View in EDS

Reaching the scalability of a distributed AMG solver.

Saved in:
Bibliographic Details
Title: Reaching the scalability of a distributed AMG solver.
Authors: Konshin, Igor N.1,2,3 (AUTHOR), Terekhov, Kirill M.1 (AUTHOR) terekhov@inm.ras.ru
Source: Russian Journal of Numerical Analysis & Mathematical Modelling. Feb2026, Vol. 41 Issue 1, p45-67. 23p.
Subjects: Algebraic multigrid methods, Scalability, Parallel programming, Linear systems, Elliptic equations, High performance computing, Computational mathematics
Abstract: In this paper, a distributed parallel version of algebraic multigrid method is proposed for linear algebraic systems arising from discretization of systems of scalar elliptic equations. The implementation demonstrates good scalability on thousands of cores and surpasses the open source packages HYPRE BoomerAMG and PETSc GAMG in performance on a number of systems. Effective scalable implementations of algorithms in the setup stage of the algebraic multigrid method are considered. The possibility of reducing the number of processors in communicator for coarse systems has been implemented. Variants of cycles of the multigrid method are considered, which make it possible to obtain independence of the number of iterations as the system size increases without losing the scalability of the method. These developments form the basis for further extension of the method for multiphysical problems. [ABSTRACT FROM AUTHOR]
Copyright of Russian Journal of Numerical Analysis & Mathematical Modelling is the property of De Gruyter and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:In this paper, a distributed parallel version of algebraic multigrid method is proposed for linear algebraic systems arising from discretization of systems of scalar elliptic equations. The implementation demonstrates good scalability on thousands of cores and surpasses the open source packages HYPRE BoomerAMG and PETSc GAMG in performance on a number of systems. Effective scalable implementations of algorithms in the setup stage of the algebraic multigrid method are considered. The possibility of reducing the number of processors in communicator for coarse systems has been implemented. Variants of cycles of the multigrid method are considered, which make it possible to obtain independence of the number of iterations as the system size increases without losing the scalability of the method. These developments form the basis for further extension of the method for multiphysical problems. [ABSTRACT FROM AUTHOR]
ISSN:09276467
DOI:10.1515/rnam-2026-0004