OpenSpeedShop is an open source multi platform Linux performance tool which is targeted to support performance analysis of applications running on both single node and large scale platforms.

This page is a tour of some of the functionality – please see the website for more details:

Basic usage

Before using, please execute the following to make the software available:

module add openspeedshop

OpenSpeedShop refer to the collection of application performance data as experiments. It supports a wide variety of different experiments, and new ones can be created by writing and adding new plugins.

No program recompilation is necessary; however, more information about source code line numbers of particular areas of the code can be provided if the application was compiled with the -g flag.

To sample where the time is being spent in an application, use the pcsamp experiment using a line in a job script of the form:

Type of program Job Script Command
Non-MPI osspcsamp “<program>”
MPI based osspcsamp “mpirun <program>”

By default, this will store raw data under /scratch . Once the application is complete, this is postprocessed into a database file with a name like <program>-pcsamp.openss

Once some experiment data has been generated, view with:

openss -f -pcsamp.openss

This should show where the application is spending its time.

Other experiments

There are many other experiments that can be done on an application, including hardware exceptions fpe , hardware sampling (i.e. PAPI counters) hwcsamp , call path profiling usertime , Lightweight MPI summary mpip , MPI summary, mpi , MPI tracing mpit , etc. Typically the command to execute is the experiment name with oss prefixed to it.

Please see the OpenSpeedShop documentation for the full list.