摘要
游戏是移动设备上功耗很高的一类应用,降低游戏功耗对提升移动设备的续航时间有重要意义.本文在Cocos2d-x-3.2游戏引擎上分析游戏画面与GPU渲染时间之间的关系,提出一种计算模型预测每一帧画面的GPU渲染时间,并利用该模型实现一种DVFS算法——"onframe"算法来降低GPU的功耗."onframe"算法根据每一帧的预测渲染时间和游戏的设计帧率,计算满足每一帧性能需求的最优GPU频率.最后通过实验将"onframe"算法与Android系统提供的系统级GPU功耗优化策略进行比较,表明"onframe"算法可以更有效地降低GPU功耗并且不对游戏性能产生明显影响.
Games are high-power applications for mobile devices, so that reducing their power consumption is important for enhancing mobile devices' battery life. This paper analyzes the relationship between game screen and corresponding GPU rendering time on the Cocos2d-x-3.2,and proposes a model to estimate the GPU rendering time of each frame. Based on this model, a DVFS algorithm named "onframe" is implemented to reduce the power consumption of GPU. The "onframe" algorithm calculates the best GPU frequency according to the predicted rendering time and the fps of game. Experiments illustrate that,compared to the system-level power optimization strategy in android GPU driver,the "onframe" algorithm can reduce GPU power consumption more efficiently without any significant impact on game performance.
出处
《小型微型计算机系统》
CSCD
北大核心
2016年第5期1112-1116,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61379040
61272131)资助
江苏省自然科学基金项目(SBK2012194)资助