【摘要】 针对SoC芯片设计中的系统划分问题,提出一种面向应用的系统结构自动划分方法——基于模板的系统划分方法,并开发了系统划分工具——TBPS.基于模板的系统划分方法通过提取应用描述中的相似结构,并使用相似结构划分系统,最终得到一个尽量复用的系统结构.系统的相似结构划分采用遗传算法实现最优覆盖的搜索,并使用贪婪算法实现资源配置和任务调度.与已有的在系统划分中事先设定目标结构方法不同,文中方法可以根据应用的特点自动生成划分的结构.采用该方法及TBPS实现了应用描述到系统结构的自动划分.实验结果表明了TBPS对数据处理为主的应用划分的有效性.更多还原

【Abstract】 A template based partition method,which is an application oriented automatic system partition method,is presented,aiming at partition problem in SoC design. Accordingly a partition tool—TBPS is also developed. The template based partition method extracts the similar structures of the description,and then uses these structures to partition the system. Finally system structure which makes the most of the resources can be obtained. In the procedure of partitioning,genetic algorithms are adopted to search the optimal covers and a greedy method is designed for resources allocation and task scheduling. Different from the previous work which partitions the system onto a pre-defined architecture,this method generates the architecture automatically according to the inherences of the applications. By this method and the tool of TBPS,the partition is carried out from application to system structure in an automatic way. The experimental results show the effectiveness of TBPS for partition problems of data intensive applications.更多还原