Parallel interpolation of high-order Runge-Kutta methods (Q2366299): Difference between revisions

The interpolation of higher-order Runge-Kutta formulas (continuous RK method) is considered. In practice, the computation of such interpolation is still unsatisfactory for its expensiveness. The interpolant is constructed for an embedded pair RK \(q(q-1)\) of Runge-Kutta formulas of order \(q\) and \(q-1\). For the computation it is supposed that a parallel machine with \(m-1\) processors \((2m>q+1)\) is available, which means that the calculation of the considered Hermite interpolating polynomial could be cheaper. The complexity investigation of the method gives a rough estimation of the speed up and the efficiency. Especially the 13-stage RK 8(7) pair of \textit{P. J. Prince} and \textit{J. R. Dormand} [J. Comput. Appl. Math. 7, 67- 75 (1981; Zbl 0449.65048)] is used for practical comparison. To test the performance of the method, 12 nonstiff problems of the package proposed by \textit{W. H. Enright} and \textit{J. D. Pryce} [ACM Trans. Math. Software 13, 1-27 (1987; Zbl 0617.65069)] are selected.