Search

Publisher Summary Computational fluid dynamics (CFD) has reached such reliability that it is now a regular industrial tool for designing aerodynamic configurations. While numerical schemes and fluid models are becoming more sophisticated and problem sizes larger, it is a constant duty for code developers in an industrial context to adapt algorithms and software for new computer architectures to decrease costs and elapsed times spent for numerical simulation. This chapter deals with a new step accomplished in the implementation of the industrial CFD code, AEROLOG developed by Matra Defense for the study of missiles aerodynamics in today's parallel supercomputers. Efficiency of this approach has been demonstrated on a complex industrial test-case consisting of the analysis of the three-dimensional unsteady flow around a rotating missile. The implementation of the AEROLOG CFD code is considered on the INTEL iPSC-860, which is based on multiple instructions, multiple data (MIMD) hypercube architecture with 128 processors and 8 Mbytes of local memory per processor. The chapter describes the extension of the multidomain method extended to the case of a large number of processors with a speed-up optimization consisting in reaching both a good load balancing and a minimal data exchange among processors.