Design And Implementation Of MPSoC Framework Based On FPGA

With the rapid development of embedded technology, embedded applications make higher demands on the performance of processor. Multi-core technology, more and more, is applied with embedded development. MPSoC (Multiprocessor Systems-on-Chip) represents the development trend of the multiprocessor of embedded applications. The most advantage of MPSoC is that with the different demand of embedded applications it can make the most suitable design so that the system performance can be made best used of.Compared with traditional single-core design, MPSoC provides designers with not only the more freedom but also the more challenge. The most important problem to be solved firstly of the MPSoC design is how to make optimal design for fixed workload.This paper reviews the Amdahl's law and the relative research about multi-core framework of Hill and Marty and gives an accelerator formula with parallel overhead based on these research. Aiming at the features of customization in embedded development, this paper proposes a novel MPSoC framework to face the embedded application and analyzes the optimization problem of MPSoC framework design in fixed multitasking workload from the theory.In order to analyze the MPSoC framework's performance, the paper establish an FPGA testing system in the Virtex-II Pro platform of Xilinx Inc. and test it using the Dijkstra program and SUSAN program from MiBench. The experiments show: the paper provides an extend MPSoC framework that it can effectively relieve bottleneck and get higher system performance in fixed multitasking workload. The result of the compare test indicate that the performance of MPSoC framework proposed in the paper is much better than the performance of traditional homogenous framework and heterogeneous framework proposed by Hill and Marty.