基于时间期望表的DPM预测策略  

Idle expectation table based prediction policy of dynamic power management

作　　者：戚隆宁[1] 胡晨[1] 张哲[1] 卜爱国[1] 

机构地区：[1]东南大学国家专用集成电路系统工程技术研究中心,江苏南京210096

出　　处：《电路与系统学报》2007年第2期89-93,123,共6页Journal of Circuits and Systems

摘　　要：系统级动态功耗管理(DPM,Dynamic Power Management)策略根据系统状态和负载的变化,动态地调整系统配置,从而能够降低系统功耗。PBALT(Probability Based Adaptive Learning Tree)预测策略以预测正确率为单一评估标准,存在高预测正确率高功耗的问题。本文提出基于空闲时间期望表(IET,Idle Expectation Table)的DPM预测策略IETBP(Idle Expectation Table Based Prediction),通过对空闲时间的分布和状态的误预测能耗的分析,以空闲时间的期望作为预测依据,从而克服了PBALT所存在的问题,并降低了算法复杂度。仿真实验表明与PBALT策略相比,IETBP策略在较低预测正确率的情况下能够更有效地降低部件的功耗。System-level dynamic power management （DPM） techniques observe the status and workload of the system, and dynamically reconfigure the system to reduce power consumption. This paper puts forward a novel prediction policy for DPM based on IET （Idle Expectation Table）, named IETBP （Idle Expectation Table Based Prediction）, which analyzes the distribution of the idle and the power consumption for mission prediction, and predicts with the expectation of the idle. IETBP resolves the problem of PBALT （Probability Based Adaptive Learning Tree） policy that high hit ratio cannot bring low power consumption, and descend the complexity. The simulation experiments show that IETBP consumes lower power than PBALT with low hit ratio.

关 键 词：动态功耗管理 预测策略 空闲时间期望表 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]