AuthorsX. Li, A. Azad, A. Buluc, P. Ghysels, J. Langguth and X. Wang
TitleExaGraph Collaboration with STRUMPACK/SuperLU: Factorization-Based Sparse Solvers and Preconditioners for Exascale
AfilliationScientific Computing
Project(s)Meeting Exascale Computing with Source-to-Source Compilers
Publication TypePublic outreach
Year of Publication2018
Date Published08/2018
PublisherExascale Computing Project (ECP)
Place PublishedECP Website

When researchers try to solve science and engineering problems, they often create systems of linear equations that need to be solved. Software libraries known as solvers provide mathematical tools that can be applied to similar problems. Direct solvers that use variants of Gaussian elimination are one of the most popular methods for solving such systems due to their robustness, especially for algebraic systems arising from multiphysics simulations.

In many situations, the system is sparse, meaning that the majority of the matrix entries are zero and ideally need neither to be stored nor operated on. The strength of SuperLU and STRUMPACK is that they can automatically determine which matrix entries are zeros and can be ignored, allowing the computer to focus its calculations on the other entries and finish the problem much faster.

Results of the effort showed that the parallel AWPM (approximate-weight perfect matching) code can run up to 2,500x faster than the sequential algorithm on 256 nodes of the Cori/KNL supercomputer.

Citation Key26375