This implements compressed wide bvh as described in the paper by Ylitie et al.
There is almost no performance gain. I am unsure as to why that is, though it seems the triangle intersections are the current bottleneck which is either due to a low quality binary bvh in the construction process or a slow triangle intersection algorithm. Anyhow, improvements will be made in the main branch as this is now functioning properly.
This implements compressed wide bvh as described in the paper by Ylitie et al.
There is almost no performance gain. I am unsure as to why that is, though it seems the triangle intersections are the current bottleneck which is either due to a low quality binary bvh in the construction process or a slow triangle intersection algorithm. Anyhow, improvements will be made in the main branch as this is now functioning properly.