A Method Of Render Pipeline Configuration Auto Tuning Optimization

Recently,the growing of game market advances the real-time rendering technology.Real-time rendering involves processing geometry data in GPU,shading and generating final pixels,which is called rendering pipeline.Game companies develop pipelines for their particular games.As rendering pipeline is a key factor to the program's speed and image quality,its optimization is an important research area.And there have been reaserches trying to optimize geometries,textures,shaders.Our focus is on the rendering effect configuration of the pipeline.Usually,one game engine has different configurations to let user set for different hardware capabilities.However,this procedure heavily depends on user's trials or suggestions from hardware vendors,therefore it is very inaccurate.In this paper,we introduce a method to address this problem,by finding the optimal rendering settings based on a time budget.Specifically,we first introduce a time,error multi-objective cost space,and then formulate it into an optimization problem.We use a two-step method to efficiently explore the rendering setting space to get the optimazied settings.We first partion the scene and compute pareto frontiers for specific views using generic algorithm,and use them at runtime.Finally we show the results of our method on our own OpenGL application.