Multiprocessing On Supercomputers for Computational Aerodynamics

Very little use is made of multiple processors available on current supercomputers (computers with a theoretical peak performance capability equal to 100 MFLOPS or more) in computational aerodynamics to significantly improve turnaround time. The productivity of a computer user is directly related to this turnaround time. In a timesharing environment, the improvement in this speed is achieved when multiple processors are used efficiently to execute an algorithm. We apply the concept of multiple instructions and multiple data (MIMD) through multitasking via a strategy which requires relatively minor modifications to an existing code for a single processor. Essentially, this approach maps the available memory to multiple processors, exploiting the C-Fortran-Unix interface. The existing single processor code is mapped without the need for developing a new algorithm. The procedure for building a code utilizing this approach is automated with the Unix stream editor. As a demonstration of this approach, a Multiple Processor Multiple Grid (MPM(-,') code is developed. It is capable of using nine processors, and can be easily extended to a larger number of processors. This code solves the three-dimensional, Reynolds averaged, thin-layer and slender-layer Navier-Stokes equations with an implicit, approximately factored and diagonalized method. The solver is applied to a generic oblique-wing aircraft problem on a four processor Cray-2 computer, using one process for data management and non-parallel computations and three processes for pseudo-time advance on three different grid systems. These grid systems are overlapped. A tricuhic interpolation scheme is developed to increase the accuracy of the grid coupling. For the oblique-wing aircraft problem, a speedup of two in elapsed (turnaround) time is observed in a saturated timesharing environment.