Vtune seems not to carry out my command line

[Doubler_THU]

TLDR

______________________________________________________________________________________

Recently, I am running VTune with a command (the long command) like this.

vtune -collect hotspots $(which mpirun) --bind-to none -np 2 -x OMP_NUM_THREAD=44 ./wrapper.sh

Though the short command "$(which mpirun) --bind-to none -np 2 -x OMP_NUM_THREAD=44 ./wrapper.sh"  needs about 30+ seconds to go. The long command seems to finish in a very short time. When I use the VTune profiler to get the report, it shows that its elapsed time is only 0.008ms and the Func/Call Stack is unknown.

So I wrote another simple C program like this:

#include <stdio.h>
int check_arr()
{
        int fx[100000];
        for(int i = 1; i < 100000; i++)
                for(int j = 1; j < 100000; j++)
                {
                        if(j % i == 0 && fx[i]) fx[j] = 1;
                                else fx[j] = 0;
                }
        return fx[99999];
}
int main()
{
        int sum;
        for(int i=0; i< 100; i++)
                sum += check_arr();
        printf("%d\n",sum);
        return 0;
}

But the result didn't change. How could I troubleshoot the problem?

[JaideepK_Intel · Accepted solution]

Hi,

Thank you for posting in Intel Analyzers Forum.

We are investigating your issue at our end, Could you please share the VTune version for better understanding.

Thanks,

Jaideep

[Doubler_THU]

Hi Jaideep,

     I think I did a wrong operation so that you didn't get my reply directly.

     But my admin has reinstalled the Vtune (to the latest version) and all the problems are fixed. If you are still interested in the original problem, you can still refer to my early reply.

Thanks,

Robin

[Doubler_THU]

Hi,

Thank you for your assistance.

My Vtune version is as followed: Intel(R) oneAPI VTune(TM) Profiler 2021.1.1 Gold (build 613804) Command Line Tool.

And the important output ( excluding the possible explanations of the short profiling time) is as followed:

Collection and Platform Info
Application Command Line: /opt/spack/opt/spack/linux-debian11-broadwell/gcc-10.2.1/openmpi-4.0.5-zdfmfdw5u4z3vkz2vehjvvvdc76e5h4y/bin/mpirun "--bind-to" "none" "-np" "2" "-x" "OMP_NUM_THREAD=44" "/home/ryanyu/SC21/cardioid/test/tests/Niederer/wrapper.sh"
Operating System: 5.10.0-8-amd64 11.0
Computer Name: cmpt
Result Size: 3.3 MB
Collection start time: 13:49:21 25/10/2021 UTC
Collection stop time: 13:49:21 25/10/2021 UTC
Collector Type: Driverless Perf per-process counting
CPU
Name: Intel(R) Xeon(R) Processor code named Broadwell
Frequency: 2.200 GHz
Logical CPU Count: 88
Cache Allocation Technology
Level 2 capability: not detected
Level 3 capability: available

[JaideepK_Intel]

Hi,

Glad to know that your issue is resolved.

These are some links which may help you to profile MPI Applications:

https://www.intel.com/content/www/us/en/develop/documentation/vtune-cookbook/top/configuration-recipes/profiling-mpi-applications.html

https://www.intel.com/content/www/us/en/develop/documentation/mpi-developer-reference-linux/top/command-reference/mpiexec-hydra/gtool-options.html

If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.

Thanks,

Jaideep