We will demonstrate feasibility of an efficient parallel implementation of the incompressible 3-dimensional Navier-Stokes solver INS3D-UP on distributed-memory MIMD computers. The work will include a relatively limited solver implementation, without turbulence model, complex boundary conditions, extensive I/O, or multi- ple-zone capabilities. Initially only point-relaxation will be provided. The target architecture is an IBM SP2 computer, but use of the Message Passing Interface standard will ensure portability to other platforms. Efficiency will be demonstrated by comparison with a single-processor version of the code, to be ported to the IBM SP2 as well. Both absolute performance and scalability will be examined. The resulting parallel code will be extended in Phase II to include a 1-equation turbulence model, realistic boundary conditions, I/O for restarts and visualization, Chimera multiple-zone capabilities, and line-relaxation. More advanced solution techniques like GMRES will also be investigated. INS3D-UP is one of the three NASA Ames production flow solvers (the others are CNSFV and OVERFLOW) for overset structured grids. Availability of an efficient parallel version will greatly enhance the usefulness of the code and increases the range of flow problems that can be solved cost-effectively.Commercial Applications:The two main target areas in the aircraft industry for use of a parallel version of INS3D-UP are that of the Advanced Subsonic Transport (AST) and of propulsion systems. Within AST, incompressible 3-dimensional flow solvers are used, among others, for the simulation of high-lift and laminar-flow-control devices. Boeing and McDonnell-Douglas would be the main beneficiaries of the parallel program. In propulsion systems Pratt\&Whitney, Rocketdyne and Aerojet constitute the most important users of high-performance incompressible-flow codes. However, parallelization of INS3D-UP will increase the cost-effectiveness of the program to such an extent that it will also become attractive for more traditional types of engineering computations, such as construction (wind noise around buildings) and automotive engineering (wind drag of cars).
