-
Notifications
You must be signed in to change notification settings - Fork 859
WeeklyTelcon_20201110
- Dialup Info: (Do not post to public mailing list or public wiki)
- NOT-YET-UPDATED
- v4.0.6 toward the end of the month.
- 7199 - ras based orted on headnode.
- No other blockers currently on
- Geoff will create milestone issue for v4.0.6
- PR 8187 - deterministic build.
- rhc will review.
- Bug in ORTE that Ralph fixed last night for AWS. This hasn't gone back to v4.1. (PR8176)
- It's technically more correct, so probably doesn't matter.
- Weren't communicating the cpuset correctly. OFI MTL needed it in v4.1.
- Jeff will see which BTL.
- Prob not critical.
- Ralph will pull it out and PR it for v4.0.x
- Did make another RC last night, Jeff will send out email this morning.
- First to include HAN and Adapt.
- Intent is that we'd activate both together.
- Jeff will email with instructions on how to enable.
- First to include HAN and Adapt.
- Warnings have gotten out of hand on v4.1 on MAC.
- Some warnings would be fixed if we just fixed some macros. A bunch of unused variables.
- In general not critical.
- /bin/sh command subst error (unexpected EOF), so that's probably a critical error.
- Coverity only runs on master - Coverity then
-
Ralph is working on updating PMIx / PRRTE submodule pointers.
- Jeff is helping with configury issues.
-
Hoping these are one-time issues, and not every time we update submodule pointers.
- All new configury.
- If this isn't a one-time thing, we should consider stop embedding this.
- Submodules are good for a number of things, but you ahve to take a change in configury, Go through CI, commit it to PMIx master, then go through OMPI CI, and find out that there's a bug. Then you have to go back to PMIx master.
-
Cisco has some of this.
- Should figure out a keyword on PMIx side, that effectively does
-
Hit a problem in AWS Amazon-Liunux1 trying to build a tarball.
-
And were hitting Mellanox CI due to Python older version.
-
IBM doesn't do dist check - bug is in there (real issue)
-
Some issues are caused by Embedding. (What do you mean by SRCDIR) (Which SRCDIR?)
-
But other issues are NOT caused by embedding pmix.
- So getting rid of embedding would not solve these issues
-
Not moment it's looking like 2Q next year.
- IBM's been pushing on PRRTE as well. We're testing the map/bind options.
- Trying to push tickets up (some are clarification of expected behavior)
- Community help with these issues would help move forward the prrte deadline.
-
What's going to be the state of the SM Cuda BTL and CUDA support in v5.0?
- What's the general state? Any known issues?
- AWS would like to get.
- Josh Ladd - Will take internally to see what they have to say.
- From nVidia/Mellanox, Cuda Support is through UCX, SM Cuda isn't tested that much.
- Hessam Mirsadeg - All Cuda awareness through UCX
- May ask George Bosilica about this.
- Don't want to remove a BTL if someone is interested in it.
- UCX also supports TCP via CUDA
- PRRTE CLI on v5.0 will have some GPU functionality that Ralph is working on
-
PR 8191 - converted all OMPI READMEs to markdown.
- Looks nicer when browsing on Github (can do formatting)
- Do things in one markdown language (easy to edit)
- Master only.
-
Should we consolidate Top level README, website FAQ (sorta googlable), All the manpages.
- If 2 of these are going to Markdown, maybe we should do FAQ in markdown, and put these all in one place like readthedocs.io
- Unless someone has an allergic reaction, Jeff's interested in working on this.
- They have a decent versioning scheme to version docs based on release(s).
- When would the docs get pushed to readthedocs.io?
- Master would be on a github hook after PRs are merged.
- Stable Release branches would go out at release time.
- LICENSE question - what license would the docs be available under? Open-MPI BSD license, or
- readthedocs.io encourages "restructured text" format over markdown.
- They also support a hybrid for projects that have both.
- Thomas Naughton has done the restructured text, and it allows
-
Ralph tried the Instant on at scale:
- 10,000 nodes x 32PPN
- Ralph verified Open-MPI could do all of that in < 5 seconds, Instant-On.
- Through MPI_Init() (if using Instant-On)
- TCP and Slingshot (OFI provider private now)
- PRRTE with PMIx v4.0 support
- SLURM has some of the integration, but hasn't taken this patch yet.
-
Discussion on:
- Draft Request Make default static https://github.com/open-mpi/ompi/pull/8132
- One con is that many providers hard link against libraries, which would then make libmpi dependent on this.
- Talking about amending to request MCAs to know if it should be slurped in.
- (if the component hard links or dlopens their libraries)
- Roadrunner experiments... The Bottleneck in launching was I/O in loading all the .sos
- spindle, and burst buffer reduce this, but still
- Still going through function pointers, no additional inlining.
- can do this today.
- Still different than STATIC (sharing this image across process), just not calling dlopen that many times.
- New proposal is to have a 3rd option where component decides it's default is to be slurped into libmpi
- It's nice to have fabric provider's not bring their dependencies into libmpi so that the main libmpi can be run on nodes that may not have the provider's dependencies installed.
- Low priority thing anyway, if we get it in for v5.0 it'd be nice, but not critical.
- George and Jeff are leading
- No new updates this week (see last week)