Research On The Parallel Programming Platform Based On PVM

Network Parallel Computing is in rapid development driven by the demand of high-performance computing, and it is widely used in the field of engineering and science. Along with Network Parallel Computing's development, some parallel platforms appear, such as PVM and MPI, etc. These platforms have a series of available API interfaces.It has some specificity and complexity on parallel program compared to serial program. The analysis and design on parallel program has to consider more things and conditions, such as task distribution, communication, synchronization and other mundane details. Parallel program needs to call the complex functions to handle the communication and data processing. Therefore, parallel program is more complicated than the serial program. It needs some good programming platform and tools to support.Nowadays the research on the Parallel Computing domestic and foreign focuses on the application, while the research on the platform and environment is fewer. There is not a good parallel platform now. The development of Network parallel computing is affected by the poor development of the parallel platform. Therefore, we build a parallel program platform based on PVM. The platform includes a parallel library, a parallel program framework App Wizard and a class manager embedded in the Visual C++ environment.Parallel library focuses to the areas of numerical computation. The library packages some typical parallel algorithms, providing a high-level application interface to the programmers. Programmer can write parallel program by calling the functions in the library.Parallel Program App Wizard is a wizard for programmer to create a new project. We write some parallel program template and package into the wizard. Programmer can easily generate parallel program framework through the Wizard.Parallel class manager is a class wizard for programmer. The design of the manager is according to the class wizard in Visual C++ environment. Programmer can easily create new class, and to manage the classes have been created in the project through the class manager.It will be easier and more efficient for programmer to program through the platform. This will not only speed up the development of the Network Parallel Computing, but also the development of relate engineering and science areas.