Gaussian

How To Use Gaussian 16 On The Cluster

Join the users group

Gaussian is a proprietary computational chemistry software package. Therefore, to use the package on the cluster you need to be part of the gaussian users group. To get into the group you can contact the Chemistry Department in the College of Science.

Slurm Example

Slurm is our cluster management and job scheduling system. You can read more about it here and here.

This is an example Slurm batch script that you can start with and modify as you see fit.

#!/bin/bash
#
#SBATCH --job-name=g16
#SBATCH --output=g16.%j.out
#SBATCH --partition=compute
#
#SBATCH --time=10:00:00

# Setting up the environment
export g16root="/data/software/packages/Gaussian-16.B.01"
export GAUSS_SCRDIR="/data/datasets/$USER/g16scr/$SLURM_JOB_ID"
source $g16root/g16/bsd/g16.profile

mkdir -p $GAUSS_SCRDIR

# Running the program
g16 input.gjf

Write the above script in a file and name it g16.sbatch for example. Submit the job using sbatch:

sbatch g16.sbatch

Processing Multiple Files on Multiple Compute Nodes

Instead of submitting multiple sbatch jobs to process multiple files. This is a single sbatch script that can launch multiple instances of g16 on different nodes.

First create the batch script:

#!/bin/bash
#
#SBATCH --job-name=g16
#SBATCH --output=g16.%j.out
#SBATCH --partition=compute
#SBATCH --nodes=4
#SBATCH --time=10:00:00

# Setting up the environment
export g16root="/data/software/packages/Gaussian-16.B.01"
source $g16root/g16/bsd/g16.profile

# Running the instances using a launger scrpit
srun ./launch_g16.sh

Then we need a launcher script, we will call it launch_g16.sh in this example:

#!/bin/sh

export GAUSS_SCRDIR="/data/datasets/${USER}/g16scr/${SLURM_JOB_ID}/job_${SLURM_NODEID}"
mkdir -p ${GAUSS_SCRDIR}

filename=job_${SLURM_NODEID}.gjf

echo "`hostname` is going to process ${filename}"
g16 ${filename}

Make the file executable by using chmod:

chmod u+x launch_g16.sh

Rename you .gjf files that you want to process in parallel following this example:

job_0.gjf
job_1.gjf
job_2.gjf
job_3.gjf

You can change this naming scheme as you want but make sure to reflect the change in the launch_g16.sh script too.

Finally, submit the batch job using sbatch:

sbatch multi_g16.sbatch