WeeklyTelcon_20230221

Open MPI Weekly Telecon ---

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

Attendees (on Web-ex)

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

---

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

• David Bernholdt

New Items

• Map-by Issue https://github.com/open-mpi/ompi/issues/11450 - Worked in RC9 and not RC10.

  • Not pleasing. Howard is trying to reproduce.
  • Is this PRTE only? It's a PMIx error message.
    • Unclear, that PMIx error is unclear.
  • Need to know severity. Need better map-by testing.

• Josh pushed the runtime harness that's designed to run runtime.

  • in ompi-tests/public, waiting for tests to be put in.
  • Right now runs in CI without MTT. Meant to exercise mpirun
  • CI for runtime testing: https://github.com/open-mpi/ompi/issues/11288#issuecomment-1439267528
  • Add mpirun tests here: https://github.com/open-mpi/ompi-tests-public/tree/master/runtime
  • Thomas Naughton is interested in looking at this.
  • How would you run this under MTT?
    • Infrastructure is meant to be run in CI, but could setup MTT to run this, but would take some work.
  • PRRTE has something like this, but doesn't run mpirun personality.
  • Could run MPI apps that have runtime dependencies.
  • put a tool to pretty-print the mapping/binding.

• MPIR Shim (https://github.com/openpmix/mpir-to-pmix-guide) is going away TODAY!

  • Howard grabbed a fork of it.
  • May be some ecosystems as well (older/current?) tools only support MPIR, not newer PMIx_Tool API
  • Where are we in the stability of the replacement?
    • Do we have a list of debugger vendors?
    • Debuggers, and profilers use it as well.
  • There are a ton of tools that have MPIR, but have they converted and released a PMIx_Tools interface.
  • This is a new feature for Open MPI v5.0.0 (MPIR is gone, now use PMIx_Tools based debuggers/profilers.
  • It would be good to determine if we need this before we release v5.0.0
    • Can we decide NOW to host this repo?
    • Shouldn't take more than a week to determine
    • Most impactful to people who debug/profile parallel applications.
  • Not packaged with Open MPI release.
  • Feels like as soon as we announce MPIR is going away, then people will complain loudly
  • Once it's in a seperate repo, might be easier for other people to support it.
  • Open MPI has CI that tests MPIR-shm with PMIx and PRRTE

• Issue #11347 Versioning is wrong in v5.0.x

  • Should be fixed in main and v5.0.x now, need to test MPI_COMM_WORLD with new .so version.

v4.1.x

• Need to pull in a PMIx v3.1.
• Demand driven bugfixes for v4.1.x
• released v4.1.5

v5.0.x

• Romio configury issue went in

• PR #11443 needs review.

• Regression testing.

• 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.
  • document in ECP -

Main branch

Administration Topics