针对以往实时低功耗策略只考虑单一任务的概率分布,而没有考虑任务间耦合关系的不足,本文提出了一种基于系统负载概率分布的实时低功耗算法(Frequency Adjustment to Reduce Power,FARP)。FARP算法分两个步骤:(1)根据系统负载的概率分布是正在运行任务概率分布的卷积的结论,在每个任务释放的时候计算当前系统负载的概率分布;(2)根据离线状态下任务释放时系统负载最长的运行时间,并结合系统负载的概率分布,得到系统负载所需的频率分配,并获得其最低的统计功耗。另外,FARP算法根据实际情况作了一定的修正以满足应用的需要。实验结果表明,FARP算法与同类算法相比,至少可以降低30%的功耗,同时可以满足系统实时性的要求。 For the past probability of real-time low-power policy is only considered a single task distribution, but no less than the coupling relationship between tasks considerations, we propose a real-time low-power algorithm based on system load probability distribution (Frequency Adjustment to Reduce Power, FARP) . FARP algorithm in two steps: (1) the distribution of the convolution operation task is concluded that the probability of the probability distribution of the system load, the probability distribution of the current system load is calculated for each task at the time of release; (2) according to the offline When the task is released, the longest running time of the system is loaded. With the probability distribution of the system load, the required frequency allocation of the system load is obtained and the lowest statistical power consumption is obtained. In addition, FARP algorithm according to the actual situation made some amendments to meet the needs of the application. Experimental results show that compared with similar algorithms, FARP algorithm can reduce power consumption by at least 30% and meet the real-time requirements of the system.