An efficient parallel finite-element-based domain decomposition iterative technique with polynomial preconditioning

Yu Liang, Ramdev Kanapady, Kumar K. Tamma

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

An efficient parallel finite element-based non-overlapping domain decomposition iterative technique formulated with polynomial preconditioned GMRES method is presented. Unlike the standard row-oriented partitioning of a matrix the finite element based domain decomposition solver with polynomial preconditioning circumvents the assembly of matrix, reordering of matrix redundant computations associated with the interface elements, numerical problems associated with local preconditioned and costly global preconditioner construction. A dramatic reduction in parallel overhead both in terms of computation and communication results in a highly scalable solver. The parallel performance results for large-scale static and dynamic problems on the IBM SP2 and the SGI Origin are presented.

Original languageEnglish (US)
Title of host publicationProceedings of the 2006 International Conference on Parallel Processing Workshops, ICPP 2006
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages6-11
Number of pages6
ISBN (Print)0769526373, 9780769526379
DOIs
StatePublished - 2006
Event2006 International Conference on Parallel Processing Workshops, ICPP 2006 - Columbus, OH, United States
Duration: Aug 14 2006Aug 18 2006

Publication series

NameProceedings of the International Conference on Parallel Processing Workshops
ISSN (Print)1530-2016

Other

Other2006 International Conference on Parallel Processing Workshops, ICPP 2006
Country/TerritoryUnited States
CityColumbus, OH
Period8/14/068/18/06

Keywords

  • Distributed format
  • Domain decomposition
  • Finite element
  • GMRES
  • Linear equations
  • Polynomial preconditioner

Fingerprint

Dive into the research topics of 'An efficient parallel finite-element-based domain decomposition iterative technique with polynomial preconditioning'. Together they form a unique fingerprint.

Cite this