throughput of "__svml_exp8"

[xwuupb]

Hi everyone,

where can I find the throughput (in terms of CPI) of "__svml_exp8" (or other math functions in Intel SVML)? Now I could only estimate it by using the following code.

#include <stdio.h>
#include <math.h>
#include <omp.h>

#define N (1 << 30)

int main(void)
{
  double sum = 0.0,
         start, stopp;

  start = omp_get_wtime();
  /*--------------------------------------------
    Instruction   CPUID Flags   Throughput (CPI)
      vcvtdq2pd       AVX512F                1.0
         vpaddd       AVX512F                0.5
         vaddpd       AVX512F                0.5
  --------------------------------------------*/
#pragma unroll(4)
  for (int i = 0; i < N; ++i) {
    sum += exp((double) i);
  }
  stopp = omp_get_wtime();
  printf("%.16f\n", sum);
  printf("%.16f\n", stopp - start);
  return 0;
}

[VidyalathaB_Intel · Accepted solution]

Hi,

Thanks for reaching out to us.

>>where can I find the throughput (in terms of CPI) of "__svml_exp8"

You can make use of the Intel VTune profiler to get the CPI rate for your program.

Please refer to the below link to get more details

https://software.intel.com/content/www/us/en/develop/documentation/vtune-help/top/reference/cpu-metrics-reference/clockticks-per-instructions-retired-cpi.html

If your Visual Studio is integrated with oneAPI Base and HPC toolkits you can directly open VTune profiler in Visual Studio.

For more details regarding the usage of VTune Profiler, please refer to the following link

https://software.intel.com/content/www/us/en/develop/documentation/vtune-help/top.html

Hope the provided information helped.

If this is not what you are looking for, please do let us know.

Regards,

Vidya.

[xwuupb]

thanks! Please close this thread.

[VidyalathaB_Intel]

Hi,

Thanks for accepting our solution.

As this issue is resolved we are closing this thread. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.

Regards,

Vidya.