摘要
针对射线追踪过程中,由于射线数目巨大、部分目标场景复杂,造成计算效率低下的问题,采用线索KD-Tree (K-dimensional tree)空间加速算法,将目标场景进行有序组织,通过对线索KD-Tree进行无堆栈遍历,加快射线与目标场景求交的计算速度。为解决传统方法中,串行计算射线与目标求交过程中造成待遍历射线多的问题,采用图形处理器(graphics processing unit, GPU)在统一计算设备架构(compute unified device architecture, CUDA)平台下并行处理所有射线,加快计算速度。实例仿真计算结果表明,基于线索KD-Tree的射线追踪并行计算相比于串行计算,计算效率提高,获得了很好的加速效果。
In the process of ray tracing,the huge number of rays and the complexity of some target scenes result in low computational efficiency.The rope KD-Tree spatial acceleration algorithm was used to organize the target scenes in an orderly way,and the computation speed of intersection between rays and target scenes was accelerated by traversing the rope KD-Tree without stack.To solve the problem that there are too many rays to be traversed during the intersection between the serial computed rays and the target in the traditional method,GPU was used to process all the rays in parallel under CUDA platform to speed up the computation.Simulation results show that the parallel ray tracing computation based on rope KD-Tree is more efficient than the serial computation and has good acceleration effects.
作者
厉夫兵
苏永琪
陈文剑
LI Fu-bing;SU Yong-qi;CHEN Wen-jian(School of Information and Communications Engineering,Beijing Information Science and Technology University,Beijing 100101,China;College of Underwater Acoustic Engineering,Harbin Engineering University,Harbin 150001,China)
出处
《计算机工程与设计》
北大核心
2023年第12期3677-3682,共6页
Computer Engineering and Design
基金
北京信息科技大学“勤信人才”培育计划基金项目(QXTCP C201905)
国家自然科学基金项目(12074089)。
