Being consistent makes sense to me. Question though, if we set:
---
OMPI_MCA_rmaps_base_schedule_policy=node
---
Would that still map -npernode allocations on a bynode basis too? I
think it should.
What's funny is how the old mpirun w/ rsh or ssh from long ago in mpich,
used to basically map by node mapping, and we've come to see us mapping
byslot now. Dang those smp systems. :-)
-cdm
> -----Original Message-----
> From: users-bounces_at_[hidden]
> [mailto:users-bounces_at_[hidden]] On Behalf Of Ralph H Castain
> Sent: Wednesday, December 13, 2006 6:57 AM
> To: Open MPI Users <users_at_[hidden]>
> Subject: Re: [OMPI users] Pernode request
>
>
>
>
> On 12/12/06 9:18 AM, "Maestas, Christopher Daniel"
> <cdmaest_at_[hidden]>
> wrote:
>
> > Ralph,
> >
> > I figured I should of run an mpi program ...here's what it
> does (seems
> > to be by-X-slot style):
> > ---
> > $ /apps/x86_64/system/mpiexec-0.82/bin/mpiexec -npernode 2 mpi_hello
> > Hello, I am node an41 with rank 0
> > Hello, I am node an41 with rank 1
> > Hello, I am node an39 with rank 4
> > Hello, I am node an40 with rank 2
> > Hello, I am node an38 with rank 6
> > Hello, I am node an39 with rank 5
> > Hello, I am node an38 with rank 7
> > Hello, I am node an40 with rank 3
> > ---
>
> Clearly mapping by slot here. Question: our default mapping
> mode is by slot,
> as this example also shows. I could turn the npernode
> behavior around from
> what I previously described - we could default to mapping by
> slot (as we do
> normally, and in concert with what mpiexec appears to be
> doing), and let
> users provide the "-bynode" option to change to mapping by node.
>
> This wouldn't have any affect on the pernode behavior since
> that option only
> maps one proc/node anyway (so mapping by slot or by node
> yields the same
> result).
>
> It seems to me this would be more consistent. Make sense?
> Ralph
>
>
>
|
![[Flag]](../../../../../community/mirrors/flag-us.gif){kind=small}