Research Article

A role for Pareto optimality in mining performance data

  1. Joël M. Malard*

Article first published online: 29 NOV 2004

DOI: 10.1002/cpe.877

Issue

Concurrency and Computation: Practice and Experience

Concurrency and Computation: Practice and Experience

Volume 17, Issue 1, pages 27–48, January 2005

Additional Information

How to Cite

Malard, J. M. (2005), A role for Pareto optimality in mining performance data. Concurrency and Computation: Practice and Experience, 17: 27–48. doi: 10.1002/cpe.877

Author Information

  1. Pacific Northwest National Laboratory, Battelle Boulevard, P.O. Box 999, Richland, WA 99352, U.S.A.

*Computational Science and Applied Mathematics, Pacific Northwest National Laboratory, Battelle Boulevard, P.O. Box 999, Richland, WA 99352, U.S.A.

Publication History

  1. Issue published online: 29 NOV 2004
  2. Article first published online: 29 NOV 2004
  3. Manuscript Accepted: 30 DEC 2003
  4. Manuscript Revised: 24 DEC 2003
  5. Manuscript Received: 31 DEC 2002

SEARCH

SEARCH BY CITATION

ARTICLE TOOLS

Get PDF (906K)

Keywords:

  • Pareto efficiency;
  • dendogram;
  • multiobjective optimization;
  • software performance;
  • hardware events

Abstract

Improvements in performance modeling and identification of computational regimes within software libraries is a critical first step in developing software libraries that are truly agile with respect to the application as well as to the hardware. It is shown here that Pareto ranking, a concept from multi-objective optimization, can be an effective tool for mining large performance datasets. The approach is illustrated using software performance data gathered using both the public domain LAPACK library and an asynchronous communication library based on IBM LAPI active message library. Copyright © 2005 John Wiley & Sons, Ltd.

Get PDF (906K)