I am experimenting with OpenMP tasks and want to write an application that runs on a 2-NUMA socket CPU and uses OpenMP's task affinity clauses which can be added to the task creation pragma. They provide a hint for where a task should be executed, by providing a variable close to whose physical location the task should be executed.

An example from the OpenMP 5.0 documentation shows how it could be used:

void task_affinity(double *A, int N)
{
 double * B;
 #pragma omp task depend(out:B) shared(B) affinity(A[0:N])
 {
 B = alloc_init_B(A,N);
 }

 #pragma omp task depend( in:B) shared(B) affinity(A[0:N])
 {
 compute_on_B(B,N);
 }

 #pragma omp taskwait
}

The compiler gcc compiler that I have in version 11.2.0, however, only provides a stub as of now, which as I understand it, means, that the functionality is not actually implemented yet.

1. Is there any compiler that has OpenMP's task affinities fully implemented yet?

2. Does the gcc implementation of OpenMP handle tasks in a way that they are assigned to threads that are physically close to the data on which they work even if no affinities are explicitly stated?