SKaMPI: a comprehensive benchmark for public benchmarking of MPI

  • Authors:

  • Ralf Reussner;Peter Sanders;Jesper Larsson Trä/ff

  • Affiliations:

  • Distributed Systems Technology Center (DSTC), Monash Univ. Caulfield Campus, VIC 3145, Australia. E-mail: reussner@acm.org (Work performed: Lehrstuhl Informatik fü/r Ingenieure und Naturwissen ...;Max-Planck-Institut fü/r Informatik, Stuhlsatzenhausweg 85, D-66123 Saarbrü/cken, Germany. E-mail: sanders@mpi-sb.mpg.de;C&C/ Research Laboratories, NEC Europe, Rathausallee 10, D-53757 Sankt Augustin, Germany. E-mail: traff@ccrl-nece.de

  • Venue:
  • Scientific Programming
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

The main objective of the MPI communication library is to enable {\it portable parallel programming} with high performance within the message-passing paradigm. Since the MPI standard has no associated performance model, and makes no performance guarantees, comprehensive, detailed and accurate performance figures for different hardware platforms and MPI implementations are important for the application programmer, both for understanding and possibly improving the behavior of a given program on a given platform, as well as for assuring a degree of predictable behavior when switching to another hardware platform and/or MPI implementation. We term this latter goal {\it performance portability}, and address the problem of attaining performance portability by benchmarking. We describe the SKaMPI benchmark which covers a large fraction of MPI, and incorporates well-accepted mechanisms for ensuring accuracy and reliability. SKaMPI is distinguished among other MPI benchmarks by an effort to maintain a public performance database with performance data from different hardware platforms and MPI implementations.