BIT Numerical Mathematics

, Volume 29, Issue 4, pp 635–657

The effect of ordering on preconditioned conjugate gradients

  • Iain S. Duff
  • Gérard A. Meurant
Preconditioned Conjugate Gradient Methods
  • 320 Downloads

Abstract

We investigate the effect of the ordering of the unknowns on the convergence of the preconditioned conjugate gradient method. We examine a wide range of ordering methods including nested dissection, minimum degree, and red-black and consider preconditionings without fill-in. We show empirically that there can be a significant difference in the number of iterations required by the conjugate gradient method and suggest reasons for this marked difference in performance.

We also consider the effect of orderings when an incomplete factorization which allows some fill-in is performed. We consider the effect of automatically controlling the sparsity of the incomplete factorization through drop tolerances and level of fill-in.

AMS Classification

65F10 

Keywords

Sparse matrices preconditioning ordering strategies conjugate gradients 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    Adams, L. M. and Jordan, H. F. (1984).Is SOR color-blind? SIAM J. Sci. Stat. Comput.7, 490–506.Google Scholar
  2. [2]
    Adams, L. M., LeVeque, R. J., and Young, D. M. (1988).Analysis of the SOR iteration for the q-point Laplacian. SIAM J. Numer. Anal.25, 1156–1180.Google Scholar
  3. [3]
    Axelsson, O. and Eijkhout, V. (1988).Robust vectorizable preconditioners for three-dimensional elliptic difference equations with anisotropy. InAlgorithms and Applications on Vector and Parallel Computers. H J J te Riele, Th J Dekker, and H A van der Vorst (editors). North-Holland, Amsterdam, New York, and London.Google Scholar
  4. [4]
    Behie, A. and Forsyth, P. A. (1984).Incomplete factorization methods for fully implicit simulation of enhanced oil recovery. SIAM J. Sci. Stat. Comput.5, 543–561.Google Scholar
  5. [5]
    Concus, P., Golub, G. H., and Meurant, G. (1985).Block preconditioning for the conjugate gradient method. SIAM J. Sci. Comput.6, 220–252.Google Scholar
  6. [6]
    Cuthill, E. and McKee, J. (1969).Reducing the bandwidth of sparse symmetric matrices. Proceedings 24th National Conference of the Association for Computing Machinery, Brandom Press, New Jersey, 157–172.Google Scholar
  7. [7]
    Duff, I. S., Erisman, A. M., and Reid, J. K. (1986).Direct Methods for Sparse Matrices. Oxford University Press, London.Google Scholar
  8. [8]
    Duff, I. S., Erisman, A. M., and Reid, J. K. (1976).On George's nested dissection method. SIAM J. Numer. Anal.13, 686–695.Google Scholar
  9. [9]
    Eisenstat, S. C., Gursky, M. C., Schultz, M. H., and Sherman, A. H. (1982).Yale sparse matrix package. 1: The symmetric codes. Int. J. Numer. Meth. Engng.18, 1145–1151.Google Scholar
  10. [10]
    Eisenstat, S. C., Elman, H. C., and Schultz, M. H. (1988).Block-preconditioned conjugate gradient-like methods for numerical reservoir simulations. SPE Reservoir Engineering. (Feb. 1988), 307–312.Google Scholar
  11. [11]
    George, A. (1971).Computer implementation of the finite-element method. Report Stan CS-71-208, Ph.D Thesis, Department of Computer Science, Stanford University, Stanford, California.Google Scholar
  12. [12]
    George, A. (1973).Nested dissection of a regular finite-element mesh. SIAM J. Numer. Anal.10, 345–363.Google Scholar
  13. [13]
    George, A. and Liu, J. W. H. (1981).Computer Solution of Large Sparse Positive-definite Systems. Prentice-Hall, Englewood Cliffs, New Jersey.Google Scholar
  14. [14]
    Golub, G. H. and Meurant, G. A. (1983).Résolution Numérique des Grandes Systèmes Linéaires. Eyrolles, Paris.Google Scholar
  15. [15]
    Gustafsson, I. (1979).Stability and rate of convergence of modified incomplete Cholesky factorization methods. PhD dissertation, Chalmers University, Göteborg, Sweden.Google Scholar
  16. [16]
    Hageman, L. A. and Young, D. M. (1981).Applied Iterative Methods. Academic Press, New York and London.Google Scholar
  17. [17]
    Kuo, C. C. J. and Chan, T. F. (1988).2-color Fourier analysis of iterative algorithms for elliptic problems with red/black ordering. CAM Report 88-15, Computational and Applied Mathematics, University of California, Los Angeles.Google Scholar
  18. [18]
    Lichnewsky, A. (1984).Some vector and parallel implementations for preconditioned conjugate gradient algorithms. InHigh-speed Computation. NATO ASI Series. Vol. F.7, edited by J. S. Kowalik. Springer-Verlag, Berlin, 343–359.Google Scholar
  19. [19]
    Meijerink, J. A. and van der Vorst, H. A. (1977).An iterative solution method for linear systems of which the coefficient matrix is a symmetric M-matrix. Math. Comp.31, 148–162.Google Scholar
  20. [20]
    O'Leary, D. P. (1982).Solving sparse matrix problems on parallel computers. Report 1234, Computer Science Center, University of Maryland, Maryland.Google Scholar
  21. [21]
    Schreiber, R. and Tang, W-P. (1982)Vectorizing the conjugate gradient method. Unpublished document. Department of Computer Science, Stanford University, Stanford, California.Google Scholar
  22. [22]
    Simon, H. D. (1985).Incomplete LU preconditioners for conjugate-gradient-type iterative methods. Paper SPE 13533, Proceedings of the 1985 Reservoir Simulation Symposium, Dallas, Feb. 10–13, 1985, 387–396.Google Scholar
  23. [23]
    Tinney, W. F. and Walker, J. W. (1967).Direct solutions of sparse network equations by optimally ordered triangular factorization. Proc. IEEE55, 1801–1809.Google Scholar

Copyright information

© BIT Foundations 1989

Authors and Affiliations

  • Iain S. Duff
    • 1
    • 2
  • Gérard A. Meurant
    • 1
    • 2
  1. 1.Computer Science and Systems DivisionHarwell LaboratoryOxonUK
  2. 2.CEA, Centre d'Etudes de Limeil-ValentonVilleneuve St GeorgesFrance

Personalised recommendations