Ray tracing is a powerful rendering technique capable of producing photorealistic images. However, it suffers from high computational costs, particularly for intersection testing. To address this issue, this study evaluates the performance of ray tracing implemented in three different environments: a CUDA-GPU environment, a GLSL-GPU environment, and an OpenMP-CPU environment. The computation time was measured, and the results showed that the CUDA-GPU implementation was the fastest, followed by the GLSL-GPU implementation. The OpenMP-CPU implementation was up to 93.3 times slower than the CUDA-GPU implementation. These results indicate that the high degree of parallelism in GPUs is extremely effective for the ray tracing process.
Matsumoto et al. (Wed,) studied this question.