RT-patches?

[Ante P]

Is there any demo (preferebly Open GL) that uses RT-patches?


Also are Quintic Béziers and B-splines synonyms for RT-patches or different HOS methods?

As of yet the only RT-patch demonstration I can find is the one in the DX8.X SDK.

NVIDIA disabled RT-patches in the 21.81 driver because they somehow caused software emulation of N-patches. The game where this first was seen was Half Life.
I don't know for a fact, but I would suspect that this emulation problem only applies for Direct3D applications as all current applications that support N-Patches do so through the ATI_PN_Triangles extention (I think I got the name right?).

So what I'm trying to figure out is this:

• Does the GF3/4Ti have a fully functional implementation of RT-patches in the hardware?
• Does the "emulation problem" effect Open GL?
• Are there any applications where RT-patches are used?
• Is someone here willing to try to create a Open GL demo that utilizes NV_Evaluators?
• Was Open GL support also disabled in the 21.81 driver, or was it just DirectX?

Any answer, theory or fact, would be appriciated!

 

[Humus]

So what I'm trying to figure out is this:

• Does the GF3/4Ti have a fully functional implementation of RT-patches in the hardware?
• Does the "emulation problem" effect Open GL?
• Are there any applications where RT-patches are used?
• Is someone here willing to try to create a Open GL demo that utilizes NV_Evaluators?
• Was Open GL support also disabled in the 21.81 driver, or was it just DirectX?

1) Yes
2) No
3) Probably, can't name any though.
4) Sure, if someone gives me a GF4
5) No

 

[Ante P]

May I ask where you get this insight? (And how can you be certain if you don't know of any demonstration?)

Anyways, go get yourself a GF3Ti200, you'll need it for testing your demos anyways.

Thanks for the swift reply. If you ever come across with any demos using RT-patches please let me know.

 

[EvilEngine]

Is this what you want?

http://www.angelfire.com/wi/EVILHEAD/nvteapot.zip
You will need to paste that into a browser to download from my account with angelfire.

 

[Simon F]

Is there any demo (preferebly Open GL) that uses RT-patches?
Also are Quintic Béziers and B-splines synonyms for RT-patches or different HOS methods?

By RT-Patches, I assume you are referring to rectangular patches, which are probably more correctly known as Tensor Product Patches. I was going to try to summarise the idea in one or two lines but decided that was tricky. Besides, there are plenty of web pages out there: (eg, Bezier Curves and Bezier Patch)

Beziers, B-Spines (and others, eg. Hermite, Catmull-Rom) are just curves (i.e. lines going through "N" dimensional space (typically 3d)) that are defined by a set of control points and their own particular set of polynomial basis functions. The basis functions "blend" the control point positions to produce the curve.

These polynomials can (often) be of whatever order you choose, eg linear, quadratic, cubic (my preference), quartic, quintic etc. The higher the order, the more degrees of freedom you have but this comes at the price of requiring more control points.

To get a surface, you can take a grid of points (eg MxM) and take the "product" of the curves; eg 2x2 points=>Bilinear, 4x4=>Bicubic, 6x6=>biquintic. These are all RT-Patches.

For example, Bicubic BSpline surfaces are a popular choice as they give lots of freedom for modelling yet don't require a stupidly large number of control points.

Has that helped, or is it just more confusing?

 

[Humus]

May I ask where you get this insight? (And how can you be certain if you don't know of any demonstration?)

I've read plenty of topics about it on opengl.org

 

[Ante P]

Ahh I just saw a demo from NVIDIA, took a while to find it. Though the bastard is refusing to run even with a reference device.

Thanks Simon F, that was quite sufficient.

So what would your guys guesses be, will it ever be used? (As in a game.)

And Humus, I hope you weren't offended by me questioning the origin of your knowledge.

 

[ERP]

Does the GF3/4Ti have a fully functional implementation of RT-patches in the hardware?

Yes, however the hardware doesn't work directly from the control points, so if the control points change repeatedly (say animation), the processor has to recompute coefficients for the hardware.

These polynomials can (often) be of whatever order you choose, eg
linear, quadratic, cubic (my preference), quartic, quintic etc. The higher the order, the more degrees of freedom you have but this comes at the price of requiring more control points.

Just in case your making hardware design decisions, can I put in my request for at least quartic, quintic is better and there are some uses for degree 8 curves.
Although it's possible to create continuous arbitary topology surfaces with cubic curves, they can only be C1 (technically G1) continuous, which tends to leave nasty looking flat spots on the model.
The same problem exists with quartics, but the surface ends up looking better.
Quintic is the minimum fror a C2 continuous surface over arbitrary topology.

[/b]

 

[Humus]

And Humus, I hope you weren't offended by me questioning the origin of your knowledge.

Not at all. Should I be?

 

[Ante P]

And Humus, I hope you weren't offended by me questioning the origin of your knowledge.

Not at all. Should I be?

Of course not. 8)