Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] btl/openib: get_ib_dev_distance doesn't see processes as bound if the job has been launched by srun
From: Ralph Castain (rhc_at_[hidden])
Date: 2012-02-17 09:42:37


I took a closer look at this, and I think we're getting ourselves confused
by the rather large differences between what is on the trunk vs the 1.5
branch. The trunk is doing the "am I bound" calculation correctly - it gets
the cpubind bitmask and compares it to the allowed/available cpus.

The 1.5 branch has a problem. Jeff and I discussed it a little more, and
agreed that I will create a minimal patch to address the issue of
direct-launched procs. We definitely don't want to back-port the logic from
the trunk.

Will try to have something next week.

On Fri, Feb 17, 2012 at 6:03 AM, Jeff Squyres <jsquyres_at_[hidden]> wrote:

> On Feb 16, 2012, at 8:16 AM, nadia.derbey_at_[hidden] wrote:
>
> > Could you please move it to v1.5 (do I need to fill a CMR)?
>
> Just to clarify - you're asking for the patch to set WHOLE_SYSTEM when we
> load the hwloc topology, right?
>
> If so, please file a CMR. Note that there's some differences between how
> hwloc is used between the trunk and the v1.5 branch; the same commit may
> not apply exactly from the trunk to v1.5.
>
> --
> Jeff Squyres
> jsquyres_at_[hidden]
> For corporate legal information go to:
> http://www.cisco.com/web/about/doing_business/legal/cri/
>
>
> _______________________________________________
> devel mailing list
> devel_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>