Abstract
To design an energy efficiency real-time system, it relates to chose appropriate applications running concurrently under the energy budget constraint from different application families. An application family is a set of software having common assets and sharing architectural properties. For low power design, each member in the same family is designed to deliver different quality with different energy consumption. Many software implementation strategies are available to provide different quality level. For example, programmers can configure the parameters of the compiler to generate a set of codes for the same application. That is, for the same application, we can generate a version of codes with high performance instructions, and another one with low performance instructions but consume less energy to meet specific energy budget. They are all regarded as members of the same application family. To satisfy the energy and timing constraint of the real-time system, we should chose a appropriate application from each application family. Since to choose the optimal combination of tasks from application families under timing and energy constraints is NP-hard, we propose a pseudo-polynomial time algorithm to find an optimal solution for the frame base task set and a heuristic algorithm for the general cases. Furthermore, if we consider the energy consumption of devices when applying real-time Dynamic Power Management (RT-DPM) policy, it becomes a more difficult problem. So, we propose another energy estimation equation to help the designers to estimate the energy consumption of a task when applying RT-DPM. After combing the heuristic algorithm and the energy estimation equation, the time complexity to find the solution of the problem when applying RT-DPM can be reduced significantly. Based on this algorithm, we developed a tool to provide constructive instructions in tuning system parameters for meeting a specific energy budget. We use an intelligent home case study to demonstrate its role in the low power design cycle. The experiment results shows that the tool can significantly reduce the design cycle of a low-power real-time embedded system.