A performance comparison of classical volume and new substructured one-and two-level Schwarz methods in PETSc - INRIA - Institut National de Recherche en Informatique et en Automatique
Journal Articles Lecture Notes in Computational Science and Engineering Year : 2024

A performance comparison of classical volume and new substructured one-and two-level Schwarz methods in PETSc

Abstract

Substructured Schwarz methods are interpretations of volume Schwarz methods as algorithms on interface variables. We compare here the Parallel Schwarz Method (PSM, equivalent to RAS) in volume to the new substructured version of PSM in [GH12, p.24] and recently extended to a two-level (i.e. coarse-corrected) framework in [CV22b] and [CV22a], using a geometric and spectral approach for the definition of the coarse space. The expected gain of substructured methods is due to the smaller size of the resulting problems, notably with Krylov-type acceleration techniques when the dimension of the subspace of approximants becomes large [Saa03].
Fichier principal
Vignette du fichier
substructuredDD27.pdf (1005.61 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

cea-04295510 , version 1 (20-11-2023)

Identifiers

  • HAL Id : cea-04295510 , version 1

Cite

Gabriele Ciaramella, Martin J. Gander, Serge Van Criekingen, Tommaso Vanzan. A performance comparison of classical volume and new substructured one-and two-level Schwarz methods in PETSc. Lecture Notes in Computational Science and Engineering, 2024, Domain Decomposition Methods in Science and Engineering XXVII. isbn:978-3-031-50768-7, 149, pp.133-140. ⟨cea-04295510⟩
111 View
55 Download

Share

More