Abstract
There are more and more portable devices nowadays. Portable devices such as notebooks、PDAs and cell phones are powered by battery. Customers care not only the function and the appearance of products, but also the battery life. Therefore, designing power saving mechanism is an important issue. The power saving mechanisms can be categorized in two dimensions. The first one is the hardware solution which making use of the material that consumes less energy or simplifying the number of components. The other one is the software approach. If the hardware supports multiple voltage levels, we can execute some operations with lower voltage to reduce energy consumption. Furthermore, the overhead produced by voltage switching should be as low as possible. Advanced microprocessors and flash memory chips support multiple voltages now. The proportion of energy consumption for operations of read/write flash memory to total energy consumption is very high. As the increasing of the size of flash memory, it’s important to design a power saving mechanism for flash memory. For embedded systems, moreover, the real-time requirement is also a critical issue. Hence, a new approach should satisfy not only the power saving but also the real-time requirements. In this thesis, we discuss the power saving problem from the viewpoint of imprecise computation. We adopt the data structure using in imprecise computation to speed up the scheduling, and proposed two algorithms. The simulation results show that our proposed scheduling algorithms perform better than Chang’s algorithm and consume less power.