EFFICIENT PARALLEL COMPUTATION OF UNSTRUCTURED FINITE-ELEMENT REACTING FLOW SOLUTIONS

A parallel unstructured finite element (FE) reacting flow solver desig ned for message passing MIMD computers is described, This implementati on employs automated partitioning algorithms for load balancing unstru ctured grids, a distributed sparse matrix representation of the global FE equations, and parallel Krylov subspace iterative solvers. In this paper, a number of issues related to the efficient implementation of parallel unstructured mesh applications are presented. These issues in clude the differences between structured and unstructured mesh paralle l applications, major communication kernels for unstructured Krylov it erative solvers, automatic mesh partitioning algorithms, and the influ ence of mesh partitioning metrics and single-node CPU performance on p arallel performance. Results are presented for example FE hear transfe r, fluid flow and full reacting flow applications on a 1024 processor nCUBE 2 hypercube and a 1904 processor Intel Paragon. Results indicate that very high computational rates and high scaled efficiencies can b e achieved for large problems despite the use of sparse matrix data st ructures and the required unstructured data communication. (C) 1997 El sevier Science B.V.