On Thu, Jan 13, 2011 at 05:34:48PM -0800, Tena Sakai wrote:
> Hi Gus,
>
> > Did you speak to the Rmpi author about this?
>
> No, I haven't, but here's what the author wrote:
> https://stat.ethz.ch/pipermail/r-sig-hpc/2009-February/000104.html
> in which he states:
> ...The way of spawning R slaves under LAM is not working
> any more under OpenMPI. Under LAM, one just uses
> R -> library(Rmpi) -> mpi.spawn.Rslaves()
> as long as host file is set. Under OpenMPI this leads only one R slave on
> the master host no matter how many remote hosts are specified in OpenMPI
> hostfile. ...
> His README file doesn't tell what I need to know. In the light of
> LAM MPI being "absorbed" into openMPI, I find this unfortunate.
Hmm. It has been a while that I had to compile Rmpi, but the following
works with openmpi-1.3.3, R-2.10.1:
# mpiexec -n 1 -hostfile mfile R --vanilla < Rmpi-hello.R
with a script Rmpi-hello.R like
library(Rmpi)
mpi.spawn.Rslaves()
mpi.remote.exec(paste("I am",mpi.comm.rank(),"of",mpi.comm.size()))
mpi.close.Rslaves()
mpi.quit()
The only unfortunate effect is that by default mpi.spawn.Rslaves()
spawns as many slaves as there are lines in the hostfile, hence you
end up with one too many processes: 1 master + N slaves. You can repair
that by using
Nprocs <- mpi.universe.size()
mpi.spawn.Rslaves(nslaves=Nprocs-1)
instead of the simple mpi.spawn.Rslaves() call.
BTW: the whole script works in the same way when submitting under torque
using the TM interface and without specifying -hostfile ... on the
mpiexec command line.
Cheers,
Martin
--
Martin Siegert
Head, Research Computing
WestGrid/ComputeCanada Site Lead
IT Services phone: 778 782-4691
Simon Fraser University fax: 778 782-4242
Burnaby, British Columbia email: sieg...@sfu.ca
Canada V5A 1S6
