一种基于统计信息的Cache漏流功耗估算模型  

A Statistics-Based Cache Leakage Power Estimation Model

作　　者：周宏伟[1] 张承义[1] 张民选[1] 

机构地区：[1]国防科技大学计算机学院,湖南长沙410073

出　　处：《计算机工程与科学》2007年第6期81-83,127,共4页Computer Engineering & Science

基　　金：国家863计划资助项目(2005AA110020)

摘　　要：本文提出了一种基于统计信息的Cache漏流功耗估算模型。该模型通过对Cache访问间隔时间的统计,估算出不同衰退间隔条件下Cache的漏流功耗。根据该模型所设计的Cache漏流功耗模拟器与Hotleakage漏流功耗模拟器相比,对于Cache漏流功耗估算的结果平均偏差小于3.46%。该模型可以用于Sleep Cache与Drowsy Cache中,估算不同衰退间隔下Cache漏流功耗比率,选取最优衰退间隔,最大程度地降低Cache漏流功耗。This paper provides a cache leakage power estimation model based on statistics. By using the statistics of cachelines＇ access intervals, this model can estimate the cache leakage power at different values of decay intervals. Compared with the Hotleakage leakage power simulator, the average variation of the estimation results from the simulator designed by our statistics-based leakage power model is smaller than 3. 46%. This model can be used to estimate the ratio of normalized leakage power at different values of decay intervals and select the best decay interval to reduce the cache leakage power at most.

关 键 词：统计 CACHE 漏流功耗 估算模型 

分 类 号：TP332[自动化与计算机技术—计算机系统结构]