Variables or registers in ISL

[bacon]

Holy Kamoses!!!
Excuse me for asking, but what kind of a troll are you Galinette ? making witchcraft like that ?

Superb material by the way

[galinette]

If you like it , then +1 all my requests for ISL improvements! I have a lot of ideas, but currently what I lack most is accessing the normal vector, the incident vector, and the ability to override surfacic R & T coefficients by shaders in the specular material. With this I could make fun stuff... I know onosendai already has this in the todo list.

Etienne

[Zom-B]

here is my big fat +1, so galinette can enhance the matdb with awesome new ISL based stuff in the future
Maybe u update of the LLVM JIT backend would be a good idea too (64bit support maybe?!? 32bit speed ups?!?)

[fused]

here is my big fat +1, so galinette can enhance the matdb with awesome new ISL based stuff in the future
Maybe u update of the LLVM JIT backend would be a good idea too (64bit support maybe?!? 32bit speed ups?!?)

iirc, we should have that in 2.2.12 already?
were building against LLVM 2.6 and the 64 bit build should use it, too (but maybe thats just for 2.4 onwards, cant really remember...)

[galinette]

What's LLVM JIT?

Etienne

[PureSpider]

It's the Just In Time compiler used for intrepreting ISL shaders in indigo.

[fused]

What's LLVM JIT?

Etienne

Im about to go to bed so I hope you dont mind if I let Wikipedia do the explaining (also, its a lot better at doing that than I am):

LLVM
JIT

[juan_irender]

Good stuff, Galinette! Yes, if ISL could be enhanced with such adds, it would be possible to create much more complex shaders, like yours, in a more close Renderman style.

[CTZn]

Pure amazingness ! I understand what you say about everything being governed by UVs, given how you make pyramids I'm pretty confident about other mapping schemes.

I need to get into ISL myself before asking anything...

I also noticed in Maya procedural shadings how it was difficult to obtain a visually smooth transition of this kind, I guess the corner function should be stretched by an order along that edge. It may actually be mathematically correct, but just too abrupt for visualisation.

[CTZn]

How about an "ISL Snippets" section of the database ?

Functions would be conserved and accessible as materials are, as opposed to: lost in the posting flow.