A Selective Nesting Approach for the Sparse Multi-threaded Cholesky Factorization

Sparse linear algebra routines are fundamental building blocks of a large variety of scientific applications. Direct solvers, which are methods for solving linear systems via the factorization of matrices into products of triangular matrices, are commonly used in many contexts. The Cholesky factoriz...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE/ACM International Workshop on Extreme Scale Programming Models and Middleware (Online) S. 1 - 9
Hauptverfasser: Le Fevre, Valentin, Usui, Tetsuzo, Casas, Marc
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.11.2022
Schlagworte:
cholesky
Context
granularity
Heuristic algorithms
Linear algebra
Linear systems
multi-threading
Parallel machines
parallel tasks
Programming
sparse matrix
Symmetric matrices
ISSN:2831-3631
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Sparse linear algebra routines are fundamental building blocks of a large variety of scientific applications. Direct solvers, which are methods for solving linear systems via the factorization of matrices into products of triangular matrices, are commonly used in many contexts. The Cholesky factorization is the fastest direct method for symmetric and positive definite matrices. This paper presents selective nesting, a method to determine the optimal task granularity for the parallel Cholesky factorization based on the structure of sparse matrices. We propose the Opt-D algorithm, which automatically and dynamically applies selective nesting. Opt-D leverages matrix sparsity to drive complex task-based parallel workloads in the context of direct solvers. We run an extensive evaluation campaign considering a heterogeneous set of 35 sparse matrices and a parallel machine featuring the A64FX processor. Opt-D delivers an average performance speedup of 1.75× with respect to the best state-of-the-art parallel methods to run direct solvers.
ISSN:2831-3631
DOI:10.1109/ESPM256814.2022.00006