Fryrender demo is up!

IanT · Post by **IanT** » Sun Sep 23, 2007 4:39 pm

Frances wrote:Did it ever occur to you that it might be a bug? WTF is with this "cheating" bullshit?

Probably because, when an identical issue was pointed out 7 months ago, the official response was...

"I wouldn't call this a bug. That is a very well-known limitation of path-tracing-based render systems that has been debated strongly in the past in other forums. The more specular a surface is, the more noise it gets when it reflects the caustics produced by itself. Which is the case here: the caustics produced by the mirror are not captured well by the mirror itself."

(A response I disagreed with at the time, in the context of that particular scene, but didn't push the point).

Ian.

Frances · Post by **Frances** » Sun Sep 23, 2007 4:46 pm

"Bi-directional path tracing is great for caustics in particular but only directly-viewed ones. It doesn't help at all for when the caustics are viewed via a specular surface (because it can't join up the light and eye portions of the path) which explains why Maxwell/Fry/Indigo/Radium et al will struggle, regardless of which technique they use for path generation. " - droid

IanT · Post by **IanT** » Sun Sep 23, 2007 4:54 pm

Frances wrote:"Bi-directional path tracing is great for caustics in particular but only directly-viewed ones. It doesn't help at all for when the caustics are viewed via a specular surface (because it can't join up the light and eye portions of the path) which explains why Maxwell/Fry/Indigo/Radium et al will struggle, regardless of which technique they use for path generation. " - droid

Nice try, but quoted out of context (the context was for indirectly-viewed sun caustics which are notoriously hard to do). Stuff like the above scene is a piece of cake, even for simple backwards path-tracing without MLT.

Just to demonstrate the point ... 30 seconds on a laptop (no MLT... all important caustics already visible):

Ian.

Frances · Post by **Frances** » Sun Sep 23, 2007 5:16 pm

IanT wrote: Stuff like the above scene is a piece of cake, even for simple backwards path-tracing without MLT.

Yes, when there isn't bug. You seem to grasp the concept of cheating so well, but you don't understand bugs?

IanT · Post by **IanT** » Sun Sep 23, 2007 5:20 pm

Yes, when there isn't bug

*shrugs* As I said before ... the official position 7 months ago was that an identical problem wasn't a bug.

Ian.

Frances · Post by **Frances** » Sun Sep 23, 2007 6:40 pm

The render of the green boxes demonstrates a limitation of bidirectional path tracing, where as I understand it, a perfect mirror cannot see it's own caustics. Your render of the spheres demonstrates a bug in vBeta 1.8 (fixed). They are two separate issues.

And that's all I have to say about that.

Anthony · Post by **Anthony** » Sun Sep 23, 2007 8:20 pm

Indigo ftw

(even though it's slower)

IanT · Post by **IanT** » Mon Sep 24, 2007 12:33 am

The render of the green boxes demonstrates a limitation of bidirectional path tracing, where as I understand it, a perfect mirror cannot see it's own caustics. Your render of the spheres demonstrates a bug in vBeta 1.8 (fixed). They are two separate issues.

I'm sorry Frances (and apologies in advance to you and everyone else for my bluntness... this IS Off-topic after all)...

...but that's utter crap.

where as I understand it, a perfect mirror cannot see it's own caustics

Who told you that? I hope to goodness it wasn't Chema because:

a) It's absolutely not true. Maybe you understood wrong?
b) That's not even the issue that was obvious in the green box render (the issue was that the caustic from the green boxes wasn't showing in the mirrored walls)

Your render of the spheres demonstrates a bug in vBeta 1.8 (fixed)

In which case, that's a pretty serious, schoolboy-error bug

Even without changing the scene at all, the bug is easily apparent (in the glass sphere) and is the kind of thing that should have been spotted months ago. Oh, hold on ... it was

They are two separate issues

Exactly the same issue ... both renders were missing ESDSL plus others. Can't state it simpler than that. I think I can guess why someone's trying to make it look like separate issues though

Ian.

Frances · Post by **Frances** » Mon Sep 24, 2007 1:16 am

I only came here to help people with the supersampling issue, and to point out to someone where the SU documentation was. I'm not here to debate your conspiracy theories.

If you want to crow about someone making a stupid mistake, and point out that the beta testers were asleep at the wheel on spotting it, then go ahead. If it makes you feel better.

This whole discussion has left me feeling rather tainted. Insulting people seems to be your main talent. I hope you have a day job though.

IanT · Post by **IanT** » Mon Sep 24, 2007 1:35 am

Sorry Frances, but when untrue statements (mirrors can't see their own caustics) are made in order to explain bias in a commercial product that's sold as "unbiased" then I might tend to get a little curt. In this case I shot the messenger and there was no excuse for that. I'm honestly really sorry.

However, seems like even glossy surfaces don't pick up the caustic, which is an even more serious issue

EDDSL broken too?

Ian.

patricks · Post by **patricks** » Mon Sep 24, 2007 1:35 am

This is one of the biggest difference i can see between Freeware or open source software and Commercial one

anybody here post a scene where you can see a mistake or a bug of any free ware render and the developer are thankful that you spot it out .....
doing the same with a commercial render and you are insulting .....

( anyway the bug or lake of implementation will not go away by such post .....)

Greetings Patrick

IanT · Post by **IanT** » Mon Sep 24, 2007 1:59 am

doing the same with a commercial render and you are insulting

Indeed, although to be fair, Frances took exception to my implication that it was a deliberate "feature" rather than a bug and also took exception to the implied insult on someone else (who I'm sure would be perfectly capable of defending themselves if they'd felt insulted *shrugs*). What else is one supposed to think when exactly the same issue was pointed out months ago? (ironically, not by me

)

Ian.

chema · Post by **chema** » Mon Sep 24, 2007 2:33 am

First of all, I would like to thank you all for trying our demo. In fact, there was a 1.8 bug (a bug that was not there in 1.7 or previous versions, only in 1.8 ) with a certain type of E..SDS...L paths. It took like 5 mins to fix. So thank you. The ball caustics will now be visible in the mirror as they always were.

patricks wrote:anybody here post a scene where you can see a mistake or a bug of any free ware render and the developer are thankful that you spot it out .....
doing the same with a commercial render and you are insulting .....

No no. Au contraire. As you can read in my statement above.

The only one being a little offensive here is Ian.

--

You know, Ian. I am not going to tolerate you calling me (or anyone at Feversoft) a liar. Your lack of vision may make you think that fryrender is 'cheating' or something like that. But the truth is that you're simply wrong.

IanT wrote:where as I understand it, a perfect mirror cannot see it's own caustics
Who told you that? I hope to goodness it wasn't Chema because:

That is true for the technology in which fryrender is based as long as the surface is perfectly specular or refractive and the solid angle of the light source as seen from the surface is small enough. That's always been like that. So, please, next time that you say that I/anyone am/is wrong, think twice before you type anything anywhere in such an arrogant way.

That comment of mine I posted 7 months ago was not regarding relatively high surface emitters, but small or distant light sources such as the Sun.

IanT wrote:Sorry Frances, but when untrue statements (mirrors can't see their own caustics) are made in order to explain bias in a commercial product that's sold as "unbiased" then I might tend to get a little curt.

That is not only a misguided assessment of yours, but also an offensive one. fryrender is sold as unbiased BECAUSE IT IS. It is still a beta product and we're fighting to kill all the remaining bugs. Period.

Regarding the bug mentioned above (indirect caustics with relatively high surface emitters), that's a mistake in vBeta1.8 not part of a conspiracy

It's fixed for the next upgrade anyway. And regarding the limitation in the type of unbiased integration techniques that fryrender is using (indirect caustics with almost-zero surface emitters), that also happens to other well-known render unbiased engines such as Maxwell. That is a limitation real users can either overcome or live with.

We will consider the creation of a demo-users section in our forum so next time that somebody finds a bug there's no need to start a 'war' anywhere else.

Again, thank you everybody for trying the demo.

And keep up the good work, Ono!

patricks · Post by **patricks** » Mon Sep 24, 2007 3:09 am

Hi chema

thank you very much for clarifying ...... i am looking forward to see the beta/demo tester forum

will there be any bonification for those that collaborate with bug finding and posting suggestions at the Fry forum ...... i mean something like a discount for a full copy of Fry ?

Greetings Patrick

IanT · Post by **IanT** » Mon Sep 24, 2007 3:11 am

That comment of mine I posted 7 months ago was not regarding relatively high surface emitters, but small or distant light sources such as the Sun.

As I implied before, the context of that particular scene didn't include the sun or a small solid angle emitter but, hey, it's a moot point anyway, it wasn't relevant to the above.

That is true for the technology in which fryrender is based as long as the surface is perfectly specular or refractive and the solid angle of the light source as seen from the surface is small enough. That's always been like that. So, please, next time that you say that I/anyone am/is wrong, think twice before you type anything anywhere in such an arrogant way.

Sorry, I was just responding to the statement "where as I understand it, a perfect mirror cannot see it's own caustics"

I just said it was untrue, not that anyone was lying (there's a subtle difference)

You might want to take a look at the following too... seems like the caustic's not reflected in the (increased roughness to 40 from default) copper wall. If the wall is implemented as specular with micro-bump then your current fix will obviously have sorted it. If it's "diffuse" then maybe not...

Ian.

Fryrender demo is up!

Who is online