MATLAB on the ACCRE Cluster
MATLAB is a commercial programming language and computing environment that is widely popular in many areas of engineering and science. Since 2018, MATLAB has been free to use on the cluster for all ACCRE users.
Versions of MATLAB on the ACCRE Cluster
To see a list of installed versions of MATLAB on the cluster, use Lmod:
[jill@vmps12 ~]$ module avail MATLAB MATLAB/2017a
Multiple versions of MATLAB are available on the cluster. We encourage users to use the most recent version of MATLAB (r2015a) installed, if possible, as we will not offer support for older versions indefinitely.
To load a particular version of MATLAB, use Lmod:
[jill@vmps12 ~]$ module load MATLAB/2017a [jill@vmps12 ~]$ which matlab /opt/easybuild/software/Core/MATLAB/2017a/bin/matlab
If you are not in the MATLAB Unix group as described above, you will get an error when trying to load MATLAB in the first “module load” command. See the introduction above for more details about obtaining a MATLAB license for the ACCRE cluster.
In order to run MATLAB interactively, simply type
matlab from the Linux command line:
[jill@vmps12 ~]$ matlab MATLAB is selecting SOFTWARE OPENGL rendering. < M A T L A B (R) > Copyright 1984-2015 The MathWorks, Inc. R2015a (126.96.36.199613) 64-bit (glnxa64) February 12, 2015 To get started, type one of these: helpwin, helpdesk, or demo. For product information, visit www.mathworks.com. Academic License >>
Note that the MATLAB command prompt may take ~30-60 seconds to completely load, depending on which version you are using and whether you’ve loaded it previously in your current shell session. You can also load the full graphical user interface (GUI) to your local machine by logging into the cluster with X11 forwarding enabled (
ssh -X <vunetid>@login.accre.vanderbilt.edu). In general, using the MATLAB GUI from the cluster will be very slow, especially if you are outside the Vanderbilt network, so we do not recommend it. Instead, most users develop their MATLAB code locally (from a laptop or desktop environment) for testing and then submit jobs to the cluster for batch (non-interactive) processing. Another option is to run MATLAB through a virtual desktop through the ACCRE Visualization Portal.
To see a list of the available Matlab toolboxes on the cluster, use the
ver command from the MATLAB command prompt:
>>ver ---------------------------------------------------------------------------------------------------- MATLAB Version: 188.8.131.52613 (R2015a) MATLAB License Number: 299681 Operating System: Linux 2.6.32-220.2.1.el6.x86_64 #1 SMP Fri Dec 23 02:21:33 CST 2011 x86_64 Java Version: Java 1.7.0_60-b19 with Oracle Corporation Java HotSpot(TM) 64-Bit Server VM mixed mode ---------------------------------------------------------------------------------------------------- MATLAB Version 8.5 (R2015a) Simulink Version 8.5 (R2015a) Aerospace Blockset Version 3.15 (R2015a) Aerospace Toolbox Version 2.15 (R2015a) Bioinformatics Toolbox Version 4.5.1 (R2015a) Communications System Toolbox Version 6.0 (R2015a) Computer Vision System Toolbox Version 6.2 (R2015a) Control System Toolbox Version 9.9 (R2015a) Curve Fitting Toolbox Version 3.5.1 (R2015a) DSP System Toolbox Version 9.0 (R2015a) Database Toolbox Version 5.2.1 (R2015a) Datafeed Toolbox Version 5.1 (R2015a) Econometrics Toolbox Version 3.2 (R2015a) Embedded Coder Version 6.8 (R2015a) Filter Design HDL Coder Version 2.9.7 (R2015a) Financial Instruments Toolbox Version 2.1 (R2015a) Financial Toolbox Version 5.5 (R2015a) Fixed-Point Designer Version 5.0 (R2015a) Fuzzy Logic Toolbox Version 2.2.21 (R2015a) Global Optimization Toolbox Version 3.3.1 (R2015a) Image Acquisition Toolbox Version 4.9 (R2015a) Image Processing Toolbox Version 9.2 (R2015a) Instrument Control Toolbox Version 3.7 (R2015a) MATLAB Coder Version 2.8 (R2015a) MATLAB Compiler Version 6.0 (R2015a) MATLAB Compiler SDK Version 6.0 (R2015a) Mapping Toolbox Version 4.1 (R2015a) Model Predictive Control Toolbox Version 5.0.1 (R2015a) Neural Network Toolbox Version 8.3 (R2015a) Optimization Toolbox Version 7.2 (R2015a) Parallel Computing Toolbox Version 6.6 (R2015a) Partial Differential Equation Toolbox Version 2.0 (R2015a) RF Toolbox Version 2.16 (R2015a) Robust Control Toolbox Version 5.3 (R2015a) Signal Processing Toolbox Version 7.0 (R2015a) SimBiology Version 5.2 (R2015a) SimDriveline Version 2.8 (R2015a) SimElectronics Version 2.7 (R2015a) SimEvents Version 4.4 (R2015a) SimMechanics Version 4.6 (R2015a) SimPowerSystems Version 6.3 (R2015a) SimRF Version 4.4 (R2015a) Simscape Version 3.13 (R2015a) Simulink 3D Animation Version 7.3 (R2015a) Simulink Coder Version 8.8 (R2015a) Simulink Control Design Version 4.2 (R2015a) Simulink Design Optimization Version 2.7 (R2015a) Stateflow Version 8.5 (R2015a) Statistics and Machine Learning Toolbox Version 10.0 (R2015a) Symbolic Math Toolbox Version 6.2 (R2015a) System Identification Toolbox Version 9.2 (R2015a) Wavelet Toolbox Version 4.14.1 (R2015a)
Note that ACCRE has a license for the Parallel Computing Toolbox , which allows for processing across multiple CPU cores and/or GPUs, and can therefore enable better performance (faster execution time) depending on the application.
Running a MATLAB script within a SLURM job is generally straightforward. Unless you are attempting to run across multiple CPU cores using MATLAB’s Parallel Computing Toolbox, you will want to request a single task, load the appropriate version of MATLAB from your SLURM script, and then launch your MATLAB job. The following example runs a simple MATLAB script that demonstrates the utility of writing vectorized Matlab code:
[jill@vmps12 run1]$ ls matlab.slurm vectorization.m
[jill@vmps12 run1]$ cat matlab.slurm #!/bin/bash #SBATCH --nodes=1 #SBATCH --ntasks=1 #SBATCH --time=00:10:00 #SBATCH --mem=500M #SBATCH --output=matlab_job_slurm.out module load MATLAB # load the default version of Matlab matlab -nodisplay -nosplash < vectorization.m
-nodisplay flag informs MATLAB you are operating in batch mode, while the
-nosplash flag will prevent the splash screen from being displayed during startup. Additionally, you might try passing the
-nojvm flag, which informs MATLAB that you do not need Java features for processing. Passing this flag often leads to faster MATLAB load-up times, but some I/O operations may depend on Java support so use this flag with caution. More information can be found on this page .
[jill@vmps12 run1]$ cat vectorization.m % surrounding a block of code with tic and toc % will time its execution % non-vectorized code tic i = 0; for t = 0:.00001:10 i = i + 1; y(i) = sin(t); end toc % vectorized code tic t = 0:.00001:10; y = sin(t); toc
[jill@vmps12 run1]$ sbatch matlab.slurm Submitted batch job 2135971
After waiting a few minutes:
[jill@vmps12 run1]$ ls matlab_job_slurm.out matlab.slurm vectorization.m
[jill@vmps12 run1]$ cat matlab_job_slurm.out < M A T L A B (R) > Copyright 1984-2015 The MathWorks, Inc. R2015a (184.108.40.206613) 64-bit (glnxa64) February 12, 2015 To get started, type one of these: helpwin, helpdesk, or demo. For product information, visit www.mathworks.com. Academic License >> >> >> >> >> >> >> >> >> Elapsed time is 0.477166 seconds. >> >> >> >> >> >> Elapsed time is 0.057248 seconds.
In this particular example no functions were defined in the vectorization.m file. Often users write scripts with functions, and need to call a function from the Linux command line. This can be accomplished by passing the MATLAB interpreter the -r option. For example:
matlab -nodisplay -nosplash -r "myFunc(1),quit()"
Here we are calling myFunc() and passing a single argument (1) to the function. It’s also necessary to call quit() afterwards to ensure your job ends once MATLAB processing has completed. You may also need to update your MATLAB path to include directories containing .m files and their function definitions. There are a few ways to accomplish this. The first way is to update your MATLABPATH Bash environment variable. Something like the following can be done within a SLURM script before MATLAB is launched:
Alternatively, the path can be updated at runtime like so:
matlab -nodisplay -nosplash -r "addpath(genpath('/home/jill/myDir')),myFunc(1),quit()"
Contributing New Examples
In order to foster collaboration and develop local MATLAB expertise at Vanderbilt, we encourage users to submit examples of their own to ACCRE’s MATLAB Github repository . Instructions for doing this can be found on this page.