WeeklyTelcon_20230321

Open MPI Weekly Telecon ---

• Dialup Info: (Do not post to public mailing list or public wiki)

Attendees (on Web-ex)

• Didn't capture today.

---

Not here today, but keep here for easy cut-n-paste for future.

• Brendan Cunningham (Cornelis Networks)
• Brian Barrett (Amazon)
• Edgar Gabriel (AMD)
• Geoffrey Paulsen (IBM)
• Howard Pritchard (LANL)
• Jeff Squyres (Cisco)
• Joseph Schuchart (UTK)
• Josh Fisher (Cornelis Networks)
• Luke Robison (Amazon)
• Quincey
• Todd Kordenbrock (Sandia)
• Tomislav Janjusic (nVidia)
• William Zhang (AWS)
• Austen Lauria (IBM)
• Christoph Nietham
• David Bernholdt
• Josh Hursey (IBM)
• Matthew Dosanjh (Sandia)
• Thomas Naughton (ORNL)

New Items

• OAC submodule

• FAQ changes - Anything else that needs progress.

  • Not a blocker for v5.0.0

• MCA attribute registration cdoe

  • Color string issue - When you call mca_register_param - it creates a string, and overwrites the pointer you gave it.
    • when you go through cleanup (MPIT_Finalize), this pointer is set to null.
    • Any time we have multiple MPIT_INIT/MPIT_Finalize together, we can get some behavior that we're not accounting for behavior. But some places in code account for this, so not unknown.
      • Do we want to fix every single invocation. Or change way the function works.
    • Is it a segv or lose param? - lose it, but then causes an issue.
    • Does valgrind help find these places? How easy is it to find other places?
      • greped for string type, and found a few places that do it correctly, and some do it wrong.
  • MPIT_pvar Issue 11492
    • on ompi 4.1 - so won't be a regression. Program didn't crash, just pvars are strange.

• MPI Sessions Issue

  • Issue - not an issue yet, will file one Some info in #general (3/20)
  • Simple example, querying a session from pset, create a group, MPI_Comm_create_from_group, since UCX didn't support it, and tried obi1.
    • UCX will have to wait, need a new thing in PMIx.
  • Don't consider anything with sessions as a show stopper.
  • May not get to until next week.

• MPIR Shim (https://github.com/openpmix/mpir-to-pmix-guide) went away.

  • Howard pushed repo to somewhere.
  • Howard will hook it up to CI for some testing.
  • MPIR shim has some docs 12.7, it just need some new URL and info.

v5.0.x

• 11510 - New Blocker because of Wrong answers - Will investigate soon.

• Amazon is investigating a significant numbers of hangs on v5.0.0rc

• What to do about 11415 - Discussed and said this is correct behavior.

  • Customer accepted new behavior for OMPI v5.0

• Runtime docs stuff should be doable by end of the month.

• We'd talked about supplying some docs about how HAN is great, and why we're enabling it for v5.0.0 by default.

  • Like to include instructions on how to reproduce as well for users.

• New MTT regression failure introduced

  • A legit issue in main
  • fixed by PR #11499
    • PR CI for Mellanox passes, but IBM CI failed (Possibly due to lab moves)
    • Give it another day or so to see if it's just

• CI topic

  • Testing a new way (not ancient Github Builder Plugin for Jenkins) using a Jenkins file Pipelines
  • Want to switch over ASAPly been more or less ready for the last month, but now we need to change.

Main branch

Administration Topics