A Dynamic Tracing Mechanism for Performance Analysis of OpenMP Applications
In this paper we present OMPtrace, a dynamic tracing mechanism that combines traditional tracing with dynamic instrumentation and access to hardware performance counters to create a powerful tool for performance analysis and optimization of OpenMP applications. Performance data collected with OMPtrace is used as input to the Paraver visualization tool for detailed analysis of the parallel behavior of the application. We demonstrate the usefulness of OMPtrace and the power of Paraver for tuning OpenMP applications with a case study running the US DOE ASCI Sweep3D benchmark on the IBM SP system at the Lawrence Livermore National Laboratory.
KeywordsCritical Section Lawrence Livermore National Laboratory Large Problem Size Parallel Loop Small Problem Size
Unable to display preview. Download preview PDF.
- 2.L. DeRose and T. H. Hoover Jr. and J. K. Hollingsworth The Dynamic Probe Class Library — An Infrastructure for Developing Instrumentation for Performance Tools. In Proceedings of 2001 International Parallel and Distributed Processing Symposium, April 2001.Google Scholar
- 3.European Center for Parallelism of Barcelona (CEPBA). Paraver — Parallel Program Visualization and Analysis Tool — Reference Manual, November 2000. http://www.cepba.upc.es/paraver.
- 4.K. R. Koch, R. S. Baker, R. E. Alcouffe. Solution of the First-Order Form of the 3-D Discrete Ordinates Equation on a Massively Parallel Processor. In Trans. Amer. Nuc. Soc. 65(198), 1992.Google Scholar
- 5.B. P. Miller, M. D. Callaghan, J. M. Cargille, J. K. Hollingsworth, R. B. Irvin, K. L. Karavanic, K. Kunchithapadam, and T. Newhall. The Paradyn Parallel Performance Measurement Tools. In IEEE Computer, 28(11):37–46, November 1995.Google Scholar