Authors: Peter Philippen
The Score-P package provides code instrumentation and performs run-time measurement for HPC applications in C, C++, or Fortran parallelized with MPI, OpenMP, CUDA, and hybrid combinations of them. It works as native data collection component for the established performance analysis tools TAU, Vampir, and Scalasca, collecting either profiles or event traces. This talk answers the questions how to prepare a parallel target application and how to execute a measurement run for the tools TAU, Vampir, and Scalasca as presented earlier in individual talks.
Peter Philippen got a masters degree in mechanical engineering from the RWTH Aachen University in 2010. Since then he has been working at the Juelich Supercomputing Centre within the Scalasca team focusing on performance measurements and analysis, especially focusing on the development of the joint measurement system Score-P.
Attachment | Size |
---|---|
Score-P.pdf | 2.49 MB |