当前位置:主页 > 科技论文 > 计算机论文 >

基于多核处理器的节能调度算法研究

发布时间:2018-08-26 18:14
【摘要】:近年来,处理器的能耗问题日益突出,单处理器到多核处理器,虽然在性能上有所提高,但同时导致了更多的能量消耗,不仅引发了难以解决的散热问题,还导致系统的可靠性下降。因此,在满足时间约束条件下,如何尽可能的降低系统能耗,已经成为多核处理器实时节能调度领域的研究热点。 已有算法的研究重点在于如何充分利用任务的实际执行时间α,远小于最坏情况执行时间ωi所产生的动态空闲时间以及任务在时限之前完成而产生的静态空闲时间,设计速率调度算法来合理降低处理器核电压/频率,达到同时满足硬实时任务时限约束与最少能耗的统一目标。本文在利用空闲时间降低速率的同时,尽量减少处理器切换电压/频率所造成的能耗。 本文的主要工作如下: (1)提出一种新的多核处理器节能调度算法,该方法基于空闲时间回收策略和尽量减少处理器频率切换造成的能耗。首先获得前面执行任务的平均任务执行比率,来估算下一个任务的执行速率,从而将其映射到速率最相近的处理器上,以减少处理器改变速率时所消耗的能耗。并通过让相邻的两个任务共享处理器上的空闲时间来降低处理器的执行速率,从而达到减少能耗的目的。通过理论推导和分析,将新算法与近三年来的算法进行对比测试,结果表明在大部分条件下,新算法能够获得更好的节能效果。 (2)根据大量测试数据,深入分析了影响算法节能效率的因素以及算法的调度约束条件和算法的适用条件,进而提出了三种改进算法TR-SS2(负载均衡策略)、ESR-SS2-CE和ESR-SS2-M,增加对速度的约束条件,分别采用估算速度和最大速度来确定单处理器上的执行速度。并将改进算法与已有算法对比测试分析,结果表明改进算法能够取得更好的节能效果。
[Abstract]:In recent years, the problem of processor energy consumption has become increasingly prominent. Although the performance of single-processor to multi-core processor has been improved, it has also led to more energy consumption, which has not only led to the difficult problem of heat dissipation. It also reduces the reliability of the system. Therefore, how to reduce the system energy consumption as much as possible has become a hotspot in the field of real-time energy-saving scheduling of multi-core processors. The emphasis of the existing algorithms is how to make full use of the actual execution time 伪, which is far less than the dynamic idle time generated by the worst case execution time 蠅 I and the static idle time generated by the completion of the task before the time limit. A rate scheduling algorithm is designed to reasonably reduce the core voltage / frequency of the processor and achieve the goal of simultaneously satisfying the time limit of hard real-time task and the minimum energy consumption. In this paper, the energy consumption caused by switching voltage / frequency of processor is minimized while the idle time is used to reduce the rate. The main work of this paper is as follows: (1) A new multi-core processor energy-saving scheduling algorithm is proposed, which is based on idle time recovery strategy and minimizes the energy consumption caused by processor frequency switching. First, the average task execution ratio of the previous task is obtained to estimate the execution rate of the next task and map it to the processor with the most similar rate to reduce the energy consumption when the processor changes the rate. The idle time on the processor is shared by two adjacent tasks to reduce the execution rate of the processor, thus reducing the energy consumption. Through theoretical deduction and analysis, the new algorithm is compared with the algorithm in recent three years. The results show that under most conditions, the new algorithm can achieve better energy saving effect. (2) according to a large number of test data, The factors that affect the efficiency of the algorithm, the scheduling constraints of the algorithm and the applicable conditions of the algorithm are analyzed in depth. Then, three improved algorithms, TR-SS2 (load balancing strategy), ESR-SS2-CE and ESR-SS2-M, are proposed to increase the speed constraints. The estimated speed and the maximum speed are used to determine the execution speed on the single processor. The results show that the improved algorithm can achieve better energy saving effect.
【学位授予单位】:武汉理工大学
【学位级别】:硕士
【学位授予年份】:2012
【分类号】:TP301.6;TP332

【引证文献】

相关硕士学位论文 前1条

1 刘宝宁;基于多核平台的实时混合任务调度算法研究[D];武汉理工大学;2013年



本文编号:2205749

资料下载
论文发表

本文链接:https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2205749.html


Copyright(c)文论论文网All Rights Reserved | 网站地图 |

版权申明:资料由用户5addc***提供,本站仅收录摘要或目录,作者需要删除请E-mail邮箱bigeng88@qq.com