Nothing Special   »   [go: up one dir, main page]
Skip to main content
Springer Nature Link
Log in

Highly stable localized ILU preconditioning for unstructured grids

  • 5. Posters
  • Conference paper
  • First Online:
High-Performance Computing and Networking (HPCN-Europe 1998)

Part of the book series: Lecture Notes in Computer Science ((LNCS,volume 1401))

Included in the following conference series:

  • 120 Accesses

Abstract

Authors have shown that “Localized” ILU(O) factorization is very effective preconditioning method for parallel iterative solvers. This method provides data locality and good parallelization effect because there are no inter-processor communications. But it's not as strong a method as the original ILU(O) preconditioner. Therefore, the convergence is getting slower if the processor number becomes larger.

In this paper, authors have developed a new method with faster and more stable convergence by considering the effects of interpartition communication during preconditioning. Thus, problem converges faster in many PE cases although slightly more calculations and communications are required.

Effect of partitioning, depth of external nodes and ordering have been investigated for various type of examples. Developed system performance has been also evaluated on workstation cluster with MPI.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Log in via an institution

Institutional subscriptions

References

  1. J.J.Dongarra et. al. Templates for the Solution, of Linear Systems: Building Blocks for Iterative Methods, SIAM, 1994.

    Google Scholar 

  2. W.Gropp et. al. Using MPI, MIT Press, 1994.

    Google Scholar 

  3. V. Venkatakrishnan. Parallel Implicit Methods for Aerodynamic Applications on Unstructured Grids, SIAM, 1994.

    Google Scholar 

  4. K.Murata et. al. Large Scale Numerical Simulation (in Japanese), Iwanami-Books, 1990.

    Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Mitsubishi Research Institute, Inc., Japan

    Kengo Nakajima (Senior Staff Researcher)

  2. Research Organization for Information Science and Technology, Japan

    Hisashi Nakamura (General Manager)

  3. Yokohama National University, Japan

    Hiroshi Okuda (Associate Professor)

Authors
  1. Kengo Nakajima
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Hisashi Nakamura
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Hiroshi Okuda
    View author publications

    You can also search for this author in PubMed Google Scholar

Editor information

Peter Sloot Marian Bubak Bob Hertzberger

Rights and permissions

Reprints and permissions

Copyright information

© 1998 Springer-Verlag Berlin Heidelberg

About this paper

Cite this paper

Nakajima, K., Nakamura, H., Okuda, H. (1998). Highly stable localized ILU preconditioning for unstructured grids. In: Sloot, P., Bubak, M., Hertzberger, B. (eds) High-Performance Computing and Networking. HPCN-Europe 1998. Lecture Notes in Computer Science, vol 1401. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0037228

Download citation

  • DOI: https://doi.org/10.1007/BFb0037228

  • Published:

  • Publisher Name: Springer, Berlin, Heidelberg

  • Print ISBN: 978-3-540-64443-9

  • Online ISBN: 978-3-540-69783-1

  • eBook Packages: Springer Book Archive

Publish with us

Policies and ethics

Search

Navigation