Implicit Navier-Stokes Codes in Parallel for Aerospace Applications

Abstract

This chapter discusses the parallel implementation of Implicit Navier-Stokes Codes to simulate two and three dimensional compressible and turbulent fluid flows. The implicit methods are based on a pre-conditioned conjugate gradient type solution of the large and sparse linear system at each time step. The pre-conditioner is based on the alternating direction implicit factorization of the linear system. An efficient parallel implementation is achieved by using data transposition to locate the data relating to complete mesh lines in each coordinate direction onto a single processor at the appropriate time in the calculation. Code timings are given for several machines and issues, such as the effectiveness of the parallel implementation are considered. The chapter illustrates the performance achieved on the Cray T3D and a workstation cluster for the transonic and turbulent flow over an aerofoil section and a wing. Results obtained on the Intel Hypercube, the Cray T3D, and a Silicon Graphics workstation cluster show that these codes can make effective use of parallel computers