Intel® Moderncode for Parallel Architectures
Support for developing parallel programming applications on Intel® Architecture.

MPI_Win_unlock_all performance

luiceur
Beginner
432 Views

I have been running some performance tests(Intel 17 and 18) and I have noticed that MPI_Win_unlock_all performance is surprisingly excessive with even a small number MPI processes. I was wondering if I could know the reasons and if this is already known.

 

0 Kudos
1 Reply
ArthurRatz
Black Belt
432 Views

luiceur, try to use MPI_Win_fence(...) instead of MPI_Win_unlock_all, and then compute the execution time.

Reply