On 1 October 2016 at 11:37, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:
On 1 October 2016 at 11:26, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:
On 1 October 2016 at 01:21, <tlinnet@xxxxxxxxxxxxx> wrote:
Author: tlinnet
Date: Sat Oct 1 01:21:30 2016
New Revision: 28234
URL: http://svn.gna.org/viewcvs/relax?rev=28234&view=rev
Log:
Added initial script for testing openmpi.
Added:
trunk/devel_scripts/openmpi_test_install.sh
Modified:
trunk/devel_scripts/deploy_google_computing_redhat_6_86_x64.sh
Hi Troels,
I would suggest adding a simple Python hello world script (e.g.
https://github.com/erdc-cm/mpi4py/blob/master/demo/helloworld.py).
That way you can test OpenMPI by itself, mpi4py in pure Python, and
mpi4py in relax. It will isolate the problem to either MPI, mpi4py or
relax.
Here is a simple way to do it from the command line:
[edward@localhost relax-trunk]$ mpirun -np 4 python -c "from mpi4py
import MPI; print('Process %d of %d on %s.' %
(MPI.COMM_WORLD.Get_rank(), MPI.COMM_WORLD.Get_size(),
MPI.Get_processor_name()))"
Process 2 of 4 on localhost.localdomain.
Process 3 of 4 on localhost.localdomain.
Process 0 of 4 on localhost.localdomain.
Process 1 of 4 on localhost.localdomain.
[edward@localhost relax-trunk]$
Or with some more info:
[edward@localhost ~]$ mpirun -np 4 python -c "import mpi4py; from
mpi4py import MPI; print('Mpi4py %s process %d of %d on %s.' %
(mpi4py.__version__, MPI.COMM_WORLD.Get_rank(),
MPI.COMM_WORLD.Get_size(), MPI.Get_processor_name()))"
Mpi4py 1.3.1 process 0 of 4 on localhost.localdomain.
Mpi4py 1.3.1 process 1 of 4 on localhost.localdomain.
Mpi4py 1.3.1 process 2 of 4 on localhost.localdomain.
Mpi4py 1.3.1 process 3 of 4 on localhost.localdomain.
[edward@localhost ~]$
Regards,
Edward
Re: r28234 - in /trunk/devel_scripts: deploy_google_computing_redhat_6_86_x64.sh openmpi_test_install.sh