ompf2.com

Re-reading your first comment, you say that another light may be visible when sampling the sky. When you select the sky, you have to shoot a shadow ray just like any other light source. If that shadow ray is occluded by another light source, it must not contribute anything, because it's the sky bein...

When your ray stumble upon a light source and you want to do MIS, you need to know the probability of it being selected. It's impossible to know that probability without having sampled random points on all other light sources in your sampling method. I'm not sure if doing that make sense and will co...

Fail trying to use sobol sequences with bidirectional path-tracing.

Coincidentally I was looking at the problem of layers in the last few days trying to improve clear coating. It's possible to account exactly the amount of light reflected by the bottom layer if you assume that the light will hit the same microfacet always, just like you can do with a thin glass mate...

Nice renders

I'm curious by what do you mean by bi-layer, multi-scattering microfacet model?

Is it GPU or CPU?

It's possible to not fail affine transformation for a sphere if you transform the ray into the local coordinate of the sphere.

Small typo in the code: search for salmpleLight.

You definitely need to calculate it before doing any shading calculation as they should be based on this normal. You should keep a copy of the surface real geometric normal along side and use that to displace the starting position of rays to avoid self-intersection.

I would use float unless you have specific needs for the precision of doubles. With floats, you can fit a vertex inside a single SSE register or you can perform 2x more operations at the same time. It'd be great if one day we can use doubles without worry but for now it's not worth the performance l...