Note that we have very few Abaqus research licenses. Please only use Abaqus if you have specifically spoken to the BisonNet team (bisonnet@bucknell.edu) about your usage.
By default, Abaqus uses /tmp as its scratch directory, but that directory is not very large and is shared by all users. It’s recommended that you create your own scratch directory with the command:
mkdir ~/abaqustmp
and then specify that directory with your Abaqus command:
abaqus scratch=$HOME/abaqustmp
You should specify the memory that Abaqus will use using the --mem-per-cpu
parameter to the SLURM workload manager and the memory=
option to Abaqus. For example:
#SBATCH --mem-per-cpu=8192 # memory per core
abaqus memory="6 gb"
Note that the --mem-per-cpu
value should always exceed the memory=
value. See the full example below for more details.
You should specify the number of CPU cores Abaqus will use using the -n
parameter to the SLURM workload manager and the cpus=
option to Abaqus. For example:
#SBATCH -n 4 # number of cores
abaqus cpus=$SLURM_NTASKS
The $SLURM_NTASKS
uses the value from #SBATCH -n
to allocate the correct number of CPU cores.
Here’s a full example job script for running Abaqus
#!/bin/bash
#SBATCH -p short # partition (queue)
#SBATCH -N 1 # number of nodes (don't change this value)
#SBATCH -n 4 # number of cores
#SBATCH --mem-per-cpu=8192 # memory per core
#SBATCH --job-name="abaqusjob" # job name
#SBATCH -o slurm.%N.%j.stdout.txt # STDOUT
#SBATCH -e slurm.%N.%j.stderr.txt # STDERR
#SBATCH --mail-user=username@bucknell.edu # address to email
#SBATCH --mail-type=ALL # mail events (NONE, BEGIN, END, FAIL, ALL)
# required for Abaqus and SLURM
unset SLURM_GTIDS
module load SIMULIA
abaqus job=jobname cpus=$SLURM_NTASKS input=file.inp scratch=$HOME/abaqustmp memory="6 gb" interactive