Boltzmann-NSGAⅡ算法的NoC映射研究  

Research of Network-on-Chip Mapping for Boltzmann-NSGAⅡ Algorithm

作　　者：易宏波[1] 罗兴国[1] 储慧琳[2] 赵凯[3] 

机构地区：[1]国家数字交换系统工程技术研究中心,郑州450002 [2]中国人民解放军61726部队,武汉430060 [3]中国人民解放军95951部队,广东汕头515065

出　　处：《计算机工程》2012年第22期283-286,共4页Computer Engineering

基　　金：国家“863”计划基金资助项目(2009AA012201);上海市科委重大科技攻关计划基金资助项目(08dz501600)

摘　　要：传统NSGAⅡ算法在解决NoC映射问题时容易出现局部最优和种群多样性的问题。为此,提出一个均衡优化时延模型和一种基于Boltzmann-NSGAⅡ的映射算法,时延模型从宏观链路负载和单个节点排队时延进行优化,映射算法引入Boltzmann机制、多点交叉算子、随机变异算子,以保持种群多样性并提高局部搜索能力。实验结果表明,Boltzmann-NSGAⅡ算法能获得更好的映射效果,最优解集的收敛性和均匀性更好。Traditional NSGA II algorithms are easy to get local optimization and less diversity of the population. Aiming at the problems, this paper proposes a balanced optimized delay model and the Boltzmann-NSGA II mapping algorithm. The delay model is constructed fi＇om the macroscopic link load distribution and the single node queue latency. The Boltzmann mechanism, multi-point crossover operator, random mutation operator are introduced into the algorithm to maintain the diversity of the population and increase the local search capability. Simulation results prove that the optimal solution set obtained by the Boltzmann-NSGA II algorithm is better in the convergence and uniformity, and the mapping effect of the Boltzmann-NSGA II algorithm is better.

关 键 词：片上网络 映射 时延模型 Boltzmann机制 NSGAⅡ算法 盒图 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]