Re: [Algorithms] Curves...

[Conor S. <cs...@tp...>]

> >Bezier patches are easy to render.
> opinion:
> imho patches are horrible, after talking with a whole lot of people (i
used
> to like patches) writing some code...

They aren't horrible - They do have their disadvantages. But they also have
their advantages.

>
> there are just to many problems with patches, it's not worth the effort.
> they get split back to triangles, artists can do more with a higher mesh
of
> triangles.
> there are "so" many problems with patches that people don't see when
> starting out :|
> someone should really right an article on the downsides of patch
rendering.

Most are solvable problems. Easily solvable. Arstist can get a higher mesh
yes, but it costs a lot more in terms of memory.

Especially when you do animations.

If you think beziers into the design of your engine/feature set, you can do
very well and not have a single problem.

>
> sigh, this will probably start a debate why patches are good, if only i
had
> the energy to discuss more about this. i think i might have already
written
> a few points about this in a previous e-mail.
>
I think they are good, but only when used right, and when you have a good
solution to handle them.

Bicubic bezier patches can be rendered blindingly quick (If you used a
bivariate forwards differencing scheme, and make sure you can collapse the
algo all the way down to adds in both the S and the T loop), they are low
memory for the look they can provide and they are biparametric which makes
them perfect to map lightmaps etc to.

One of the things I'm doing at the moment is a biparametric geometry
composite pipeline. Beziers (both curves and patches) fit well into the
scheme, and I think I'm going to have to make regular use of them. I think
the system is a big success so far - I've gotta implement some more
functions, and clean some stuff up though before I see the true worth of the
system.

Conor Stokes