基于遗传算法的软硬件划分方法

针对SOC软硬件协同设计中的软硬件划分问题,首先构建了软硬件划分的系统模型,讨论了影响SOC性能的价格、执行时间、功耗、面积等因素及其相互关系。在此基础上,提出了一个旨在实现最优性价比的目标函数,并结合软硬件划分问题自身的特点对传统遗传算法的遗传操作进行了改进,在Matlab中进行了软硬件划分方法的仿真,仿真实验获得的结果有力地证明了算法的稳定性和有效性。同时,该算法在MPEG-2视频解码芯片设计中得到了实际应用,芯片设计的结果良好地反映了设计目标,证明了算法的实用性。

Hardware-Software Partitioning Using Genetic Algorithms

Hardware-software partitioning is an important issue in system-on-chip designing.This paper discussed many factors related to the performance of the partitioning,such as the price,execution time,power consumption,size and so on.To achieve the minimum designing cost,this paper proposed an objective and fitness function,and improved genetic algorithms according to the characteristic of hardware-software partitioning.Also the algorithm was simulated in Matlab.The results show the stability,efficiency and quality of such method.At last,It is used to design the MPEG-2 video decoder.The high performance of the design proves the practicability of the algorithm.