面向代码的软件能耗优化研究进展

Research Advance on Code Oriented Optimization of Software Energy Consumption

面向代码的软件能耗优化从程序设计和编码角度优化软件系统的能耗,能够很好地弥补面向硬件和面向资源的能耗优化方法过多依赖硬件环境、普适性较差、粒度过大,且难以在软件开发过程中应用等缺点.该文综述了近年来面向代码软件能耗优化领域的主要研究成果,总结了能耗优化的基本方法和技术层次以及面向代码的软件能耗优化基本思路;随后从面向代码的软件能耗估算方法和优化方法两个方面对现有工作加以梳理,逐一介绍了相关优化工具.该文提出了若干进一步研究的问题.首先,该文重定义面向代码的软件能耗评估模型和方法应该满足的特性,并提出代码的运行时能耗(Runtime Energy Consumption)和视在能耗(Apparent Energy Consumption)的概念;其次,该文认为现有面向代码的能耗优化技术过于具体,或针对具体的代码,或针对具体的功能,缺乏抽象层次的优化技术,缺乏算法层面的能耗优化方法,且没有充分考虑编程语言特征,尤其是面向对象语言特征;最后,该文提出算法能耗复杂度这一新观点,指出仍然存在的问题和可能的解决办法.

Code oriented optimization of software energy consumption focuses on aspects of programming and coding.It is a good complement to the hardware oriented and resource oriented optimizations because the latter two have the disadvantages of hardware environment depended,weak universality,coast granularity and inapplicable in the software development processes.In this paper,the primary research achievements of code oriented energy consumption optimization are summarized.Firstly,we introduce the basic approaches of energy consumption optimization,the layers of optimization techniques,and the basic principles of code oriented optimization.Then,the related works are explained from two aspects:estimation and optimization.Moreover,challenges in both aspects are reviewed as well as future research trends are predicted.In the paper,we propose several challenges for the further study.First,we redefine the characters of code oriented estimation model and approaches for software energy consumption,then propose the concepts of runtime energy consumption and apparent energy consumption of codes.Second,we think the existing code oriented energy consumption optimization approaches are more specific,andthey focus on fixed codes or functions.There is few abstract-level and generalized optimizations,few algorithm or optimizations,and few optimizations which taken the features of programming language,especially the object oriented language,into consideration.Finally,we propose the new challenge of this research topic,which is energy consumption complexity of algorithm,and then,the existing problems and possible solutions,which can be referred by researchers,are given.