Computational fluid dynamics on the transputer: system organization and performance

Abstract

A parallel processing system consisting of an array of commercial microprocessors known as transputers is utilized to solve the Navier Stokes partial differential equations for incompressible laminar flow. Two parallel algorithms are given for the solution of these equations. The structure and organization of the system hardware is described, and a method is presented for mapping the problem domain onto the processor array. The performance of the two algorithms is analyzed for a wide range of system configurations and grid sizes, showing the limitations of this system.<>