Abstract
The effect of two soft fault error models on the convergence of the parallel flexible GMRES (FGMRES) iterative method solving an elliptical PDE problem on a regular grid is evaluated. We consider two types of preconditioners: an incomplete LU factorization with dual threshold (ILUT), and an algebraic recursive multilevel solver (ARMS) combined with random butterfly transformation (RBT). The experiments quantify the difference between two soft fault error models considered in this study and compare their potential impact on the convergence.
Author supplied keywords
Cite
CITATION STYLE
Coleman, E., Jamal, A., Baboulin, M., Khabou, A., & Sosonkina, M. (2018). A comparison of soft-fault error models in the parallel preconditioned flexible GMRES. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10777 LNCS, pp. 36–46). Springer Verlag. https://doi.org/10.1007/978-3-319-78024-5_4
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
