Parallel HPF-MPI implementation of the TBSCM Poisson solver

  • Jean-Yves Berthou
  • Laurent Plagne
2. Computational Science
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1401)


The aim of this paper is to describe the parallel implementation of the Tensorial Basis Spline Collocation Method for solving Poisson's equation. The TBSCM Poisson solver has been developed mainly in High Performance Fortran(HPF) [For97]. Some small parts, which were inefficiently compiled by the pghpf2.3 HPF compiler, have been written with the Message Passing Interface library (MPI)[The94] using the HPF-LOCAL extrinsic feature. 90% of parallelization efforts have been concentrated on the most computationally intensive part of the code. This kernel achieves very good scalability and it attains 85 Gflops (85 109 floating points operations per second) on 256 Cray T3E-750 processors with a 5123 mesh which represents 332 Mflops per processor. Once the kernel was efficient, the whole code was ported in HPF-MPI. The whole code also scales well and achieves good performance since nearly 210 Mflops per processor are obtained with a 2563 mesh on 128 Cray T3E-750 processors.


Poisson Solver Vlasov Equation High Performance Fortran Message Passing Interface library Semi-Automatic Parallelization 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [BC97]
    Jean-Yves Berthou and Laurent Colombet. Which approach to parallelizing scientific codes — that is the question. Parallel Computing, (23):165–179, 1997.Google Scholar
  2. [BFB92]
    B. Bialecki, G. Fairweather, and K.R. Bennett. Fast direct solvers for piecewise hermite bicubic orthogonal spline collocation equations. SIAM J. Numerical Analysis, 29:156, 1992.Google Scholar
  3. [BMN+97]
    Z. Bozkus, L. Meadows, S. Nakamoto, V. Schuster, and M. Young. Pghpf-an optimizing high performance fortran compiler for distributed memory machines. Scientific Programming, 6(1):29–40, Spring 1997.Google Scholar
  4. [For97]
    High Performance Fortran Forum. High Performance Fortran Language Specification, version 2.0. Technical report, Center for Research on Parallel Computation, Rice University, Houston, TX, January 1997.Google Scholar
  5. [Jac75]
    J.D. Jackson. Classical Electrodynamics. Wiley, New York, 1975.Google Scholar
  6. [Res]
    Cray Research. Message passing toolkit: Release overview, ro-5290, 1.1. Technical report, July.Google Scholar
  7. [Res95]
    Applied Parallel Research. Forge high performance fortran, xhpf version 2.1, user's guide. Technical report, October 1995.Google Scholar
  8. [The94]
    The MPI Forum. Document for a Standard Message-Passing Interface, April 1994.Google Scholar
  9. [TPG96]
    Inc The Portland Group, pghpf user's guide, version 2.1. Technical report, May 1996.Google Scholar
  10. [UWSB91]
    A.S. Umar, J. Wu, M.R. Strayer, and C. Bottcher. Basis-spline collocation method for the lattice solution of boundary value problems. Computational Physics, 93:426, 1991.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1998

Authors and Affiliations

  • Jean-Yves Berthou
    • 1
  • Laurent Plagne
    • 2
  1. 1.EDF-DER/IMA/MMN/ISAClamart CedexFrance
  2. 2.Dèpartement de Recherche Fondamentale sur la Matière Condensèe CEA GrenobleGrenoble Cedex 9France

Personalised recommendations