Question

I have a function compute() that has parallelized matrix multiplication inside of it using OpenMP

#pragma omp parallel for

This function is called many times in a loop – which I would like to run in parallel.
Will there be any issues in running parallel code inside other parallel code?

This is c++ compiled on Ubuntu.

Answer 1

It will work fine, but you’ll need to enable OpenMP nesting for it to work.

Call

omp_set_nested(1);

at the start of the program and it will allow you to have nested parallel regions.

However: Be aware, that you could end up running many more threads than what you want. So you will want to limit the # of threads of both the top and inner parallel regions.