On Algorithmic Variants of Parallel Gaussian Elimination: Comparison of Implementations in Terms of Performance and Numerical Properties

Abstract : Gaussian elimination is a canonical linear algebra procedure for solving linear systems of equations. In the last few years, the algorithm received a lot of attention in an attempt to improve its parallel performance. This article surveys recent developments in parallel implementations of the Gaussian elimination. Five different flavors are investigated. Three of them are based on different strategies for pivoting: partial pivoting, incremental pivoting, and tournament pivoting. The fourth one replaces pivoting with the Random Butterfly Transformation, and finally, an implementation without pivoting is used as a performance baseline. The technique of iterative refinement is applied to recover numerical accuracy when necessary. All parallel implementations are produced using dynamic, superscalar, runtime scheduling and tile matrix layout. Results on two multi-socket multicore systems are presented. Performance and numerical accuracy is analyzed.
Type de document :
Rapport
[Research Report] 2013
Liste complète des métadonnées

Littérature citée [47 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-00867837
Contributeur : Mathieu Faverge <>
Soumis le : vendredi 15 novembre 2013 - 09:35:22
Dernière modification le : jeudi 11 janvier 2018 - 06:22:35
Document(s) archivé(s) le : dimanche 16 février 2014 - 04:28:26

Fichier

lawn280.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-00867837, version 1

Collections

Citation

Simplice Donfack, Jack Dongarra, Mathieu Faverge, Mark Gates, Jakub Kurzak, et al.. On Algorithmic Variants of Parallel Gaussian Elimination: Comparison of Implementations in Terms of Performance and Numerical Properties. [Research Report] 2013. 〈hal-00867837〉

Partager

Métriques

Consultations de la notice

407

Téléchargements de fichiers

306