Experimental optimization of parallel 3D overlapping domain decomposition schemes

0Citations
Citations of this article
2Readers
Mendeley users who have this article in their library.
Get full text

Abstract

Overlapping domain decomposition is a possible convenient technique for solving complex problems described by Partial Differential Equations in a parallel framework. The performance of this approach strongly depends on the size and the position of the overlap since the overlapping has a positive impact on the number of iterations required by the numerical scheme and the relatively flexible and judicious choice of the interface may lead to a reduction of the communication time. In this paper we test the overlapping domain decomposition method on the finite element discretization of a diffusion reaction problem in both idealized and real 3D geometries. Results confirm that the detection of the optimal overlapping in real cases is not trivial but has the potential to significantly reduce the computational costs of the entire solution process.

Cite

CITATION STYLE

APA

Guzzetti, S., Veneziani, A., & Sunderam, V. (2016). Experimental optimization of parallel 3D overlapping domain decomposition schemes. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9573, pp. 138–149). Springer Verlag. https://doi.org/10.1007/978-3-319-32149-3_14

Register to see more suggestions

Mendeley helps you to discover research relevant for your work.

Already have an account?

Save time finding and organizing research with Mendeley

Sign up for free