Publication Details

Fast Sparse Matrix Multiplication on GPU

POLOK, L.; ILA, V.; SMRŽ, P. Fast Sparse Matrix Multiplication on GPU. In Proceedings of the 23rd High Performance Computing Symposium (HPC'15). Alexandria, Virginia: Association for Computing Machinery, 2015. p. 1-8. ISBN: 978-1-5108-0101-1.
Czech title
Rychlé násobení řídkých matic na GPU
Type
conference paper
Language
English
Authors
Polok Lukáš, Ing., Ph.D.
Ila Viorela Simona, Ph.D.
Smrž Pavel, doc. RNDr., Ph.D. (DCGM)
URL
Keywords

parallel sparse matrix multiplication, parallel linear algebra, matrix-matrix multiplication, GPGPU

Abstract

Sparse matrix multiplication is an important algorithm in a wide variety of problems, including graph algorithms, simulations and linear solving to name a few. Yet, there are but a few works related to acceleration of sparse matrix multiplication on a GPU. We present a fast, novel algorithm for sparse matrix multiplication, outperforming the previous algorithm on GPU up to 3x and CPU up to 30x. The principal improvements include more efficient load balancing strategy, and a faster sorting algorithm. The main contribution is design and implementation of efficient sparse matrix multiplication algorithm and extending it to sparse block matrices, which is to our best knowledge the first implementation of this kind.

Published
2015
Pages
1–8
Proceedings
Proceedings of the 23rd High Performance Computing Symposium (HPC'15)
Conference
23rd High Performance Computing Symposia, Alexandria, Virginia, US
ISBN
978-1-5108-0101-1
Publisher
Association for Computing Machinery
Place
Alexandria, Virginia
EID Scopus
2-s2.0-84937400532
BibTeX 
@inproceedings{BUT119833,
  author="Lukáš {Polok} and Viorela Simona {Ila} and Pavel {Smrž}",
  title="Fast Sparse Matrix Multiplication on GPU",
  booktitle="Proceedings of the 23rd High Performance Computing Symposium (HPC'15)",
  year="2015",
  pages="1--8",
  publisher="Association for Computing Machinery",
  address="Alexandria, Virginia",
  isbn="978-1-5108-0101-1",
  url="http://dl.acm.org/citation.cfm?id=2872604"
}
Back to top