Molpro#
Setting up the environment#
To set up the environment load the version of the software that you want to
use. For example for version 2010.1.25
use:
$ module add molpro/2010.1.25
Scratch files#
Molpro writes a larges amount of temporary or scratch files. By default these
are written to the /tmp
directory on compute nodes. The /tmp
directories
are usually fairly small and may fill up quickly causing jobs to fail. In this
case, to avoid these failures you may need to change the location of the
scratch files. You can redirect these scratch files to another directory, such
as the local SSD, or an area on /nobackup using the -d
argument, e.g.:
# Use locally allocated space
$ molpro -d $TMPDIR
# or if this is insufficient
$ molpro -d /nobackup/your_scratch_directory
The directory /nobackup/your_scratch_directory
is a directory that you create
in the /nobackup
area. Read more about the /nobackup
filesystem
here.
Example job scripts#
Serial job submission#
To run jobs through the batch scheduler, you need to submit them by using the
command qsub <script name>
. Below is an example job script. The example
script below requests 1Gb RAM per core, and a single core for 2 hours, using
the current environment and running from the current working directory. It
runs the molpro script called example.com
.
#!/bin/bash
#$ -l h_rt=2:0:0
#$ -l h_vmem=1G
#$ -V
#$ -cwd
molpro -d $TMPDIR example.com
MPP job submission#
This is the same as the serial version, but asks for 8 cores and uses the MPP mode.
#!/bin/bash
#$ -l h_rt=2:0:0
#$ -l h_vmem=1G
#$ -pe smp 8
#$ -V
#$ -cwd
molprop -n $NSLOTS -d $TMPDIR example.com
MPPX job submission#
This is the same as the serial version, but asks for 8 cores and uses the MPPX mode.
#!/bin/bash
#$ -l h_rt=2:0:0
#$ -l h_vmem=1G
#$ -pe smp 8
#$ -V
#$ -cwd
molprox -t $NSLOTS -d $TMPDIR example.com