Abstract: A load-balancing algorithm based on greedy strategy for multi-node parallel ray tracing was proposed to improve the parallelism degree of applying ray tracing cluster to producing high-definition pictures of complex scenes. Firstly, according to parallel property of GPU, the screen space was divided into a number of square image blocks. Based on the projection of the spherical bounding box of moving objects on the screen, the binary rendering time influence map was constructed. Then on the basis of the spatiotemporal correlation, the rendering task queue was established by combining the time consumed by the previous image block and the binary rendering time influence map. Then the dynamic allocation of multi-rendering node tasks was realized through two-step load-balancing. Finally, the method was verified and analyzed based on experimental data. The results showed that this method presents good load-balancing effect, and the rendering efficiency can be improved by up to 4.96 times.

Key words: ray tracing, high-definition picture, load balancing, multi-node rendering, greedy strategy