Recent advances on sparse direct solver using Block Low-Rank and mixed precision for large scale applications

Description

The solution of sparse systems of linear equations is a key computational kernel in scientific computing. It often represents the most time, memory and energy consuming part of the whole numerical simulation process. In this talk, we focus on sparse direct solvers, and more particularly on MUMPS sparse multifrontal solver (https://mumps-solver.org/). We present recent features that enable to efficiently solve large real and complex systems, and that have led to a strong gain of interest of these methods in many applications. We also illustrate how today challenges related to heterogeneity of computers, mixed-precision arithmetics, new models and applications, motivate activity in the field of sparse direct solvers.

Abstract

International audience

Additional details