[Question]What is the parallelization threshold value for Caxpy benchmark?

[tianyizhangcs]

Hello, I am writing a paper using caxpy benchmark, could someone tell me what is the parallelization threshold value for Caxpy benchmark or somewhere I can find the info?
I feel that OpenMP is really called when the size of the vector is really small.

[brada4]

Any considerate improvements are more than welcome, especially if they help your research...
I measured just once and it sort of pointed to something bigger than 10000
interface/axpy.c

#define MULTI_THREAD_MINIMAL  10000

  //Temporarily work-around the low performance issue with small imput size &
  //multithreads.
  if (incx == 0 || incy == 0 || n <= MULTI_THREAD_MINIMAL)
          nthreads = 1;

[brada4]

Also see #1886 for other things help is needed.

[tianyizhangcs]

Very helpful! Thank you so much

[brada4]

add your research result as a pull request later if possible.

[tianyizhangcs]

Sure, my research is mainly comparing hpxMP

 https://github.com/STEllAR-GROUP/hpxMP

with llvm-OpenMP and GOMP.

As the benchmark is using OpenMP, so I employed as a performance comparison.