ShyLU node: On-node Scalable Solvers and Preconditioners Recent Progresses and Current Performance

TL;DR

This paper discusses recent advances in ShyLU-node, an open-source package for scalable linear solvers and preconditioners on multicore CPUs and GPUs, highlighting its performance on real-world applications.

Contribution

It introduces and evaluates new sparse direct solvers Basker and Tacho, and the FastILU preconditioner within the ShyLU-node framework for large-scale linear systems.

Findings

Basker effectively solves circuit simulation problems.

Tacho performs well on land-ice simulation models.

FastILU shows promising results on 3D model problems.

Abstract

ShyLU-node is an open-source software package that implements linear solvers and preconditioners on shared-memory multicore CPUs or on a GPU. It is part of the Trilinos software framework and designed to provide a robust and efficient solution of large-scale linear systems from real-world applications on the current and emerging computers. In this paper, we discuss two sparse direct solvers, Basker and Tacho, and an algebraic preconditioner, FastILU, in ShyLU-node package. These ShyLU solvers and preconditioner can be used as a stand-alone global problem solver, as a local subdomain solver for domain decomposition (DD) preconditioner, or as the coarse-problem solver in algebraic multi-grid preconditioner. We present performance results with the sparse direct solvers for real application problems, namely, Basker for Xyce Circuit Simulations and Tacho for Albany Land-Ice Simulation of Antarctica. FastILU has been also used in real-world applications, but in this paper, we illustrate its performance using 3D model problems.