Skip to content

Instructions for using MATLAB software

MATLAB is programming platform for analysing and designing systems, primary for numeric computing. Supports matrix manipulations, data plotting, implementation of algorithms, development of user interfaces, etc. Programs written in MATLAB, can use libraries written in C, C++, Java and Fortran. Developed by MathWorks, read more.

License

MATLAB is a licensed software, HPC Vega offers Standard Concurrent license, which is intended to be used by multiple users simultaneously.

  • MATLAB (Total of 200 licenses issued)
  • MATLAB Parallel Server (Total of 4000 licenses issued)

Academic users can freely use MATLAB on HPC Vega, commercial users have to bring their own license. For more information ask support@sling.si.

Alternative: use your own license

An alternative option for using your own license is supported. If you want to use your own license, you have to export environment variable for license file.

export MLM_LICENSE_FILE=/path/to/file/license.dat

Various MATLAB versions are available on HPC Vega, through modules. 

module load matlab

MATLAB user interfaces

For the graphical user interface (GUI), X11 forwarding must be enabled on the client site, otherwise an interactive command line session will begin. The easiest way to check this is echo $DISPLAY, if value is not specified, the forwarding does not work.

Graphical user interface (GUI)

matlab -display Xdisplay

To start interactive session, first you need to request an interactive allocation from Slurm. 

salloc --job-name=matlab-int \
       --time=00:30:00 \
       --partition=cpu \
       --nodes=1 \
       --ntasks=1 \
       --cpus-per-task=8 \
       --x11

Launch tasks on allocated resources:

srun --pty bash

Load module: 

module load matlab

Run the script with graphical components (for example plots):

matlab -nodesktop -nosplash -r "run('matlabtest.m');"

Command line interface (CLI) only:

matlab -nodisplay

For additional information about MATLAB commands, load module from your shell and check the help section.

matlab -h

Example of a Slurm job

The following example shows a task that will be run on 1 node and 8 cores. The job requires 4 GB RAM per core and will be carried out for 5 minutes on cpu partition.

Content of inputfile matlabtest.m:

function matlabtest
   disp('Hello from MATLAB!')
end

Content of SBATCH script:

#!/bin/bash

#SBATCH --job-name=matlab-test
#SBATCH --time=00:05:00
#SBATCH --partition=cpu
#SBATCH --nodes=1
#SBATCH --mem=4G
#SBATCH --cpus-per-task=8

module load matlab

matlab -batch matlabtest

Input file, e.g. shell or sbatch script must be exclude .m, output will be written inside slurm-<jobid>.log.

Example of an Apptainer job

Content of SBATCH script:

#!/bin/bash

#SBATCH --mem-per-cpu=512M
#SBATCH --job-name=matlab-test
#SBATCH --time=00:05:00
#SBATCH --partition=cpu
#SBATCH --nodes=1
#SBATCH --mem=4G
#SBATCH --cpus-per-task=8

apptainer exec /ceph/hpc/software/containers/singularity/images/matlab-<version>.sif matlab -batch matlabscript

Example of scaling your MATLAB code to run in parallel

To learn more about running MATLAB parallel jobs, launched from your local desktop, read here

Example of sending with ARC

Firstly you will need ARC Client and a .xrsl file in which you state which resources you need to start the job, and contains two additional input files.

&
(count = 8)
(countpernode = 1)
(jobname = matlab-test)
(inputfiles = 
  ("matlabscript.sh" "")
  ("matlabscript.m" "")
)
(executable = matlabscript.sh)
(outputfiles = ("example.out" ""))
(stdout=example.log)
(join=yes)
(gmlog=log)
(memory=8000)
(walltime="5 minutes")

Example of matlabscript.sh:

#!/bin/bash

#Setup environment
. /ceph/hpc/software/cvmfs_env.sh

#Load module
module --ignore-cache load matlab

#Execute script
matlab -batch matlabscript

For input script examples e.g. matlabscript.m, check the official documentation.

Documentation