[Bf-cycles] Volume cell marching work
kartochka22 at yandex.ru
Thu Dec 29 19:16:37 CET 2011
Another update, I fix bugs related to cell marching code, my 3 test
scenes now rendering fine and almost equal compared to Woodcock
algorithm. I added related vars to Integrator property GUI, now you can
change on fly marching algorithm "volume_sampling_algorithm" and control
cell step and iteration limit. Of course it makes sense only if
"Homogeneous" check box set to off.
Most OpenCL related change reverted, I give it a try later after another
month driver update.
Sorry for many stupid junk comments all over path_integrate.c, I need
time to be sure I chose proper algorithm and corner case tests, I clear
I think that my "work" is done, my goal was to have something renderable
on my home box, that can handle volume data as refernce to help when
somebody debugging other algorithms, maybe much faster and less
accurate, learn Cycles internals, light transport related theory (just
because it cool and modern), and I do not see any changes that can
significally improve quality w/o loss of unbiasness or accuracy. Of
course, patch require many improvements, and I will keep updates, but it
not that interesting as other more bright attractive features like MLT
About GPU support, as I once managed to get it work on my GPU (after big
cut of other features, but anyway), it _MAYBE_ work on CUDA, I keep data
types as GPU friendly as possible.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 96544 bytes
Desc: not available
Url : http://lists.blender.org/pipermail/bf-cycles/attachments/20111229/d7b06a10/attachment.bin
More information about the Bf-cycles