WeeklyTelcon_20160315

Open MPI Weekly Telcon

---

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

Attendees

• Jeff Squyres
• Geoff Paulsen
• Brad Benton
• Howard
• Josh Hursey
• Nathan Hjelm
• ralph
• Ryan Grant
• Sylvain Jequgey
• Todd Kordenbrock
• Yohann Burette

---

Agenda

Review 1.10

• Milestones: https://github.com/open-mpi/ompi-release/milestones/v1.10.3
  • PR 1004 - MPI_Ineighbor_alltoallw
    • Needs reviewer
  • PR 1002 - Memory allocation hooks
    • Jeff to review
  • PR 1006 - Ralph will review.
  • PR 1008 - Jeff already reviewed.
  • Anyone testing on PSM2 or Omnipath? Intel guys have been. Howard - Ralph check this.
    • Symbol confusion fix requires 1.10.2, and need newer Omnipath (PSM2) library/driver.
• 1.10.3 in Late April - Some smaller fixes accumulating in the branch
  • Nothing critical at the moment

Review 2.0.x

• Wiki: https://github.com/open-mpi/ompi/wiki/Releasev20
• Blocker Issues: https://github.com/open-mpi/ompi/issues?utf8=%E2%9C%93&q=is%3Aopen+milestone%3Av2.0.0+label%3Ablocker
  • Issue 1425 - External PMIx server support
    • Ralph working on a fix - should be quick.
  • Issue 1418 - fix MPI process suicide code
    • Delayed until 2.1 or 3.0
  • Issue 1406 - TCP BTL THREAD_MULTIPLE deadlock
    • Nathan - George is working on a fix, but it is a rewrite. So might take some time.
      • If old and new rewrite of TCP BTL are "compatible", then we can switch based on threaded state of MPI_Init.
      • OR could require two different TCP components "tcp" / "tcpmt", and expose this issue to users.
  • Issue 1353 - -host behavior
    • Ticket has been updated with new commits. Jeff to test.
    • Ralph is out of time to work on.
    • Need to document behavior in different releases, then close this for v2.0.0.
• Milestones: https://github.com/open-mpi/ompi-release/milestones/v2.0.0
  • PR 1003 - Race condition in process matching thread
    • Nathan to review the patch and sign off.
  • PR 1000 - misc warnings and missing include files
    • Ralph to review and update
  • PR 977 - span on heterogeneous clusters
    • Ralph to review and update
  • PR 973 - Parsing of envvars in MCA
    • Nathan pushing update, Jeff to review after the call.
• Reviewed all Pull Requests, and pinged a few for comments.
• Really need more people doing testing on v2.0 branch
• Decision to shoot for April 24th for RC0 of 2.0.
• Jeff saw someone is getting OPAL-FIFO is failing on 2.0.
• Need more thread safety testing.
  • Jeff added his thread safety tests are a 2 night cycle to do all the tests.
• Jeff seeing SIGPIPEs ONLY on master with usNIC - Ralph wonders if it's valgrind issue Mellanox was seeing.

Misc

• MPI Forum - MPI_Info under discussion
  • Don't propagate infos with MPI_Comm_dup - use MPI_Comm_dup_with_info to propagate infos
  • Still discussion about MPI_Info_get/MPI_Info_set behavior
• Open MPI Developer's Meeting
  • General Notes
  • Geoff Additional Notes
• Nathan - Enabling thread_multiple all the time
  • PR 1397 - always enable MPI_THREAD_MULTIPLE support
  • Should we turn this on for everyone? Generally feeling is to accept this
  • Send another note to the devel list to give folks one last change to comment before commit.
• Need to do some performance testing
  • v2.0.0 better MPI_THREAD_MULTIPLE correctness
  • focus on performance improvements in next v2.X series release

Review Master?

• https://github.com/open-mpi/ompi/pull/1417: "RFC: change default build to always be optimized (even for developers)" If no one has any further comments, it's time to merge.
  • Jeff thinks we are in consensus, but wants to check with developers.
  • To turn on --enable-debug, --enable-memdebug, --enable-picky.
• Nathan - heads up that mpool re-write is ready to go.
  • will get merged in afternoon today.

MTT status:

• Really need more people doing testing on v2.0 branch

Status Updates:

• delayed until next week.

---

Status Update Rotation

1. Cisco, ORNL, UTK, NVIDIA
2. Mellanox, Sandia, Intel
3. LANL, Houston, IBM

---

Back to 2016 WeeklyTelcon-2016