基于多目标优化的SoC软硬件划分方法研究  

Multi-Objective Optimization Based SoC Hardware-software Partition

作　　者：陆忆[1] 罗胜钦[1] 王长慧[1] 

机构地区：[1]同济大学电子科学与技术系,上海200092

出　　处：《电子与封装》2007年第2期30-33,47,共5页Electronics & Packaging

摘　　要：集成电路不断发展,SoC已经成为电子系统设计的主流,软硬件的划分又是其中的一个重要部分。文章采用基于多目标优化的遗传算法,对从任务级进行抽象建模所得到的系统模型进行软硬件划分。将Pareto最优概念与多目标优化问题相结合,应用于遗传算法中,从而实现了兼顾系统面积、功耗、成本等参数的软硬件划分方法。With the development of IC industry, SoC has taken an important part in electronic designing, while the partition of Hardware/Software is a significant part. This paper introduces a genetic algorithm based on multi-objective optimization, which departs the system by the DFG（Data Flow Graphic）. Combining the charac- ters of the two algorithms into the Pareto concept of multi-objective optimization problems, it can get an accurate approximation of the power/cost and pareto-front.

关 键 词：软硬件划分 遗传算法 PARETO最优 多目标优化 

分 类 号：TN702[电子电信—电路与系统]