Reproducible Triangular Solvers for High-Performance Computing

Roman Iakymchuk; David Defour; Sylvain Collange; Stef Graillat

doi:10.1109/ITNG.2015.63

Reproducible Triangular Solvers for High-Performance Computing

Roman Iakymchuk ^{1, 2} David Defour ³ Sylvain Collange ⁴ Stef Graillat ¹

1 PEQUAN - Performance et Qualité des Algorithmes Numériques

LIP6 - Laboratoire d'Informatique de Paris 6

2 ICS - Institut du Calcul et de la Simulation

3 DALI - Digits, Architectures et Logiciels Informatiques

LIRMM - Laboratoire d'Informatique de Robotique et de Microélectronique de Montpellier, UPVD - Université de Perpignan Via Domitia

4 ALF - Amdahl's Law is Forever

Inria Rennes – Bretagne Atlantique , IRISA-D3 - ARCHITECTURE

Abstract : On modern parallel architectures, floating-point computations may become non-deterministic and, therefore, non-reproducible mainly due to non-associativity of floating-point operations. We propose an algorithm to solve dense triangular systems by leveraging the standard parallel triangular solver and our, recently introduced, multi-level exact summation approach. Finally, we present implementations of the proposed fast reproducible triangular solver and results on recent NVIDIA GPUs.

Type de document :

Communication dans un congrès

ITNG: Information Technology - New Generations, Apr 2015, Las Vegas, NV, United States. 12th International Conference on Information Technology - New Generations, pp.353-358, 2015, <10.1109/ITNG.2015.63>

Domaine :

Informatique [cs] / Architectures Matérielles [cs.AR]

Liste complète des métadonnées

https://hal-lirmm.ccsd.cnrs.fr/lirmm-01206371
Contributeur : David Defour <>
Soumis le : lundi 28 septembre 2015 - 21:34:54
Dernière modification le : mercredi 2 août 2017 - 10:10:32

Reproducible Triangular Solvers for High-Performance Computing

Identifiants

Collections

Citation

Exporter

Métriques

264

Contact

Reproducible Triangular Solvers for High-Performance Computing

Identifiants

Collections

Citation

Exporter

Partager

Métriques

264

Contact