On the performance of an algebraic multigrid solver on multicore clusters

Allison H. Baker; Martin Schulz; Ulrike M. Yang

doi:10.1007/978-3-642-19328-6_12

On the performance of an algebraic multigrid solver on multicore clusters

Allison H. Baker, Martin Schulz, Ulrike M. Yang

Application Scalability and Performance Group

Lawrence Livermore Natl. Laboratory

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

20 Scopus citations

Abstract

Algebraic multigrid (AMG) solvers have proven to be extremely efficient on distributed-memory architectures. However, when executed on modern multicore cluster architectures, we face new challenges that can significantly harm AMG's performance. We discuss our experiences on such an architecture and present a set of techniques that help users to overcome the associated problems, including thread and process pinning and correct memory associations. We have implemented most of the techniques in a MultiCore SUPport library (MCSup), which helps to map OpenMP applications to multicore machines. We present results using both an MPI-only and a hybrid MPI/OpenMP model.

Original language	English
Title of host publication	High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers
Pages	102-115
Number of pages	14
DOIs	https://doi.org/10.1007/978-3-642-19328-6_12
State	Published - 2011
Event	9th International Conference on High Performance Computing for Computational Science, VECPAR 2010 - Berkeley, CA, United States Duration: Jun 22 2010 → Jun 25 2010

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	6449 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	9th International Conference on High Performance Computing for Computational Science, VECPAR 2010
Country/Territory	United States
City	Berkeley, CA
Period	06/22/10 → 06/25/10

Access to Document

10.1007/978-3-642-19328-6_12

Cite this

Baker, A. H., Schulz, M., & Yang, U. M. (2011). On the performance of an algebraic multigrid solver on multicore clusters. In High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers (pp. 102-115). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6449 LNCS). https://doi.org/10.1007/978-3-642-19328-6_12

Baker, Allison H. ; Schulz, Martin ; Yang, Ulrike M. / On the performance of an algebraic multigrid solver on multicore clusters. High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers. 2011. pp. 102-115 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{c9a3f01736544fa1aecc17c9019c1807,

title = "On the performance of an algebraic multigrid solver on multicore clusters",

abstract = "Algebraic multigrid (AMG) solvers have proven to be extremely efficient on distributed-memory architectures. However, when executed on modern multicore cluster architectures, we face new challenges that can significantly harm AMG's performance. We discuss our experiences on such an architecture and present a set of techniques that help users to overcome the associated problems, including thread and process pinning and correct memory associations. We have implemented most of the techniques in a MultiCore SUPport library (MCSup), which helps to map OpenMP applications to multicore machines. We present results using both an MPI-only and a hybrid MPI/OpenMP model.",

author = "Baker, \{Allison H.\} and Martin Schulz and Yang, \{Ulrike M.\}",

year = "2011",

doi = "10.1007/978-3-642-19328-6\_12",

language = "English",

isbn = "9783642193279",

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

pages = "102--115",

booktitle = "High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers",

note = "9th International Conference on High Performance Computing for Computational Science, VECPAR 2010 ; Conference date: 22-06-2010 Through 25-06-2010",

}

Baker, AH, Schulz, M & Yang, UM 2011, On the performance of an algebraic multigrid solver on multicore clusters. in High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 6449 LNCS, pp. 102-115, 9th International Conference on High Performance Computing for Computational Science, VECPAR 2010, Berkeley, CA, United States, 06/22/10. https://doi.org/10.1007/978-3-642-19328-6_12

On the performance of an algebraic multigrid solver on multicore clusters. / Baker, Allison H.; Schulz, Martin; Yang, Ulrike M.
High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers. 2011. p. 102-115 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6449 LNCS).

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

TY - GEN

T1 - On the performance of an algebraic multigrid solver on multicore clusters

AU - Baker, Allison H.

AU - Schulz, Martin

AU - Yang, Ulrike M.

PY - 2011

Y1 - 2011

N2 - Algebraic multigrid (AMG) solvers have proven to be extremely efficient on distributed-memory architectures. However, when executed on modern multicore cluster architectures, we face new challenges that can significantly harm AMG's performance. We discuss our experiences on such an architecture and present a set of techniques that help users to overcome the associated problems, including thread and process pinning and correct memory associations. We have implemented most of the techniques in a MultiCore SUPport library (MCSup), which helps to map OpenMP applications to multicore machines. We present results using both an MPI-only and a hybrid MPI/OpenMP model.

AB - Algebraic multigrid (AMG) solvers have proven to be extremely efficient on distributed-memory architectures. However, when executed on modern multicore cluster architectures, we face new challenges that can significantly harm AMG's performance. We discuss our experiences on such an architecture and present a set of techniques that help users to overcome the associated problems, including thread and process pinning and correct memory associations. We have implemented most of the techniques in a MultiCore SUPport library (MCSup), which helps to map OpenMP applications to multicore machines. We present results using both an MPI-only and a hybrid MPI/OpenMP model.

UR - https://www.scopus.com/pages/publications/79952587273

U2 - 10.1007/978-3-642-19328-6_12

DO - 10.1007/978-3-642-19328-6_12

M3 - Conference contribution

AN - SCOPUS:79952587273

SN - 9783642193279

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

SP - 102

EP - 115

BT - High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers

T2 - 9th International Conference on High Performance Computing for Computational Science, VECPAR 2010

Y2 - 22 June 2010 through 25 June 2010

ER -

Baker AH, Schulz M, Yang UM. On the performance of an algebraic multigrid solver on multicore clusters. In High Performance Computing for Computational Science, VECPAR 2010 - 9th International Conference, Revised Selected Papers. 2011. p. 102-115. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-19328-6_12

On the performance of an algebraic multigrid solver on multicore clusters

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this