Multidimensional dynamic-programming on massively-parallel computers

Massively parallel computers have become widely available for the solution of a wide range of numerical problems. However, methods that work well on single processor architectures are often not the best algorithms when multiple processors are available. In this paper, we present a dynamic programming approach to the numerical solution of elliptic partial differential equations. This method, based upon the earlier work of Angel and others, indicates that reconsideration of direct (noniterative) approaches to the numerical solution of partial differential equations can lead to competitive algorithms.