| Along with the development of computer system, low-power techniques have been given more and more attention. Low-power optimization for real-time systems is one of the important low-power research domains. Furthermore, dynamic voltage scaling is one of the effective low-power techniques for real-time systems. Based on the result of the WCET (Worst Case Execution Time) analysis, assisted with software such as the operating system or the compiler, real-time dynamic voltage scaling makes use of the dynamic variability of the execution time of a program, and effectively reduces the power consumption of real-time system.In this paper, we investigate the static WCET analysis technique, and using the technique, we improve the energy efficiency of some real-time applications. The static WCET analysis is the basis of real-time dynamic voltage scaling, and real-time dynamic voltage scaling is an important application of the static WCET analysis in the low power optimization domain.In this paper, we firstly study two WCET analysis methods (control flow graph based WCET analysis method and syntax tree based WCET analysis method), and get integrated comprehension on the static WCET analysis.Then, we propose a low-power optimization algorithm based on WCET analysis. The algorithm inserts the dynamic voltage scaling code before the condition statements and after the loop statements, which affect the program execution time. The algorithm makes full use of the slack time for the running program and effectively reduces system power consumption.Finally, we detailedly analyze two different WCET analysis tools, i.e. Cinderella and HEPTANE, and compare them with each other. From the analysis of the two tools, we choose HEPTANE as the timing system and develop a low-power experiment environment for real-time systems named RTLPower(Real-Time Low-Power). Using the RTLPower, we do some simulations, and the simulation results show that the proposed algorithm can reduce up to 50 percent energy consumption. |
PDF Full Text Request