Share Email Print

Proceedings Paper

Dealing with performance/portability and performance/accuracy trade-offs in heterogeneous computing systems: a case study with matrix multiplication modulo primes
Author(s): Matthew Wezowicz; B. David Saunder; Michela Taufer
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

We present the study of two important trade-offs in heterogeneous systems (i.e., between performance versus portability and between performance and accuracy) for a relevant linear algebra problem, matrix multiplication modulo primes. Integer matrix linear algebra methods rely heavily on matrix multiplication modulo primes. Double precision is necessary for exact representation of sufficiently many primes. We examine the performance losses due to the use of OpenCL versus CUDA and the use of double versus single precision. Our results indicate that performance losses from the former are minimal with the benefit of cross-platform portability and from the latter are acceptable when double precision is required.

Paper Details

Date Published: 4 May 2012
PDF: 10 pages
Proc. SPIE 8403, Modeling and Simulation for Defense Systems and Applications VII, 840308 (4 May 2012); doi: 10.1117/12.919323
Show Author Affiliations
Matthew Wezowicz, Univ. of Delaware (United States)
B. David Saunder, Univ. of Delaware (United States)
Michela Taufer, Univ. of Delaware (United States)

Published in SPIE Proceedings Vol. 8403:
Modeling and Simulation for Defense Systems and Applications VII
Eric J. Kelmelis, Editor(s)

© SPIE. Terms of Use
Back to Top