Lumerical Software on WestGrid Systems
Introduction
Lumerical FDTD Solutions Engine is available on the WestGrid Orcinus cluster. The Lumerical Engine was formerly available to WestGrid users on the glacier.westgrid.ca system, with a limit of 10 concurrent simulations. Through a donation from Lumerical, this license has been extended to the Orcinus cluster.
For details how to create the input file and analyze the data, please visit the Lumerical web site (http://www.lumerical.com/support.php).
User Responsibilities
WestGrid no longer has the Shared User Interface license. It is up to the users to obtain a client license.
The CAD - FDTD Lumerical User Interface is not supported, please contact http://www.lumerical.com/contact.php for details.
Submitting Lumerical jobs on Orcinus
An example job submission script is available on Orcinus at
/global/system/info/notes/script-examples/FDTD/FDTD-7.5.2-parallel.pbs
An alternative to creating your own batch job script is to run fdtd-run-pbs.sh, which is a script that automatically creates and submits a batch job script for each of one or more specified Lumerical .fsp files. Use the module command to set up the environment for the particular version of the FDTD Solutions software you want to use prior to running fdtd-run-pbs.sh:
module load lumerical/7.5.2
fdtd-run-pbs.sh -n number_of_processors project1.fsp project2.fsp ...
The script reads the .fsp files to extract information about memory and walltime requirements and substitutes these in a template batch job script (/global/software/FDTD_Solutions-7.5.2/templates/fdtd-pbs-template.sh), to produce corresponding batch job files project1.sh, project2.sh, etc. in your directory. These scripts are then submitted as individual batch jobs, each to be run on the number of cores specified with the -n argument.
The -n argument to the fdtd-run-pbs.sh script is optional. If it is omitted, the jobs will be run using 8 cores. Also, with the default template file, the script uses the line
#PBS -l qos=parallel
The qos=parallel resource request is appropriate only when the number of cores requested is four or more. The job will generate an error if you try to use the fdtd-run-pbs.sh request for jobs with smaller numbers of cores. You could edit the job scripts to remove that line if you wanted to use less than four cores and then manually submit the job with qsub:
qsub project1.fsp
qsub project2.fsp
...
A disadvantage of specifying more than one .fsp file on the fdtd-run-pbs.sh command line is that the files from all the jobs will be written to the same directory. You may prefer to keep each job in a separate directory.
Updated 2011-03-08.
