Software: Gaussian

In order to use this software, you must be a member of the “gaussian” Linux group, due to licensing restrictions. If you need to be added, please submit a ticket.

Gaussian cannot use multiple hosts, but it can use multiple CPUs on the same host. The Gaussian input file needs to include information about the number of CPUs and amount of memory available (per your SLURM job requests).

Your SLURM (sbatch) input file might look like this

#!/bin/bash #SBATCH --partition=short #SBATCH --mem=4G #SBATCH --job-name=yourjob #SBATCH --output=yourjob.slurm.out #SBATCH --nodes=1 #SBATCH --ntasks-per-node=1 #SBATCH --cpus-per-task=28 #SBATCH --account=youraccount module load gaussian export OMP_NUM_THREADS=${SLURM_CPUS_PER_TASK:-1} # The CPU and memory parameters requested from SLURM should match those in the # Gaussian input file. For the above, use # %NProcShared=28 # %Mem=4GB # Gaussian doesn't seem to use much memory, but you may as well give it at # least 4GB. (The default seems to be 800MB.) \time g09 < yourinput > youroutput

Note again that the file ‘yourinput’ should include the two mentioned “%” lines at the top, like so

%NProcShared=28 %Mem=4GB

You can choose to use fewer or more CPUs, but not more than the hosts in the partition you select have.

If your jobs crashes (or produces an “oom” or “out of memory” error), you should increase the memory request. Gaussian doesn’t seem to use a lot of memory, but for some jobs, you might have to double or quadruple this value.