With advancements in computing and communication technologies on mobile devices, the performance requirements of embedded processors have significantly increased, resulting in a corresponding increase in its energy consumption. Dynamic scaling of operating voltage and operating frequency has a strong correlation to energy minimization in CMOS real-time circuits. Simultaneous optimization of (v, f) pairs under dynamic activity levels is thus extensively investigated over several years. The supply voltage is tuned dynamically during runtime (DVS), with a fixed threshold voltage, to achieve energy minimization. This work addresses the issue of maximizing the energy efficiency of real-time periodic, aperiodic and mixed task sets, in a uniprocessor system, by developing a novel task feasibility methodology, with a novel processor performance-based constraint, to generate the optimal operating supply voltage to the individual task of task sets. The energy minimization of real-time mixed task sets is formulated as Geometric Programming (GP) problem, by varying frequency for periodic tasks sets and keeping fixed frequency for aperiodic tasks set, over a range of task sets and hence computing optimal operating voltages. Simulation experiments show energy savings on the cumulative basis of 50%, 38% and 29% for periodic, aperiodic and mixed task sets, respectively, based on the processing timing constraints of task sets.

中文翻译：

---

DVS-enabled单处理器系统实时混合任务集能量最小化的新可行性测试

随着移动设备上计算和通信技术的进步，嵌入式处理器的性能要求显着提高，导致其能耗相应增加。工作电压和工作频率的动态缩放与 CMOS 实时电路中的能量最小化有很强的相关性。同时优化 (v,F)因此，在动态活动水平下的配对在几年内得到了广泛的研究。电源电压在运行期间 (DVS) 动态调整，具有固定的阈值电压，以实现能量最小化。这项工作解决了在单处理器系统中最大化实时周期性、非周期性和混合任务集的能量效率的问题，通过开发一种新的任务可行性方法，具有新的基于处理器性能的约束，以产生最佳的操作供应电压给任务集的单个任务。实时混合任务集的能量最小化被表述为几何规划（GP）问题，通过在一系列任务集上改变周期性任务集的频率并为非周期性任务集保持固定频率，从而计算最佳工作电压。