Profiling Parallel Application Performance

08/01/2014 - 10:35am
Mesa Lab Main Seminar Room

Thomas Gowan portrait

Thomas Gowan, SIParCS Intern
(Pennsylvania State University)

Small percentage gains in the FLOP performance of parallel MPI-based Scientific applications can cause significant gains in their performance.  Using tools such as profilers, one can determine and assess where changes can be made to applications to increase their performance.  Efforts are underway to profile the MPI-based scientific applications, such as CESM and WRF that run on Yellowstone.  I plan to utilize and explore some of the performance analysis tools available on Yellowstone, such as TAU, Allinea’s MAP, and Intel’s ITAC to investigate their utility for profiling CESM and WRF.

Video Presentation

Gowan, T_SIParCS2014.ppsx