DMCA
Parallel Sparse Matrix-Vector Multiplication (1997)
| Citations: | 2 - 0 self |
BibTeX
@MISC{Tavakoli97parallelsparse,
author = {Faroogh Tavakoli and Faroogh Tavakoli},
title = {Parallel Sparse Matrix-Vector Multiplication},
year = {1997}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
In this paper we describe an algorithm for unstructured sparse matrix-vector multiplication on distributed memory parallel computers. We focus on both local and global computational efficiency, i.e. single processor computational performance and inter-processor communication efficiency. Numerical experiments on the IBM SP "Strindberg" show that our algorithm is very effective for sparse matrices with some data locality, i.e. sparse matrices that can be renumbered such that all non-zero elements lie within a certain bandwidth. Contents 1 Introduction 4 2 Sparse matrix formats 4 2.1 Storage schemes . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Compressed Row Storage . . . . . . . . . . . . . . . . . . . . 5 2.3 Compressed Columns Storage . . . . . . . . . . . . . . . . . . 5 2.4 Diagonal Storage . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.5 Ellpack-Itpack . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.6 Jagged-Diagonal . . . . . . . . . . . . . . . . . ...
Keyphrases
sparse matrix parallel sparse matrix-vector multiplication single processor computational performance numerical experiment certain bandwidth compressed column storage storage scheme compressed row storage distributed memory parallel computer non-zero element unstructured sparse matrix-vector multiplication inter-processor communication efficiency global computational efficiency diagonal storage ibm sp strindberg data locality
