Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication

Grey Ballard; Nicholas Knight; Alex Druinsky; Oded Schwartz

doi:10.1145/2755573.2755613

Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication

Grey Ballard, Nicholas Knight, Alex Druinsky, Oded Schwartz

The Rachel and Selim Benin School of Engineering and Computer Science

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

20 Scopus citations

Abstract

The performance of parallel algorithms for sparse matrixmatrix multiplication is typically determined by the amount of interprocessor communication performed, which in turn depends on the nonzero structure of the input matrices. In this paper, we characterize the communication cost of a sparse matrix-matrix multiplication algorithm in terms of the size of a cut of an associated hypergraph that encodes the computation for a given input nonzero structure. Obtaining an optimal algorithm corresponds to solving a hypergraph partitioning problem. Our hypergraph model generalizes several existing models for sparse matrix-vector multiplication, and we can leverage hypergraph partitioners developed for that computation to improve applicationspecific algorithms for multiplying sparse matrices.

Original language	English
Title of host publication	SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures
Publisher	Association for Computing Machinery
Pages	86-88
Number of pages	3
ISBN (Electronic)	9781450335881
DOIs	https://doi.org/10.1145/2755573.2755613
State	Published - 13 Jun 2015
Event	27th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 2015 - Portland, United States Duration: 13 Jun 2015 → 15 Jun 2015

Publication series

Name	Annual ACM Symposium on Parallelism in Algorithms and Architectures
Volume	2015-June

Conference

Conference	27th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 2015
Country/Territory	United States
City	Portland
Period	13/06/15 → 15/06/15

Bibliographical note

Publisher Copyright:
Copyright © 2015 ACM.

Access to Document

10.1145/2755573.2755613

Cite this

Ballard, G., Knight, N., Druinsky, A., & Schwartz, O. (2015). Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication. In SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures (pp. 86-88). (Annual ACM Symposium on Parallelism in Algorithms and Architectures; Vol. 2015-June). Association for Computing Machinery. https://doi.org/10.1145/2755573.2755613

Ballard, Grey ; Knight, Nicholas ; Druinsky, Alex et al. / Brief announcement : Hypergraph partitioning for parallel sparse matrix-matrix multiplication. SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures. Association for Computing Machinery, 2015. pp. 86-88 (Annual ACM Symposium on Parallelism in Algorithms and Architectures).

@inproceedings{40453b5bd6ba4e47a34811687512dd12,

title = "Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication",

abstract = "The performance of parallel algorithms for sparse matrixmatrix multiplication is typically determined by the amount of interprocessor communication performed, which in turn depends on the nonzero structure of the input matrices. In this paper, we characterize the communication cost of a sparse matrix-matrix multiplication algorithm in terms of the size of a cut of an associated hypergraph that encodes the computation for a given input nonzero structure. Obtaining an optimal algorithm corresponds to solving a hypergraph partitioning problem. Our hypergraph model generalizes several existing models for sparse matrix-vector multiplication, and we can leverage hypergraph partitioners developed for that computation to improve applicationspecific algorithms for multiplying sparse matrices.",

author = "Grey Ballard and Nicholas Knight and Alex Druinsky and Oded Schwartz",

note = "Publisher Copyright: Copyright {\textcopyright} 2015 ACM.; 27th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 2015 ; Conference date: 13-06-2015 Through 15-06-2015",

year = "2015",

month = jun,

day = "13",

doi = "10.1145/2755573.2755613",

language = "אנגלית",

series = "Annual ACM Symposium on Parallelism in Algorithms and Architectures",

publisher = "Association for Computing Machinery",

pages = "86--88",

booktitle = "SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures",

}

Ballard, G, Knight, N, Druinsky, A & Schwartz, O 2015, Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication. in SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures. Annual ACM Symposium on Parallelism in Algorithms and Architectures, vol. 2015-June, Association for Computing Machinery, pp. 86-88, 27th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 2015, Portland, United States, 13/06/15. https://doi.org/10.1145/2755573.2755613

Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication. / Ballard, Grey; Knight, Nicholas; Druinsky, Alex et al.
SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures. Association for Computing Machinery, 2015. p. 86-88 (Annual ACM Symposium on Parallelism in Algorithms and Architectures; Vol. 2015-June).

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

TY - GEN

T1 - Brief announcement

T2 - 27th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 2015

AU - Ballard, Grey

AU - Knight, Nicholas

AU - Druinsky, Alex

AU - Schwartz, Oded

PY - 2015/6/13

Y1 - 2015/6/13

N2 - The performance of parallel algorithms for sparse matrixmatrix multiplication is typically determined by the amount of interprocessor communication performed, which in turn depends on the nonzero structure of the input matrices. In this paper, we characterize the communication cost of a sparse matrix-matrix multiplication algorithm in terms of the size of a cut of an associated hypergraph that encodes the computation for a given input nonzero structure. Obtaining an optimal algorithm corresponds to solving a hypergraph partitioning problem. Our hypergraph model generalizes several existing models for sparse matrix-vector multiplication, and we can leverage hypergraph partitioners developed for that computation to improve applicationspecific algorithms for multiplying sparse matrices.

AB - The performance of parallel algorithms for sparse matrixmatrix multiplication is typically determined by the amount of interprocessor communication performed, which in turn depends on the nonzero structure of the input matrices. In this paper, we characterize the communication cost of a sparse matrix-matrix multiplication algorithm in terms of the size of a cut of an associated hypergraph that encodes the computation for a given input nonzero structure. Obtaining an optimal algorithm corresponds to solving a hypergraph partitioning problem. Our hypergraph model generalizes several existing models for sparse matrix-vector multiplication, and we can leverage hypergraph partitioners developed for that computation to improve applicationspecific algorithms for multiplying sparse matrices.

UR - http://www.scopus.com/inward/record.url?scp=84950239912&partnerID=8YFLogxK

U2 - 10.1145/2755573.2755613

DO - 10.1145/2755573.2755613

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:84950239912

T3 - Annual ACM Symposium on Parallelism in Algorithms and Architectures

SP - 86

EP - 88

BT - SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures

PB - Association for Computing Machinery

Y2 - 13 June 2015 through 15 June 2015

ER -

Ballard G, Knight N, Druinsky A, Schwartz O. Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication. In SPAA 2015 - Proceedings of the 27th ACM Symposium on Parallelism in Algorithms and Architectures. Association for Computing Machinery. 2015. p. 86-88. (Annual ACM Symposium on Parallelism in Algorithms and Architectures). doi: 10.1145/2755573.2755613

Brief announcement: Hypergraph partitioning for parallel sparse matrix-matrix multiplication

Abstract

Publication series

Conference

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this