[Bf-cycles] Volumetric patch update
kartochka22 at yandex.ru
Sun Apr 22 00:30:11 CEST 2012
Another step close to bi-directional support, i think i have done proper
MIS between indirect, direct and light chains. There is big missing part
left, MIS with contribution from many everything-to everything
connections between eye and light subpaths, but since on simple scenes
they tend to have little brightness, it not very noticeable.
No difference for scenes w/o point or mesh lights, you better stay with
uni-directional PT, it produce absolutely same image using more
But for scenes with mesh and point lights there is big advantage, now
Cycles sample more paths that cannot be found before, and show more
details (as caustics from point light) that cannot possible with
uni-directional PT by definition.
Only Perspective camera supported, without DOF.
wrong ray generation from point light with size > 0, use size=0 for now
Only diffusion shader get adjacent BSDF (not sure it done right at all,
but as least it produce smooth image even from light only chain). It
mean other shaders will work wrong when smooth normal used, producing
wrong colors, mostly looks as flat faces.
Some test scenes make many ugly black dots (infinity or negative colors
with very high value, as result by division by very tiny number). I must
to find that places and insert extra threshold comparison with some
Code is mess, extra duplications, unused variables, enormous number of
commented out code lines with failed attempts, I must clear all that but
only after i get some sign that it work right, i fear to touch even 1
line after that messing with probability theory.
As previous patches, do not save scenes with patched Blender, it can
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 43403 bytes
Desc: not available
Url : http://lists.blender.org/pipermail/bf-cycles/attachments/20120422/3b58957a/attachment.bin
More information about the Bf-cycles