[mpiwg-hybridpm] Hybrid telecon fiasco
balaji at anl.gov
Mon Apr 7 13:36:21 CDT 2014
On Apr 7, 2014, at 1:32 PM, Jim Dinan <james.dinan at gmail.com> wrote:
> That is another option... Consider a case with cooperatively scheduled user-level threads. It could be hard to meet the concurrency requirement for collectives even if you have MPI_THREAD_MULTIPLE.
FWIW, MPI_THREAD_MULTIPLE is required to work correctly even with cooperatively scheduled user-level threads. The MPI implementation is required to yield appropriately within each blocking call. E.g., both threads do MPI_RECV.
More information about the mpiwg-hybridpm