PMaCToolkit for IPM
View results of profiling applications on different systems
IPM is a portable profiling infrastructure for parallel codes. It provides a low-overhead performance summary of the computation and communication in a parallel program. The amount of detailed reported is selectable at runtime via environment variables or through a MPI_Pcontrol interface. IPM has extremely low overhead, is scalable and easy to use requiring no source code modification. More details are available at the IPM Sourceforge page.
You can view a summary of results.
You can also filter the results.
IPM is collaborative project between PMaC at SDSC and David Skinner and Katherine Yelick of UC Berkeley and LBNL/NERSC.
Current work on IPM is focused on extending its capabilties to petascale-era machines as well as using IPM to further understand and predict performance of scientific codes.
For more information about IPM or for assistance with installing it on your HPC resource, please contact Nick Wright.
