MLT working - 9000 series or higher only

[understand]

nice! ..ill be back then

[pedrofelipe]

Radiance,
Any change to squeeze in 'architectural glass' ?

[Chris]

btw,

the new MLT algorithm won't be MLT Radiance

What exactly do you mean by that?
It will be a Monte Carlo method offcourse? Perhaps a hybrid of both MLT and BPT perhaps ?

[Sam]

btw,

the new MLT algorithm won't be MLT, i'm developing a new algorithm that's much better.
it also does'nt work with passes anymore, so you don't have any more UI lag when using octane on single GPU systems.

users can render finals with it, much more efficient, with pronounced caustics, and windows / GPU will respond as normal.

Radiance

What ?
No lag on single GPU ? Awesome

[Whichway]

Would you help relieve my ignorance by giving some pointers or something to what MLT is and how it works? I'm still a beginner at these things. Thanks.

[radiance]

Would you help relieve my ignorance by giving some pointers or something to what MLT is and how it works? I'm still a beginner at these things. Thanks.

you will see what it does when we release beta2.1.
remember, i can't talk about engine internals...

MLT makes the fireflies go away and handles caustics very well.

Radiance

[Whichway]

I don't necessarily mean how yours works, but you seem to be assuming people know something about MLT in general. I haven't been able to whittle my Google searches for "MLT" down enough to get anything that seems even remotely relevent. What do the letters stand for, for example?

[Chris]

Do a google search for Metropolis Light Transport. You can also search for Monte Carlo methods wich MLT is based on

Edit: This is one of the best places to read about it, its a bit technical.
http://graphics.stanford.edu/papers/veach_thesis/

[Whichway]

Ahh! Much cleaner search results! Thank you, Chris.

[Sam]

Just quoting Wiki here

Metropolis light transport obtains more important samples first, by slightly modifying previously-traced successful paths. This can result in a lower-noise image with fewer samples.