This is an old revision of the document!
Matlab is a universal interactive numerical application system with advanced graphical user interface. You can find some documentation of the provider here.
On the HPC resources at GWDG there are 5 network licences available for Matlab. Also we have following extensions: Simulink, Optimization Toolbox, Parallel Computing Toolbox and Statistics and Machine Learning Toolbox.
For users from MPG we offer the flat rate License from MPG , which covers all toolboxes including MATLAB Parallel Server (formerly Distributed computing). However, you need to apply for the access to the license. You can do it by writing to email@example.com.
The dialog server and frontend nodes of the HPC resources at GWDG can be reached directly only inside the GoeNet network. Login to gwdu101 (or alternatively to gwdu102) via:
ssh -Y gwdu101
For the connection from Microsoft clients use an appropriate software like X-Win32.
If your connection to gwdu101 is refused you might not have the permission to use the HPC resource. You can request this permission by a short email to mailto:firstname.lastname@example.org.
If you want to use the service from outside GoeNet, first login to login.gwdg.de and then proceed as above.
First prepare the necessary environment with:
module load matlab/2020a
The use of Matlab must be be scheduled by the batch system in the interactive queue onto an available node. Therefore the following command to the batch system is necessary:
srun --x11 -c 20 -N 1 -p int --pty bash
after a short time period you will get a Shell prompt and you can call Matlab with the command:
The current version can be found on the main matlab screen under 'Help - About Matlab'.
The cluster currently has only Parallel Computing Toolbox for Matlab (without Matlab Distributed Computing Server), it means that the opportunities of parallelization are limited in one node. You can only use multiple processors of a single computing node.
Parallel Computing Toolbox provides following commands and structures for parallel programs:
- parfor - parallel for loop
- gpuArray - to work with GPU
- tall arrays
Dr. Ulrich Schwardmann, Tel.: 0551 201-1542, E-Mail: uschwar1 [at] gwdg.de