If you have a code (which is parallelized by OpenMP) that uses some random function inside of a parallel region, is it possible that changing the number of threads changes the outcome?
Aucun commentaire:
Enregistrer un commentaire