Parallel performance of algebraic multigrid domain decomposition
Abstract
Algebraic multigrid (AMG) is a widely used scalable solver and preconditioner for large-scale linear systems resulting from the discretization of a wide class of elliptic PDEs. While AMG has optimal computational complexity, the cost of communication has become a significant bottleneck that limits its scalability as processor counts continue to grow on modern machines. This article examines the design, implementation, and parallel performance of a novel algorithm, algebraic multigrid domain decomposition (AMG-DD), designed specifically to limit communication. The goal of AMG-DD is to provide a low-communication alternative to standard AMG V-cycles by trading some additional computational overhead for a significant reduction in communication cost. Numerical results show that AMG-DD achieves superior accuracy per communication cost compared with AMG, and speedup over AMG is demonstrated on a large GPU cluster.
- Authors:
-
- Universität Heidelberg (Germany)
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Publication Date:
- Research Org.:
- Lawrence Livermore National Laboratory (LLNL), Livermore, CA (United States)
- Sponsoring Org.:
- USDOE National Nuclear Security Administration (NNSA); German Research Foundation (DFG)
- OSTI Identifier:
- 1776659
- Report Number(s):
- LLNL-JRNL-801850
Journal ID: ISSN 1070-5325; 1005848
- Grant/Contract Number:
- AC52-07NA27344; INST 35/1134/1
- Resource Type:
- Accepted Manuscript
- Journal Name:
- Numerical Linear Algebra with Applications
- Additional Journal Information:
- Journal Volume: 28; Journal Issue: 3; Journal ID: ISSN 1070-5325
- Publisher:
- Wiley
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 97 MATHEMATICS AND COMPUTING; algebraic multigrid; low-communication algorithms; parallel performance
Citation Formats
Mitchell, Wayne B., Strzodka, Robert, and Falgout, Robert D. Parallel performance of algebraic multigrid domain decomposition. United States: N. p., 2020.
Web. doi:10.1002/nla.2342.
Mitchell, Wayne B., Strzodka, Robert, & Falgout, Robert D. Parallel performance of algebraic multigrid domain decomposition. United States. https://doi.org/10.1002/nla.2342
Mitchell, Wayne B., Strzodka, Robert, and Falgout, Robert D. Mon .
"Parallel performance of algebraic multigrid domain decomposition". United States. https://doi.org/10.1002/nla.2342. https://www.osti.gov/servlets/purl/1776659.
@article{osti_1776659,
title = {Parallel performance of algebraic multigrid domain decomposition},
author = {Mitchell, Wayne B. and Strzodka, Robert and Falgout, Robert D.},
abstractNote = {Algebraic multigrid (AMG) is a widely used scalable solver and preconditioner for large-scale linear systems resulting from the discretization of a wide class of elliptic PDEs. While AMG has optimal computational complexity, the cost of communication has become a significant bottleneck that limits its scalability as processor counts continue to grow on modern machines. This article examines the design, implementation, and parallel performance of a novel algorithm, algebraic multigrid domain decomposition (AMG-DD), designed specifically to limit communication. The goal of AMG-DD is to provide a low-communication alternative to standard AMG V-cycles by trading some additional computational overhead for a significant reduction in communication cost. Numerical results show that AMG-DD achieves superior accuracy per communication cost compared with AMG, and speedup over AMG is demonstrated on a large GPU cluster.},
doi = {10.1002/nla.2342},
journal = {Numerical Linear Algebra with Applications},
number = 3,
volume = 28,
place = {United States},
year = {Mon Oct 12 00:00:00 EDT 2020},
month = {Mon Oct 12 00:00:00 EDT 2020}
}
Works referenced in this record:
A Multigrid Tutorial, Second Edition
book, January 2000
- Briggs, William L.; Henson, Van Emden; McCormick, Steve F.
- Other Titles in Applied Mathematics
A low-communication, parallel algorithm for solving PDEs based on range decomposition: RANGE DECOMPOSITION: A LOW COMMUNICATION ALGORITHM FOR SOLVING PDES
journal, March 2016
- Appelhans, David J.; Manteuffel, Tom; McCormick, Steve
- Numerical Linear Algebra with Applications, Vol. 24, Issue 3
Reducing Parallel Communication in Algebraic Multigrid through Sparsification
journal, January 2016
- Bienz, Amanda; Falgout, Robert D.; Gropp, William
- SIAM Journal on Scientific Computing, Vol. 38, Issue 5
Algebraic multigrid by smoothed aggregation for second and fourth order elliptic problems
journal, September 1996
- Vaněk, P.; Mandel, J.; Brezina, M.
- Computing, Vol. 56, Issue 3
Node aware sparse matrix–vector multiplication
journal, August 2019
- Bienz, Amanda; Gropp, William D.; Olson, Luke N.
- Journal of Parallel and Distributed Computing, Vol. 130
An assumed partition algorithm for determining processor inter-communication
journal, June 2006
- Baker, A. H.; Falgout, R. D.; Yang, U. M.
- Parallel Computing, Vol. 32, Issue 5-6
Modeling the performance of an algebraic multigrid cycle on HPC platforms
conference, January 2011
- Gahvari, Hormozd; Baker, Allison H.; Schulz, Martin
- Proceedings of the international conference on Supercomputing - ICS '11
Parallel Adaptive Multilevel Methods with Full Domain Partitions
journal, March 2004
- Mitchell, William F.
- Applied Numerical Analysis & Computational Mathematics, Vol. 1, Issue 1
Non-Galerkin Coarse Grids for Algebraic Multigrid
journal, January 2014
- Falgout, Robert D.; Schroder, Jacob B.
- SIAM Journal on Scientific Computing, Vol. 36, Issue 3
Multigrid Smoothers for Ultraparallel Computing
journal, January 2011
- Baker, Allison H.; Falgout, Robert D.; Kolev, Tzanio V.
- SIAM Journal on Scientific Computing, Vol. 33, Issue 5
Reducing communication in algebraic multigrid using additive variants: REDUCING COMMUNICATION IN AMG WITH ADDITIVE VARIANTS
journal, February 2014
- Vassilevski, Panayot S.; Yang, Ulrike Meier
- Numerical Linear Algebra with Applications, Vol. 21, Issue 2
Reducing Complexity in Parallel Algebraic Multigrid Preconditioners
journal, January 2006
- De Sterck, Hans; Yang, Ulrike Meier; Heys, Jeffrey J.
- SIAM Journal on Matrix Analysis and Applications, Vol. 27, Issue 4
Asynchronous Fast Adaptive Composite-Grid Methods for Elliptic Problems: Theoretical Foundations
journal, January 2004
- Lee, Barry; McCormick, Stephen F.; Philip, Bobby
- SIAM Journal on Numerical Analysis, Vol. 42, Issue 1
Algebraic Multigrid Domain and Range Decomposition (AMG-DD/AMG-RD)
journal, January 2015
- Bank, R.; Falgout, R.; Jones, T.
- SIAM Journal on Scientific Computing, Vol. 37, Issue 5
Distance-two interpolation for parallel algebraic multigrid
journal, January 2008
- De Sterck, Hans; Falgout, Robert D.; Nolting, Joshua W.
- Numerical Linear Algebra with Applications, Vol. 15, Issue 2-3
Advances in implementation, theoretical motivation, and numerical results for the nested iteration with range decomposition algorithm: Advances in implementation, theoretical motivation, and numerical results for the nested iteration with range decomposition algorithm
journal, February 2018
- Mitchell, Wayne; Manteuffel, Tom
- Numerical Linear Algebra with Applications, Vol. 25, Issue 3
On long-range interpolation operators for aggressive coarsening: INTERPOLATION FOR AGGRESSIVE COARSENING
journal, December 2009
- Yang, Ulrike Meier
- Numerical Linear Algebra with Applications, Vol. 17, Issue 2-3