Abstract: Taking full consideration of that CMOS multi-core embedded processor only provides global Dynamic Voltage Scaling (DVS) and its power leakage is serious, this paper proposed a new power-aware scheduling algorithm GRR&CS for hard real-time tasks in multi-core embedded environment. The power saving was achieved by three steps, including greedy-based static tasks partition, global resource reclamation based dynamic load balance and dynamic core scaling. The algorithm also keeps the schedulability of tasks. The experimental results show that the proposed algorithm saves more about 14.8%-41.2% energy than other existing works.

Key words: multi-core processor, hard real-time tasks, Dynamic Voltage Scaling (DVS), task partition, task migration, core scaling

摘要： 充分考虑当前CMOS多核嵌入式处理器片上仅提供全局动态电压缩放(DVS)支持以及亚纳米时代后CMOS处理器泄露功耗不可忽视的现状，提出一种新的多核嵌入式环境中的硬实时任务感功调度算法GRR&CS。算法通过基于贪心法的静态任务划分，基于全局资源回收利用和任务迁移的动态负载均衡，以及动态核缩放三个步骤实现整体能耗的降低，并同时保证实时任务的可调度性约束。实验表明，提出的算法相比较现有算法多节省14.8%~41.2%的能耗。

关键词: 多核处理器, 硬实时任务, 动态电压缩放, 任务划分, 任务迁移, 核缩放