Luxblend25 preview render-mode

[J the Ninja]

If you haven't tried it yet, attempting to preview a material that uses the homogeneous volume in 2.5 results in the entire preview going turtle slow.
On a note that was initially unrelated (I started this to see what would happen when I had an hour to kill on campus today) I hacked up the preview scene to use photon mapping instead of the distributed path (which it used before). The results were....well...

Pros:
-Clears up faster, debatably even faster than dp
-Doesn't choke when a homogeneous medium shows up
-Doesn't need as many s/px

Cons:
-Can occasionally choke on roughglass if the roughness is too high (caustic map fails, and the glass renders casts a shadow like it was solid)
-This one is the kicker: When you update the preview, there is a short lag before anything noticeably happens (1-5 sec usually, will depend on the mat and your machine).

Some I'm putting it out here for you all to try and see what you think. You can download the modified file here: http://dl.dropbox.com/u/1706676/preview_scene.py

Swap that in for the existing one in the "export" folder of luxblend25 (that would be inside the "src/luxrender" folder). Back up the old one first!

Let me know what you think.

Related developments: LordCrc was also experimenting with using bidir for previews. It works better than you might think, although the grain can take awhile to go away. Jensverwiebe did a few tests and found it might be possible to preview homogeneous with dpath, with some parameter tweaks. Stay tuned.

[jeanphi]

-Can occasionally choke on roughglass if the roughness is too high (caustic map fails, and the glass renders casts a shadow like it was solid)
-This one is the kicker: When you update the preview, there is a short lag before anything noticeably happens (1-5 sec usually, will depend on the mat and your machine).

If roughness is high then it becomes diffuse and the caustic map is rightfully not used.
The second issue is due to the photon shooting phase.

Jeanphi

[Lord Crc]

Another thing which occurred to me is that we could use the outlier rejection for the preview. Since the resolution is so small, speed and memory shouldn't be much of an issue.