多核处理器环境下功耗感知的任务调度算法研究  

Research on Power-aware Task Scheduling Algorithm in Multi-core Processor Environment

作　　者：朱金彬 肖利民[1] 王继业 秦广军 张锐[1] 左礼威 ZHU Jinbin;XIAO Limin;WANG Jiye;QIN Guangjun;ZHANG Rui;ZUO Liwei(School of Computer Science and Engineering,Beihang University,Beijing 100191,China;Big Data Center,State Grid Corporation of China,Beijing 100031,China;Smart City College,Beijing Union University,Beijing 100191,China)

机构地区：[1]北京航空航天大学计算机学院,北京100191 [2]国家电网有限公司大数据中心,北京100031 [3]北京联合大学智慧城市学院,北京100191

出　　处：《电力信息与通信技术》2021年第3期19-25,共7页Electric Power Information and Communication Technology

基　　金：国家重点研发计划项目(2017YFB1010000)。

摘　　要：采用big.LITTLE技术的ARM架构已经成为高效能处理器领域的热点研究对象,文章基于遗传基因算法设计了一种功耗感知的实时任务调度算法。算法基于动态电压频率调整技术(Dynamic Voltage and Frequency Scaling,DVFS)动态调整处理器电压以控制处理器动态功耗,在已知任务执行开销和通信开销的情况下实现功耗优化的实时任务调度。利用伪随机任务图生成器(Task Graphs for Free,TGFF)进行了Benchmark实验,实验结果表明,与现有调度算法相比,新算法在满足任务的完成时间约束下平均降低30%的处理器功耗。ARM architecture based on big.LITTLE technology has become a hot research object in the field of high-performance processor.In this paper,a power-aware real-time task scheduling algorithm is designed based on genetic algorithm.Based on DVFS(dynamic voltage and frequency scaling)technology,the algorithm dynamically adjusts the processor voltage to control the dynamic power consumption of the processor,and realizes the real-time task scheduling with power optimization when the task execution cost and communication cost are known.In order to verify the effectiveness of the proposed algorithm,Benchmark experiments are carried out with TGFF(Task Graphs for Free).Compared with the existing optimization algorithms,the proposed algorithm reduces the power consumption by an average of 30%under the constraint of task completion time.

关 键 词：ARM big.LITTLE 处理器功耗 遗传算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]