Difference between revisions of "Matlab"

From UFRC
Jump to navigation Jump to search
Line 90: Line 90:
 
----
 
----
 
Documentation for the MatLab compiler can be viewed on the [http://www.mathworks.com/access/helpdesk/help/toolbox/compiler MatLab Website].
 
Documentation for the MatLab compiler can be viewed on the [http://www.mathworks.com/access/helpdesk/help/toolbox/compiler MatLab Website].
 
  
 
===Is the Distributed Computing Toolbox available?===
 
===Is the Distributed Computing Toolbox available?===

Revision as of 18:30, 7 October 2010

Installation

Currently matlab is located here:

Version 2009b: /apps/matlab/2009b/

Policies

  1. You may not run MatLab nor compiled MatLab executables interactively on submit1 or submit2.
  2. You must use the test nodes (test01 - test05) to test and compile your MatLab programs.
  3. Long running (more than 30 minutes) MatLab jobs must be submitted to the batch system.
  4. Large numbers of jobs must not be run on the test nodes but must be submitted to the batch system.
  5. The UF HPC Center has no budget for application software and licensing. All MatLab licenses (including the compiler) come from, and are shared with, all of campus. For this reason it is sometimes difficult to obtain a license. If your research depends on MatLab and you do not want to be delayed by licensing issues, you should consider purchasing a MatLab license for your personal or group use.

Compiling MatLab Programs

MatLab programs must be compiled via the MatLab compiler before they can be run in batch mode on the HPC Center cluster. In order to compile MatLab programs you should

1. Once on submit1 or submit2, log into one of the test nodes (test01 - test05) via ssh. Due to user memory constraints on submit1 and submit2, the MatLab compiler cannot be used on them.

submit1% ssh test05

2. Set up your environment so that the libraries needed by the compiler can be found.

 test05% export LD_LIBRARY_PATH=/apps/matlab/2009b/bin/glnxa64:/apps/matlab/2009b/sys/os/glnxa64
 test05% export PATH=${PATH}:/apps/matlab/2009b/bin

Note: These environment settings should be included in your batch submission script since they will also be needed by the resulting MatLab executable.

3a. For batch jobs, you should also set the MCR_CACHE_ROOT environment variable in your PBS job script.

 test05% export MCR_CACHE_ROOT=${TMPDIR}

Please see the PBS Sample Job Scripts for a sample MatLab submission script.

3b. For an interactive run (only on the test nodes), you should set MCR_CACHE_ROOT to a directory /scratch/local.

 test05% mkdir -p /scratch/local/$USER/mcr_cache
 test05% export MCR_CACHE_ROOT=/scratch/local/$USER/mcr_cache

4. Compile your matlab program (.m) files using the MatLab compiler (mcc).

 test05% mcc -R -singleCompThread -m some_prog.m

Example

Let us say that we have a program such as that below in a source file named houdini.m

function [m,t] = houdini();
% Doc example, chapter 3.

% Copyright 1997 The MathWorks, Inc.
% $Revision: 1.1.6.1 $

m = magic(2); % Assign 2x2 matrix to m. 
t = m .^ 3;   % Cube each element of m. 
disp(t);      % Display the value of t.

Since we already have the environment setup, we can compile this program. It is recommended that this be done in a separate directory as a number of files are produced.

 test05% mkdir tmp
 test05% cd tmp
 test05% cp ../houdini.m .
 test05% ls
houdini.m
 test05% export LD_LIBRARY_PATH=/apps/matlab/2009b/bin/glnxa64:/apps/matlab/2009b/sys/os/glnxa64
 test05% export PATH=${PATH}:/apps/matlab/2009b/bin
 test05% mcc -R -singleCompThread -m houdini.m 
Warning: Unable to open display , MATLAB is starting without a display.
  You will not be able to display graphics on the screen.
 test05% ls -l
total 124
-rwxr-xr-x  1 root root 12299 Apr  3 10:08 houdini
-rw-r--r--  1 root root 63146 Apr  3 10:08 houdini.ctf
-rw-r--r--  1 root root   239 Apr  3 10:06 houdini.m
-rw-r--r--  1 root root  2900 Apr  3 10:08 houdini_main.c
-rw-r--r--  1 root root  6328 Apr  3 10:08 houdini_mcc_component_data.c
-rw-r--r--  1 root root  5461 Apr  3 10:08 houdini.prj
-rw-r--r--  1 root root  1010 Apr  3 10:08 mccExcludedFiles.log
-rw-r--r--  1 root root  6757 Apr  3 10:08 readme.txt
-rwxr--r--  1 root root  1446 Apr  3 10:08 run_houdini.sh

The MatLab executable houdini has been generated.

Note: The "-R -singleCompThread" options on the mcc command line will result in a single-threaded executable. By default, multi-threading is enabled in executables produced by mcc. The result is that one computational thread will be started for each processor in a machine that runs your executable. That's not usually what you want. Unless you know what you are doing and intend to take advantage of multithreading (including requesting the appropriate resources in your submission script) you should use these options and generate a single-threaded executable. Otherwise, your MatLab jobs will oversubscribe the compute nodes and interfere with other users' programs.


Documentation for the MatLab compiler can be viewed on the MatLab Website.

Is the Distributed Computing Toolbox available?

No, not at this time. The University of Florida has not purchased a license for that product. The HPC Center has no budget for application software or licensing. Users and/or groups must purchase commercial software on their own.

Compiled DLL

If you have compiled a DLL to use as an executable, it will not run on the cluster. DLL's are Microsoft Windows based dynamically linked libraries that will only work under the Windows operating system.

Setting a Path

In order to set a path in your MatLab program should add the following code.

path(path, '/path/to/wherever/you/want')

This would add a search path to the MatLab system. For more information see the Matlab web site.

Known Issues

No Licenses Available

You may run into an error when using the MatLab compiler such as

Licensed number of users already reached.

If so, the number of licenses available for the MatLab compiler has been exceeded. There are only two shared among the entire university so you should expect to see this problem from time to time if you use the MatLab compiler frequently.

MCR_CACHE_ROOT Directory

Could not access the MCR component cache

This error message means that MatLab is not able to access the MCR cache directory which, by default, is created by MatLab in your home directory. If you are running a large number of MatLab jobs, the file locking used by MatLab to manage concurrent access to files in the MCR_CACHE_ROOT directory can generate a significant load on the file system on which these files reside. Thus, we ask that you use a local scratch directory for your MCR_CACHE_ROOT directory. To do so, set the value of MCR_CACHE_ROOT to $TMPDIR in your job submission script.

 export MCR_CACHE_ROOT=$TMPDIR