基于减少CPU功耗的目的,采用了机器学习SVM算法为CPU的动态频率电压调节技术(DVFS)制定决策模型,采用EDP(Energy Delay Product)作为最终优化指标.通过GEM5和McPAT工具进行仿真实验,发现新建的模型按照执行程序不同最多可以节省20%的EDP.%Power management of processor is always an important research field. In this paper , we take advantage of Support Vector Machine (SVM) Algorithm in Machine Learning to train and get the decision model for Dynamic Voltage and Frequency Scaling (DVFS) technology. We set the Energy Delay Product (EDP) as our optimization goal, to get a better trade-off between energy and performance. Experiment result shows that it can reduce as much as 20%EDP with different applications.
展开▼
