-
Notifications
You must be signed in to change notification settings - Fork 1
2021 10 18 webex
#10/18/21 webex notes
Attending: Howard Pritchard, Thomas Hines, Dan Holmes, Isaias Urena, Grace Nansambe(UTC), Trupeshkumar Patel, Martin Schreiber, Martin Schulz
- Followup on issue https://github.com/mpi-forum/mpi-issues/issues/511
- Review where we are with dynamic process sets in the context of next week
Reviewed PR in current state. Dan sees a comma where there should be a period. Howard will fix.
Possibly bring up issues in section 9.3 about error handlers and MPI calls not directlyr related to communicators, windows, files, and sesions.
Next week we'll want to bring up growing/shrinking and how this can be handled between dynamic process sets concepts, excluding dead processes, and recovering. Maybe determining dead processes would be part of FT, and then use that info to create a new pset/group and hence a new communicator. May be able to prune methods out of proposals etc. if we leverage methods from other proposals (e.g. dynamic process sets). Could we use the sessions delta functionality to weed out dead/isolate processes.
Sessions has a more general approach to processes coming and going, whereas FT WG has focused on a more particular set of use cases. May be that FT methods might be designed to expedite app lookup of new process sets.
Discuss some of the issues with ULFM that have kept it from getting voted in to the Standard. Martin Schulz mentioned their interrupting error handler concept address some deadlock issues raised by the Forum.
Reinit work? How does this fit in.
Links:
Google doc we're using - https://docs.google.com/document/d/1l7LQ8eeVOUW69TDVG9LjKJUuerfE3S3teaMFG5DOudM/edit#heading=h.voobxhw94rt3 Miro document - https://miro.com/app/board/o9J_l_Rxe9Q=/
- Howard will update wiki with the above links