-
Notifications
You must be signed in to change notification settings - Fork 859
WeeklyTelcon_20230620
Geoffrey Paulsen edited this page Jun 20, 2023
·
1 revision
- Dialup Info: (Do not post to public mailing list or public wiki)
- Christoph Niethammer (HLRS)
- Edgar Gabriel (AMD)
- Geoff Paulsen (IBM)
- Joseph Schuchart
- Josh Fisher
- Luke Robison (Amazon)
- Quincey Koziol (AWS)
- Todd Kordenbrok
- Issue # with OMPI v4.1.5 - and latest PMIx (v4.2.5???)
- Just got a patch to fix it.
- Will drive a v4.1.x release
- fix is on the branch, so works in latest nightlies
- Closed Issue #11749 - OMPI won't spawn
- Doc Issues: https://github.com/open-mpi/ompi/projects/3
- Documentation meeting on doc-review
- Quincey and Luke 3pm Central
- MCA Params issues are biggest issues now
-
https://github.com/openmpi/ompi/issues/11532
- PMIX command line parsing issue fixed the first stage completed, next stage fix over the next few days.
- https://github.com/openpmix/prrte/issues/1731
- Plan is to have 2 of the 3 fixes for v5.0.0, 3rd issue can wait for 5.0.x
- Quincy assigned, working on docs first.
-
https://github.com/openmpi/ompi/issues/11532
- async modex issues - just need to increase the timeout one
- Trying to make the timeout directly related to scale.
- very hard to know what this will look out
- Please update help message
- Tommy is working.
- NIC selection coming back from main PR #11739
- Somewhat large change, but really want this back, but it broke EFA
- review needed, and then we can merge!
- #11726 -N bind ppr:X:node, map by package (socket), or core
- Should OMPI binding defaults be a "reliable"?
- Users shouldn't have to depend on this.
- Just note that the default changed between v4.0 and v5.0
- Consensus is that it's okay and that we should document this.
- Should OMPI binding defaults be a "reliable"?