-
Notifications
You must be signed in to change notification settings - Fork 1
2023 12 18 webex
- Discuss proposed changes to https://github.com/mpi-forum/mpi-standard/pull/938 (the relax provenance of pset names)
Howard opened a test PR at https://github.com/mpiwg-sessions/mpi-standard/pull/50 to discuss additions to PR 938. These additions would be responses to feedback at the reading at the December Forum. Any changes would have to be handled as a no-no vote if we wished to continue with the original 938 and its passing first vote.
Martin Schreiber added some comments to the PR. Dan responded to some of the points Martin raised on the PR as well.
The WG discussed how the bubbles concept fit in to this. Dan sees bubbles as a connectivity model with non-local state, whereas sessions themselves are purely local, as the MPI_GROUP_FROM_SESSION_PSET function. There was considerable discussion about whether or not the change in 938 is sufficient for the dynamic process set activities being envisioned/designed/implemented currently. Martin said no but that this was a good first step. Howard and Dan said we can't expect much more than this small type of change for the 4.2 standard. Anything bigger, for example, an expanded API for better support for dynamic process sets, would need to wait for 5.0.
Dan expressed some concern that any no-no votes added to 938 may risk it getting voted down on the second vote.
The WG decided to work on the word smithing currently in 50 at the next meeting.