MultiThreads Concurrent Object-Oriented Model Based On MPS

Traditional parallel and concurrent programming took sequence programming to program, programming and debugging was complex and difficult, for this developing the parallel software was at low efficiency. In the meantime, adopting OO to programming could reduce the semantic gap between problems field and solutions field. OOP makes the software development more effective. The object parallel and concurrent systems was the outcome when combining the parallel and concurrent model with OO model, these systems not only had the superiority of OO, but also made parallel software modeling easy and bettered the maintainability and reuseability of parallel software system. In the procedure of combination of both models, a COO model must be built up.According the analysis of usual COO models characteristics, a summarization of key technology used to build a COO model was drawn. On basis of this, a multithreads COO model which designed with framework pattern was built up based on MPS, it could run on heterogeneous parallel architectures and had two layer parallel capability. Inside the model, the Process Model was adopted to map a parallel object to a parallel task, runtime support system was used to distribute parallel objects to distributed environments. In favor of the MP system effective message corresponding manner, it could easily send remote method requests and receive the returned results. The methods in parallel object were encapsulated with concurrent threads, threads must satisfied with vertical conditions. Mini C/S framework and body in object was introduced to control the threads concurrent execution. Furthermore, States-Methods Set was used to resolve inheritance abnormity. Finally, the COO model was implemented by means of the library, a RTS including a preprocess framework was designed above PVM, it provided fine runtime support for the upper COO model. It also introduced much Thread-Safe mechanism into RTS to settle the Thread-Safe questions in MP system.According test, this model worked availably, and run the test example written in concurrent class library already implemented on the Beowulf environment, it showed a better speedup performance.