tricky scene, Indigo stops sampling

FakeShamus · Post by **FakeShamus** » Fri Nov 01, 2013 2:57 am

so, this is a very complex set up in this scene (ies lighting, fog, cauchy materials...) but I have never seen Indigo just give up on a render before. the scene starts out rendering fine and then grinds to a halt pretty quickly - depending on what render method you use it gets a little farther in the spp. but inevitably the spp stops increasing. but the render progress bar keeps updating and according to my activity monitor Indigo is still going full-bore. to make things even more odd, Indigo must then be force-closed every time, it won't allow you to quit normally or even to close the scene and open a new one. it just hangs. I have tried a few things, removing different objects, etc, but couldn't find out a way to get it to work. this same scene used to render just fine (on some old version of Indigo, not sure which) with even more ies lights and basically the same overall set up. if anybody from the team can check it out and see what might be wrong I'd appreciate it. this is on mac, version 3.6.23 , osx 10.6.
thanks!

here's a link to the scene: https://dl.dropboxusercontent.com/u/128 ... lic_2.pigs

Voytech · Post by **Voytech** » Fri Nov 01, 2013 4:16 am

Ran this on RT (no SSS and no IES) and it drops from 300k s/s to about 5k s/s in 50 minutes.

Comp Specs:
Indigo RT v3.6.22, OS X 64-bit build.
OS X 10.9

ior · Post by **ior** » Fri Nov 01, 2013 9:56 am

Same here in indigo 3.6.23 on windows, it just stops sampling and if you update the image nothing happen, but the CPU is 100% and it uses almost one giga of memory. The progress bar is working and if you stop indigo and close it, it freezes and you have to kill it.

FakeShamus · Post by **FakeShamus** » Sat Nov 02, 2013 3:49 am

right, exactly. anybody from Glare care to check it out?
maybe you guys can tell me what's going on...

Zom-B · Post by **Zom-B** » Sat Nov 02, 2013 4:51 am

Such "hangs" in more complicated scenes are not new...
I remember to encounter them once in a while in extreme scenarios.
Glare does know, they do care and for sure will reply at some point,
hopefully with the message that its fixed for the next build

Post by **OnoSendai** » Sat Nov 02, 2013 8:00 am

It's some kind of media issue. Will have a closer look at it soon.

yonosoy · Post by **yonosoy** » Sun Nov 03, 2013 2:14 am

Hi. Please check this; if you define the SSS part of the quartz medium you got the scene (maybe; running here).

FakeShamus · Post by **FakeShamus** » Wed Nov 06, 2013 7:58 am

ah, yes that fixes it!
it keeps running as long as the quartz medium has some scattering defined, I just set it to 0.
way to figure it out, thanks!

yonosoy · Post by **yonosoy** » Wed Nov 06, 2013 3:08 pm

Hi Fakeshamus, thanks for exposse the issue.
Tthere are other two points that leads you to better result.

An ior of 1 is an exclusive property of vacuum,; To tge Air medium give an ior of 0,0001 (5) due to ths SSS part is non 0.
Speedups and more stable/correct.

Precedence is an internal concept for indigo than don't apport any kind of physic popertie's
1-2-3 instead of 1-10.100 seems to speeds up too.

So, many thanks in advance. A couple of things was learned whit ithis...

CTZn · Post by **CTZn** » Thu Nov 07, 2013 2:40 am

yonosoy wrote:An ior of 1 is an exclusive property of vacuum,; To tge Air medium give an ior of 0,0001 (5) due to ths SSS part is non 0.
Speedups and more stable/correct.

This is maybe more correct but normally slower than using an ior of exactly 1.

Stumbled upon the same issue yesterday with a scattering shader using a condition to define a threshold between zero and non-zero scattering values.

FakeShamus · Post by **FakeShamus** » Thu Nov 07, 2013 7:19 am

yonosoy wrote:An ior of 1 is an exclusive property of vacuum,; To tge Air medium give an ior of 0,0001 (5) due to ths SSS part is non 0.
Speedups and more stable/correct.

Precedence is an internal concept for indigo than don't apport any kind of physic popertie's
1-2-3 instead of 1-10.100 seems to speeds up too.

I just tried both of these changes and the difference is pretty much non-existant. setting the ior to a tiny bit larger than 1 showed a very slight slow-down as CTZn mentions. and from what I understand of the precedence system in Indigo it's basically just comparing numbers, the higher precedence of 2 mediums occupying the same space gets rendered in your image, so I don't think precedence of 1 vs 2 would be any faster than 1 vs 50 or 100. in any case I saw no difference in my tests.

CTZn · Post by **CTZn** » Thu Nov 07, 2013 12:47 pm

For reference on IOR, at the question

should the IOR be 1.0 or 1.0029 ?

OnoSendai answered

You want to use IOR 1.0 exactly. it's very important for fast rendering.

http://www.indigorenderer.com/forum/vie ... 10#p124110
On a side note, I'm loving sss rendering speed with PT+MLT... with ior 1 for the least

ior · Post by **ior** » Thu Nov 07, 2013 1:05 pm

see here: http://www.indigorenderer.com/forum/vie ... =5&t=11876

yonosoy · Post by **yonosoy** » Thu Nov 07, 2013 2:11 pm

So many thanks for the info.
Almost on an i3 the experience was that a definition of ior 1 for AIR medium and a SSS 0 for quartz produces an "overcome" of memory in a short time. For sure the fastest one situation. For this reason other values was tested and gives more stability and a bit slower count. In fact here finally don`t run properly !!!

Sorry for the inconveniencies.

tricky scene, Indigo stops sampling

tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Re: tricky scene, Indigo stops sampling

Who is online