A parallel implementation of the block preconditioned GCR method

C. Vuik; J. Frank

doi:10.1007/bfb0100666

A parallel implementation of the block preconditioned GCR method

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

The parallel implementation of GCR is addressed, with particular focus on communication costs associated with orthogonalization processes. This consideration brings up questions concerning the use of Householder reflections with GCR. To precondition the GCR method a block Gauss-Jacobi method is used. Approximate solvers are used to obtain a solution of the diagonal blocks. Experiments on a cluster of HP workstations and on a Cray T3E are given.

Original language	English
Title of host publication	High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings
Editors	Peter Sloot, Alfons Hoekstra, Bob Hertzberger, Marian Bubak
Publisher	Springer
Pages	1052-1060
Number of pages	9
ISBN (Print)	3540658211, 9783540658214
DOIs	https://doi.org/10.1007/bfb0100666
Publication status	Published - 1999
Externally published	Yes
Event	7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999 - Amsterdam, Netherlands Duration: 12 Apr 1999 → 14 Apr 1999

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	1593
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999
Country/Territory	Netherlands
City	Amsterdam
Period	12/04/99 → 14/04/99

Bibliographical note

Publisher Copyright:
© Springer-Verlag Berlin Heidelberg 1999.

Keywords

Approximate subdomain solution
Orthogonalization methods
Parallel Krylov subspace methods

Access to Document

10.1007/bfb0100666

Cite this

Vuik, C., & Frank, J. (1999). A parallel implementation of the block preconditioned GCR method. In P. Sloot, A. Hoekstra, B. Hertzberger, & M. Bubak (Eds.), High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings (pp. 1052-1060). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1593). Springer. https://doi.org/10.1007/bfb0100666

Vuik, C. ; Frank, J. / A parallel implementation of the block preconditioned GCR method. High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings. editor / Peter Sloot ; Alfons Hoekstra ; Bob Hertzberger ; Marian Bubak. Springer, 1999. pp. 1052-1060 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{45178d57570e4534a0b41c6a51ea5b97,

title = "A parallel implementation of the block preconditioned GCR method",

abstract = "The parallel implementation of GCR is addressed, with particular focus on communication costs associated with orthogonalization processes. This consideration brings up questions concerning the use of Householder reflections with GCR. To precondition the GCR method a block Gauss-Jacobi method is used. Approximate solvers are used to obtain a solution of the diagonal blocks. Experiments on a cluster of HP workstations and on a Cray T3E are given.",

keywords = "Approximate subdomain solution, Orthogonalization methods, Parallel Krylov subspace methods",

author = "C. Vuik and J. Frank",

note = "Publisher Copyright: {\textcopyright} Springer-Verlag Berlin Heidelberg 1999.; 7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999 ; Conference date: 12-04-1999 Through 14-04-1999",

year = "1999",

doi = "10.1007/bfb0100666",

language = "English",

isbn = "3540658211",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "1052--1060",

editor = "Peter Sloot and Alfons Hoekstra and Bob Hertzberger and Marian Bubak",

booktitle = "High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings",

}

Vuik, C & Frank, J 1999, A parallel implementation of the block preconditioned GCR method. in P Sloot, A Hoekstra, B Hertzberger & M Bubak (eds), High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1593, Springer, pp. 1052-1060, 7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999, Amsterdam, Netherlands, 12/04/99. https://doi.org/10.1007/bfb0100666

A parallel implementation of the block preconditioned GCR method. / Vuik, C.; Frank, J.
High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings. ed. / Peter Sloot; Alfons Hoekstra; Bob Hertzberger; Marian Bubak. Springer, 1999. p. 1052-1060 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1593).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - A parallel implementation of the block preconditioned GCR method

AU - Vuik, C.

AU - Frank, J.

N1 - Publisher Copyright: © Springer-Verlag Berlin Heidelberg 1999.

PY - 1999

Y1 - 1999

N2 - The parallel implementation of GCR is addressed, with particular focus on communication costs associated with orthogonalization processes. This consideration brings up questions concerning the use of Householder reflections with GCR. To precondition the GCR method a block Gauss-Jacobi method is used. Approximate solvers are used to obtain a solution of the diagonal blocks. Experiments on a cluster of HP workstations and on a Cray T3E are given.

AB - The parallel implementation of GCR is addressed, with particular focus on communication costs associated with orthogonalization processes. This consideration brings up questions concerning the use of Householder reflections with GCR. To precondition the GCR method a block Gauss-Jacobi method is used. Approximate solvers are used to obtain a solution of the diagonal blocks. Experiments on a cluster of HP workstations and on a Cray T3E are given.

KW - Approximate subdomain solution

KW - Orthogonalization methods

KW - Parallel Krylov subspace methods

UR - http://www.scopus.com/inward/record.url?scp=84956864774&partnerID=8YFLogxK

U2 - 10.1007/bfb0100666

DO - 10.1007/bfb0100666

M3 - Conference contribution

AN - SCOPUS:84956864774

SN - 3540658211

SN - 9783540658214

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 1052

EP - 1060

BT - High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings

A2 - Sloot, Peter

A2 - Hoekstra, Alfons

A2 - Hertzberger, Bob

A2 - Bubak, Marian

PB - Springer

T2 - 7th International Conference on High-Performance Computing and Networking, HPCN Europe 1999

Y2 - 12 April 1999 through 14 April 1999

ER -

Vuik C, Frank J. A parallel implementation of the block preconditioned GCR method. In Sloot P, Hoekstra A, Hertzberger B, Bubak M, editors, High-Performance Computing and Networking - 7th International Conference, HPCN Europe 1999, Proceedings. Springer. 1999. p. 1052-1060. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/bfb0100666

A parallel implementation of the block preconditioned GCR method

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this