摘要： 提取应用描述中的相似运算结构并使用相似结构划分系统可以有效实现片上系统划分。提出一种基于生长的相似结构自动提取方法，其中的单模板匹配算法实现任意结构模板的提取，多模板生成算法采用模板和子图同步生长的方法。实验结果表明，该方法适用于包含扇出和汇聚结构的模板提取，计算时间与传统方法相比可减少30%~70%。

关键词: 片上系统, 划分, 相似性

Abstract: Extracting similar structure and using the structure to cover the system is an effective method for the partition of System-on-Chip(SoC). This paper presents an automatic extraction method of similar structure based on growth, of which Single Template Matching Algorithm(STMA) extracts any kinds of structure of templates, and Multiple Templates Generation Algorithms(MTGAs) uses the incremental and synchronous growth method for all the templates and sub-graphs. Experimental results show that the template containing fan-in and fan-out branches can be supported by the new method, while computing time is reduced by 30%~70% compared with traditional methods.

Key words: System-on-Chip(SoC), partition, similarity

中图分类号: 

• N945