Numerical solution of partial differential equations by associative processing

Abstract

In a number of recent articles the application of associative processing to a variety of data processing problems has been considered. In this paper we consider the application of the parallel arithmetic capability offered by associative array processors to the numerical solution of partial differential equations. A set of equations concerned with weather forecasting is selected as a representative problem to show the methodology used in applying associative processing technology. An associative array processor implementation of a numerical solution to the equations by a time-marching process is developed for a 50X50 mesh and corresponding execution times are given. The solution yields the time-dependent behavior of three time and space-dependent variables which represent x and y components of wind velocity and height of a constant pressure surface. The associative array processor employed is the Goodyear Aerospace STARAN IV. Since its organization and operation are not widely known, the next section of this paper is devoted to a description of the associative array processor and its operation.