[Bf-cycles] Volumetric patch update

[storm]

TL;DR:

Volumetric part not changed, Bidir part get some promising fix related
to MIS, small property GUI rewrite, not worth attantion if you not
coder.


Get almost proper MIS between light and path samples, not sure it work
right for number of bounces > 1, but testing pictures looks good, so
this post. Not ready for testing yet, more like "patch is still alive".

My obfucsating skills improved even more, now with all that endless
commented out lines i can easy forget what is going on in next page of
kernel_mlt.h after pressing by accident PgUP ir blink twice, and need to
recover minute or so :) But i fear it will not change, logic of
obtaining ray pdf too differ in details, and can not make real one
common function (bidir_recalc_backward_cumulative_ray_pdf() madness),
and will be even worse after proper "both subchain connect" MIS
implementation.

For those brave who dare to look at kernel_mlt_c code, it work only for
perspective, non DoF camera, emission light or zero size point light. No
adjoint BSDF implemented, so you will get wrong picture of on objects
that have interpolated normals (smooth), looking as they are flat and
have weird gradient colors in faces. Transparent or perfect mirrors not
supported yet, will produce wrong overbright.

 You better skip this update, or turn off "bidirectinal sampling"
checkbox on slightly updated integrator property panel, to get
almost-safe Cycles with volumetric addition.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: blender_cycles_volume_mlt_51746.patch.bz2
Type: application/x-bzip
Size: 56160 bytes
Desc: not available
Url : http://lists.blender.org/pipermail/bf-cycles/attachments/20121030/62373aad/attachment.bin