Difference between revisions of "Submit Matlab jobs on Abel"
From mn/geo/geoit
(Created page with "<div class="vrtx-introduction"> This is an example of a Matlab submit script for use on the Abel cluster. Abel usage is documented at [http://www.uio.no/english/services/it/re...") |
|||
| Line 6: | Line 6: | ||
For normal off-line non-parallell Matlab running on any Linux server, you can simply do e.g. | For normal off-line non-parallell Matlab running on any Linux server, you can simply do e.g. | ||
<pre>nohup matlab -nodisplay -nojvm -nodesktop -nosplash < Matlabprog.m > outdata.txt & | <pre>nohup matlab -nodisplay -nojvm -nodesktop -nosplash < Matlabprog.m > outdata.txt & | ||
| − | </pre> | + | </pre> |
For large jobs, this script should set you up OK at the Abel cluster: | For large jobs, this script should set you up OK at the Abel cluster: | ||
<pre>#!/bin/bash | <pre>#!/bin/bash | ||
| Line 12: | Line 12: | ||
#SBATCH --job-name=Matlab | #SBATCH --job-name=Matlab | ||
# | # | ||
| − | # Project: | + | # Project (change to your : |
| − | #SBATCH --account= | + | #SBATCH --account=geofag |
# | # | ||
# Wall clock limit: | # Wall clock limit: | ||
| Line 20: | Line 20: | ||
# Max memory usage per core (MB): | # Max memory usage per core (MB): | ||
#SBATCH --mem-per-cpu=2000M | #SBATCH --mem-per-cpu=2000M | ||
| − | + | # | |
# Number of tasks (cores): | # Number of tasks (cores): | ||
#SBATCH --ntasks=1 | #SBATCH --ntasks=1 | ||
| + | # | ||
| − | + | ## Setup job environment | |
| − | ## | + | source /cluster/bin/jobsetup |
| − | source / | ||
# Check if we have enough input arguments | # Check if we have enough input arguments | ||
| − | if [ $# -lt 2 ]; then | + | </pre><pre>if [ $# -lt 2 ]; then |
# No, print a usage message and exit script | # No, print a usage message and exit script | ||
echo echo -e "\nUsage: sbatch submitscript.sh program dataset\n" | echo echo -e "\nUsage: sbatch submitscript.sh program dataset\n" | ||
| Line 45: | Line 45: | ||
PROG=$1 | PROG=$1 | ||
DATA=$2 | DATA=$2 | ||
| − | + | </pre><pre># Copy files to work directory: | |
| − | # Copy files to work directory: | ||
cp $SUBMITDIR/$PROG $SCRATCH | cp $SUBMITDIR/$PROG $SCRATCH | ||
cp $SUBMITDIR/$DATA $SCRATCH | cp $SUBMITDIR/$DATA $SCRATCH | ||
Revision as of 11:49, 5 January 2015
This is an example of a Matlab submit script for use on the Abel cluster. Abel usage is documented at UiOs HPC pages
For normal off-line non-parallell Matlab running on any Linux server, you can simply do e.g.
nohup matlab -nodisplay -nojvm -nodesktop -nosplash < Matlabprog.m > outdata.txt &
For large jobs, this script should set you up OK at the Abel cluster:
#!/bin/bash # Job name: #SBATCH --job-name=Matlab # # Project (change to your : #SBATCH --account=geofag # # Wall clock limit: #SBATCH --time=1000:0:0 # # Max memory usage per core (MB): #SBATCH --mem-per-cpu=2000M # # Number of tasks (cores): #SBATCH --ntasks=1 # ## Setup job environment source /cluster/bin/jobsetup # Check if we have enough input arguments
if [ $# -lt 2 ]; then# No, print a usage message and exit script echo echo -e "\nUsage: sbatch submitscript.sh program dataset\n" exitfi
- Load matlab module
module load matlab
- Make a result directory
RESULT=$SUBMITDIR/Result_`date +%d_%b_%H.%M.%S` mkdir -p $RESULT
- Save the name of the matlab program and the dataset
PROG=$1 DATA=$2
# Copy files to work directory:cp $SUBMITDIR/$PROG $SCRATCH cp $SUBMITDIR/$DATA $SCRATCH
- Change directory to the work directory
cd $SCRATCH
- Start matlab
matlab -nodisplay -nojvm -nodesktop -nosplash < $PROG
- Copy the necessary files from $SCRATCH to the Result directory
cp $SCRATCH/*.txt $RESULT cp $SCRATCH/*.eps $RESULT
- End of script
