High Performance Computing

Versions and Availability

About the Software

MATLAB is a high-level language and interactive environment developed by Mathworks. It is widely u sed in science and engineering across industry and academia. Users need to provide their own license file, which should have a .lic extension and be located under /home/$U SER/.matlab/R2020a_licenses.

Usage

Licensing

Matlab is installed on most of the HPC clusters. It is, however, license software, which means that it is only availabe to users or user groups who have access to a license, purchased either individually or by the users' institution.

If you do have lots of similar jobs, we recommend you to compile your Mablab scripts using the Matlab compiler, and the jobs using the compiled executable will not take license seats, If you need help in this regard, please contact us at sys-help@loni.org.

If you are not enrolled at LSU, you are not allowed to use Matlab on our LSU HPC clusters according to the current terms of our Matlab license agreement and the program offering guide.

Note: For LSU and ULL users on QB2, a license file has been created, so the steps described below are not necessary for them.

For LONI users who want to use Matlab on QB2, you need to provide your own license file. The license file should have a .lic extension and be located under the "/home/$USER/.matlab/R<matlab version>_licenses" directory. For example, in order to user Matlab R2013a, you need to create the directory ".matlab/R2013a_licenses" under your home directory and put the license file there.

Users should note that most Matlab license servers are restrictive about the IP addresses from which licenses can be checked out. In another word, a user will not be able to run Matlab on QB2 until the IP addresses of QB2 nodes are added to the allowed list of the license server specified in the license file. In most cases, users need to contact the admistrator of their license server and request the IP addresses of QB2 nodes to be added. If you need help in this regard, please contact us at sys-help@loni.org.

Running Matalb

Matlab can be run interactively, but requires use of X-windows for any graphical capability.

Running Matlab Parallel Server

Note: The content of this section applies to LSU HPC users only.

Matlab Parallel Server (MPS) allows users to run their Matlab codes on multiple nodes, potentially accelerating the execution significantly. To use MPS on Super Mike 2 or SuperMIC, follow the steps below:

Step 1: Make sure the module "matlab/r2019b" is loaded. MPS will not work with other versions of Matlab.

Step 2: Import the cluster profile in your Matlab code:

    profile_master = parallel.importProfile('/usr/local/packages/license/matlab/r2019b/lsuhpcparserver/SuperMike2.mlsettings');

    parallel.defaultClusterProfile(profile_master);
    

For SuperMIC, change the "SuperMike2.mlsettings" to "SuperMIC.mlsettings";

Step 3: Create a pool of workers before any other parallel functions:

    parpool(<name of cluster>,<number of workers>)
    

, where the <name of cluster> should be 'SuperMike2' on Super Mike 2 and 'SuperMIC' on SuperMIC, and the <number of workeres> should be equal to the number of cores available to your job. For instance, for 4 nodes on SuperMIC, the line should look like:

    parpool('SuperMIC',80)
    

Step 4: Now all implicitly parallel functions such as parfor can launch workers on multiple hosts.

MPS Example

The example below is based on the one described on this web page.

It runs integer factorization on 1 to 64 cores on 4 Super Mike 2 nodes, records the timing information and plots the speedup (as compared to running on 1 core).

Content of the code:

$ cat composite.m

profile_master = parallel.importProfile('/usr/local/packages/license/matlab/r2019b/lsuhpcparserver/SuperMike2.mlsettings');
parallel.defaultClusterProfile(profile_master);

primeNumbers = primes(uint64(2^21));
compositeNumbers = primeNumbers.*primeNumbers(randperm(numel(primeNumbers)));
factors = zeros(numel(primeNumbers),2);

parpool('SuperMike2',64);

numWorkers = [1 2 4 6 16 32 64];
tCluster = zeros(size(numWorkers));
for w = 1:numel(numWorkers)
    tic;
    parfor (idx = 1:numel(compositeNumbers), numWorkers(w))
        factors(idx,:) = factor(compositeNumbers(idx));
    end
    tCluster(w) = toc;
end
f = figure;
speedup = tCluster(1)./tCluster;
plot(numWorkers, speedup);
title('Speedup with the number of workers');
xlabel('Number of workers');
xticks(numWorkers(2:end));
ylabel('Speedup');
saveas(gcf, 'composite_speedup.png');
delete(gcf);
delete(gcp);

Content of the job script:

$ cat mps.pbs

#!/bin/bash

#PBS -l nodes=4:ppn=16
#PBS -l walltime=4:00:00
#PBS -j oe
#PBS -N matlab.parallel.server
#PBS -A hpc_my_allocation
#PBS -q checkpt

cd $PBS_O_WORKDIR
matlab -nodesktop -r "try; run composite; catch; end; quit"

A user can submit this job script as any other jobs:

$ qsub mps.pbs

% ssh -X -Y username@remote.host.tdl

% xterm&

% ssh -X -Y username@remote.host.tdl

% xterm&

Matlab can also be run in batch mode using a PBS script (say matlab.pbs). The execution command line would look like:

matlab -nodisplay -nojvm -nosplash -r matlab_cmd

#!/bin/bash
#PBS -q single 
#PBS -A CHANGE_TO_YOUR_ALLOCATION
#PBS -l nodes=1:ppn=4 
#PBS -l walltime=HH:MM:SS 
#PBS -o myoutput
#PBS -j oe 
#PBS -N MATLAB_JOB

cd /path/to/your/matlab/files
matlab -nodisplay -nojvm -nosplash -r matlab_cmd

matlab_cmd can be the name of a Matlab function, or a command string, exactly as you would type interactively. For instance you could add a path name to Matlab's internal path and run fname.m like so:

matlab -nodisplay -nojvm -nosplash -r "addpath /my/custom/path; fname"

The script would then be submitted to PBS via qsub:

$ qsub matlab.pbs

$ qsub job_script

#PBS -X value

 #!/bin/bash
 #
 # Use "workq" as the job queue, and specify the allocation code.
 #
 #PBS -q workq
 #PBS -A your_allocation_code
 # 
 # Assuming you want to run 16 processes, and each node supports 4 processes, 
 # you need to ask for a total of 4 nodes. The number of processes per node 
 # will vary from machine to machine, so double-check that your have the right 
 # values before submitting the job.
 #
 #PBS -l nodes=4:ppn=4
 # 
 # Set the maximum wall-clock time. In this case, 10 minutes.
 #
 #PBS -l walltime=00:10:00
 # 
 # Specify the name of a file which will receive all standard output,
 # and merge standard error with standard output.
 #
 #PBS -o /scratch/myName/parallel/output
 #PBS -j oe
 # 
 # Give the job a name so it can be easily tracked with qstat.
 #
 #PBS -N MyParJob
 #
 # That is it for PBS instructions. The rest of the file is a shell script.
 # 
 # PLEASE ADOPT THE EXECUTION SCHEME USED HERE IN YOUR OWN PBS SCRIPTS:
 #
 #   1. Copy the necessary files from your home directory to your scratch directory.
 #   2. Execute in your scratch directory.
 #   3. Copy any necessary files back to your home directory.

 # Let's mark the time things get started.

 date

 # Set some handy environment variables.

 export HOME_DIR=/home/$USER/parallel
 export WORK_DIR=/scratch/myName/parallel
 
 # Set a variable that will be used to tell MPI how many processes will be run.
 # This makes sure MPI gets the same information provided to PBS above.

 export NPROCS=`wc -l $PBS_NODEFILE |gawk '//{print $1}'`

 # Copy the files, jump to WORK_DIR, and execute! The program is named "hydro".

 cp $HOME_DIR/hydro $WORK_DIR
 cd $WORK_DIR
 mpirun -machinefile $PBS_NODEFILE -np $NPROCS $WORK_DIR/hydro

 # Mark the time processing ends.

 date
 
 # And we're out'a here!

 exit 0

Note: Clusters supported by HPC@LSU do not have parallel Matlab installed due to license issues.

Resources

• Extensive on-line documentation is available at Mathworks TechDocs.