登录 | 注册 | 充值 | 退出 | 公司首页 | 繁体中文 | 满意度调查
综合馆
一种基于并行度分析模型的GPU功耗优化技术
  • 摘要

    随着硬件功能的不断丰富和软件开发环境的逐渐成熟,GPU开始被应用于通用计算领域,协助CPU加速程序的运行.为了追求高性能,GPU往往包含成百上千个核心运算单元.高密度的计算资源,使得其在性能远高于CPU的同时功耗也高于CPU.功耗问题已经成为制约GPU发展的重要问题之一.DVFS技术被广泛应用于处理器的低功耗优化,而对GPU进行相应研究的前提是对其程序运行过程进行分析和建模,从而可以根据应用程序的特征来确定优化策略.此外,GPU主要由图形处理器芯片和片外的DRAM组成,有研究指出针对这类系统的功耗优化应当综合考虑处理器和存储器,使二者可以互相协调以达到更好的优化效果.文中在一个已有的基于程序并行度分析的GPU性能模型的基础上,综合考虑计算部件与存储部件的功耗,建立了性能约束条件下的GPU功耗优化模型.对于给定的程序,在满足性能约束的前提下,以功耗最优为目标分别给出处理器和存储器的DVFS优化策略.作者选取了9个测试用例在3种模拟平台上进行了实验验证,结果表明文中的方法可以在满足性能约束条件10%的误差范围内获得最优的GPU能量消耗.

  • 作者

    林一松  杨学军  唐滔  王桂彬  徐新海  LIN Yi-Song  YANG Xue-Jun  TANG Tao  WANG Gui-Bin  XU Xin-Hai 

  • 作者单位

    国防科学技术大学并行与分布处理国家重点实验室,长沙,410073

  • 刊期

    2011年4期 ISTIC EI PKU

  • 关键词

    GPU  并行度模型  功耗模型  功耗优化 

参考文献
  • [1] Simunic T;Benini L;De Micheli G. Cycle-accurate simulation of energy consumption in embedded systems. Atlanta,Georgia, 1999
  • [2] Govil K;Chan E;Wasserman H. Comparing algorithms for dynamic speedsetting of a low-power CPU. Claremont Hotel,Berkeley,California,USA, 1995
  • [3] Lorch J;Smith A. Improving dynamic voltage algorithms with PACE. Cambridge,Massachusetts,USA, 2001
  • [4] Hsu C;Kremer U;Hsiao M. Compiler-directed dynamic voltage/frequency scheduling for energy reduction in microprocessors. Huntington Beach,CA, 2001
  • [5] Hsu Chung-Hsing;Kremer U. The design,implementation,and evaluation of a compiler algorithm for CPU energy reduction. San Diego,California,USA, 2003
  • [6] Collange S;Defour D;Tisserand A;Power consumption of GPUs from a software perspective. . Berlin,Heidelberg:Springer Verlag, 2009
  • [7] Huang S;Xiao S;Feng W. On the energy efficiency of graphics processing units for scientific computing. Rome,Italy, 2009
  • [8] Rofouei Mahsan;Stathopoulos Thanos;Ryffel Sebi;Kaiser William,Sarrafzadeh Majid. Energy-aware high performance computing with graphic processing units. San Diego,California,USA, 2008
  • [9] Weiser M;Welch B;Demers A;Shenker S. Scheduling for reduced CPU energy. Monterey,California,USA, 1994
  • [10] . http://ati.amd.com/technology/streamcomputing/product_FireStream_9250.html
  • [11] Takizawa H;Satol K;Kobayashi H. SPRAT:Runtime processor selection for energy-aware computing. Tsukuba International Congress Center,Epochal Tsukuba, 2008
  • [12] Luebke D;Harris M;Govindaraju N;Lefohn A,Houston M,Owens J,Segal M,Papakipos M,Buck I. GPGPU:General-purpose computation on graphics hardware. Tampa,Florida, 2006
  • [13] Ma Xiaohan;Dong Mian;Zhong Lin;Deng Zhigang. Statistical power consumption analysis and modeling for GPUbased computing. Big Sky Resort in Big Sky,Montana, 2009
  • [14] Fan Xiaobo;Ellis Carla S;Lebeck Alvin R. The synergy between power-aware memory systems and processor voltage scaling. New York,USA, 2003
  • [15] Hong S;Kim H. An analytical model for a GPU architecture with memory-level and thread-level parallelism awareness. Austin,Texas,USA, 2009
  • [16] NVIDIA Corporation. CUDA Programming Guide,Version 2.1.
  • [17] Burd T;Brodersen R. Design issues for dynamic voltage scaling. Rapallo,Italy, 2000
  • [18] Bakhoda Ali;Yuan George;Fung Wilson W L;Wong Henry,Aamodt Tor M. Analyzing CUDA workloads using a detailed GPU simulator. Boston,Massachusetts, 2009
  • [19] Brooks D;Tiwari V;Martonosi M. Wattch:A framework for architectural-level power analysis and optimizations. Vancouver,British Columbia,Canada, 2000
  • [20] Ramani K;Ibrahim A;Shimizu D. PowerRed:A flexible modeling framework for power efficiency exploration in GPUs. New York,USA, 2007
查看更多︾
相似文献 查看更多>>
34.238.192.150