Getting Started
Supported Releases of MATLAB and MATLAB Compiler Runtime
MATLAB Release (Patch level) | MATLAB Compiler Runtime | ||
|---|---|---|---|
The software stack on CoolMUC-4 is still incomplete. Matlab is not yet available in the default stack. As a workaround a temporary installation of R2024a is available. Use the following commands to load matlab: module use -p /lrz/sys/share/modules/extfiles | |||
| List available modules | > module avail matlab | > module avail matlab-mcr | |
R2024a | > module remove intel-mpi > module load intel-mpi/2018.4.274 > module load matlab/R2024a_Update4-generic | > module remove intel-mpi > module load intel-mpi/2018.4.274 > module load matlab-mcr/R2024a | R2024a |
| R2023b (Update 7) | > module remove intel-mpi > module load intel-mpi/2018.4.274 > module load matlab/R2023b_Update7-generic | > module remove intel-mpi > module load intel-mpi/2018.4.274 > module load matlab-mcr/R2023b_Update7 | R2023b (Update 7) |
| Availability | Linux Cluster (CoolMUC-4) | Linux Cluster (CoolMUC-4) SuperMUC-NG | |
Using the default Intel-MPI module (Version 2019 or newer versions), unintended crashes of parallel MATLAB jobs might occur. Please switch to Intel MPI 2018 as shown above.
Useful MATLAB Command-line Options
Run MATLAB via the "matlab" command only or add command-line arguments. Please consider: All MATLAB command-line arguments are case-sensitive!
> matlab arg_1 ... arg_N
| Command-line argument | Meaning |
|---|---|
| no argument | Start MATLAB GUI. |
-nodesktop | Start MATLAB without desktop but allow GUI and graphics output (figures). |
-nodisplay | Start MATLAB without any GUI support. Graphics output (figures) is also not possible. |
-singleCompThread | Many intrinsic MATLAB functions automatically exploit multithreading. MATLAB can be forced to disable this feature. Use this option for any work on the login nodes! |
-r myfunc | Run a MATLAB script or function, e. g. myfunc.m. |
Further information:
- https://de.mathworks.com/help/matlab/matlab_env/startup-options.html
- https://de.mathworks.com/help/matlab/matlab_env/commonly-used-startup-options.html
Interactive MATLAB Jobs
LRZ policies for using MATLAB on Login Nodes
MATLAB is a very resource-hungry application with respect to both memory and CPU resources! Please run MATLAB on login nodes with care!
Especially, involving multiple threads (many MATLAB functions/operators use multithreading by default) or multiple processes might cause a high load on the login node and handicapping other users!
MATLAB applications, which handicap other users or have a negative impact on the general operation of the login nodes, will be terminated by system administrators!
Our rules using MATLAB on the login nodes:
- Only run a maximum of one MATLAB instance!
- Disable any kind of parallelization, i.e. multithreading (see command-line options above) or the Parallel Computing Toolbox!
- Check the load and memory consumption of your own MATLAB session. Usually, you can do this via the "top" command, e.g.:
top -u $USER
- We recommend NOT to use the graphical user interface (GUI) of MATLAB, as this unnecessarily increases the load (resource consumption). Furthermore, using GUI decreases the overall performance of MATLAB.
- If the MATLAB GUI is needed, then you may run MATLAB via a VNC session (VNC Server on Login-Nodes) to increase the performance and responsiveness of the GUI!
- If the MATLAB GUI is not needed, it is better to run MATLAB via an interactive Slurm job (see next section). This job runs on a compute node. Parallelization is explicitly allowed here!
- Repeated violation of the above mentioned restrictions on login nodes might result in a ban of the affected user account and notification to the responsible master user or PI.
Interactive MATLAB Jobs via Slurm Job
Usecases
Depending on the purpose, there are different possibilities to use MATLAB interactively:
- MATLAB computations with focus on visualisation: We recommend to run MATLAB via a VNC session.
- Pure MATLAB computations: Interactive MATLAB sessions may be started on compute nodes of CoolMUC- by employing interactive Slurm jobs (see Slurm documentation).
Constraints
- Interactive jobs depend on the availability of compute resources. Matlab may not start immediately.
- An interactive Matlab session should be limited to 1 compute node only.
- Consider the restrictions on interactive jobs: Job Processing on the Linux-Cluster
Common Batch Jobs
Batch jobs are used for all MATLAB production runs. Usually, the resources, consumed by MATLAB applications, are limited to a single CPU core (pure serial job) or a single compute node (parallel job involving either multithreading or the Parallel Computing Toolbox [PCT]). Following table lists job examples for various cases which can be used on CoolMUC-4.
| Slurm job script | Matlab script |
|---|---|
Serial batch job No parallelization at all, MATLAB is intended to run on a single core. | |
Parallel job using multithreading MATLAB will run on a single compute node. | |
Parallel batch job using Parallel Computing Toolbox (PCT) MATLAB will run on a single compute node. | |