PARALLEL CFD BENCHMARKS ON CRAY COMPUTERS
Constantinos Evangelinos, George Em Karniadakis · International Journal of Parallel Emergent and Distributed Systems · 1996
In this paper we present benchmark results from the parallel implementation of the three-dimensional Navier-Stokes solver Prism [1] on the Cray T3D, We compare the single processor performance with other Cray computers, namely the Cray C90, J90 and EL98, as well as Digital Equipment Corporation's DEC 3000/500 (which uses the same processor as the T3D) and AlphaServer 8400 5/300 (which uses the same processor as the Cray T3E, the successor system to the T3D). The numerical method used in the solver is based on mixed spectral element-Fourier expansions in (x − y) and z-directions, respectively. Each (or a group) of Fourier modes can be computed on a separate processor as the linear contributions (Helmholtz solves) in Navier-Stokes equations are completely uncoupled. Coupling is obtained via the nonlinear contributions (adveaion terms) and requires a global transpose of the data and one-dimensional multiple-point FFTs. We first analyze the computational complexity of Prism identifying basic computational kernels and communication bottlenecks. These kernels are benchmarked separately on each of the CRAY computers. Next, we define a three-dimensional flow past a circular cylinder as a benchmark flow problem, and measure parallel scalability and performance using different message passing libraries. Similar bench mark results on other parallel distributed and shared memory platforms have been reported recently in [2].