关键词: 动态电压缩放, 节能, 动态规划, 可变电压处理器

Abstract:

For battery-based embedded systems, how to save energy and extend the systems' life time is one of the most important problems. Existing research work considers the ideal processors that have a continuous variable voltage. However, practical processors only have discrete voltage levels. Therefore, for hard real-time embedded applications runing on practical dynamic voltage-scaling processors, this paper proposes an optimal voltage selection algorithm to obtain the minimum energy consumption of these applications without violating the given time constraint. Different from those existing heuristic algorithms, the energy-aware voltage selection problem is first formulized as a variation of the multiple-choice knapsack problem, and then an algorithm based on the dynamic programming technique is proposed to solve this problem. By comparing several voltage selection strategies on real embedding systems, experimental results show that our algorithm can obtain the minimum energy consumption without violating the given time constraint of the emmbedding application.

Key words: dynamic voltage scaling (DVS), energy efficiency, dynamic programming, variable voltage processor