Limits of parallelism in explicit ODE methods
- 32 Downloads
Numerical methods for ordinary initial value problems that do not depend on special properties of the system are usually found in the class of linear multistage multivalue methods, first formulated by J.C. Butcher. Among these the explicit methods are easiest to implement. For these reasons there has been considerable research activity devoted to generating methods of this class which utilize independent function evaluations that can be performed in parallel. Each such group of concurrent function evaluations can be regarded as a stage of the method. However, it turns out that parallelism affords only limited opportunity for reducing the computing time with such methods. This is most evident for the simple linear homogeneous constant-coefficient test problem, whose solution is essentially a matter of approximating the exponential by an algebraic function. For a given number of stages and a given number of saved values, parallelism offers a somewhat enlarged set of algebraic functions from which to choose. However, there is absolutely no benefit in having the degree of parallelism (number of processors) exceed the number of saved values of the method. Thus, in particular, parallel one-step methods offer no speedup over serial one-step methods for the standard linear test problem. Although the implication of this result for general nonlinear problems is unclear, there are indications that dramatic speedups are not possible in general. Also given are some results relevant to the construction of methods.
Subject classificationAMS(MOS) 65L05 CR G.1.7, G.1.0
Unable to display preview. Download preview PDF.
- A. Bellen, Parallelism across the steps for difference and differential equations, in:Numerical Methods for Ordinary Differential Equations, eds. A. Bellen, C.W. Gear and E. Russo, Proceedings, L'Aquila, 1987, Lecture Notes in Mathematics no. 1386 (Springer, Berlin, 1989) pp. 22–35.CrossRefGoogle Scholar
- C.W. Gear, The potential for parallelism in ordinary differential equations, in:Computational Mathematics II, ed. S.O. Fatunla (Boole Press, Dublin, 1987) pp. 33–48.Google Scholar
- S. Horiguchi, Y. Kawazoe and H. Nara, A parallel algorithm for the integration of ordinary differential equations,Proc. Int. Conf. on Parallel Processing (1984) pp. 465–469.Google Scholar
- K.R. Jackson and S.P. Norsett, The potential for parallelism in Runge-Kutta methods. Part I: RK formulas in standard form, in preparation.Google Scholar
- R.D. Skeel and H.-W. Tam, Potential for parallelism in explicit linear methods,Proc. IMA Conf. on Computational ODEs, eds. J.R. Cash and I. Gladwell (Oxford Univ. Press), to appear.Google Scholar
- H.-W. Tam, Parallel methods for the numerical solution of ordinary differential equations, Univ. of Illinois at Urbana-Champaign, Report No. UIUCDCS-R-89-1516 (1989); also, Ph.D. thesis.Google Scholar
- H.-W. Tam, Parallel predictor-corrector block methods for ordinary differential equations,Proc. 4th SIAM Conf. on Parallel Processing for Scientific Computing eds. J. Dongarra, P. Messina, D. Sorensen and R. Voigt (SIAM, Philadelphia, 1990) pp. 204–209.Google Scholar
- H.-W. Tam, A new parallel algorithm for the numerical solution of ordinary differential equations,Proc. IMA Conf. on Computational ODEs, eds. J.R. Cash and I. Gladwell (Oxford Univ. Press), to appear.Google Scholar
- H.-W. Tam, One-stage parallel methods for the numerical solution of ordinary differential equations, SIAM J. Sci. Statist. Comput 13 (1992), to appear.Google Scholar
- H.-W. Tam, Two-stage parallel methods for the numerical solution of ordinary differential equations, SIAM J. Sci. Statist. Comput 13 (1992), to appear.Google Scholar
- H.-W. Tam and R.D. Skeel, Stability of parallel explicit ODE methods, manuscript (1991).Google Scholar