Abstract:
This paper presents an optimized and fine grained strategy LPDVFS(DVFS based on Linear Prediction) aiming at mobile operating systems based on Linux kernel. This strategy uses the linear prediction mechanism based on historical data to adjust the range and orientation of voltage and frequency. The parameters are obtained through regression method. Experimental results show that the optimized LPDVFS strategy can save 13.55% energy consumption compared with the default coarse grained frequency scaling strategy. The LPDVFS strategy can reach better power saving effect and enhance the cruising ability of mobile terminals.
Key words:
low energy consumption technology,
dynamic voltage frequency,
linear prediction,
fine grained,
benchmark program
摘要: 针对以Linux为内核的移动操作系统,提出一种细粒度的DVFS策略LPDVFS。该策略基于历史数据,使用线性预测的方法,指导电压频率调整方向和幅度。线性预测中的参数通过回归方法确定。实验结果表明,LPDVFS策略相比Linux内核默认使用的粗粒度调频策略,能降低系统13.55%的功耗,延长移动终端的续航时间。
关键词:
低功耗技术,
动态电压频率,
线性预测,
细粒度,
benchmark程序
CLC Number:
ZHANG Li, YUAN Xiao-Long, HAN Yin-He. LPDVFS Strategy of Mobile Terminal Based on Linear Prediction[J]. Computer Engineering, 2012, 38(12): 239-242.
张立, 袁小龙, 韩银和. 基于线性预测的移动终端LPDVFS策略[J]. 计算机工程, 2012, 38(12): 239-242.