A Multi-Platform Shared- Or Distributed-Memory Navier-Stokes Code

This paper describes the implementation of a finite-element Navier-Stokes code on two parallel architectures: the IBM SP2 and the NEC SX-4. although these implementations are based on two different memory representations, the shared-and the distributed-memory paradigms, the actual source codes are extremely similar, thanks to the finite-element structure of the program. Two industrial applications exemplify the use of parallel computers in the Aerodynamic Modelization Department at Dassault Aviation.