基于CUDA的渐进式光子映射并行渲染实现

Implementation of Progressive Photon Mapping Parallel Rendering Based on CUDA

对于复杂场景的高质量图像渲染，可能需花费几个小时，而使用图形处理器（GPU）来加速渲染进程是一种很好的选择。针对该问题，通过修改渐进式光子算法的实现过程，使得该算法能够通过统一设备计算架构（CUDA）和光线追踪引擎OptiX完全运行在图像处理器（GPU）上；从而充分利用GPU强大的并行计算能力，加速光子映射的实现。并提出了渐进式光子映射的分布式渲染实现方法，同时使用多个GPU高效率地执行改进的光子映射算法。实验结果证明了采用分布式系统中6个GPU进行渲染，经过1000次迭代，加速比提高到5．7，得到接近线性的加速。

It took a few hours to render high-quality images in complex scenes. So, it was a good choice using Graphics Processing Unit （GPU） accelerate the rendering process. We modified the implementation process of progressive photon algorithm, and let the algorithm runs entirely in the GPU by Compute Unified Device Architecture （CUDA） and ray tracing engine OptiX. So, we could take full advantage of the powerful parallel computing capabilities of GPU to accelerate the photon mapping implementation. Then we proposed the distributed rendering implementation of progressive photon mapping, while executing the improved progressive photon mapping implementation algorithms using multiple GPUs. The results show that the speedup increased to 5.7 after 1000 iterations rendering in six GPUs of distributed system, and it gets close to linear acceleration.