Research On Hardware/Software Partitioning Method Based On GA And PSO

With development of microelectronic techniques, integrated circuit goes into the stage of SoC (system on chip). Due to the increasing of design complexity, traditional design methodology can not meet the need of SoC any more. To solve the new problem, hardware/software co-design methodology has been proposed, and the hardware/software partitioning is one of the key techniques in it. Therefore, it is critical to study on the hardware/software automatic partitioning method and to use it for determining the best implementation of system. It can bring about many boons to both theory and application.First, this dissertation introduces the basic concepts and development situations of hardware/software co-design, analyzes the current hardware/software co-design models, algorithms and their advantages and disadvantages. It then describes the Genetic Algorithm (GA) and the Particle Swarm Optimization (PSO), improves them respectively to GAnew and PSOnew. Based on these two algorithms, a combined algorithm GAPSO, which utilized the advantages of GAnew and PSOnew, has been proposed for further improvement. Task flow graph was used as mathematical system model in experimental, and the result told that GAPSO was better than GAnew and PSOnew.Finally, to solve hardware/software partitioning problem of large systems, this dissertation points out a direction that large embedded systems should be pretreated before their hardware/software partition. After brief introduction to the classical clustering method, flow chart of process about hardware/software partition of large embedded systems was given.