Matlab

Matlab is installed locally on all nodes. Currently, the releases R2026a and R2025b are available.

Do not even try opening Matlab on a login node!

Just opening Matlab needs ~4GB of RAM. This is way above what you may use on a login node!

Therefore, always use an interactive session when working with Matlab!

How do I access Matlab?

Desktop Session

If you are using an Interactive Desktop Session, you are going to find the Matlab entries in the Applications Menu on the top right. It is in the "Others" category.

Visual Studio Code via a tunnel session

The most convenient way would be to use a Visual Studio Code Tunnel and then install the official Matlab extension.

You are going to have to tell the extension where it can find the so-called matlabroot. We strongly advise you to do this on a per project basis.

  1. Create a folder called .vscode in the root folder of your project, if it does not exists.
  2. Create or edit the file called settings.json in the .vscode folder.

The file should look like this:

{
    "MATLAB.installPath": "/opt/matlab/R2026a"
}

If you want to use another release, just change the last part of the path.

If the file already existed, you might see some more entries between the brackets. In this case make sure not to duplicate the outer brackets but just add the "MATLAB.installPath" entry between the brackets!

When this is done, VSCode tries to start and connect to Matlab as soon as you open a .m file. You can run and debug your scripts just as normal.

You can also open a Matlab Terminal. Try clicking the "Plus" sign and choose MATLAB if you already have a terminal open. If not, Click Terminal -> New Terminal in the menu.

If you are connected from Linux or Mac (or have managed to get X11 forwarding to work in Windows), you can get a desktop Window of the Matlab session by typing desktop in the Matlab Terminal.

Terminal

Of course, you can also access Matlab from a terminal session.

Symlinks have been created to /usr/local/bin So you can start the matlab release of your choice with:

matlab-r2026a

and so on.

Again, do not even try this from a login node!

How do I submit jobs?

At the moment, your best option is PLUS Slurm Matlab, a package that lets you create and submit jobs quite conveniently from Matlab. Take a look at the examples, to get a first impression.

Parallel Server, which should make the whole thing a lot more convenient, is on our todo list.