reflection glossiness

[Kakashi]

Hi.

Are there different by technical with between reflection glossiness 1.0 with number input and 1.0 with a texture.
A render time rendered with a texture is very slow compare with number input.

[ivaylo.ivanov]

Yes, of course. VRay doesn't know that your texture is pure white.

[vlado]

I think we actually have some bug to fix there; when the glossiness is 1.0 with a texture, we seem to be doing a lot of unnecessary light calculations.

Best regards,
Vlado

[re:FORM]

Interesting.
On a slightly related issue. I've just discovered you can drive the number of reflection subdivs with a map; eg where glossiness = 1, subdivs = 8, where glossiness <0.8, subdivs = 64 (using a gradient to smooth the transition).
This appears to work, but it's not something I thought we could do (as it's not possible in max).
Is this an acceptable approach or does it introduce unpleasant issues I'm not aware of?

[vlado]

I don't think this actually works, even if you find a way to connect a texture to the parameter in Maya.

Best regards,
Vlado

[re:FORM]

Could you try it? My initial tests were favourable... but I agree, it's more likely I'm mis-interpreting the results.
I created a checker texture (white = 1, black = 0.7) and plugged it in to the glossiness channel of a mirror material. I piped a copy through a colour adjustment to the subdivs channel (R only so it's a single value not RGB), and used the colour remap to set the min and max subdivs to correlate to the mirror (1) and glossy (0.7) tiles of the checker. (ie 8 for glossiness 1 and 32 for glossiness 0.7).

It appears to change the sampling when I change the remapping... but I'm unsure whether the value is being applied across the entire shader rather than the specific checker tile...

[re:FORM]

Actually, Scrap that! I did a proper test and of course, the value takes the average of the two checker squares and applies the same subdivs to the entire shader.
I'm curious. Would it be desirable to expose the subdivs parameter to texturemapping? That way we could set the number of samples on a per-pixel basis? Or would this be a superfluous feature given the DMC should undersample where glossiness approaches 1.0 ?