Oprofile is a flexible, kernel level application profiling and instrumentation tool available on the DRP Cluster. Its use can impact application performance, so it must be requested through the scheduler when submitting a job for the profiling environment to be available on nodes allocated to the job. This is done by adding the string "oprofile" (case sensitive) anywhere in the job's user defined name (set with the -J or --job-name flag to sbatch, srun, and salloc). Here's a brief example:

sbatch -J myoprofilejob -n 1 ~/myjob.sh

The myjob.sh script can either pause to allow interactive use of the tool on the node (ssh access to nodes is allowed while one's jobs are running) or contain scripted use of the tool. For more information on Oprofile, see the Oprofile project's online documentation: http://oprofile.sourceforge.net/docs/ or example guide: http://oprofile.sourceforge.net/examples/