ioam / topographica

A general-purpose neural simulator focusing on topographic maps.
topographica.org
BSD 3-Clause "New" or "Revised" License
53 stars 32 forks source link

using openmp or mvapich2 library to process gcal.py #660

Closed windform2015 closed 8 years ago

windform2015 commented 8 years ago

Hi, I would like to ask you the help check : I am trying to use openmp queues in the cluster to run gcal.py and I put the command about the number of threads to use in the head of the script like this : "

!/usr/bin/env python

import os

os.environ['OMP_NUM_THREADS']='10'

" But I am not sure this is correct, since when I set OMP_NUM_THREADS=10, the number of cores the job use is actually about 6. When I set OMP_NUM_THREADS=4, the actually number of cores the job use is about 3.

And to use openmp, it only requires that the jobs are submitted to the queues for openmp, right ? Of course, I can set the number of threads as environmental variable like stated above.

In addition, could gcal.py be parallel processed by mvapich2 ? I tested a job in mvapich2 queue and it seems that mvapich2 process the same job faster than openmp does, but I am not sure the job run properly.

Thanks for your attention.