Research Of Parallel Computing In Computational Fluid Dynamics

With the increasingly development of computer science, numerical computing, which has be juxtaposed by experimentation, has been one of the most significant methods in solving fluid problem. But the numerical computing of fluid dynamics is so huge or complex that the most excellent computer could not be precise simulation. Therefore, parallel computing, a great efficient means, could solve the fluid numerical simulation very well. In narrow sense, parallel computing is running a work, which is achieved formerly by running serially, simultaneously in the parallel mode on several processors. The aim of parallel computing is enhancing the efficiency of running a program largely.The purpose of the dissertation is: Applying parallel language to set up a steady and efficient parallel computing environment; realizing some parallel computing by using parallel language; proving that parallel computing could solve the problem that is computing large-scale numerical simulation difficultly due to the limited source of a computer.The dissertation aims at the parallel system-PC Cluster, which has distributed shared-memory multiprocessor, in order to introduce and apply two parallel computing models and their realizing tools. There are message transferring model (the realizing tool is MPI), and share memory model (the realizing tool is OpenMP). Moreover, the dissertation has discussed the parallel capability of the parallel computing programs.Furthermore, by combining MPI and OpenMP parallel language programming, and using the finite element method with unstructured grids, the dissertation has accomplished a parallel realization of numerical simulating temperature distribution with steady-state in a complex zone.As a result, combining MPI and OpenMP could solve intricate fluid problem in PC Cluster more efficient than solve it serially in one computer. Further, the shape of computing zone and equation form, etc, would not limit this parallel means. More significantly, this combine method could settle the problems that are difficult to describe large-scale issue or describe a issue most exactly. The parallel computing function of MPI and OpenMP is huge and has important meaning for scientific computing. Via the discussion of the dissertation, it could predict that parallel computing of combining MPI and OpenMP would be applied widely in various fields.