Applications Of Parallel Computing On CFD Problems

A study in applications,computer architecture and networking shows that parallel programming is becoming the center and direction of development to the programming enterprises. As to parallel programming,there have been a number of rather mature design methods. At the same time, network computation technologies which can be used to parallel programming have been developed rapidly. Among those technologies,message-passing technology is a kind of widely-used technology in professional area. One of message-passing technologies,MPI (Message Passing Interface) is especially widely-used due to its powerful ability and also because it is easy to use. This paper tries to give a solution to a practical problem in engineering—the CFD problem using MPI. The purpose of this paper is to show the general design method of parallel programming and the possibility of solving problem using MPI. TEAMKE1 is a 2D steady state flow prediction code from the University of Manchester Institute of Science and Technology (UMIST). Laminar and turbulent predictions can be made on a structured grid, and the linear equation solver used is bi-directional LSOR. The code contains upwind style computations in setting up the coefficients of the systems to be solved requiring communication splitting to allow communication migration out of loops and thus efficient parallel code. This paper starts with a brief introduction of MPI which includes the discussion of message-send & receive mechanism , global communication mechanism ,asynchronous communication mechanism and modularity. After that,this paper introduces the background of CFD problem briefly. Finally, this paper gives an approximate solution to the CFD problem.Something like general consideration of parallel programming, design process of CFD problem and part of program code are included in the solution.