Research Computing

Please note the following requirements when running Matlab on the Nobel clusters:

• To run Matlab interactively with its graphic interface, invoke it with the -singleCompThread command-line argument to prevent the creation of multiple threads.
• To run Matlab in command line mode, invoke it with the -singleCompThread -nodisplay -nosplash -nojvm command-line arguments to prevent the use of X Windows, and the creation of multiple threads:
  • /usr/licensed/bin/matlab -singleCompThread -nodisplay -nosplash -nojvm -r <m_script_name>
    (e.g., if your *.m script is 'myscript.m', then you must specify 'myscript' as <m_script_name>
    without the quotes and without the '.m')
• NOTE: In the spirit of fairness to all users sharing these servers, running Matlab in multi-threaded mode on Nobel is not permitted.

Please note the following requirements when running Matlab on the Adroit or Della clusters:

• Invoke matlab with the -singleCompThread -nodisplay -nosplash -nojvm options to prevent the use of X Windows, and the creation of multiple threads.
• Your PBS script file must include the following lines:
  • #PBS -l nodes=1:ppn=1
  • cd  <directory_containing_m_script>
  • /usr/licensed/bin/matlab -singleCompThread -nodisplay -nosplash -nojvm -r <m_script_name>
    (e.g., if your *.m script is 'myscript.m', then you must specify 'myscript' as <m_script_name>
    without the quotes and without the '.m')

Most of the time, running Matlab in single-threaded mode (as described above) will meet your needs. But if you have very mathematically intense computations that might benefit from the built-in multi-threading provided by Matlab's BLAS implementation, then you can experiment with running in multi-threaded mode as follows:

• You will need to invoke Matlab as above, but without the -singleCompThread option.
• Your *.m script must include the following line:
  
  • maxNumCompThreads(N)
    (Note that you'll probably get a deprecation warning, which you can ignore.)
• Your PBS script must include the following lines:
  • #PBS -l nodes=1:ppn=N
  • cd  <directory_containing_m_script>
  • /usr/licensed/bin/matlab -nodisplay -nosplash -nojvm -r <m_script_name>
    (e.g., if your *.m script is 'myscript.m', then you must specify 'myscript' as <m_script_name>
    without the quotes and without the '.m')

NOTE: Both in your *.m script and in your PBS script, 'N' must be a number from 2 to the maximum number of cores per node on the cluster you are running. Use the smallest 'N' that gives you a significant performance boost.