Open jalberse opened 4 months ago
This is the issue. I would expect the shape of the light cast to match the random walk more closely; but we veer to the right. This seems like it could be due to how we're weighting samples on the triangular light?
This isn't really something I notice in other scenes with large triangular meshes; I think it's only apparent in some scenarios. Low-ish priority for right now.
Check if this is still an issue now that the most recent commit is in fixing uniform hemisphere PDF sampling
While I've recently improved correctness by a lot, I still see an issue with this scene. The sphere is too illuminated; I think it's possibly an issue with how we are sampling the triangular light?
As of 204524ed2