Conference contribution
(Conference Contribution)


Efficiency of general Krylov methods on GPUs - An experimental study


Publication Details
Author(s): Wellein G, Anzt H, Dongarra J, Kreutzer M, Köhler M
Publisher: IEEE Computer Society
Publication year: 2016
Pages range: 683-691
ISBN: 9781509021406
Event: 30th IEEE International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2016

Abstract

This paper compares different Krylov methods based on short recurrences with respect to their efficiency whenimplemented on GPUs. The comparison includes BiCGSTAB, CGS, QMR, and IDR using different shadow space dimensions. These methods are known for their good convergencecharacteristics. For a large set of test matrices taken from theUniversity of Florida Matrix Collection, we evaluate the methods'performance against different target metrics: convergence, number of sparse matrix-vector multiplications, and executiontime. We also analyze whether the methods are «orthogonal»in terms of problem suitability. We propose best practicesfor choosing methods in a «black box» scenario, where noinformation about the optimal solver is available.



Focus Area of Individual Faculties


How to cite
APA: Wellein, G., Anzt, H., Dongarra, J., Kreutzer, M., & Köhler, M. (2016). Efficiency of general Krylov methods on GPUs - An experimental study. (pp. 683-691). IEEE Computer Society.

MLA: Wellein, Gerhard, et al. "Efficiency of general Krylov methods on GPUs - An experimental study." Proceedings of the 30th IEEE International Parallel and Distributed Processing Symposium Workshops, IPDPSW 2016 IEEE Computer Society, 2016. 683-691.

BibTeX: Download
Share link
Last updated on 2017-07-26 at 02:38
PDF downloaded successfully