Advertisement

Application of Mixed MPI/OpenMP Programming in a Multi SMP Cluster Computer

  • Adam Smyk
  • Marek Tudruj
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 2328)

Abstract

The paper presents results of experiments concerned with parallel program implementation based on mixed mode communication. Parallel programs were implemented with the simultaneous use of communication library based on message passing at the level of parallel processes (MPI) and shared memory communication at the level of computing threads (OpenMP). The experiments were done in the executive computer environment that was composed of several shared memory multiprocessor clusters connected by external inter-processor communication network. Different parallelisation methods (program decomposition into processes and threads) have been studied and experimentally evaluated with the use of communication networks with drastically different bandwidths: FastEthernet and Myrinet. The measurements revealed the superiority of the mixed communication mode programming. The experiments enabled to estimate the required bandwidth of the message passing network that provides program execution efficiency equivalent to mixed mode communication. The examined application concerned the FDTD method used for electromagnetic wave propagation modeling.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    B. Wilkinson, M. Allen: Parallel Programming Techniques and Applications using Networked Clusters of Workstations and Parallel Comp., Prentice Hall 1999, pp 450.Google Scholar
  2. 2.
    Y. Kanaka, M. Matsuda, M. Ando, K. Kazuto, M. Sato: COMPaS: A Pentium Pro PC-based SMP Cluster and its Experience, IPPS Workshop on Personal Computer Based Networks of Workstations, LNCS 1388, pp. 486–497. 1998.Google Scholar
  3. 3.
    Pentium Pro Cluster Workshop, http://www.scl.ameslab.gov/workshops/.
  4. 4.
    Scalable Clusters of Commodity Computers, http://www.csag.cs.uiuc.edu/projects/clusters.html.
  5. 5.
    K.S. Kunz, R.J. Luebers: The Finite Difference Time Domain Method for Electromagnetics, CRC Press Inc., 1993.Google Scholar
  6. 6.
  7. 7.
    M. Snir et al.: MPI, The Complete Reference, The MIT Press, 1998.Google Scholar
  8. 8.
    L. Smith, M. Bulk: Development of Mixed Mode MPI/OpenMP Applications, WOMPAT 2000, San Diego Supercomputer Center, California, July 6th-7th, 2000.Google Scholar
  9. 9.
    P. Lanucara, S. Rovida: Conjugate-Gradient alghoritms: An MPI-OpenMP implementation on distributed-shared memory systems, Proceedings of the 1st European Workshop on OpenMP, Lund, Sweden, 1999, pp 76–78.Google Scholar
  10. 10.
    Allan J. Wallcraft: SPMD OpenMP vs MPI for Ocean Models, Proceeding of the 1st European Workshop on OpenMP, Lund, Sweden, 1999.Google Scholar
  11. 11.
    T. Boku, K. Itakura, S. Yoshikawa, M. Kondo, M. Sato: Performance Analysis of PC-CLUMP Based on SMP Bus Utilization, Proceedings of the Second Workshop on Cluster-Based Computing, Santa Fe, New Mexico, 2000.Google Scholar
  12. 12.
    K.S. Yee: Numerical Solution of Initial Boundary Value Problems Maxwell’s Equations in Isotropic Media, IEEE Trans. On Antennas and Propagation, Vol. AP-14, N. 3, pp. 302–307, (1966).Google Scholar
  13. 13.
    P-GRADE, A Professional Graphical Parallel Programming Environment, http://www.lpds.sztaki.hu/projects/p-grade/.

Copyright information

© Springer-Verlag Berlin Heidelberg 2002

Authors and Affiliations

  • Adam Smyk
    • 1
  • Marek Tudruj
    • 1
  1. 1.Polish - Japanese Institute of Information TechnologyWarsawPoland

Personalised recommendations