WeeklyTelcon_20210504

Open MPI Weekly Telecon ---

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

Attendees (on Web-ex)

• Aurelien Bouteiller (UTK)
• Austen Lauria (IBM)
• Brian Barrett (AWS)
• David Bernhold (ORNL)
• Edgar Gabriel (UH)
• Geoffrey Paulsen (IBM)
• Harumi Kuno (HPE)
• Hessam Mirsadeghi (UCX/nVidia)
• Howard Pritchard (LANL)
• Jeff Squyres (Cisco)
• Joseph Schuchart
• Josh Hursey (IBM)
• Matthew Dosanjh (Sandia)
• Raghu Raja
• Sam Gutierrez (LANL)
• Todd Kordenbrock (Sandia)
• Tomislav Janjusic
• William Zhang (AWS)

not there today (I keep this for easy cut-n-paste for future notes)

• Akshay Venkatesh (NVIDIA)
• Artem Polyakov (nVidia/Mellanox)
• Brandon Yates (Intel)
• Brendan Cunningham (Cornelis Networks)
• Charles Shereda (LLNL)
• Christoph Niethammer (HLRS)
• Erik Zeiske
• Geoffroy Vallee (ARM)
• George Bosilca (UTK)
• Joshua Ladd (nVidia/Mellanox)
• Marisa Roman (Cornelius)
• Mark Allen (IBM)
• Matias Cabral (Intel)
• Michael Heinz (Cornelis Networks)
• Nathan Hjelm (Google)
• Naughton III, Thomas (ORNL)
• Noah Evans (Sandia)
• Ralph Castain (Intel)
• Scott Breyer (Sandia?)
• Shintaro iwasaki
• Xin Zhao (nVidia/Mellanox)

---

New Items

• Tommy is taking over for Josh Ladd for short-term.
  • Please send Mellanox items to him.
  • He will also help with v5 RM work.
• Howard was trying to build OSU benchmark (most recent) doesn't build simply against master and v5
  • Howard didn't have mpicxx or mpicpp
  • If this is an actual issue, assign this to Jeff.
  • Also, Joseph set CC not CCX env, and C++ wraper wasn't being built.
    • This Could be correct behavior even if it's unexpected.

4.0.x

• We're still waiting on Datatype issues now reported in v4.1.1
  • If others can replicate tests/datatype/partial - make check
    • Jeff and George can not get it to fail.
    • If can make it fail with original, then try debug test with lots of output.
  • Two users have reported it with two different environments.
    • concerned.
  • Also run by CI.
  • Test we're talking about is on master (partial.c) This test was not cherry-picked back to release branches.
    • Test is in PRs merged into v4.1.x, but we haven't merged PR to v4.0.x yet.
    • Jeff will check the test on v4.1.x branch.
• Issue 8918 - Another datatype issue we need to look at.
• Need a review for 8898 (and equivalent v4.1.x)

v4.1.x

• In holding pattern waiting for Datatype issue.
• Not taking too many more PRs in case we decide to spin a v4.1.2 with datatype regression fixes

v5.0.0

• Went through a bunch of stuff last week.
• At least 3 PRs pending for v5.0
• Got ROMIO 3.4.1 sync in.
• Bringing Tommy in for nVidia RM.

Master

• Examples and tests directory need to get done.
• Code Refactoring needs to get done.
  • ompi PR 8816 is still open. Need rebasing
  • Could be as easy as running clang-format on HEAD, and merging quickly.
  • Any volunteer?
    • Joseph saw Opal code, some copyright headers got scrambled.
      • Fixed in master and v5.0
    • Macros might need "don't reformat" tags around some macros.
    • includes might need reordering to build properly.
  • May need to stop committing other PRs until this gets done.
  • Nathan responded to a ping during the call and will try to get it done Thursday.
• Should eventually do oshem eventually
• Some folks didn't like the results
  • Macro was one area and that can be address with tags.
• Do we want to set a date to close master if this doesn't get done?
  • Not really, someone should just do it.
• Scope should only be an hour.
• May 14th turn on CI.

Reformatting master

• PR 8816
• Would like Nathan to rebase and merge to master.
• Certain blocks we don't want to format (specifically some in datatype)
  • clang format trips over

PMIx

• Pmix is trying to maintain standards and library versions that are in sync with each other.
  • There is a PMIx standard version and an open PMIx library version.
• Added some PMIx v4.1 standard items to the PMIx v4.0 branch
• Rest the PMIx v4.0 branch without all of v4.1 functionality.
• Open-MPI v5.0 will ship open PMIx v4.1 submodule
  • Will require Prrte 2.0 will require open PMIx v4.1
  • So if running with Open PMIx v4.0 or older, just can't use PRRTE
• Has anyone checked how far back Open-MPI v5 can work with PMIx?
  • At one point verified it worked with open pmix v3.1, but there had been some work on top and need to reverify.

PRRTE v2.0

• No update

Some outstanding work for the way that OMPI calls PRRTE configure.

• Also some changes with libcurl, especially since this breaks OMPI built.
• PMIx can interface with REST interfaces (used by libcurl)
  • JSON
  • Build system issue in PMIx when we changed to static DSOs.
  • Think this has been resolved
• Ralph was looking at this (private messaged Geoff)

issue 8801 - mpirun and prefixing.

• Jeff and Ralph and Yosif had a good conversation
• Lengthy discussion, Summary is, that it's a work in progress.

MTT

• Need to look at the public tests repo for merging in both ULFM and Sessions tests.
• Howard and Geoff will look at this week.
• ULFM is built in by default.
  • Since we don't test it, then it degenerates quickly.
  • At this moment the latest changes to PRRTE has broken ULFM.
  • May be easier to integrate into somewhere else.
  • Some tests put into OMPI-public - this test ran for 4 minutes on 4 nodes
• Would MTT be sufficient for the ULFM testing?
  • It would be a step in the right direction.
  • It WOULD be good to get things into CI.
  • How do we do it without adding more time to CI.
• If someone has one physical box.
  • Open MPI CI is not machines, it's someone needs to set this up and maintain it.
• It's be great if someone in the community could extend the Open-MPI Infrastructure and maintain this.
  • Our CI tests are currently running on a single node.
  • Could be extended, just need volunteers to learn and maintain.

Longer Term discussions

Doc update

• Converting docs to Readthedocs.io
• https://github.com/open-mpi/ompi/pull/8329
• PRRTE issue https://github.com/openpmix/prrte/issues/931 on how to document personalities