基于遗传算法的片上网络低功耗映射

随着片上网络(Network-on-Chip)集成度的提高,功耗逐渐成为设计的焦点.本文提出了一种在延时约束条件下,基于遗传算法的片上网络通信链路的低功耗映射算法.该算法使用数组方式编码染色体,并采用非常规码的交叉和变异运算因子.它充分利用遗传算法的群体优势,能快速有效地对通信功耗作优化.实验表明,该算法能平均减少50%左右的通信功耗.

Power-aware mapping based-on genetic algorithm for Network-on-Chip

As the number of components in NoC（Network-on-Chip） increases,power becomes the focus of the design.In this paper,a low communication power mapping algorithm was proposed,which is based-on genetic algorithm and subject to latency constraint.The algorithm encodes the chromosome as an array of integers,and uses irregular crossover and mutation operations.It takes the advantage of population property of genetic algorithm to find low communication power solution efficiently.The results showed that,the proposed algorithm could effectively reduce the communication power to approximately 50% in average.