使用OpenCL實現蒙地卡羅路徑追蹤於Android手機平台

Abstract

隨著科技的發展手機的進步速度越趨明顯，手機遊戲也變成開發手機產品中一個不可或缺的角色。隨著手機晶片上的運算能力快速進步，讓手機遊戲可以從2D快速進入到3D的手機遊戲時代,而相對於畫面品質和效能也是越來越苛求。為了使3D場景的繪製達到更逼真的畫質，至今已經發展了許多不同的演算法以模擬真實環境的光照情形，其中以光線追蹤方法擁有物理運算的正確性並能夠直覺性地計算光照效果，但因光線追蹤的運算速度相對於其他手機遊戲渲染效果是較慢的所以幾乎沒有人將其運用於手機遊戲當中。 本研究搭配OpenCL異質系統架構中平行且良好地運作之特性，對於光線追蹤龐大的計算下平行處理能讓畫面的渲染速度更有效率且畫面品質將有不錯的效果。本渲染對象以動態場景為主，會分別對於場景的複雜度和其渲染速度上做效能分析，將重點著重於可行性的探討與提升程式效能，針對所使用的平台進行實驗結果比較以及影響效能的相關探討，最後，我們將因應平台特性將演算法與之結合，試圖找出在多核心平行化架構下提升各平台程式效能的實驗依據並以其研究出適合於手機平台上的光線追蹤演算法架構。

With development of mobile chip computing technique increasingly, it enhances the growth of a range of related to frame quality especially mobile game. It progresses the mobile game transform from two-dimensional space generation to three-dimensional (3D) space game. Graphics Processing Unit (GPU) of mobile phone offers powerful computing ability, it is applied to 3D scene compute. Majority of 3D game scene are not making user to experience realistic. In order to improve 3D sense realistic, that developed a number of different algorithms to compute ray in the scene. This research focus on ray tracing algorithm. Ray tracing uses mathematical formulas, it can calculate lighting realistically and correctly. However ray tracing need a lot of compute, it spends too much time to render. If ray tracing compare with others mobile’s render technique, it will too slowly. This research uses OpenCL to accelerate ray tracing in mobile phone. It will make scene not only realistically but also efficiently. We will use other platform to experiment our research, and find out good ray tracing algorithm for mobile phone.