You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Ken verified that MPICH does not hang in MPI_Comm_disconnect if inactive persistent requests associated with a communicator had not been freed.
However, Open MPI does hang in this case. We need to check about MPI_Finalize in this scenario. If there would be an impact by adopting
changes proposed in #732 it would be the change in behaviour for MPI_Finalize.
The proposal in the issue is associated with the 'C' option in issue 710.
Discuss as well the datatype issue that came in with PR #832 (see related issue https://github.com/mpi-forum/mpi-issues/issues/710).
Dan says the changes in PR reflect what is implied with the current state of the sessions and MPI handles that are not directly associated with
MPI_COMM_WORLD or a particular session. Howard will open a new issue specific to this datatype issue, including some possible proposals.