Description
MATLAB is a high-level language and interactive environment for numerical computation, visualization, and programming. Using MATLAB, you can analyze data, develop algorithms, and create models and applications.
Home Page
http://www.mathworks.se/products/matlab/
Documentation
http://www.mathworks.se/help/matlab/
License
UiO has a site license for MATLAB.
Usage
Use
module avail MATLAB
to see which versions of matlab are available. Use
module load MATLAB
or
module load MATLAB/version
to get access to matlab. To use the graphical interface of Matlab you need a X-server installed on your local machine (for example Xming or Xwin32 for Windows machines).
To run matlab in a job script, use
matlab -nodesktop < script.m
Warning: By default, Matlab will use all CPUs on the node! This means that one must either ask for all the CPUs on the node in the job script, or add the switch -singleCompThread to the matlab command line.
To run Matlab in parallel using up to 12 threads within a single node one need to put
#SBATCH --cpus-per-task=12
in the jobscript to ask for 12 cores within the same node. Then, for a parallel job in the Matlab-script one should do something like
pc = parcluster
pc.JobStorageLocation = getenv('SCRATCH')
to set up a cluster using the job specific scratch space issued by the queuing system. Now,
parpool(pc,12)will set up a pool of 12 workers on the cluster for using 'parfor' or 'spmd'. See http://www.mathworks.se/help/distcomp/parpool.html
To set up an independent job on the cluster one can use
j = createJob(pc)
and issue this job with tasks. Have a look at http://www.mathworks.se/help/distcomp/createjob.html
We currently do not have a license for MATLAB Distributed Computing Server, hence multi-node jobs are not supported.