Re: xfree will be the best....

On Sun, Apr 06, 2003 at 10:24:48PM +0200, Marco Bubke wrote:
| Allen Akin wrote:
| 
| >Other very interesting questions include whether hinting can be
| >implemented by vertex programs, and whether performance/quality can be
| >enhanced by breaking glyphs into curve-bounded "trapezoids" that can be
| >processed by vertex programs or evaluators without triangulation by the
| >host CPU.
| > 
| >
| There are AFAIK not very much GPU's(Geforce 3/4) outwhere which are 
| doing elevators in hardware.  And you can't use vertex array for elevators.

It's true that evaluators tend to be accelerated only on high-end
hardware.  I thought NVIDIA had released an extension for curve and
surface acceleration that was intended to replace evaluators on consumer
hardware, but I can't find it at the moment.

It doesn't matter for this situation, though.  All you need to do is put
the control points in parameter registers, then feed the parametric
coordinate to the vertex program using vertex arrays or any other method
of your choice.  Then a simple computation in the vertex program
generates the points along the edge curves.  Generate them alternately
(first one edge, then the other) and you've got a triangle strip.

I should emphasize that I don't *know* this will work; I just think it's
worth the experiment.  It would be a good project for someone wanting to
learn more about vertex programming.

| >It can be handled by a fragment program.  The multiple alphas reside in
| >a texture that the program accesses.
| > 
| >
| Interesting idea. You could maybe also use multipasses on old hardware 
| with locking of the components but that would be ugly and slow.

The hardware that predates fragment programs often has some
vendor-specific OpenGL extension for programmability (like Register
Combiners), so you can probably run pretty fast on hardware even four
generations old.  For hardware older than that, we might have to do as
you suggest.

Allen