Automatic Parallelization And Implementation Of Loop Nest Based On MPI Platform

With the rapid development of high performance computing technology, high performance computers have been widely used in many fields. The rapid development of grid computing technology has greatly improved the applications of high performance computers. The key software, such as parallel operating system and parallel application program, is necessary to make good use of high performance computers. But very few programmers can write parallel program, and many parallel programs are still made at expert level. This situation has made it very difficult to use widely high performance computers. Automatic parallization technology will greatly change this situation. It will help programmer make parallel program easily.The technology of automatic parallelization has made great progress with the development of optimizing compilers technology. The related theory based on sharing memory is becoming perfectly. But the current automatic parallelizer is still realized at lower layer. The parallel program made by means of traditional automatic parallizer can not transplant easily. MPI has been widely used as a really standard and a message transposed interface, which has good transplanting.The main contributions of this paper are summarized as follows.(1) Construct a testing platform for automatic parallelization;(2) Implement program architecture analysis, dependence analysis, data dividing and translating the AST to MPI parallel program on part of serial c language source programs on the platform;(3) Implement the SPMD model of automatic translate from serial source program to parallel program;(4) Test the performance of parallel program made by the automatic parallel platform.