• DocumentCode
    2985381
  • Title

    An Embedded Software Fusion Power Model Based on Algorithm and Micro-architectural Level

  • Author

    Guo, Bing ; Shen, Yan ; Liu, XiaoBin ; Li, Qi ; Wang, JiHe ; Wu, YuanSheng

  • Author_Institution
    Coll. of Comput. Sci., SiChuan Univ., Chengdu, China
  • fYear
    2011
  • fDate
    3-4 Dec. 2011
  • Firstpage
    431
  • Lastpage
    435
  • Abstract
    According to the power model based on algorithm complexity, we analyze the power characteristic on both the algorithm level and micro-architectural level. Therefore, we propose a fusion power model that combines both algorithm level and micro-architectural level. We extract the time complexity and space complexity on the algorithm level, the CPI(Cycles Per Instruction) on the micro-architectural level as the variables of this fusion model. We use HMSim, a high accurate power simulator based on ARM7TDMI instruction set, as the experimental platform. After measuring the power of the some selected benchmarks from HMSim, we develop a linear regression method to get the fusion model´s coefficient. Simulation results show that the power is a linear function of both time complexity and CPI of a algorithm, but no direct relations with the space complexity of the algorithm, the relative error is below 4%, which proves the accurate of this fusion model.
  • Keywords
    computational complexity; instruction sets; regression analysis; software architecture; ARM7TDMI instruction set; CPI; HMSim; algorithm complexity; cycles per instruction; embedded software fusion power model; linear regression method; microarchitectural level; power characteristic; space complexity; time complexity; Complexity theory; Computational modeling; Embedded software; Mathematical model; Power demand; Software algorithms; CPI; algorithm complexity; fusion power model; software power consumption;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Computational Intelligence and Security (CIS), 2011 Seventh International Conference on
  • Conference_Location
    Hainan
  • Print_ISBN
    978-1-4577-2008-6
  • Type

    conf

  • DOI
    10.1109/CIS.2011.102
  • Filename
    6128061