A Study Of MPI Parallel Programming With Software Patterns Based On Object Technology

The technology of modern computer has stimulated the rapid growth of the computational science. In practice, the limit of speed and technology has embarrassed the PC to be used widely and the single processor can not meet the demand of computation in many challenge tasks any more. So parallel computing becomes a hotspot. One of the key to perform the parallel computing is to design effective parallel programs. While, parallel programming must face the more complex condition than serial programming. Such as task assigning, communication and synchronization between processes. As one of the most important tool of design parallel program, MPI is the same. So parallel programming with MPI is very difficult, including design, debugging, migrating and maintenance. To improve the methods of parallel programming, the software patterns based on object technology, which is prevailed in software engineering is introduced into MPI parallel programming in this research. Software patterns enhance the virtues of object-oriented technology and offers parallel programming a powerful intelligence tools. In the research of this paper, parallel programming technology of MPI is studied and some typical difficulty and characteristic is abstracted from it. And for these intrinsic difficulty and characteristic, this paper provides the solution of software patterns based on object t echnology. I n which, i nclude e ncapsulating n on-object-oriented i nterface i n M PI with object technology, implementing SPMD & MPMD model with strategy pattern and handling the interactive problem in parallel programming with MVC pattern, etc. At last, the statistic data collected from the practical experiment testify the feasibility and e fficiency o f t he r esolution using software p atterns b ased o n o bject t echnology i n parallel programming. By the way, the experiment in this research is run on the PC cluster platform which is constructed independently. Its theory and practice is instructive to perform researches on larger scale or complex cluster systems.