标题: 异质双核心嵌入式即时系统之有效低能源消耗排程演算法
An Efficient Low Power Scheduling for Heterogeneous Dual-Core Embedded Real-Time Systems
作者: 林柏君
Pochun Lin
王国祯
Kuochen Wang
网路工程研究所
关键字: 动态电压调整;嵌入式即时系统;异质双核心;低能源消耗排程;总执行时间;Dynamic voltage scaling;embedded real-time system;heterogeneous dual-core;low power scheduling;total execution time
公开日期: 2007
摘要: 近年来,异质双核心嵌入式即时系统,如个人数位助理及手机越来越受到欢迎。为了达到即时以及低能源的耗损,低能源消耗排程成为一个值得重视的研究课题。大多数现有低能源消耗排程的动态电压调整演算法,其目标多在单一CPU或是在同质性的多核心系统上。本论文中,在异质性的双核心嵌入式即时系统上,我们提出一个较长共同执行时间(LCET)演算法,以适用在具动态电压调整能力的异质双核心。本演算法包含两个阶段:第一,运用所提出的排班演算法来缩短在异质双核心嵌入式即时系统上的工作总执行时间;第二,我们更进一步探讨利用缩短过后的总执行时间来调整电压及频率以达到低能量的耗损。模拟结果显示,相较于Kim et al. 方法,在使用(不使用)动态电压调整之下,我们所提出的方法分别在可插队的LCET (P-LCET)及不可插队的LCET (NP-LCET),各节省了8%及16% ~ 25% (13%及33% ~ 38%)的总能量耗损。
In recent years, heterogeneous dual-core embedded real-time systems, such as personal digital assistants (PDAs) and cellular phones, have become more and more popular. In order to achieve real time performance and low energy consumption, low power scheduling becomes a critical issue. Most researches on low power scheduling with dynamic voltage scaling (DVS) were targeted at only one CPU or homogeneous multi-core systems. In this thesis, we propose a low power scheduling algorithm called Longer Common Execution Time (LCET) for DVS enabled heterogeneous dual-core embedded real-time systems, which includes two steps. First, we reduce total execution time of tasks by using LCET in heterogeneous dual-core embedded real-time systems. Second, we further exploit the reduced total execution time to adjust voltage and frequency levels in order to reduce the total energy consumption. Simulation results show that the proposed P-LCET (a preemptive version) and NP-LCET (a non-preemptive version) can effectively reduce the total energy consumption by 8% and 16% ~ 25% (13% and 33% ~ 38%) compared with the work by Kim et al. with (without) dynamic voltage scaling.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009556527
http://hdl.handle.net/11536/39622
显示于类别:Thesis


文件中的档案:

  1. 652701.pdf

If it is a zip file, please download the file and unzip it, then open index.html in a browser to view the full text content.