Performance Evaluation of Acceleration Structures with Stack for Ray Tracing in OpenCL

• Main Authors: ZOU, Yu-Wei, 鄒毓偉
• Other Authors: Chang, Chun-Fa
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/39547408781783473244

碩士 === 國立臺灣師範大學 === 資訊工程學系 === 105 === Ray tracing is an Algorithm that used to render picture in computer graphics. It is based on ray in physical properties and imitates the natural phenomena of ray, which makes the rendered scene becomes more realistic. However, the cost of improving realistic is must compute a large number of operations in ray tracing. In order to improve this disadvantage, OpenCL is used to perform parallel computing on multi-core platform, and the object in the scene is constructed as an acceleration structure that called Bounding volume hierarchy tree. The acceleration method can reduce rendering time. After using the OpenCL API, A large number of operations in ray tracing is parallel. However, there are some limitations programming kernel of OpenCL, especially tThe function can’t be called recursively in kernel. The program uses an acceleration structure in tree structure, so it must have the way to traverse the tree structure. OpenCL has its own implementation of the memory space, the programmer can choose which type of the memory space to simulate the array and declare the stack in the Kernel function. The purpose of this paper is to study the performance to change some parameters and implement the stack in a variety of OpenCL memory.