[gmx-users] Replica exchange shell script

[Sukit Leekumjorn]

Dear GMX users;

I have a question about running replica exchange MD on parallel 
machines.  Current, I am able to run a single job in parallel using 
Gromacs 3.3 official version.  To submitted and run this job I have to 
have a shell script.  My question is  how to write a shell script to 
submitted a job in parallel using REMD.  Below is the shell script that 
I use to submitted a job running with 12 nodes.  I have seen several 
emails on replica exchange but none had mentioned using a shell script.  
I have four md#.tpr files "grompp" at different temperatures (did not 
use -np option) and I want to run using 4 nodes.  I tried to modify the 
line "export EXEC=" line to match what others have posted on mailing 
list but it does not seem to work.  If anyone had experience with this, 
please give me some insight on how to summited REMD job in parallel 
using a shell script.

Thank you very much,
Sukit

#!/bin/bash

## The time limit for the job.  If it runs over this it will be killed.
## -lwalltime=0:05:0 would be five minutes.
#PBS -lwalltime=150:00:00

#PBS -W group_list=heroes
#PBS -q production_q
#PBS -lnodes=6:ppn=2

#$ -notify

#PBS -A jobx

#PBS -N pe128_330

NUM_NODES=`cat $PBS_NODEFILE | wc -l | sed "s/ //g"`

## Get in to the directory qsub was run from
cd $PBS_O_WORKDIR

export PATH=/nfs/software/bin:$PATH

export FF=/nfs/storage1/users/gromacs3.3_022006/bin
export EXEC="$FF/mdrun_3.3_gcc_mpi -s 1pe330_2.tpr -np 12 -deffnm 1pe330_2"

echo starting

date
echo "/// Running on:"
cat $PBS_NODEFILE
echo "///"
echo
jmdrun -printhostname -np $NUM_NODES -hostfile $PBS_NODEFILE $EXEC
echo
date

echo ending

exit;