Fast SSS2 tip

[jop]

You know that SSS2 is a little bit tricky when it comes to DR rendering.

In short, if you plan to use it, you'de better rendering low-res on a single worksation...
Anyway that's observation I've made.

It's a bit unpredictable when you decide to render highres after set it up in lowres for fast preview. The result differs. You can try to render preview regioned highres, but the result will differs too, cause sampled area will not be the same as in full render....
So, it's a bit hard to use under production circumstance I think.

Anyway, I had problems when I rendered using DR and backed LC and IRR.
Indeed, even with backed lighting, SSS2 renders a prepass before final rendering, wich crash my max... But it will not crash if I don't use the IRR cache. The SSS2 prepass appears to take part of the IRR prepass calculation, and will not crash this way.

Strange behaviour, isn'it ?

To end, if you experiment problems with FastSSS2 and DR, try to not cache the IRR.

Just wanted to share this.

[vlado]

All these issues will be solved when we add brute-force method for calculating the sss2 effect (although it will be somewhat slower).

I've made a note to look into the crashes.

Best regards,
Vlado

[alexander.kazandzhiev]

Hi,
I'm investigating the crash issue with saved irradiance map and DR, but I wasn't able to reproduce the problem. Is it possible to provide some additional details about the renderer setup and/or simple scene that reproduces the problem? If yes please send to vray@chaosgroup.com.

Best regards

[jop]

Hi,

it's a quite huge scene, but I'll try to find some time to make a simpler one that will reproduce the problem.
Thanks.

[jop]

I tried to reproduce the problem on an simplified scene, but wasn't able to reproduce it.
I can't post the scene, it's too heavy.

I wonder if some geometry in the original scene is involved.
The strange thing is that it only occurs at high resolution, maybe a detail in an incorrect geometry isn't sampled in lowres but in the high one ?

[alexander.kazandzhiev]

It might be geometry. I think if it is we should be able to isolate it, by removing the other parts of the scene.
Is it possible for you to provide the scene settings for Indirect illumination, resolution and what seems to be relevant to the problem either by quoting them or emailing the scene with removed contents?

[jop]

I'm working on it

[jop]

I'm not able to reproduce the crash, even with the same scene !
Maybe it was some slave node related isolated problem.

Nevermind... it works now...

Maybe some day it will be back ! (and me too....)

[alexander.kazandzhiev]

Maybe now the DR setup is different, anyway in case that something appears, please let us know. Thanks for the efforts.

Best regards