Varying the s in Your s-step GMRES

1 FLUMINANCE - Fluid Flow Analysis, Description and Control from Image Sequences
IRMAR - Institut de Recherche Mathématique de Rennes, IRSTEA - Institut national de recherche en sciences et technologies pour l'environnement et l'agriculture, Inria Rennes – Bretagne Atlantique
Abstract : Krylov subspace methods are commonly used iterative methods for solving large sparse linear systems, however they suffer from communication bottlenecks on parallel computers. Therefore, $s$-step methods have been developed where the Krylov subspace is built block by block, so that $s$ matrix-vector multiplications can be done before orthonormalizing the block. Then Communication-Avoiding algorithms can be used for both kernels. This paper introduces a new variation on $s$-step GMRES in order to reduce the number of iterations necessary to ensure convergence, with a small overhead in the number of communications. Namely, we develop a $s$-step GMRES algorithm, where the block size is variable and increases gradually. Our numerical experiments show a good agreement with our analysis of condition numbers and demonstrate the efficiency of our variable $s$-step approach.
Keywords :
Document type :
Journal articles
Domain :

Cited literature [36 references]

https://hal.inria.fr/hal-01299652
Contributor : Jocelyne Erhel Connect in order to contact the contributor
Submitted on : Thursday, December 7, 2017 - 10:18:31 AM
Last modification on : Friday, May 20, 2022 - 9:04:52 AM

File

FibGMRES.pdf
Files produced by the author(s)

Identifiers

• HAL Id : hal-01299652, version 3

Citation

David Imberti, Jocelyne Erhel. Varying the s in Your s-step GMRES. Electronic Transactions on Numerical Analysis, Kent State University Library, 2017, 47, pp.206-230. ⟨hal-01299652v3⟩

Record views