#
MATLAB

MATLAB

MATLAB is a powerful scripting language and computational environment. It is designed for numerical computing, visualization and high-level programming and simulations. MATLAB also has parallel processing capabilities.

MATLAB is installed on all HPC compute and login nodes, and on all Spear nodes.

## Limited Licenses

We maintain a limited number of licenses for MATLAB use at the RCC. Sometimes, all available licenses are in-use by other users and you must wait to checkout a license. To see how many licenses are available, you can run the following command from inside MATLAB:

```
license('checkout','MATLAB')
```

Non-interactive jobs submitted using the scripts shown in this page will check for available MATLAB licenses before running. However, if you want to use large number of MATLAB jobs simultaneously, use MATLAB Compiler as described below to create executables from your code. This will compile your MATLAB code to C code and avoid all license restrictions.

## Don't run MATLAB on Login Nodes

The MATLAB Graphical User Interface (GUI) depends on the Java Virtual Machine in order to start and is very memory-intensive. Due to the the limited amount of RAM available on the login nodes, it is probable that you will run into a hard-to-diagnose situations where your MATLAB GUI crashes every time you try to start the program. If you encounter this issue, try running MATLAB through our Open OnDemand service, or by starting an interactive session on a compute node.

## Versions

We have multiple versions of MATLAB available on our systems. The default is version **2021b**, however we also have several other versions available via kernel modules. These can be loaded by using the following command:

```
module load matlab # <-- Load default version
module load matlab/VERSION # <-- Load specific version
```

Version | Load Command |
---|---|

2021b | module load matlab/2021b |

2020a | module load matlab/2020a |

2018b | module load matlab |

2017a | module load matlab/2017a |

2015b | module load matlab/2015b |

2014b | module load matlab/2014b |

2013b | module load matlab/2013b |

## Interactively running MATLAB

You can work with MATLAB interactively on our servers using our Open OnDemand App. To get started, follow these steps:

- Navigate to Open OnDemand
- Note: If you are on campus, make sure you are connected to FSUSecure and if you are off-campus, make sure you are connected to the FSU VPN)

- Now Log in using your HPC Credentials (these are usually the same as your FSUID Credentials)
- Now click on "Interactive Apps" and select "MATLAB" under the GUIs heading in the drop-down list
- Now set up your interactive session by selecting:
- The SLURM Queue that you want to use (the default is
**genacc_q**if you leave this text field blank) - The number of Cores you want to use
- The Version of MATLAB you want to use (the default for this will appear in this field automatically)
- The total amount of RAM you think your job will need (advice on this can be found on our HPC Job Resource Planning page.
- The total amount of time you think you will need to complete your work on MATLAB.
- You may also want to check the
**"I would like to recieve an email when the session starts"**box too. This will send you an email when resources have been allocated for your session. Since HPC is a shared resource, you may have to wait for enough resources to become available. - An example of the job setup screen follows:
- Click LAUNCH when you have set up the job options as desired.
- When the job is ready, click on the "My Interactive Sessions" tab. Your session will appear as a green box with a "Launch MATLAB" button in the bottom left-hand corner. Click this button to access the MATLAB GUI. The box will appear like this:

- The SLURM Queue that you want to use (the default is

## Parallel Computing in MATLAB on the HPC

MATLAB has a very powerful Parallel Computing Toolbox which is available by default on the HPC system. In order to learn how to use it, follow the tutorials and documentation provided by MATLAB themselves at the following links (you may want to use an Open OnDemand session for learning):

- The MATLAB Parallel Computing Toolbox Documentation Landing Page
- Getting Started with Parallel Computing
- Parallel Computing Fundamentals
- Parallel FOR Loops in MATLAB
- Parallel Computing Toolbox Examples

### Using pmode interactively

Below is an example of how to invoke interactive *pmode* with four workers:

`pmode start local 4`

This will open a Parallel Command Window (PCW). The workers then receive commands entered in PCW (at the *P>>* prompt), process them, and send the command output back to the PCW. You can transfer variables between the MATLAB client and the workers. For example, to copy the variable *x* in Lab 2 to *xc* on the client, use:

```
pmode lab2client x 2 xc
```

Similiarly, to copy the variable *xc* on the client to the variable *x *on Lab 2, use:

```
pmode client2lab xc 2 x
```

You can perform plotting and all other operations from inside PCW.

You can distribute values among workers as well. For example, to distribute the array *x *among workers, use:

`codistributed(x,'convert')`

Use `numlabs`

, `labindex`

, `labSend`

, `labReceive`

, `labProbe`

, `labBroadcast`

, `labBarrier`

functions similar to MPI commands for parallelizing. Please refer to MATLAB manual for a full discussion of commands. Entering the following command in the PCW will end the session and release the licenses for other users:

```
pmode quit
```

### Interactive parallel computing using *parpool*

The `matlabpool`

utility has been replaced by the `parpool`

utility in current versions of MATLAB. The syntax for `parpool`

is

```
parpool
parpool(poolsize)
parpool('profile',poolsize)
parpool('cluster',poolsize)
ph = parpool(...)
```

...where `poosize`

, `profile`

, and `cluster`

are respectively the size of the MATLAB pool of workers and the profile or the cluster you created. The last line creates a handle * ph* for the pool.

`parpool`

enables the full functionality of the parallel language features (`parfor`

and `spmd`

) in MATLAB by creating a special job on a pool of workers, and connecting the MATLAB client to the parallel pool.

The following example creates a pool of 4 workers, and runs a `parfor-loop`

using this pool:

```
>>parpool(4)
>> parfor i = 1:10
feature getpid;
disp(ans)
end
1172
1172
1171
1171
1169
1169
1170
1172
1171
1169
>> delete(gcp)
```

`The following example creates a pool of 4 workers, and runs a simple ``spmd`

code block using this pool:

```
>> ph = parpool('local',4) % ph is the handle of the pool
>> spmd
>> a = labindex
>> b = a.^2
>> end
Lab 1:
a = 1
b = 1
Lab 2:
a = 2
b = 4
Lab 3:
a = 3
b = 9
Lab 4:
a = 4
b = 16
>> delete(ph)
```

You cannot simultaenously run more than one interactiveNote.`parpool`

session. You must to delete your current parpool session before starting a new one. To delete the current session, use:

```
delete(gcp)
delete(ph)
```

...where `gcp`

utility returns the current pool, and * ph* is the handle of the pool.

## Non-interactive job submission

The following examples demonstrate how to submit MATLAB jobs to the HPC. You should already be familiar with how to connect and submit jobs in order to submit MATLAB jobs.

### A Note about MATLAB 2017a

You may notice several warning messages related to Java in the output files or on the terminal when running non-interactive jobs using MATLAB 2017a. While the cause of these warnings is not clear, they do not appear to cause any errors in the job runs themselves. For that reason, these warnings may be ignored.

### Single core jobs

The following example is a sample submit script (*test1.sh*) to run the MATLAB program *test1.m* that uses a single core.

* Note*:

**test1.m should be a function, not a script. This can be easily done by simply enclosing your script with a dummy function.**

```
#!/bin/bash
#SBATCH -N 1
#SBATCH --ntasks-per-node=1
#SBATCH -p genacc_q
#SBATCH -t 01:00:00 #Change the walltime as necessary
module load matlab
matlab -nosplash -nojvm -nodesktop -r "test1; exit"
```

Then, you can submit your job to Slurm normally:

```
sbatch test1.sh
```

* Note:* The Parallel Computing Toolkit (PCT) cannot be used within your function; you cannot use parfor or any other command that utilizes more than one core.

### Multiple Core Jobs

You can submit MATLAB jobs to run on multiple cores by changing the ppn value in the MATLAB submit script or by using *parpool* with the **-n** parameter adjusted in the MATLAB Slurm Submit Script. The maximum number of cores available for any single job is currently **eight**. The earlier *matlabpool* utility has been deprecated and removed in recent versions. For current versions of MATLAB, you will need to modify your MATLAB code to include the following lines:

```
n_cores = str2num(getenv('SLURM_NTASKS'));
pool = parpool('local', n_cores);
... your matlab code should go here ...
delete(pool)
```

With these modifications, your code will now get the number of cores you requested in your Slurm submit script (from the **-n **parameter) and will create a parallel pool which has as many workers as the number of processes you have requested. Then, your code can take advantage of typical parallel computing constructs like **parfor** and others. When your job finishes, the last line will delete the parallel pool of workers before exiting.

Once your code has been modified, your submit script should be similar to the serial non-interactive job described above but with more tasks requested.

```
#!/bin/bash
#SBATCH -N 1
#SBATCH -n 4
#SBATCH -p genacc_q
#SBATCH -t 01:00:00 #Change the walltime as necessary
module load matlab
matlab -nosplash -nodesktop -r "test1; exit"
```

Then, you can submit your job to Slurm normally:

```
sbatch test1.sh
```

### Multiple node jobs

Multiple node jobs require the *MATLAB_DCS* package. This is currently not available on the HPC. If your job needs access to MATLAB_DCS, please submit a ticket or consider rewriting your code to take advantage of MPI capabilities (C, C++, Fortran or Python).

## Using the MATLAB Compiler

If you need to use a large number of simultaneous MATLAB workers, we advise that you compile your code into an executable. This allows you to write and test your code in MATLAB, and compile it to C when you are ready to run a production job. One major advantage of compiling your MATLAB code is that your job will not be restricted by license limitations.

You can use the MATLAB compiler, `mcc`

, on Spear or any of the HPC login nodes in order to create a binary executable of the code. Be sure to compile in whichever environment (HPC or Spear) you intend to run the code.

To compile the non-parallel code *test1.m,* use the MATLAB command:

```
mcc -R -nodisplay -R -nojvm -R -nosplash -R -singleCompThread -m test1.m
```

* Note: *This works only with the

*matlab*module, not the

*matlab_dcs*module.

The above command will create the script *run_test1.sh* and the executable *test1*. A brief description of these compiler flags follows.

```
-R: Specifies runtime options, and must be used with the other runtime flags (nodisplay, nosplash, etc)
-R -nodisplay: Any functions that provide a display will be disabled
-R -nojvm: Disable the Java Virtual Machine
-R -nosplash: Starts MATLAB, but does not display the splash screen
-R -singleCompThread: Runs only a single thread in the runtime environment
-m: Generates a C binary (-p would generate a C++ binary)
```

After successful completion, `mcc`

creates the following files:

- a binary file,
- a script to load necessary environment variables and run the binary,
- a readme.txt, and,
- a log file.

The binary file can be run via the generated script with the following command:

`<full-path-to-script>run_test1.sh $MCCROOT <input-arguments>`

You can place this command in your Slurm submit script in order to run it as part of an HPC job.

Note that input arguments will be interpreted as string values, so any code that utilizes these arguments must convert these strings to the correct data type.

One concern with this method of compiling and running a MATLAB program is that the binaries generated will contain all of the toolboxes available in the user's MATLAB environment, resulting in large binary files. To avoid this, use the `-N`

compiler flag. This will remove all but essential toolboxes, and other tools or `.m`

files required can be attached with the `-a`

option. The recommended syntax for generating serial binaries is:

```
mcc -N -v -R -nodisplay -R -nojvm -R -nosplash -R -singleCompThread -m test1.m
```

In this example, the `-v`

option enables verbose mode. To run a MATLAB program in parallel using the Parallel Computing Toolbox, add the `-p distcomp`

option to the command line arguments,

```
mcc -N -v -p distcomp -m test1.m
```

To run this binary, you must provide a parallel profile (see *Creating a Profile* above):

```
# the following assuming you are using matlab r2018b,
# revise it to reflect the version you are using
./run_test1.sh $MCCROOT -mcruserdata ParallelProfile:/gpfs/research/software/matlab/r2018b/toolbox/distcomp/parallel.settings
```

In this example, the default `local`

profile is used. This can be used in any SLURM submit script without any modification, and no MATLAB licenses will be used when the job runs.

The generation of binaries is the same for MATLAB Distributed Computing Engine jobs, but a different SLURM-aware profile needs to be provided. You should always used the supplied `HPC`

profile for these types of jobs. See the section above on submitting paraellel MATLAB jobs to see what lines should be added to a MATLAB script when using the Distributed Computing Engine. The submit command for a distributed job is:

`./run_test1.sh $MCCROOT -mcruserdata ParallelProfile:/gpfs/research/software/userfiles/hpc.settings`

### Using MATLAB with GPU Processors

MATLAB is capable of using GPUs to accelerate calculations. Most built-in functions have alternative GPU versions. In order to take advantage of GPUs, you will need to submit your jobs to the HPC **backfill2, backfill **or **genacc_q **partitions which currently have GPUs available. This will schedule your jobs to run on our GPU-enabled compute nodes.

You can try the following example, which is a mandelbrot program, `f_mandelbrot:`

```
function [mbset, t] = gpu_mandelbrot(niter, steps, xmin, xmax, ymin, ymax)
t0 = tic();
x = gpuArray.linspace(xmin, xmax, steps);
y = gpuArray.linspace(ymin, ymax, steps);
[xGrid,yGrid] = meshgrid(x, y);
c = xGrid + 1i * yGrid;
z = zeros(size(c));
mbset = zeros(size(c));
for ii = 1:niter
z = z.*z + c;
mbset(abs(z) > 2 & mbset == 0) = niter - ii;
end
t = toc(t0);
```

Run the program and display the results with the following commands:

```
[mandelSet, time] = gpu_mandelbrot(3600,100,-2,1,-1.5,1.5)
surface(mandelSet)
```

The arrays `x`

and `y`

are generated on the GPU, utilizing its massively parallel architecture, which also handles the remainder of the computations that involve these arrays.

## Submitting HPC jobs from your own copy of MATLAB

This section describes how to submit MATLAB jobs from your own workstation (desktop/laptop) to the HPC cluster. Use VPN if you are off campus.

The following figure illustrates the scheme for remote job submissions:

To submit a MATLAB job to the HPC cluster from your computer, a *generic scheduler interface* must be used. You must create a *generic *cluster object in MATLAB before you can submit jobs:

Function | Description |
---|---|

CancelJobFcn | Function to run when cancelling job |

CancelTaskFcn | Function to run when cancelling task |

CommunicatingSubmitFcn | Function to run when submitting communicating job |

DeleteJobFcn | Function to run when deleting job |

DeleteTaskFcn | Function to run when deleting task |

GetJobStateFcn | Function to run when querying job state |

IndependentSubmitFcn | Function to run when submitting independent job |

We have provided these functions in the tarball `matlab.tar`

located at:

```
/gpfs/research/software/userfiles/matlab.tar
```

These functions should be copied to:

```
[LOCAL_MATLAB_ROOT_DIRECTORY]/toolbox/local/
```

Substitute the path to MATLAB on your workstation for [`LOCAL_MATLAB_ROOT_DIRECTORY]. `

You can use the following commands to download and extract the files (assuming you are in your MATLAB root directory) on your workstation:

```
$ scp you@hpc-login.rcc.fsu.edu:/gpfs/research/software/userfiles/matlab.tar .
$ tar -xvf matlab.tar
matlab/
matlab/communicatingSubmitFcn.m
matlab/independentSubmitFcn.m
matlab/getSubmitString.m
matlab/communicatingJobWrapper.sh
matlab/createSubmitScript.m
matlab/extractJobId.m
matlab/deleteJobFcn.m
matlab/getRemoteConnection.m
matlab/getJobStateFcn.m
matlab/independentJobWrapper.sh
matlab/getCluster.m
```

Next, you must setup PKI authentication (ssh-key-based instead of password) to the HPC. Refer to our "Using SSH Keys" documentation for how to do this. Make sure that you do NOT use a passphrase when setting up your keypair.

Now, move all the files except `getCluster.m`

to the *toolbox/local* directory on your machine. For example, if you installed MATLAB to* /usr/local* directory, this can be done with:

```
mv matlab/!(getCluster.m) /usr/local/MATLAB/R2013b/toolbox/local/
```

Make sure the path is correct before you copy and that you have administrative privileges on your computer.

Next, update the `getCluster.m` file. `matlab.tar`

provides several functions you need to create a `generic`

cluster object. The function `getCluster()`

use these functions to configure a cluster object for you.

Here is the content of the `getCluster.m:`

```
function [ cluster ] = getCluster(ppn, queue, rtime, LocalDataLocation, RemoteDataLocation)
%Find the path to id_rsa key file in YOUR system and update the following line
username = 'YOUR HPC USER NAME'
keyfile = '/home/USER/.ssh/id_rsa'; %Your actual path may be DIFFERENT!
%Do not change anything below this line
if (strcmp(username, 'YOUR HPC USER NAME') == 1)
disp('You need to put your RCC user name in line 4!')
return
end
if (exist(keyfile, 'file') == 0)
disp('Key file path does not exist. Did you configure password-less login to HPC?');
return
end
ClusterMatlabRoot = '/opt/matlab/current';
clusterHost='submit.hpc.fsu.edu';
cluster = parcluster('hpc');
set(cluster,'HasSharedFilesystem',false);
set(cluster,'JobStorageLocation',LocalDataLocation);
set(cluster,'OperatingSystem','unix');
set(cluster,'ClusterMatlabRoot',ClusterMatlabRoot);
set(cluster,'IndependentSubmitFcn',{@independentSubmitFcn,clusterHost, ...
RemoteDataLocation,username,keyfile,rtime,queue});
set(cluster,'CommunicatingSubmitFcn',{@communicatingSubmitFcn,clusterHost, ...
RemoteDataLocation,username,keyfile,rtime,queue,ppn});
set(cluster,'GetJobStateFcn',{@getJobStateFcn,username,keyfile});
set(cluster,'DeleteJobFcn',{@deleteJobFcn,username,keyfile});
```

`The five input arguments of the function ``getCluster`

are

Argument | Description |
---|---|

ppn | processor/core per node |

queue | queue you want to submit job to (e.g., backfill, genacc_q) |

rtime | wall time |

LocalDataLocation | directory to store job data on your workstation |

RemoteDataLocation | directory to store job data on your HPC disk space |

The first time you download, edit the getCluster.m file to include the following:

- your RCC user name in line 4, and
- the correct path to your
*id_rsa*key file in line 5 (usually*~/.ssh/id_rsa*)

Before you call this function, create a separate folder (to be used as "RemoteDataLocation" in the `getCluster.m`

script) in your HPC disk space to store runtime MATLAB files, for example,

```
[user@hpc-login.rcc.fsu.edu] $ mkdir -p $HOME/matlab/work
```

Also create a folder in your workstation to be used as the `LocalDataLocation`

. Clean these folders regularly after finishing jobs; they tend to fill up.

### Submitting Jobs

The following lines can be used as a template to create a `generic`

cluster object which you will use to submit jobs to HPC from your local copy of MATLAB:

```
processors = 4; % Number of processors used. MUST BE LESS THAN OR EQUAL TO 32
ppn = 4; % Number of cores used per processor
queue = 'genacc_q'; % Replace this with your choice of partition
time = '01:00:00'; % Run time
LocalDataLocation = ''; % Full path to the MATLAB job folder on your workstation (not the HPC)
% Full path to a MATLAB scratch folder on HPC (replace USER with your RCC username)
RemoteDataLocation = '/gpfs/home/USER/matlab/work';
cluster = getCluster(ppn, queue, time, LocalDataLocation, RemoteDataLocation);
```

The following is an example to create a`communicating job`

for the`cluster`

:

```
j1 = createCommunicatingJob(cluster); % This example creates a communicating job (eg:parfor)
j1.AttachedFiles = {'testparfor2.m'}; % Send all scripts and data files needed for the job
set(j1, 'NumWorkersRange', [1 processors]); % Number of processors
set(j1, 'Name', 'Test'); % Give a name for your job
t1 = createTask(j1, @testparfor2, 1, {processors-1});
submit(j1);
%wait(j1); % MATLAB will wait for the completion of the job
%o=j1.fetchOutputs; % Collect outputs after job is done
```

Only use the last two lines in this example for testing small jobs. Production jobs may wait in the HPC queue for a long time and will cause your copy of MATLAB to wait until the job has completed before being usable again. For production jobs, you should write data directly to the filesystem from your MATLAB script instead of collecting it when the script is complete.Note.

The following is an example to create an `independent job`

for the `cluster`

:

```
j2 = createJob(cluster); % create an independent job
t2 = createTask(j2, @rand, 1, {{10,10},{10,10},{10,10},{10,10}}); % create an array of 4 tasks
submit(j2)
wait(j2)
o2 = fetchOutputs(j2) % fetch the results
o2{1:4} % display the results
```

** Note.** A

`communicating`

job contains only one task. However, this task can run on multiple workers. Also, the task can contain `parfor-loop`

or `spmd`

code block to improve the performance. Conversely, an `independent`

job can contain multiple tasks. These tasks do not communicate with each other and each task runs on a single worker.## MATLAB Workshops

We regularly offer MATLAB workshops. Keep an eye on our event schedule or send us a message to find out when the next one will be. You can also check out our slides and materials from past workshops.