Anyone made a wrapper for Nanu and the Nvidia demos?

[Myth]

Anyone made a wrapper for Nanu and the Nvidia demos?

[Joe DeFuria]

Doesn't nVidia have to actually release the demos first? ;)

[Myth]

Just saw the add, and thought they were there..

[Chalnoth]

Anyone made a wrapper for Nanu and the Nvidia demos?
It would be rather challenging, given that Nalu uses one huge shader with static branching.

Given the somewhat complex border between skin and scales, I would expect performance to plummet if one attempted to split that "uber shader" unto multiple shaders without branching.

[Myth]

When it comes to Nvidia, i'll wait and see what and how it's made.. they've been pulling a leg og two before.. but if it really is one huge shader it'll be hard..

[Ostsol]

Wasn't the instruction limit something like 1024 on the GeforceFX?

[Chalnoth]

Wasn't the instruction limit something like 1024 on the GeforceFX?
You could have 1024 total instructions + constants. In Direct3D the limit was 512 instructions (this is the limit for PS 2.x).

[Hyp-X]

It would be rather challenging, given that Nalu uses one huge shader with static branching.

If the branch is static it's not a problem.

[Chalnoth]

It would be rather challenging, given that Nalu uses one huge shader with static branching.
If the branch is static it's not a problem.
1. It would take a fair bit of work to split the shader into multiple smaller ones.
2. The border between skin and scales looked rather complex, meaning that there may be a lot of state changes if one splits the shader.

[Gold]

Hell I'm still waiting on Vulcan, and Last Chance Gas.

[Chalnoth]

It would be rather challenging, given that Nalu uses one huge shader with static branching.

If the branch is static it's not a problem.
By the way, from reading the recent interview at nV News (http://www.nvnews.net/articles/geforce_6_interview/), I'm not so sure it was actually static branching:

Yes, this is a state change (and shader recompile in pixel shader). Uber shaders can be used to avoid state changes, but uber shaders use extremely coherent dynamic branching (branch condition supplied as a vertex attribute), rather than static branching.
So, if this was used in Nalu, then it could be quite challenging to unroll the branches, as it may require a number of state changes.

[Sworkhard]

I'm not a graphics programmer, but in that case, it would still be quite easy to make it run on the gForce FX series of cards as it supports dynamic branching (albeit slowly).So, if this was used in Nalu, then it could be quite challenging to unroll the branches, as it may require a number of state changes.

[Chalnoth]

The GeForce FX series does not support branching in the pixel shader. One could always execute all branches and use compares, but that's been an option since DX8 hardware, and it'd be far less efficient than branching if the shaders are reasonably long (which I'm sure they are).