MATLAB

MATLAB is a high-level language and interactive environment that enables you to perform computationally intensive tasks faster than with traditional programming languages.

MATLAB is available on PSC systems for academic users.   Academic users must complete the MATLAB Request form to request access to MATLAB. Once access has been granted, follow the steps below.

Non-academic users should contact help@psc.edu to discuss how they can get access to MATLAB at PSC.

Sample scripts for MATLAB use are available in directory /opt/packages/examples/matlab on Bridges.

Documentation

• http://www.mathworks.com/products/matlab/

Usage

Use –egress on the interact command

Use the --egress option to interact to ensure that your job can reach the MATLAB license server.  See the Running Jobs section of the Bridges User Guide for more information on interact and its options.

For any nodes except the DGX-2, load the MATLAB module

To see which versions of MATLAB are available

module avail matlab

Start an interactive session, then load the MATLAB module. To load the default MATLAB module

module load matlab

To load a non-default version, if more than one is available, use its full name, e.g.,

module load matlab/R2018b

Then start up MATLAB. You can run your MATLAB job on as many cores as the node has on which you are running. Use a command like

matlab -options   < input-file  > output-file

On the DGX-2, use the supplied MATLAB Singularity container

 NVIDIA provides a Docker container for MATLAB in the NVIDIA GPU Cloud site (see more information at https://www.nvidia.com/en-us/gpu-cloud/).  We  converted this to Singularity and provide it on Bridges for use with MATLAB on the DGX-2.

It is available at /pylon5/containers/mdl/matlab_r2019a.sif.  To use MATLAB on the DGX-2, start an interactive session using a command  like:

[user@login005 ~]$ interact -p GPU-AI --gres=gpu:volta32:1 --egress

Once your interactive session has started, use these commands:

[user@gpu058 ~]$ module load singularity
[user@gpu058 ~]$ singularity exec /pylon5/containers/mdl/matlab_r2019a.sif matlab -c 28000@matlablic.andrew.cmu.edu

See more information on

• Containers on Bridges
• Singularity on Bridges

See what toolkits are available

Once you have started MATLAB, typing

ver

lists all the installed toolkits.  PSC has the PCT toolkit, which is how you can run parallel MATLAB jobs. This command also gives information on your license number, operating system, Java version, etc., which may be useful in diagnosing problems.