8483 articles  [version française]

inria-00593306, version 1

Accelerating linear system solutions using randomization techniques

Marc Baboulin (Author to contact preferably) a1, Jack Dongarra () b2, Julien Herrmann () c1, Stanimire Tomov () b2

N° RR-7616 (2011)

Abstract: We illustrate how linear algebra calculations can be enhanced by statistical techniques in the case of a square linear system Ax = b. We study a random transformation of A that enables us to avoid pivoting and then to reduce the amount of communication. Numerical experiments show that this randomization can be performed at a very affordable computational price while providing us with a satisfying accuracy when compared to partial pivoting. This random transformation called Partial Random Butter y Transformation (PRBT) is optimized in terms of data storage and flops count. We propose a solver where PRBT and the LU factorization with no pivoting take advantage of the latest generation of hybrid multicore/GPU machines and we compare its Gfl op/s performance with a solver implemented in a current parallel library.

  • a –  INRIA/University Paris-Sud
  • b –  University of Tennessee
  • c –  École Normale Supérieure de Lyon
  • 1:  Laboratoire de Recherche en Informatique (LRI)
  • CNRS : UMR8623 – Université Paris XI - Paris Sud
  • 2:  Innovative Computing Laboratory (ICL)
  • University of Tennessee
  • Domain : Computer Science/Numerical Analysis
  • Keywords : dense linear algebra – linear systems – LU factorization – randomization – multiplicative preconditioning – Graphics Processing Units
  • Internal note : RR-7616
 
  • inria-00593306, version 1
  • oai:hal.inria.fr:inria-00593306
  • From: 
  • Submitted on: Friday, 13 May 2011 20:16:25
  • Updated on: Tuesday, 17 May 2011 09:27:27