Ralph,
what i had in mind is ignore dynamically spawned orted for the time being,
and only consider the number of orted spawned by mpirun at MPI_Init() time.
the rationale is that "in the real world", mpirun is invoked via a
batch manager (that does not resize jobs) , and/or with a machinefile
or
Hi Matias,
i read putenv() instead of setenv(), so the code is correct as it is,
sorry about the noise
as far as i am concerned, i'd rather add a FAQ in the open-mpi.org web site.
you can simply issue a PR vs the https://github.com/open-mpi/ompi-www.git repo
Cheers,
Gilles
On Fri, Sep 30, 2016
No simple solution, I fear. I know Matias et al are looking at the dynamic
situation. Getting the number of orted, especially if/when they can be
dynamically spawned, requires use of the notification method - in the plans,
but not yet implemented.
> On Sep 29, 2016, at 6:13 PM, Gilles Gouailla
Hey Gilles,
Quick answer on the first part until I read a little more about num_max_procs :O
Being the third parameter of setenv a 0 means: do not override if found in the
env. So the workaround does work today. Moreover, I would like to know if
there is a place in some OMPI wiki to document t
This is a follow-up of
https://mail-archive.com/users@lists.open-mpi.org/msg30055.html
Thanks Matias for the lengthy explanation.
currently, PSM2_DEVICES is overwritten, so i do not think setting it
before invoking mpirun will help
also, in this specific case
- the user is running within
Hi Nick,
> I agree that symmetric objects in the data segment are a challenging issue.
> Could dynamically-allocated symmetric objects be supported much more easily?
Indeed. At least on the technical level you don't depend on a 3rd
party kernel module.
> Does Open MPI / OSHMEM already have cross
Hi Pasha,
I agree that symmetric objects in the data segment are a challenging issue.
Could dynamically-allocated symmetric objects be supported much more
easily? Does Open MPI / OSHMEM already have cross-process mappings for
symmetric allocations (e.g., for local put/get via memcpy)?
Thanks,
Ni
