基于并行模拟退火算法的NoC映射  

A Mapping Algorithm based on Parallel Simulated Annealing on NoC

作　　者：刘蓓[1] 

机构地区：[1]安徽工程大学现代教育技术中心,安徽芜湖241000

出　　处：《洛阳理工学院学报（自然科学版）》2015年第3期61-64,共4页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

基　　金：安徽省自然科学基金项目(KJ2012B022);安徽省教育厅教研基金资助项目(2012jyxm280);安徽工程大学青年科研基金资助项目(2013YQ32)

摘　　要：片上网络是一种多核系统互连的有效解决方案,片上网络的映射问题是在芯片设计阶段的关键问题,映射的结果直接影响系统的功耗。为降低功耗且减少执行时间,提出一种并行模拟退火的映射方法,采用多种群群体优化机制,对每个子问题并行求解,找出了一个最优解作为全局最优解。实验结果表明,该方法与自适应模拟退火算法相比,可以降低9.1%的平均功耗,并能减少32%的执行时间。NoC architecture is viewed as a potential solution for interconnect demands of multi-core systems. Mapping tasks onto different cores of the network is a critical phase in NoC design because it determines the energy consumption of the system. In order to reduce the energy consumption of applications running on the system, a multi-population optimization mapping algorithm based on parallel simulated annealing is proposed. Compared with the adaptive genetic annealing algorithm the proposed method can reduce average energy consumption and execution time by 9.1% and 32% respectively.

关 键 词：并行模拟退火 片上网络 映射算法 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]