Abstract : The linear system solver plays a key role in scientific computing. This paper evaluates the performances of PETSc (Portable, Extensible Toolkit for Scientific Computation) and HIPS (Portable, Extensible Toolkit for Scientific Computation) solving a singular matrix arising from a bucking problem from UF matrix collection. We employ preconditioned (Level-based ILU) iterative methods (GMRES) for PETSc while HIPS adopts the Schur complement method, named iterative or hybrid mode. Moreover, HIPS proposed HID (Hierarchical interface decomposition) to improve the parallel efficiency. We hope to transfer these results into industrial applications.
Type de document :
Communication dans un congrès
Ching-Hsien Hsu; Xuanhua Shi; Valentina Salapura. 11th IFIP International Conference on Network and Parallel Computing (NPC), Sep 2014, Ilan, Taiwan. Springer, Lecture Notes in Computer Science, LNCS-8707, pp.583-586, 2014, Network and Parallel Computing. 〈10.1007/978-3-662-44917-2_58〉
https://hal.inria.fr/hal-01403151
Contributeur : Hal Ifip
<>
Soumis le : vendredi 25 novembre 2016 - 14:48:05
Dernière modification le : vendredi 1 décembre 2017 - 01:10:10
Document(s) archivé(s) le : lundi 20 mars 2017 - 19:11:22
