Parallel multigrid using PVM Original Research Article

This paper discusses an implementation of a parallel multigrid method using Parallel Virtual Machine (PVM). The basics described here apply to general parallel computers, either with shared memory or distributed memory systems. The actual implementation has been performed on both type of systems, although we will focus on distributed memory systems in this paper. The distributed memory implementation has been done using PVM on Convex MetaSeries machines. The speed-up results will be analyzed with respect to computational work and communicational overhead.