[sldev] P2P/Squid Web Textures: Enabling Greater Quality Images - draft 2

[Dirk Moerenhout]

Current subsampling and LOD routines expect it to be a power of 2.

I hate the current LOD handling too as it considers the texture no
different than an image and hence LOD downscales the image and
interpolates the vertices.

I'd change the LOD handling but there's little point before we have an
agreement with LL on losslessly compressed sculpt textures.

Dirk aka Blakar Ogre

On 7/9/07, Callum Lerwick <seg at haxxed.com> wrote:
> On Sun, 2007-07-08 at 01:14 +0200, Dirk Moerenhout wrote:
> > Sort of. By design you can actually do more in a 64x64. In a 64x64 you
> > get 2 lines for poles and 31 lines for vertices. In a 32x32 you don't
> > get a seperate 33rd line for the pole, your pole is picked from your
> > 31st line of vertices.
> >
> > Not a huge difference but it's there. Personally I'm not fond of the
> > current way poles are done. I'd rather see 32 lines of vertices plus
> > poles that are calculated by taking the middle point between vertex 0
> > and 16 of top and bottom line.
>
> You know, this isn't actually being used as an OpenGL texture, there's
> really no reason a sculpt map has to be power of two dimentioned. Why
> not just use a 32x33 map?
>
> _______________________________________________
> Click here to unsubscribe or manage your list subscription:
> /index.html
>
>
>