Design Of MPSoC And Parallel Programming Model

With the further improvement of chip design complexity, the single-processor-system can not meet the need of processor industry's quick development. In order to meet the complex needs of the application, it is the time to introduce an innovative way to design processor. The MPSoC (Multiprocessor-system-on-chip) represented the next generation of multi-core technology has become the mainstream of IC design technology.There are challenges on the study of MPSoC at the aspects of architecture, sharing and synchronization technology between processors, software design model, operating system design, security, et al. On the other hand, there is great potential. Because the design of MPSoC depend on it's application, so we can custom and optimize it on the core instruction set, IP, bus, multi-core communications technology and software model.In this paper, an MPSoC is designed and it makes use of open source IP core which supplied by OpenCores organizations, including the processor core OpenRISC 1200, UART IP core and Conbus bus IP core. Since we focuses on the establishment of multi-core system building, the cache and MMU are not in the processor's configuration. In order to be compatible to the OpenRISC architecture, the designed IP cores are according to the WISHBONE criterion including the memory IP core and the component controlling the Program synchronization. Also, we design a hardware-semaphore method to solve the multi-processors' synchronization.On the other hand, because all the traditional programs are wrote for the single processor, they can not be accelerated on multi-processor system directly. To solve this problem, we extend the OpenMP parallel programming model on our MPSoC Platform, including: the clauses describe variable attributes, the global shared variables, the non-global shared of variable, parallel instructions, synchronize instructions and work-sharing directives.The result shows that as the calculation increases the acceleration between efficiency of MPSoC and SPSoC grows nearby 1.9. Compared to the written parallel programs, the OpenMP parallel programming model implementation on our MPSoC platform simplified the parallel programming without high overhead. Finally we talk about the shortage in the system, and then figure out the future research.