Re: mpirun problem using pypar

[Ole Nielsen]

Hi Juan Carlos.

I think the different underlying MPI implementations differ slightly there. Some require a host file listing the nodes (and the number of cpus on each) to be used.
I have attached the file we use for our cluster.

We usually use OpenMPI on Linux with pypar and a typical invokation (for one of our codes) would be as follows:

mpirun -x FALL3DHOME -x PYTHONPATH -hostfile /etc/mpihosts -host node<#>,node<#> python runscript.py

For more details on hostfiles see
e.g. http://linux.die.net/man/1/mpirun or http://www.open-mpi.org/faq/?category=running

Hope this helps
Cheers
Ole

On Thu, Sep 1, 2011 at 9:09 PM, juan carlos beamin <jcbe...@gmail.com> wrote:

Hi

I'm trying to use pypar but I have a problem and i didn't found the answer on the web.

Pypar detects just one processor in my computer(i have a intel core i7)

When I run a program in c using mpirun -np <> just detect one processor nevermind how many processors i say it to use, but when I use explicit mpirun.mpich2 -np <> it uses as many processors as i asked.

Do you think that there is a way to fix these (in general or in my particular case?) 

I have tried to uninstall every version of mpi(open mpi. mpich mpich2) and start from zero but I always get the same resuts.

thanks a lot for your time

best regards

--
Juan Carlos Beamin M.