Comparison of BVH and KD-Tree for the GPGPU Acceleration on Real Mobile Devices

摘要

Portability and thermal issues significantly limit the performance of mobile devices, hindering the use of the global illumination that enables high-quality rendering. For a long time, the BVH and KD-tree algorithms through GPU acceleration have been widely used to accelerate global illumination in the desktop environment. However, these algorithms have not been sufficiently studied in the mobile environment, and therefore, the use of either of them presents many challenges. This paper describes the implementation of BVH and KD-tree algorithms that utilize GPU acceleration on mobile devices. The experimental results show that the path tracing performance can be increased 2.3 times on average when KD-tree is used and that KD-tree can improve the performance more than BVH. To the best of our knowledge, this study is the first to examine BVH and KD-tree in the real mobile environment.