Mapping enthalpy-based solidification algorithms onto vector and parallel architectures

Abstract

Effective methods are required to solve three-dimensional solidification problems with fine meshes. This paper considers some of the problems associated with mapping control volume-based enthalpy algorithms onto vector processors and transputer-based parallel networks. Although vector processors may bring speedups of a factor of 5 or so, it appears that transputer-based networks have the facility to yield speedups that increase linearly with the number of transputer chips used. © 1989.