On Aug 24, 1:32=A0pm, "jbwest" <jbw...@[EMAIL PROTECTED]
> wrote:
> "uuilly" <willy.li...@[EMAIL PROTECTED]
> wrote in message
>
>
news:45365034-b72c-4076-b8c8-a3c2940ee44e@[EMAIL PROTECTED]
> On Aug 22, 3:15 pm, uuilly <willy.li...@[EMAIL PROTECTED]
> wrote:
>
> [...]
>
> ...A little more info
> ..With this signature:>void combineCallback(GLfloat coords[ 3 ],
>
> GLfloat *vertex_data[ 4 ],
> GLfloat weight[4], GLfloat **dataOut )
>
> The variable "coords," that is supposed to contain the location of the
> newly created point, is printing total junk. =A0Ie: numbers that are way
> too big and seemingly out of left field. =A0So somehow I'm giving GLU
> something that is causing it to puke.
>
> >Is it a problem that I'm using GLfloats instead of GLdoubles? =A0I was
> >always under the impression that GPU's don't have double precision so
> >they're no point in using them.
> >Thanks,
> >Willy
>
> That's not the right signature for the combine callback, coords must be
> double. There's no GPU involved here. the signature is defined by the
glu
> source code and requires double.
> (if gpu's could do tesselation, we wouldn't need tesselation routines...
> right ?).
>
> jbw
I gave up. It took me a while to shake the kinks out but I wrote my
own tessellator.
A few notes, no conclusions.
1) The verts I copy pasted to you were truncated. I began printing
every significant digit and the data I'm rendering had ~15 per vert.
Not sure if that makes a difference in GluTess.
2) I know that GLU is building the Tess on the CPU. I just use
GLfloats to store all my verts. As I understand it, GPU's aren't even
designed to handle the extra resolution of GLdouble so why use them?
I tried converting all my GLfloats to a GLdouble array just before
building the tess but it didn't help.
3) Since the point of intersection in combineCallback was always
crazy I wrote my own intersect code. It worked for many of the
previous problem counties but sometimes the intersection was REALLY
far away. In other words glu was telling me that two almost parallel
lines on opposite sides of a county were intersecting.
4) I debugged everything coming through the combineCallback. Some of
the vertex_data was NAN.
Thank you so much for all your help. Really thank you. I have
certainly learned a lot and I'm sure it will come in handy down the
road. Should you ever have a nasty issue and you need some help
please email me directly. I hope one day I can return the favor.
Have a good one...
-Willy
|
