f -o perf.nvvp ../../../build/bin/bfs market belgium_osm.mtx

in order to profile an application. The output file size is 86M file, but when I open it via /usr/local/cuda/NsightCompute-2019.3/nv-nsight-cu->FILE->OPEN FILE, I see the picture below which shows both metrics are n/a.

https://pasteboard.co/IxaRuVo.png

That is weird because, the same program with nvprof on older devices works fine and I can see the information for kernels used in the program.

Hi,

Did you copy the metric names from “Nsight Compute CLI” document page (https://docs.nvidia.com/nsight-compute/NsightComputeCli/index.html#nvprof-metric-comparison) and paste in command-line? If yes, note that there is an issue copying and pasting zero-width space character. Due to this issue, metric name is not interpreted properly and it is being treated as an invalid metric.

Can you please type metric name manually on command line and try profiling? Please let me know if it works for you.


Thanks,
Ramesh