Pc Cluster System Of Parallel Development Environment Mpi Application Performance Analysis

The technology of modern computer has stimulated the rapid growth of the computational science. In practice, the limit of the speed of the physical equipment and technology has embarrassed the PC to be used widely, so it is important to study the high performance computing technology. :In order to implement high performance computing, it is necessary to design more efficient parallel programs. The tools to design parallel program include MPI, PVM, Linda and so on. Due to the advantage of MPI (Message Passing Interface), which is portal, powerful, and highly efficient, MPI became the most important tool to design parallel programs.To gain high performance is one of the purpose of parallel processing, in which effective speed is one important factor. The development of sequential software is increased by the implementation of the technology of structure, process and object through more than twenty years. Parallel program is more difficult to design than sequential one, and it is just more difficult to analyze than to write parallel program. The complex of parallel computers decides the difficulty of programming, debugging and analyzing of parallel programs. Although now the peak speed of parallel computer is surprising, it is disappointing that the actual speed is one forth of peak speed in practice. Now it is important to implement the performance test of parallel programs, analyze the performance of parallel programs, according to which we can instruct the optimization of programs, provide the performance factor, and improve the designing of the system. This paper analyzes the methods that can be used to test the performance of parallel program.The tools of performance analyze includes static and dynamic method. Static method gains performance data of source programs, which is given to users. This method can estimate the performance of different programs and running environments, but its accuracy is low. Dynamic method collects every performance data through measuring, which is given to users immediately or afterwards. The accuracy of this method is higher than static method. This paper mainly uses dynamic method, using MPI environment and the technology of embedded MPE to implement the visualization of program performance.