FullSWOF_Paral: Comparison of two parallelization strategies (MPI and SKELGIS) on a software designed for hydrology applications

Stéphane Cordier (1) , Hélène Coullon (2) , Olivier Delestre (3) , Christian Laguerre (1) , Minh Hoang Le (1) , Daniel Pierre (4) , Georges Sadaka (5)

1 MAPMO - Mathématiques - Analyse, Probabilités, Modélisation - Orléans
2 LIFO - Laboratoire d'Informatique Fondamentale d'Orléans
3 JAD - Laboratoire Jean Alexandre Dieudonné
4 Geo-Hyd
5 LAMFA - Laboratoire Amiénois de Mathématique Fondamentale et Appliquée - UMR CNRS 7352

In this paper, we perform a comparison of two approaches for the parallelization of an existing, free software, FullSWOF 2D (http://www. univ-orleans.fr/mapmo/soft/FullSWOF/ that solves shallow water equations for applications in hydrology) based on a domain decomposition strategy. The first approach is based on the classical MPI library while the second approach uses Parallel Algorithmic Skeletons and more precisely a library named SkelGIS (Skeletons for Geographical Information Systems). The first results presented in this article show that the two approaches are similar in terms of performance and scalability. The two implementation strategies are however very different and we discuss the advantages of each one.

Mots clés: Shallow-Water equation; Saint-Venant; parallelization; MPI; SKELGIS; skeleton; domain decomposition; Malpasset; finite volume; well-balanced; hydrostatic reconstruction; HLL numerical flux; MUSCL reconstruction

Domaines: Analyse numérique [math.NA]; Calcul parallèle, distribué et partagé [cs.DC]; Analyse numérique [cs.NA]

Stéphane Cordier, Hélène Coullon, Olivier Delestre, Christian Laguerre, Minh Hoang Le, et al.. FullSWOF_Paral: Comparison of two parallelization strategies (MPI and SKELGIS) on a software designed for hydrology applications. ESAIM: Proceedings, 2013, 43, pp.59-79. ⟨10.1051/proc/201343004⟩. ⟨hal-00845812⟩