Indigo Renderer 3.4.1

[OnoSendai]

Hi All,

There's a lot of new stuff and changes in this one, so it's quite possible something may be broken.

I've added scatter and parametric surface test scenes if anyone wants to have a play with it.


This is a Beta release.
If you spot any bugs or problems, please make a post about them in this thread.
Thanks!

Indigo for Windows 32-bit:
IndigoRenderer_3.4.1_Setup.exe

Indigo for Windows 64-bit:
IndigoRenderer_x64_3.4.1_Setup.exe


Indigo for Linux 32-bit:
IndigoRenderer_v3.4.1.tar.gz

Indigo for Linux 64-bit:
IndigoRenderer_x64_v3.4.1.tar.gz

Indigo for Mac OSX (10.5 - 10.7):
IndigoRenderer3.4.1.dmg


Indigo RT for Windows 32-bit:
IndigoRT_3.4.1_Setup.exe

Indigo RT for Windows 64-bit:
IndigoRT_x64_3.4.1_Setup.exe

Indigo RT for Linux 32-bit:
IndigoRT_v3.4.1.tar.gz

Indigo RT for Linux 64-bit:
IndigoRT_x64_v3.4.1.tar.gz

Indigo RT for Mac OSX (10.5 - 10.7):
IndigoRT3.4.1.dmg

Changelog:

3.4.1
* Improved bidirectional path tracing. Should now converge significantly faster.
* Fixed rare crash with Glass Acceleration
* Fixed issue where blends of diffuse and diffuse transmitter were rendering incorrectly in some cases.
* Added experimental Parametric surface support.
* Added experimental scatter support.
* Fixed crash when render queue fails to parse (for example, an </item> to much)
* Fixed normal smoothing on diffuse transmitter material.
* Updated bundled Blendigo installer to blendigo-2.5-3.2.14.0.
* Updated Indigo for Cinema 4D to 3.2.13
* Various core optimisations.
* Slave GUI now remembers screen position
* Fixed crash when creating blend material and rendering without assigning a or b.
* Texture editor window now says "gamma" instead of "exponent"
* Default texture ABC for exponent maps is now C = 100, B = 1000 instead of C = 0, B = 1
* Added wrapping of strings written to XML in a CDATA section, when XML parsing may remove whitespace from the strings.

[fused]

Just a quick test, also with teapots.

[CTZn]

Did the 3d mandelbrot set example use the same eval(vec2) technique ? Is it relevant to request that test scene ?

Indigo Renderer Win 64 is crashing in attempting to scatter a cube, I just set a valid path to a local obj file for the test scene. The pigs below is still crashing Indigo for me.

[StompinTom]

Just a quick test, also with teapots.

kitchen_scatter_3.png

Looks wicked :) Would be great to see a small tutorial on doing this!

(or better yet, have it in the exporter options somehow ;) )

[pixie]

Code: Select all

* Improved bidirectional path tracing. Should now converge significantly faster.

While comparing 3.4 to 3.4.1 I find the caterpiller to be way slow on 3.4 and significantly faster on 3.4.1, still it seems to be quite slower then I remembered from past versions...

[OnoSendai]

Did the 3d mandelbrot set example use the same eval(vec2) technique ? Is it relevant to request that test scene ?

Indigo Renderer Win 64 is crashing in attempting to scatter a cube, I just set a valid path to a local obj file for the test scene. The pigs below is still crashing Indigo for me.

It's a problem with having an object with only quads. I will fix this at some point.

[CTZn]

Oh right, I considered that eventuality later yesterday; I remember you telling me about it.

[lycium]

Code: Select all

* Improved bidirectional path tracing. Should now converge significantly faster.

While comparing 3.4 to 3.4.1 I find the caterpiller to be way slow on 3.4 and significantly faster on 3.4.1, still it seems to be quite slower then I remembered from past versions...

The caterpillar test is a little "broken" or has some numerical issues in the thin leaf and/or SSS that makes it an unreliable performance test! Please compare with another, somewhat more "normal" scene

[lycium]

Here are some images from Zom-B (thx again mate!) comparing the new and old bi-directional path tracing modes. The JPEG images are very high quality so there isn't any undue loss of image quality for comparing.

Both rendered at the same time so the raw numbers are about half what they should be, and note that although the samples per second is lower now, the actual image quality for the same render time is much better.

This much noise reduction is actually a pretty big deal (best seen on the floor and on the wardrobe), because to be so visually noticeable the raw numerical variance would have to decrease by many times. So great job Nick on those improvements!

I suggest you view the images 1:1, and use a browser that allows you to flick between the image tabs easily (eg Ctrl + Page Up and Page Down).

[pixie]

the samples per second is lower now

neo0. won't be happy about it...

[lycium]

the samples per second is lower now

neo0. won't be happy about it...

trololololol, you're so right, and i avoided writing this myself

[CTZn]

Nice ! I would also underline the improvements over the foreground, indirectly lit side of the bed.

Do you have visual estimators for this ? I'd bet that the timewise improvement is above 20%. And, that pools will benefit !

note that although the samples per second is lower now, the actual image quality for the same render time is much better.

Quote, just a reminder that numbers are only relevant to the Indigo version used. If you observe a discrepancy in samples per second between different versions, look for a visual improvement in the newest. Actually recent improvements, IES in mind, are spectacular.

I'm close getting something parametric at all... I mean, here's a landscape I did with it !

[lycium]

I have no idea what that is, but I wanna render it biiiiiiig :O

[ENSLAVER]

I'm close getting something parametric at all... I mean, here's a landscape I did with it !

more!

[CTZn]

I have no idea what that is, but I wanna render it biiiiiiig :O

me neither and me too >:] Well I guess there is some coplanarity involved, the features seem to relate with the surface resolution.

more!

Alas that accident left no furter traces, at this stage I can not pretend to be able to reproduce it. It is a different approach than with material parameters, more difficult to me.

I expect to see more myself, those guys know better ;)