抄録
A parallel predictor-corrector method, which consists of an explicit block method and an implicit Runge-Kutta-Nystrom method, is developed for solving second order initial value problems of the form y″=f(x, y), y(x_0)=η, y′(x_0)=ζ. A stepsize strategy based on Milne's device and an adaptive scheme for the predictor-corrector iteration are proposed. The method is implemented on a KSR1 parallel computer, which is a distributed memory system with 32 processors. The numerical experiment on the computer shows that the most successful implementation achieves a peak performance of speed-up 7.6 when the number of processors is 18.