Hi all,
Does any one know how to profile multiple kernel names with nvprof.
I can use nvprof --kernels “kernelName” to specify one kernel name but I don’t know how to add another name under this option.
Thanks!
Hi all,
Does any one know how to profile multiple kernel names with nvprof.
I can use nvprof --kernels “kernelName” to specify one kernel name but I don’t know how to add another name under this option.
Thanks!
Oh I found the solution. I need to specify kernelName1|kernelName2|Kernelnamen.
https://docs.nvidia.com/cuda/profiler-users-guide/index.html#profiling-scope
–kernels “kernel1” --metrics gld_efficiency --kernels “kernel2” --metrics gld_efficiency
example:
$ nvprof --kernels "ewMulti" --metrics gld_efficiency --kernels "SwapKernel" --metrics gld_efficiency ./t96
==26396== NVPROF is profiling process 26396, command: ./t96
==26396== Some kernel(s) will be replayed on device 0 in order to collect all events/metrics.
Replaying kernel "ewMulti(double2*, double2*)" (done)
Replaying kernel "SwapKernel(double2*, double2*)" (done)
y = x'*x = 7.000000 5.000000i
==26396== Profiling application: ./t96
==26396== Profiling result:
==26396== Metric result:
Invocations Metric Name Metric Description Min Max Avg
Device "Tesla P100-PCIE-16GB (0)"
Kernel: SwapKernel(double2*, double2*)
1 gld_efficiency Global Memory Load Efficiency 100.00% 100.00% 100.00%
Kernel: ewMulti(double2*, double2*)
1 gld_efficiency Global Memory Load Efficiency 100.00% 100.00% 100.00%
Thanks! Robert,
Now I understand that nvprof can specify multiple kernels,metrics,event config.