移动平台游戏应用功耗优化

发布时间：2018-03-07 13:47

本文选题：游戏状态　切入点：功耗优化　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：目前,低功耗已经成为计算机体系结构的重要研究主题。特别是在移动平台上,功耗问题已经成为制约移动平台发展的瓶颈。游戏应用占据移动应用超过50%的下载量,且由于其交互密集特性或计算密集特性或图形密集特性,往往比其他应用更耗功耗。为了更加有效的在游戏场景下降低移动平台的功耗,学术界提出了不少针对游戏应用的功耗优化研究,但其中针对交互特性的研究非常有限。本文从游戏应用的特点出发,结合游戏应用需要与用户进行交互这一特点,引入了游戏状态这一信息来指导移动平台下不同物理部件的功耗优化。同时,由于目前针对移动平台的相关基准程序没有体现用户交互,因此,本文引入用户交互,提出了一个游戏专用的基准程序。论文的主要工作包括:1.根据游戏程序执行的不同阶段具有不同的用户交互特性以及不同的用户体验和性能需求,定义了不同的游戏状态。为了能够在线的检测闭源游戏的不同游戏状态,提出和实现了基于特征和基于分类算法的两种游戏状态检测方法,并比对了它们的准确率和开销。2.提出了基于帧率自调节的CPU功耗管理算法。不同游戏和不同游戏状态具有不同的性能需求。该算法实时检测游戏的帧率需求,动态调整CPU频率,实现高效的CPU功耗管理。3.提出了基于游戏状态的动态背光调整策略。目前游戏过程中屏幕亮度采用静态背光值,而实际上不同游戏状态下用户有不同的性能需求。该策略通过权衡背光变化对用户体验的影响和背光变化对屏幕功耗的影响,在不影响用户体验的情况下,尽量的降低屏幕功耗。4.提出了基于用户交互特征的测试基准程序PUBench。基于本文提出的移动平台要素剖析框架FPA(Factors Profiling Architecture),PUBench以排队论为基础模拟用户交互行为,并对APP-OS交互和CPU-GPU交互进行量化分析。实验表明,PUBench很好地展现了移动平台的人机交互、负载和功耗管理特征,消除了"测试者依赖性"。最后,给出PUBench的一个使用案例。
[Abstract]:At present, the low power consumption has become an important research topic in computer architecture. Especially in the mobile platform, the problem of power consumption has become a bottleneck restricting the development of mobile platforms. The game application downloads occupy more than 50% of mobile applications, and because of its interaction characteristics or properties or intensive computing intensive graphics intensive characteristics, often more than other power consumption application. In order to more effectively reduce the power consumption of mobile platform in the game scene, the academic circles have put forward many applications for game power optimization research, but the research on the characteristics of the interactions is very limited. In this paper, from the characteristics of the game application of game applications need to interact with the user and the characteristics, the introduction of this information the game state to guide the power optimization of different physical components of the mobile platform. At the same time, as the benchmark for the mobile platform has no body The user interaction, therefore, this paper introduces user interaction, proposed a game dedicated benchmark program. The main work includes: 1. according to the different stages of the game program has different characteristics and different user interactive user experience and performance requirements, define a different game state. In order to detect online closed source different game game state, is proposed and realized based on the characteristics and two kinds of game state detection method based on classification algorithm, and compare their accuracy and overhead.2. proposed CPU power consumption management frame rate self regulating algorithm based on different games and different games have different performance requirements. The algorithm of real-time detection of the game the rate of demand, adjust CPU frequency, to achieve efficient CPU power management.3. proposed a dynamic backlight adjustment strategy based on the state of the game. The course of the game screen Screen brightness backlight using static value, but in fact different game state users have different performance requirements. This strategy by weighing the change on the screen backlight power consumption affects the user experience and the backlight changes, without affecting the user experience, try to reduce the power consumption of the.4. screen based on interactive feature test the benchmark PUBench. mobile platform elements based on the analysis framework of FPA (Factors Profiling Architecture), PUBench based on the queuing theory simulation of user interaction, and the APP-OS interaction and CPU-GPU interaction for quantitative analysis. Experimental results show that the PUBench is very good to show the mobile platform for human-computer interaction, load and power management features, eliminate "the test of dependence. Finally, a case of PUBench is given.

【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP332

【参考文献】