[Bf-cycles] Volume patch rebased, added bidirectional experimental core.

storm kartochka22 at yandex.ru
Fri Jan 20 22:57:40 CET 2012


Minor change to volume patch, I enabled direct light, it help a lot in
case my outdoor "fog street lights" test scenes, and now point lights
work. Bug was related to MULTI_CLOSURE case that I handle wrong way, now
it wrong too but work for simple cases. But direct light algorithm
cannot handle shadow probe rays well (ok, I cannot integrate it proper
way, and going to use bi-dir tracer for that) and point light cannot
penetrate object boundary, even Transparent material or Glass. So, you
can test point light only with Background volume. No "point light near
wax canlde", only "glowing point light on myst street", sorry.

I have first half(quarter?)-working bidirectional intergator, so share
that code as well as promised. It can be selected by changing
"integrator type" in Integrator settings panel to 1. I keep all in
kernel_mlt.h. Please do not use it at all, it plain wrong, much slower
(orders of magnitude) than path tracer and produce wrong colors (of
course in a time it is not busy eating cats, hard disk data, you know).
Brecht, do even open it pease, it more to myself, to stimulate myself to
fix it when other ppl can see that sh*t.

For all other curious : "Woring" part is light contribution
connect_background(), it 1:1 as PT, and more interesting
connect_both_paths(), it cost me a lot of nerve cells and some weeks
until I figure that I not need to divide by (eye_bsdf_pdf *
light_bsdf_pdf).

Patch grow to 150+KB, I use bzip2.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: blender_cycles_volume_mlt_43572.patch.bz2
Type: application/x-bzip
Size: 30653 bytes
Desc: not available
Url : http://lists.blender.org/pipermail/bf-cycles/attachments/20120121/f3b4a53b/attachment.bin 


More information about the Bf-cycles mailing list