摘要
特定应用片上网络的拓扑优化是一种NP难的问题,吸引着很多研究人员对其进行研究,G.Leary等在VLSI刊物上提出的三级遗传算法进行拓扑制定,但其存在遗传个体数量大,仿真时间长的缺点。针对G.Leary的三级遗传算法的遗传个体数量巨大,仿真速度慢的不足,提出一种改进的三级遗传算法来进行拓扑优化定制。实验结果表明,我们的改进算法虽然在能耗方面改进不大,仅有3.74%,但是仿真时间却有较大的减少,由309.125秒减少至254.2秒,平均提高17.4%,因此我们的GA算法具有较好的实用价值。
Topology optimization for application-specific Network-on-chip is an NP-hard problem which abstracts lots of researchers to study it.G.Leary et al.made the topological development of the three-level genetic algorithm proposed in the VLSI publication,but there is a shortcoming of large number of genetic individuals and long simulation time.In this paper,aiming at G.Leary s three-level genetic algorithm,which has a large number of genetic individuals and slow simulation speed,an improved three-level genetic algorithm was proposed to optimize topology.Experimental results showed that our GA had greatly reduced the simulation time from 309.125 s of Leary s GA to 254.2 s,averagely 17.4%,though in terms of energy consumption,only 3.74%improvement.As a result,our improved GA algorithm had good practical value.
作者
熊昕
赖国明
Xiong Xin;Lai Guoming(Modern Education Technology and Information Center,Guangzhou Panyu Polytechnic,Guangzhou 511483,Guangdong,China;College of Information Science and Technology,Huizhou University,Huizhou 516007,Guangdong,China)
出处
《计算机应用与软件》
北大核心
2018年第3期241-246,共6页
Computer Applications and Software
基金
国家自然科学基金项目(11247310
11505042)
广东省自然科学基金项目(2015A030313807)
广东省教育厅学科与专业建设专项基金项目(KJCX20130128)
广州市属高校科研课题(1201630601)
惠州学院教授
博士启动项目(2017JB005)
关键词
片上网络
片上系统
网络拓扑
遗传算法
Network-on-chip
System-on-chip
Network topology
Genetic algorithm
