A class of communication-avoiding algorithms for solving general dense linear systems on CPU/GPU parallel machines

Abstract : We study several solvers for the solution of general linear systems where the main objective is to reduce the communication overhead due to pivoting. We first describe two existing algorithms for the LU factorization on hybrid CPU/GPU architectures. The first one is based on partial pivoting and the second uses a random preconditioning of the original matrix to avoid pivoting. Then we introduce a solver where the panel factorization is performed using a communication-avoiding pivoting heuristic while the update of the trailing submatrix is performed by the GPU. We provide performance comparisons for these solvers on current hybrid multicore-GPU parallel machines.
Type de document :
Rapport
[Research Report] RR-7854, INRIA. 2012
Liste complète des métadonnées

Littérature citée [20 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-00656457
Contributeur : Marc Baboulin <>
Soumis le : mercredi 29 février 2012 - 11:27:57
Dernière modification le : mardi 24 avril 2018 - 13:37:48
Document(s) archivé(s) le : mercredi 14 décembre 2016 - 09:17:28

Fichier

RR-7854.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-00656457, version 3

Collections

Citation

Marc Baboulin, Simplice Donfack, Jack Dongarra, Laura Grigori, Adrien Rémy, et al.. A class of communication-avoiding algorithms for solving general dense linear systems on CPU/GPU parallel machines. [Research Report] RR-7854, INRIA. 2012. 〈hal-00656457v3〉

Partager

Métriques

Consultations de la notice

562

Téléchargements de fichiers

295