A Vectorized Traversal Algorithm for Ray Tracing

This article presents an optimized ray tracing algorithm which improves standard existing algorithms by processing simultaneously a large set of rays and carrying out a single traversal of the spatial indexing of the scene with all rays. This allows hardware SIMD functionality to be used efficiently and produces coherent memory accesses. Furthermore, during the single traversal, rays are grouped in such a way that these advantages are maintained for non coherent ray sets. The algorithm was observed to reduce the computation costs with respect to other standard solutions, especially for non coherent ray sets. It is worth noting that its characteristics make it especially suitable for graphics hardware.