Event-based performance analysis aims at modeling the behavior of parallel applications through a series of state transitions during execution. Different approaches to obtain such transition points for OpenMP programs include source-level instrumentation (e.g., OPARI) and callback-driven runtime support (e.g., OMPT). In this paper, we revisit a previous evaluation and comparison of OPARI and an LLVM OMPT implementation—now updated to the OpenMP 5.0 specification—in the context of Score-P. We describe the challenges faced while trying to use OMPT as a drop-in replacement for the existing instrumentation-based approach and the changes in event order that could not be avoided. Furthermore, we provide details on Score-P measurements using OPARI and OMPT as event sources with the EPCC and SPEC OpenMP benchmark suites.
CITATION STYLE
Feld, C., Convent, S., Hermanns, M. A., Protze, J., Geimer, M., & Mohr, B. (2019). Score-P and OMPT: Navigating the perils of callback-driven parallel runtime introspection. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11718 LNCS, pp. 21–35). Springer Verlag. https://doi.org/10.1007/978-3-030-28596-8_2
Mendeley helps you to discover research relevant for your work.