Skip to main navigation Skip to search Skip to main content

Understanding performance variability in standard and pipelined parallel Krylov solvers

  • Argonne National Laboratory
  • Swiss Federal Institute of Technology Zurich

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

In this work, we collect data from runs of Krylov subspace methods and pipelined Krylov algorithms in an effort to understand and model the impact of machine noise and other sources of variability on performance. We find large variability of Krylov iterations between compute nodes for standard methods that is reduced in pipelined algorithms, directly supporting conjecture, as well as large variation between statistical distributions of runtimes across iterations. Based on these results, we improve upon a previously introduced nondeterministic performance model by allowing iterations to fluctuate over time. We present our data from runs of various Krylov algorithms across multiple platforms as well as our updated non-stationary model that provides good agreement with observations. We also suggest how it can be used as a predictive tool.

Original languageEnglish
Pages (from-to)47-59
Number of pages13
JournalInternational Journal of High Performance Computing Applications
Volume35
Issue number1
DOIs
StatePublished - Jan 2021

Keywords

  • BiCGStab
  • GMRES
  • Krylov
  • PGMRES
  • noise
  • performance model
  • pipelining

Fingerprint

Dive into the research topics of 'Understanding performance variability in standard and pipelined parallel Krylov solvers'. Together they form a unique fingerprint.

Cite this