Accelerating the Conjugate Gradient Method by Solving Multiple GPU-Parallelized Duplicate Systems

DOI10.48550/arXiv.2608.07327MaRDI QIDQ7284728zbMATH OpenarXivFDO

Authors Shaked Regev, Andrew Xu

Publication date 7 August 2026

Published in arXiv

Tldr We accelerate the conjugate gradient (CG) method on parallel computing hardware by solving multiple copies of a linear system simultaneously, even without additional hardware resources, reducing solve time by up to 6 and 5 times for dense and sparse systems respectively.

Full work available at URL https://doi.org/10.48550/arXiv.2608.07327

Mathematics Subject Classification ID

Iterative numerical methods for linear systems (65F10)

arXiv classification

math.NA

Cites work

• A breakdown-free block conjugate gradient method
• An eigen-based high-order expansion basis for structured spectral elements
• An Empirical Study of Conjugate Gradient Preconditioners for Solving Symmetric Positive Definite Systems of Linear Equations
• Block CG algorithms revisited
• Block conjugate gradient algorithms for least squares problems
• CUSPARSE
• Fast randomized matrix and tensor interpolative decomposition using countsketch
• Methods of conjugate gradients for solving linear systems
• Numerical Linear Algebra
• Optimizing Sparse Linear Algebra Through Automatic Format Selection and Machine Learning
• Parallel preconditioned conjugate gradient optimization of the Rayleigh quotient for the solution of sparse eigenproblems
• s-step iterative methods for symmetric linear systems
• The numerical stability analysis of pipelined conjugate gradient methods: historical context and methodology
• The University of Florida sparse matrix collection

This page was built for publication: Accelerating the Conjugate Gradient Method by Solving Multiple GPU-Parallelized Duplicate Systems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284728)