Gaussian

Important

Access to Gaussian requires permission. Please open a help ticket by emailing help@smu.edu with “[HPC]” in the subject line.

Gaussian Gaussian is a series of electronic structure programs, used by chemists, chemical engineers, biochemists, physicists and other scientists worldwide.

Gaussian homepage: gaussian.com/

Official Gaussian Manual: gaussian.com/man/

See also

For examples and tips on submitting jobs, see our SLURM documentation and Best Practices for Jobs

For compute resources, see HPC Queues

Using Gaussian on M3

Files used in the example are available:

• on github

• on M3 at /hpc/m3/examples/gaussian/

Example submission script

The following job script can be submitted using sbatch guassian_cpu_example.sbatch.

This example will run in about 10 seconds and use less that 1GB of memory. In general, more complicated simulations will take much longer and should typically use more cores and memory. Note, we know this from running the job. It is always a good idea to review the resources your jobs use and adjust future jobs to more accurately request resources.

This uses the dev queue which has a 2 hour time limit. In general, the dev queue should only be used for testing code and running interactive sessions. Most normal jobs should be submitted to the standard queues such as standard-s where longer runtimes are allowed.

This example also writes output to $SCRATCH. $SCRATCH is a high performance file system designed to be used for temporary job files and data. After a job has finished running, any data you need to keep should be moved to a project directory or your $HOME directory. Files in SCRATCH are subject to a 60 day purge policy where files older than 60 days may be automatically deleted without warning.

Note

This job script will not run without modification. In particular, you must change the account from peruna_project_0001 to the account name for an allocation you have access to.

#!/bin/bash
#SBATCH -J gaussian            # job name shown in queue
#SBATCH -o gaussian_%j.out     # output file, %j is the job id number
#SBATCH -p dev                 # request the dev queue
#SBATCH -c 16                  # request 16 CPU cores
#SBATCH --mem=1G               # Request 1GB of memory
#SBATCH -t 00:01:00            # request 1 minute
#SBATCH -A peruna_project_0001 # example account

# specify the input file
input_file=gaussian_example.cpu

# module purge clears any existing modules so only the modules
# requested will be loaded. This improves reproducibility
module purge

# load gaussian. It is a good practice to include the version
# the default modules without a version may change when software
# is updated
module load gaussian/g16c/zen3

# create a temporary job directory in scratch
# copy the input file and cd into it
# Note: scratch is intended for temporary job storage
# you should move any files you need to keep when the job
# completes and delete unneeded files. Files in $SCRATCH
# are subject to automatic deletion after 60 days.
job_dir=${SCRATCH}/${SLURM_JOB_ID}
mkdir -p ${job_dir}
cp ${input_file} ${job_dir}
cd ${job_dir}
GAUSS_SCRDIR=${job_dir}

# gets requested mem in GB
sleep 5 # to make sure the job info is available
mem=$(sacct -o "ReqMem" --units=G -j ${SLURM_JOB_ID} -n | xargs)

# get the cores assigned to the job
# This is M3 specific, though something similar will work on many systems
cpus=$(cat /sys/fs/cgroup/system.slice/slurmstepd.scope/job_${SLURM_JOB_ID}/cpuset.cpus)

# make sure mem is an int with correct units
mem=${mem%.*}
mem=${mem//[!0-9]/}
mem="${mem}GB"

# this function fills in the cpu and memory information into the
# input file
cpu_mem() {
  sed -i -e "/^%CPU/c\%CPU=${1}" ${input_file}
  sed -i -e "/^%Mem/c\%Mem=${2}" ${input_file}
}

cpu_mem ${cpus} ${mem}

# run gaussian
g16 < ${input_file}

# clean up temp $SCRATCH directory
# this moves the rwf file from Gaussian to the
# directory you submitted the job from. 
# Note, if the job runs out of time, memory, or
# fails in some other way, these commands may not be reached.

# You should make sure you are keeping the files you need
# this is just an example of a plausible workflow.
# For instance, you may want to keep the checkpoint files
mv *.rwf ${SLURM_SUBMIT_DIR}/
cd ${SLURM_SUBMIT_DIR}
rm -rf ${job_dir}