EOFDM:一种面向众核架构的最低能耗搜索方法  

EOFDM:A Search Method for Energy-Efficient Optimization in Many-Core Architecture

作　　者：朱亚涛[1,2,3] 张帅[4] 王达[1] 叶笑春[1] 张洋[1,2] 胡九川[5] 张志敏[1] 范东睿[1] 李宏亮 

机构地区：[1]计算机体系结构国家重点实验室(中国科学院计算技术研究所),北京100190 [2]中国科学院大学计算机与控制学院,北京100049 [3]河北农业大学信息科学与技术学院,河北保定071001 [4]国家计算机网络应急技术处理协调中心,北京100029 [5]北京交通大学计算机与信息技术学院,北京100044 [6]数学工程与先进计算国家重点实验室,江苏无锡214125

出　　处：《计算机研究与发展》2015年第6期1303-1315,共13页Journal of Computer Research and Development

基　　金：国家"九七三"重点基础研究发展计划基金项目(2011CB302501);"核高基"国家科技重大专项基金项目(2013ZX0102-8001-001-001);国家"八六三"高技术研究发展计划基金项目(2015AA011204;2012AA010901);国家自然科学基金项目(61173007;61204047;61332009)

摘　　要：面向能耗优化的面积(核数)-功率(频率)分配问题是当前众核处理器研究热点之一.通过性能-功耗模型了解其在核数-频率空间的分布规律,然后在核数和频率级别这2个维度上通过实测执行逐步搜索,可以获取"核数-频率"配置的最优解,从而达到能耗优化的目的;然而本领域现有方法在核数-频率空间内实测搜索最低能耗时收敛速度慢、搜索开销大、可扩展性差.针对此问题,提出了一种基于求解最优化问题的经典数学方法——可行方向法的最低能耗搜索方法(energy-efficient optimization based on feasible direction method,EOFDM),每次执行都能从核数和频率2个维度上同时减小搜索空间,在迭代执行中快速收敛至最低能耗点.该方法与现有研究中最优的启发式爬山法(hill-climbing heuristic,HCH)进行了对比实验,平均执行次数、执行时间和能耗分别降低39.5%,46.8%,48.3%,提高了收敛速度,降低了搜索开销;当核数增加一倍时,平均执行次数、执行时间和能耗分别降低48.8%,51.6%,50.9%;当频率级数增加一倍时,平均执行次数、执行时间和能耗分别降低45.5%,49.8%,54.4%,在收敛速度、搜索开销和可扩展性方面均有提高.Based on the optimization of energy consumption, ＂area-power＂ assignment is one of research issues in many-core processors. The distribution of area-power in space of core number and frequency level can be obtained form energy-performance model. Then the progressive search for optimal solutions of ＂core number and frequency level＂ configuration can be implemented in two dimensions. However, the existing methods of searching for energy-efficient optimization have slow convergence speed and great overhead of search in the space of core number and frequency level. Moreover, though searching for optimal core number and frequency level in the space composed by an analytical energy-performance model can reduce the overhead of real execution, the accuracy of optimal solution greatly depends on the misprediction of the model. Therefore, a search method based on FDM（EOFDM） is developed to reduce the dimensions of core number and frequency, and to involve the real energy and the performance of each feasible point to correct the model computation. The experimental results show that, compared with hill-climbing heuristic（HCH） in the execution times, the performance overhead and the energy overhead, our method makes an average reduction by 39.5%, 46.8%, 48.3 %, and 48.8%, 51.6%, 50.9% in doubling the number of cores, and 45.5%, 49.8%, 54.4%in doubling the number of frequency levels. Our method is improved in convergence, search cost and scalability.

关 键 词：能耗优化 众核 最优解 可行方向法 启发式爬山法 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]