Keith Whitwell wrote:
>
> Brian Paul wrote:
> >
> > CVSROOT: /cvsroot/dri
> > Module name: xc
> > Repository: xc/xc/lib/GL/mesa/src/drv/tdfx/
> > Changes by: brianp@usw-pr-cvs1. 01/06/07 09:46:17
> >
> > Log message:
> > add divide by zero checks to fix occasional clipping bugs in Q3
> >
> > Modified files:
> > xc/xc/lib/GL/mesa/src/drv/tdfx/: Tag: mesa-3-5-branch
> > tdfx_vbtmp.h
> >
>
> Have you traced the source of these zero oow values? Any vertex with clip w
> == 0 should have clipmask != 0, therefore in tdfx_vbtmp.h
>
> if (mask[i] == 0) {
> /* unclipped */
> v->v.x = s[0] * proj[0][0] + s[12];
> v->v.y = s[5] * proj[0][1] + s[13];
> v->v.z = s[10] * proj[0][2] + s[14];
> v->v.rhw = proj[0][3];
> } else {
> /* clipped */
> v->v.rhw = 1.0;
> }
>
> the vertex oow value will be assigned 1.0. Any ideas what's going wrong?
I found that dstclip[3] in the tdfx interp function was sometimes zero.
This value is interpolated from the 'in' and 'out' coordinates by the
INTERP_4F macro.
The divide by zero resulted in 'inf' values and later 'nan' texture
coordinates.
I also found that the 'out' vertex rhw was also zero sometimes so I
applied the same divide by zero check to the wout computation. It
does seem strange that out->v.rhw would be zero, as you said. I've
already spent too much time on this but I might take a second look.
-Brian
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/dri-devel
