基于GINI指数分类的嵌入式CPU功耗预测方法

A Method of the Power Consumption Prediction of Embedded CPU Based on the GINI Index Classification Method

文中提出了一种基于嵌入式系统CPU功耗预测并对其进行低功耗优化的方法.引入GINI指数的构建训练分类器,利用PowerTop工具对系统CPU进行监测,并以此作为训练数据集,将该分类器封装到系统中,对嵌入式系统的CPU频率、电压及所处的状态进行预测.通过仿真实验表明,该方法在系统负载较小的情况下,对嵌入式CPU功耗的优化的效果更好.

In this paper, a method to predict and optimise the embedded CPU power consumption is put forward. In the method, GINI index method is introduced to construct the classifier. PowerTop tools monitor the CPU and collect the training data set, then encapsulate the classifier to the system as prediction module to predict the CPU frequency, voltage and the state of the system CPU. Through the simulation experiment shows that the method under the condition of the system load is small, the optimization of embedded CPU power optimization effect is better.