[Soc-2018-dev] Weekly Report #02 - Many Light Sampling

[Stefan Werner]

Hi Erik,

> To the best of my knowledge, the construction and sampling of the light BVH are both using top-down algorithms. So hopefully, a bottom-up traversal algorithm should not be necessary. 

I believe it is necessary for multiple importance sampling. When a BRDF ray hits a light source, we need to get the probability of having selected that light source via light sampling. You can take a look at Appleseed's LightTree::evaluate_node_pdf() for reference.

-Stefan