Parallel methods for integrating ordinary differential equations

Abstract

This paper is dedicated to the proposition that, in order to take full advantage for real-time computations of highly parallel computers as can be expected to be available in the near future, much of numerical analysis will have to be recast in a more “parallel” form. By this is meant that serial algorithms ought to be replaced by algorithms which consist of several subtasks which can be computed without knowledge of the results of the other subtasks. As an example, a method is proposed for “parallelizing” the numerical integration of an ordinary differential equation, which process, by all standard methods, is entirely serial.