改进的三级遗传算法定制片上网络拓扑仿真

Simulation of Customized On-chip Network Topology Based on Improved Three-level Genetic Algorithm

针对G.Leary的3级遗传算法的遗传个体数量巨大,仿真速度慢的不足,提出了一种改进的3级遗传算法,从第1级路由器选择级数据结构,第2级节点/路由器映射级数据结构,第3级路由器/路由器映射级别数据结构来改进3级GA的拓扑结构优化技术。实验结果表明,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 s减少至254.2 s,平均提高17.4%,因此新的GA算法具有较好的实用价值。

An improved algorithm is proposed for G. Leafy＇ s three-level genetic algorithm which has the weakness of large amount of GA individuals and long simulation time. The topology optimization technology of the three level GA is improved. The first-level router is with selection-level data structure, the second-level node / router is with mappinglevel data structure, the third-level router / router has mapping-level data structure. The experimental resuhs show that the improved algorithm has little improvement in energy consumption, only 3.74% , but the simulation time is greatly reduced from 309. 125 s to 254.2 s, the average increase is 17.4% , so the new GA algorithm has a good practical value.