Why do we not apply Russian roulette also to the direct lighting?
I'm confused...what is the difference between terminateProbability and f.rho?
@PandaX russian roulette is a way to sample arbitrary-length paths. Direct lighting is a length-1 path (you either reach the light or there's something in the way), so we don't need to use monte carlo for it.
@ak-47 rho is the percentage of light that would be reflected along the sample direction, and is defined over [0, 1]. We use this to choose a terminate probability according to the following rule: the more light would reflect, the more worthwhile it would be to trace this new ray, so the less likely we should be to terminate the path here. If 0% of the light reflects (rho = 0) then we should terminate with 100% probability, because no amount of light traveling along the new ray will change the amount of color we return.