cpu_para.pbs

#!/bin/bash

###specify the queue of job
#PBS -q cpu

###specify the name of job
#PBS -N benchmark

###Submit to specified 2 nodes and 4 cores per node
#PBS -l nodes=4:ppn=24
#PBS -l walltime=00:03:30

###prepare env for computing

export  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/intel/mic/coi/host-linux-release/lib/

cd $PBS_O_WORKDIR
cat $PBS_NODEFILE > $PBS_O_WORKDIR/hosts.${PBS_JOBID}
NP=`cat $PBS_NODEFILE | wc -l`
###just replace cpi-mpich with your input file,optional argument lists.

mpirun -np $NP -machinefile $PBS_O_WORKDIR/hosts.${PBS_JOBID} ./Pi.out > Pihex-1M

rm -f $PBS_O_WORKDIR/hosts.${PBS_JOBID}