摘要
针对采用2D-Torus拓扑结构且支持电压频率岛(VFI)的异步片上网络能耗优化问题,提出了具有可靠性的、基于电压频率岛的划分和分配及片上网络任务映射的能耗优化方法.该方法采用递进优化的方式,根据IP核的动态处理能耗,不同电压频率岛之间的转换能耗和可靠性带来的能耗开销定义了IP核在电压频率岛之间移动的阈值函数,并通过对阈值函数进行判断完成电压频率岛的划分和分配,应用基于三元相关性量子粒子群优化算法完成处理单元到资源节点的映射,在映射中考虑保证系统可靠性的通信开销,对异步片上网络系统的可靠性进行优化.实验结果表明,该算法可以在不过多消耗能耗的情况下显著的改善片上网络系统的可靠性,且可有效降低NOC系统的能耗.
A method for reliability and energy optimization using voltage-frequency island （VFI） partitioning, assignment, and task mapping is proposed for a network on chip （ NoC ） which uses 2D-Torus topology and supports voltage-frequency islands. The method adopts a progressive optimization approach. A moving threshold function between different voltage-frequency islands is defined according to the dynamic processing energy of IP cores, energy overhead of VFIs, and reliability costs. The division and allocation of the voltage-frequency island is based on the moving threshold function. A ternary correlation quantum-behavior particle swarm optimization （PSO） algorithm is used for the IP core mapping, during which, the reliability cost is fully considered to increase the reliability of the NoC system. The experimental result shows that the reliability of the NoC system is increased without consuming too much energy, and the energy consumption of the whole NoC system can be reduced effectively.
出处
《信息与控制》
CSCD
北大核心
2016年第6期671-676,共6页
Information and Control
基金
国家自然科学基金资助项目(61370153)
关键词
异步片上网络
电压频率岛
可靠性
映射算法
低功耗
asychronous NoC
voltage-frequency island
reliability
mapping algorithm
low power
