Bug found, 3d cloud crash...

Started by odd, January 05, 2007, 11:28:36 AM

Previous topic - Next topic

odd


..if you do the following:

1. Start terragen

2. go to atmosphere and add any 3D cloud layer

3. Go to the quality tab of that layer and enter 1.4 in the jitter value

4. Crash (TG2 just dies)

Im on Windows XP sp2, 2Gb of ram, Dual Xeon 2.8Ghz.

Sorry if this is not the place for bug reports...just let me know if you want them some other way.

cheers

//O.

3DGuy

I think it's that way for any value > 1

Tangled-Universe

Yes indeed, you should think more or less in factors about the quality settings.

1 = 100%
0.5 = 50%

1.4 = 140%....doesn't make sense huh  ;)

Njen

I kind of disagree a little. With enough atmosphere samples, I get a perfect image with 0.65 for Detail. I have done tests above 0.65, and the difference is virtually indistinguishable.

3DGuy

Hmm njen, I don't really see how your answer relates...

Njen

My point being that I don't see 0.6 in my Detail setting as only 60% detail in my scene. Based on what JavaJones has explained, all the quality settings (samples, detail, AA, etc). are so interrelated that it's hard to say that 1 = 100% for any particular setting.

3DGuy

Ok, I see what you mean. Maybe it's the difference between actual detail and percieved detail. Like difference in jpg compression percentages. Or TV. An actual TV image isn't really that sharp and low res, but the arrangement of phosphor cells makes it  look sharp. Perhaps something similar is going on in the details. The detail may increase, but it isn't seen, or the resolution is just to low to actually see the detailing. Maybe the higher detail settings are meant for very high resolutions, but that's something the devs really know.

Njen

#7
Well I remember what JavaJones was saying something like in that, for example, a difference between a Detail setting of 0.6 and 0.7 is not 10%, that it's not linear. But I don't know for sure.

What I do know that increasing the detail from 0.5 to 0.6 increases your render a lot more than 10%
http://www.motionmagnetic.com/terragen2/render.html

3DGuy

Well, thinking back to other 3D packages, that's probably true.  If the percentages influence the amount of subdivisions then every subdivision increases the amount of triangles exponentially.

Tangled-Universe

Well Njen I think you've been taking my answer a little bit to literally.
Of course I know all the quality settings are interrelated and yes, therefore my explanation isn't really correct, but I was trying to keep things simple.

[stubborn]
Still, the setting "1" is the highest and therefore 100% quality for that particular setting, so still 1.4 means 140% quality....ghehe lol  ;D
[/stubborn]

Martin

Njen

Haha, putting on your stubborn hat, eh?

But like I was saying (and I can't find the exact post), I think JavaJones was saying something like the difference of 0.5 and 0.6 is double the detail - 200%, not 10%.

And the highest setting for detail is not 1. It can be anything you want ;)

Oshyan

Most settings intentionally allow values greater than 1. The "jitter" setting would seem to not like values above 1, though I'm not sure why that specific setting would have such an issue. It probably has something to do with how it works internally. There are other settings that cause unpredictable and odd results with values outside the normal slider range, but few I know of that crash. TG2's underlying systems are fairly robust, which is good. But in any case you wouldn't want a jitter above 1, so just avoid setting it that way. ;)

Thanks for the report. :)

- Oshyan

buzzzzz

Speaking of Jitter, what does the jitter setting do anyway?  All I remember is the "Jitter Bug" Oops! showing my age!  ;)

odd

oops - didnt think my tiny bug report would cause such havoc  :o

..well - in case of the jitter setting..if I understand it correctly, this is a setting that moves around the position of samples to avoid things like banding and moire pattern interference that cannot be solved with AA or other tricks.

As the Jitter settings default value seems to be 1, I would very much like to be able to set it to higher values...

A typical case of usage would be having a very unsaturated sky (or heavily tinted in one colour) with just a slight gradient I.E having few colours I.E being prone to banding...a case where I would prefer having extreme values of jitter giving me a very noisy and diffused sky...as opposed to a sky with banding.

Makes sense?

..it could also have artistic usage...Im using the Brazil render system and sometimes I use high contrast lighting, set the renderer to a low number of samples with loads of jitter and unsaturated colours...and the image then gives the impression of a thin emulsion film stock from the 60's.  ;D

..and I agree with Javajones...I think the TG2 render engine is pretty stable for a beta. In my experience it copes very well with extreme values...so the jitter thing is probably just a bug. Guess we have to ask the code maker himself  ;)

//O.

Oshyan

Quote from: odd on January 06, 2007, 09:23:33 AM
..well - in case of the jitter setting..if I understand it correctly, this is a setting that moves around the position of samples to avoid things like banding and moire pattern interference that cannot be solved with AA or other tricks.

Precisely.

Quote from: odd on January 06, 2007, 09:23:33 AM
A typical case of usage would be having a very unsaturated sky (or heavily tinted in one colour) with just a slight gradient I.E having few colours I.E being prone to banding...a case where I would prefer having extreme values of jitter giving me a very noisy and diffused sky...as opposed to a sky with banding.

Makes sense?

Absolutely. Fortunately you *can* set the jitter higher than 1 in the atmosphere.  :) It's just clouds that don't allow it, and since you seldom - if ever - get banding in the clouds, this doesn't seem like a big problem. Although again it is a bug that should either be dealt with by allowing greater than 1, or simply disallowing entry of larger values (assuming the internal jitter function for cloud samples can't cope with >1 values).

- Oshyan