forked from open-mpi/ompi
-
Notifications
You must be signed in to change notification settings - Fork 4
WeeklyTelcon_20160501
Geoff Paulsen edited this page May 3, 2016
·
20 revisions
- Dialup Info: (Do not post to public mailing list or public wiki)
- Geoff Paulsen
- Jeff Squyres
- Brad Benton
- Howard
- Josh Hursey
- Joshua Ladd
- Nathan Hjelm
- ralph
- Todd Kordenbrock
- Milestones: https://github.com/open-mpi/ompi-release/milestones/v1.10.3
- Two pull requests that have not been reviewed:
- NULL datatypes - Jeff's looking at some oddities to see if we care.
- distinct namespaces for OSC for pt2pt. - Nathan thinks it's fine.
- Blocker for "PSM stuff" - Howard reviewed this. Good with this.
- Jeff owes a library versioning review. already updated for 1.10.3.
- Two pull requests that have not been reviewed:
- 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 *
- Milestones: https://github.com/open-mpi/ompi-release/milestones/v2.0.0
*
- Looking pretty good, until Paul found a bunch of obscure things.
- have most of them either fixes, or have issues or PRs to fix them.
- Nathan has - can't clobber EBX - would hope the compiler would put store/restore around it.
- Fix against master in.
- 32bit powerpc issue in hook.
- PR 1129 - ralph pulled the fix, waiting for paul's test result.
- PR 1133 - trivial.
- PR 1134 - OMPIO comp on netbsd - Paul queued tests.
- Howard queued up some Readme changes
- PR 1051 - marked for 2.1, but is annoying, would like to pull back to 2.0
- Howard is okay.
- Nathan kinda wants PR1127 in v2.x - OSC correctness fixes. Fixes map-by node for Graph500.
- Important for Mellanox, v2.0.0
- Howard is concerned about the change churn to put this into v2.0.0, and would prefer this in v2.0.1
- master PR1617 - hcoll, hang in Finalize with srun - Mellanox would prefer v2.0.0
- Fix on 1.10, but not on master or 2.x, but haven't opened PR for v2.x yet (today).
- Looking pretty good, until Paul found a bunch of obscure things.
- Discussion:
- What "gotchas" do we need to communicate to users? I.e., what will people upgrading from v1.8.x/v1.10.x be surprised by?
- The most obvious one I can think of is mpirun requiring -np when slots are not specified somehow.
- OMPIO is not default for Luster - Edgar will writeup blurb.
- Discuss version numbering change.
- What else do we need to communicate? It would be nice to avoid the confusion users experienced regarding affinity functionality/options when upgrading from v1.6 -> v1.8 (because we didn't communicate these changes well, IMHO).
- https://github.com/open-mpi/ompi/wiki/User-Migration-Guide:-1.8.x-and-v1.10.x-to-v2.0.0
- https://github.com/open-mpi/ompi/wiki/Developer-Migration-Guide:-v1.8.x-and-v1.10.x-to-v2.x
- Want it to be googlable.
- A couple of paragraphs or 3 on biggest changes.
- Removed support.
- We need to collectively edit on wiki, and then we'll put it up on the open-mpi website.
- new OSHMEM interfaces added, but still not implemented until 2.1
- Biggest change is job launch / stuff to support (Josh)
- PMI support changed, it's a framework now, expect orte_info components.
- New RMA capabilities (Nathan)
- What "gotchas" do we need to communicate to users? I.e., what will people upgrading from v1.8.x/v1.10.x be surprised by?
Review Master MTT testing (https://mtt.open-mpi.org/)
- Mellanox, Sandia, Intel
- LANL, Houston, IBM
- Cisco, ORNL, UTK, NVIDIA