It seems like a really cool and fast solution especially since you can separate the sampled shader (use a simplified shader to plug into F).
Unfortunately I've come up against a couple of issues, I dont know if they are bugs or limitations.
Area Lights:
So in my shader I have a lambert's “f” output plugged into “F” and the “clr” added to pathtrace's “gillum” and then plugged into “Cf” (simplest possible setup). The first problem I noticed is that area lights are treated as if it were a point light (sharp shadows), and that only one light would be visible to the Pathtracer.
I looked around the mantra settings and found that if I increased the “min ray samples” to 2, my second light would be picked up, 3 the third light and so on… then I noticed that if I increase this value until it is greater than the number of lights in the scene, I would start to get additional light samples appearing, but they are not jittered… it looks like several point lights scattered on the surface of the area light so there's alot of banding and very slow renders due to so many min ray samples.
Then I converted the Lambert's “clr” to bsdf and plugged that into “F”, the area light interactions now look as expected but I supect I'm negating any benefit from using the pathtrace as a result.
The other big problem is reflections… If I add any of the flavours of reflection shader to “clr”, “gillum” and finally “Cf” the reflection stops working properly, not vanished completely just very screwed.
Hope It's going to get fixed soon or someone tell me I'm stupid
