Hello everyone, One of my users has an hybrid MPI+OpenMP job, for which he launches three MPI processes. However, and this is where it gets interesting, when it comes to the number of threads per process he needs: 1, x, x
with x being set as optimally as possible. In this case, the cluster has 16 cores, so he wants 7 threads for each (with one of the cores being unused). Is there a way to do this? It seems to me that --cpu_bind may offer some options (with map_cpu), but I may be misunderstanding the purpose of cpu_bind, hence my question. I'm waiting for some feedback from the user, so right now I can't actually run any tests. Thank you all in advance, Daniel Jana
