The Application Research Of Serial Program Automatic Parallelization Based On MPI

With the fast development of parallel computer,More and more field demand the improvement of large-scale problem solving’s speed actively.Serial programs’ paralleli-zation can shorten time of problem solving.With realizing fast parallelization, Serial program automatic parallelization software’ has been designed and realized.MPI (message passing interface) information transfer programm is foundation of p-arallel program. MPI procedural basic flow scheme and programming mode are applied in system software. Massively parallel computation algorithm of matrix and array are researched.Parallel algorithm is evaluated and complexity analyzed according to releva-nt theory. Interprocess parallel execution needs for message communication. Interproc-ess’communication cooperation is solved by function and pattern theory which is point-to-point and set communication.Automatic parallelization software is designed by programming languages.Two modes which are serial program analyses and automatic transformation are mainly researched. AI theory is foundation of serial program analysis procedure. Artificial Intelligence’searching algorithms which are researched and reformed are applied to serial program’s reading-in and analyses. Procedural high efficiency read and accuracy analyses can be realized by algorithm.Automatic switching mode is present according to variable definition, sequence statement, looping statement, branch statement as well as large-scale matrix and array operation by principle of automatic switching. At last,system performance test is undertook through the example of proximate calculation of PI value.Conclusion is carrie-d out through comparison with serial computing’runtime and acceleration ratio.Parallel computation’performance can get better and better when calculation amountis more and more. Serial program automatic parallelization software not only realizes fast parallelizat-ion but also relief programmer’burden and improve efficiency in time and space.