Am 21.03.2013 15:48, schrieb James Holsenback:

>> what's up with blotches? change #5819/5820 (octree) or maybe something
>> still with area lights?

Octree problems look way different.

> OK ... turning area_illumination off got rid of the blotches!

Crap - I guess I have an idea what's happening there. I happen to know 
that in a portion of code that is called both for each individual point 
light as well as each "lightlet" of standard area lights, and which 
dispatches to the computations of the diffuse, specular, phong and ior 
components, there is some test to make sure that highlights are not 
computed for radiosity secondary rays. There exist two other, very 
similar pieces of code to handle photons and the "lightlets" of 
area_illumination lights respectively, which both /should/ have the same 
test as well, but I bet the area_illumination variant of the code 
doesn't. The result would be that a few radiosity samples would pick up 
that highlight as an extremely bright light (boosting the sample's total 
brightness a lot), while most samples would just miss it entirely.

(Without the test, if you'd crank the radiosity count up to insane 
values you could actually use this mechanism to compute reflective 
caustics. It would be extremely ineffective though, so we leave that up 
to photons in combination with proper reflection.)

It'll take me a few days to look into this; in the meantime, could you 
please file a bug report, and maybe already have a first glance at the 
code to check whether I'm right?

Post a reply to this message