This HOWTO describes how to run an MPI job on the cluster. Wikipedia would call this article a stub. As I have very little experience with running MPI jobs, I am more or less pasting the script from http://csc.cnsi.ucsb.edu/docs/running-jobs-torque.
The submission script
1 #!/bin/bash
2 #PBS -l nodes=2:ppn=4
3
4 # Make sure that we are in the same subdirectory as where the qsub command
5 # is issued.
6 cd $PBS_O_WORKDIR
7
8 # make a list of allocated nodes(cores)
9 cat $PBS_NODEFILE > nodes
10
11 # How many cores total do we have?
12 NO_OF_CORES=`cat $PBS_NODEFILE | egrep -v '^#'\|'^$' | wc -l | awk '{print $1}'`
13 NODE_LIST=`cat $PBS_NODEFILE `
14
15 # Just for kicks, see which nodes we got.
16 echo $NODE_LIST
17
18 # Run the executable. *DO NOT PUT* a '&' at the end!!
19 mpirun -np $NO_OF_CORES -machinefile nodes ./pi3 >& log