Exploiting Task-Based Parallelism for the Red-Black Gauss-Seidel Method on 2D Grids

DOI10.48550/arXiv.2607.01735MaRDI QIDQ7270489arXivFDO

Authors Andreas Frommer, Dirk Pleiter, Shiting Long, Gustavo Ramirez-Hidalgo

Publication date 2 July 2026

Published in arXiv

Tldr We explore fine-grained parallelism in Gauss-Seidel methods through two task-based programming models and demonstrate their potential for increased efficiency on 2D grid problems, achieving comparable performance to conventional divide-and-conquer parallelization while mitigating global synchronization requirements and load imbalances.

Full work available at URL https://doi.org/10.48550/arXiv.2607.01735

arXiv classification

cs.DCmath.NA

Cites work

• A class of parallel tiled linear algebra algorithms for multicore architectures
• Task-Based FMM for Multicore Architectures

This page was built for publication: Exploiting Task-Based Parallelism for the Red-Black Gauss-Seidel Method on 2D Grids

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7270489)