ncu -o ./data/"$output" --metrics \ gpu__time_duration.sum,lts__t_sectors.avg.pct_of_peak_sustained_elapsed \ trtexec --loadEngine=../"$engine" --shapes=actual_input_1:1x"$shape" --avgRuns=1 --warmUp=0 --duration=0
I don't get any detail info of CUDA kernel, the error message is No kernels were profiled
and it is no use when I add --target-processes all
.
I want to get the detail message of CUDA kernel runtime.