一种基于遗传算法的片上网络电压岛划分方法  

Genetic algorithm based NoC voltage-frequency island partition method

作　　者：刘斌[1] 常振超[1] 张兴明[1] 闫佳佳[2] 易洪波[1] 

机构地区：[1]解放军信息工程大学信息工程学院,郑州450002 [2]郑州大学信息工程学院,郑州450000

出　　处：《计算机应用研究》2012年第10期3740-3743,共4页Application Research of Computers

基　　金：国家"863"计划重点资助项目(2009AA012201)

摘　　要：针对片上网络电压岛划分的低能耗问题,提出一种基于遗传模拟退火算法的低能耗电压岛划分方法。该方法通过改进遗传算法的编码方法将电压岛划分融入到IP核映射中,综合考虑电压岛划分和IP核映射对片上总能耗的影响,通过遗传算法罚函数的设计保证了算法准确运行。仿真分析表明,在满足时间约束的条件下,相比于经典的方法,该方法的总能耗更低。To deal with the energy problem in voltage-frequency island partitioning,this paper proposed a low-power voltage-frequency island partition method based on genetic simulated annealing algorithm.After considering the effect on total energy consumption from both voltage-frequency island partitioning and IP core mapping,it re-designed the operator of genetic algorithm and penalty function.By improving coding method of genetic algorithm,it divided voltage islands into the IP core mapping.The experimental result shows that compared to classic methods,the proposed method has lower total energy consumption.

关 键 词：片上网络 电压岛 IP核映射 遗传算法 低能耗 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]