Abstract
Multigrid methods are powerful techniques to accelerate the solution of computationally-intensive problems arising in a broad range of applications. Used in conjunction with iterative processes for solving partial differential equations, multigrid methods speed up iterative methods by moving the computation from the original mesh covering the problem domain through a series of coarser meshes. But this hierarchical structure leaves domain-parallel versions of the standard multigrid algorithms with a deficiency of parallelism on coarser grids. To compensate, several parallel multigrid strategies with more parallelism, but also more work, have been designed. We examine these parallel strategies and compare them to simpler standard algorithms to try to determine which techniques are more efficient and practical. We consider three parallel multigrid strategies: (1) domain-parallel versions of the standard V-cycle and F-cycle algorithms; (2) a multiple coarse grid algorithm, proposed by Fredrickson and McBryan, which generates several coarse grids for each fine grid; and (3) two Rosendale algorithm, which allow computation on all grids simultaneously. We study an elliptic model problem on simple domains, discretized with finite difference techniques on block-structured meshes in two or three dimensions with up to 106 or 109 points, respectively. We analyze performance using three models of parallel computation: the PRAM and two bridging models. The bridging models reflect the salient characteristics of two kinds of parallel computers: SIMD fine-grain computers, which contain a large number of small (bitserial) processors, and SPMD medium-grain computers, which have a more modest number of powerful (single chip) processors. Our analysis suggests that the standard algorithms are substantially more efficient than algorithms utilizing either parallel strategy. Both parallel strategies need too much extra work to compensate for their extra parallelism. They require a highly impractical number of processors to be competitive with simpler, standard algorithms. The analysis also suggests that the F-cycle, with the appropriate optimization techniques, is more efficient than the V-cycle under a broad range of problem, implementation, and machine characteristics, despite the fact that it exhibits even less parallelism than the V-cycle.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
R. Fedorenko, Relaksacionnyi Metod Resenija Raznostynch Ellipticeskich Uravenija,CISL Matem i Matem Fiz,1:922–927 (1961).
L. Matheson, Multigrid Algorithms on Massively Parallel Computers, Ph.D. Dissertation, Department of Computer Science, Princeton University (1994).
L. Matheson and R. Tarjan, Analysis of Multigrid Algorithms on Massively Parallel Computers: Architectural Implications,Journal of Parallel and Distributed Computing,33:33–43 (1996).
W. Briggs,A Multigrid Tutorial, Society for Industrial and Applied Mathematics, Philadelphia, Pennsylvania (1987).
D. Jespersen, Multigrid Methods for Partial Differential Equations,Studies in Numerical Analysis, E. Golub, pp. 270–318 (1984).
W. Hackbush,Multi-Grid Methods and Applications, Springer-Verlag, Berlin, Germany (1985).
G. Forsythe and W. Wasow,Finite Difference Methods for Partial Differential Equations, John Wiley and Sons, New York (1960).
C. Desai,Elliptic Finite Element Methods, Prentice-Hall, Englewood Cliffs, New Jersey (1979).
J. Botha and G. Pinder,Fundamental Concepts in the Numerical Solution of Differential Equations, John Wiley and Sons, New Yor (1963).
P. Fredrickson and O. McBryan, Parallel Superconvergent Multigrid. In:Proc. 3rd Copper Mountain Conf. on Multigrid Methods, S. McCormick (ed.), Marcel Dekker, New York, pp. 195–210 (1987).
T. Chan and R. Tuminaro, Design and Implementation of Prallel Multigrid Algorithms. In:Proc. of the 3rd Copper Mountain Conf. on Multigrid Methods, S. McCormick (ed.), Marcel Dekker, New York, pp. 101–115 (1987).
R. Tuminaro, Multigrid Algorithms on Parallel Processing Systems, Ph.D. Dissertation, Department of Computer Science, Standford University (1989).
T. Chan and R. Tuminaro, Analysis of a Parallel Multigrid Algorithm. In:Proc. of the 4th Copper Mountain Conf. of Multigrid Methods, S. McCormick (ed.) Marcel Dekker, New York, pp. 69–86 (1989).
D. Gannon and J. van Rosendale, On the Structure of Parallelism in a Highly Concurrent PDE Solver,Journal of Parallel and Distributed Computing,3:106–135 (1986).
J. Jaja,An Introduction to Parallel Algorithms, Addison-Wesley, Reading, Massachusetts (1992).
L. Snyder, Type Architectures, Shared Memory and the Corollary of Modest Potential,Ann. Review of Computer Science, Annual Review, Inc., pp. 289–318 (1986).
W. Hightower, J. Prins, and J. Reif, Implementations of Randomized Sorting Algorithms on Large Parallel Machines,Proc. 4th Ann. ACM Symp. on Parallel Algorithms and Architectures, pp. 158–167 (1992).
G. Blelloch, C. Leiserson, B. Maggs, G. Plaxton, S. Smith and M. Zagha, A Comparison of Sorting Algorithms for the Connection Machine CM-2,Proc. of the 3rd Ann. ACM Symp. on Parallel Algorithms and Architectures, pp. 3–16 (1991).
T. Leighton,Introduction to Parallel Architectures: Arrays. Trees, Hypercubes, Morgan-Kaufmann, San Mateo, California (1992).
L. Valiant, A Bridging Model for Parallel Computation,Comm. of the ACM,33:103–111 (1990).
Technical Summary, The MasPar MPI System, MasPar Computer Corporation, Sunnyvale, California (1992).
The Connection Machine CM-200 Series, Technical Summary, Thinking Machines Corporation, Cambridge, Massachusetts (1992).
J. McDonnell, Private Communication, MasPar Computer Corporation (1992).
D. Culler, R. Karp, D. Patterson, A. Sahay, K. Schauser, E. Santos and T. von Eicken, LogP: Towards a Realistic Model of Parallel Computation,Proc. of the ACM SIGPLAN Symp. on Principles and Practices of Parallel Programming and ACM SIGPLAN Notices.28:1–12 (1993).
B. Maggs, L. Matheson and R. Tarjan, Models of Parallel Computation: A Survey and Synthesis.Proc. 28th Hawaii Int. Conf. on Syst. Sci., IEEE Computer Society Press.2:61–70 (1995).
Z. Bozkus, S. Ranka and G. Fox, Benchmarking the CM-5 Multicomputer,Proc. of the 4th Symp. on the Frontiers of Massively Parallel Computation, pp. 100–107 (1992).
R. Ponnusamy, A. Choudhary and G. Fox, Communications on the CM-5: An Experimental Performance Evaluation,Proc. of the 4th Symp. on the Frontiers of Massively Parallel Computation, pp. 108–117 (1992).
Pablo Tomayo, Primate Communication, Thinking Machines Inc. (1992).
L. Martinelli, Private Communication, Department of Mechanical and Aerospace Engineering, Princeton University, 1994.
K. Stuban and U. Trottenberg, Muttigrid Methods: Fundamental Algorithms, Model Problem Analysis and Applications. In:Multigrid Methods, Proceedings of the Koln-Porz Conference, W. Hackbusch and U. Trottenberg (eds.),Lecture Notes in Mathematics 960, Springer-Verlag, Berlin, Germany, pp. 1–176 (1982).
P. Fredrickson and O. McBryan, Normalized Convergence Rates for the PSMG Methods,SIAM Journal of Scientific and Statitical Computing,12:221–229 (1991).
N. Naik, Timely Communications: A Note on the Parallel Efficiency of the Fredrickson-McBryan Algorithm,SIAM Journal of Scientific and Statistical Computing,12:208–220 (1991).
C. Douglas and W. Miranker, Constructive Interference in Parallel Algorithms,SIAM Journal of Numerical Analysis 25:376–398 (1988).
J; Dendy, Private Communication, Los Alamos National Laboratory (1993).
R. Tuminaro and D. Womble, Analysis of the FMV Cycle for Massively Parallel Machines (abstract),Proc. Fifth Copper Mountain Conf. on Multigrid. Methods, Vol. 1. (1991).
J. Dendy, M. Ida and J. Rutledge, A Semi-Coarsening Multigrid Algorithm for SIMD Machines,SIAM J. Sci. Stat. Comput. 13:1460–1469 (1992).
N. Naik and J. van Rosendale, The Improved Robustness of Multigrid Elliptic Solvers Based on Multiple Semi-Coarsened Grids.SIAM Journal of Numerical Analysis 30:215–229 (1993).
M. Greibel, Parallel Domain-Oriented Multilevel Methods,SIAM Journal on Scientific Computing,9:1105–1125 (1995).
Author information
Authors and Affiliations
Additional information
Research at Princeton University partially supported by the National Science Foundation, Grant No. CCR-8920505, and the Office of Naval Research, Contract No. N0014-91-J-1463.
Rights and permissions
About this article
Cite this article
Matheson, L.R., Tarjan, R.E. Parallelism in multigrid methods: How much is too much?. Int J Parallel Prog 24, 397–432 (1996). https://doi.org/10.1007/BF02583022
Received:
Revised:
Accepted:
Issue Date:
DOI: https://doi.org/10.1007/BF02583022