Re: [Mesa3d-dev] shadow extensions

[Allen A. <ak...@po...>]

On Tue, Feb 20, 2001 at 10:22:28AM -0700, Brian Paul wrote:
| I want to experiment with shadow map sampling.  Linear filtering
| won't necessarily give better results than nearest filtering.
| Consider the discontinuities along object profiles.  I might try
| taking 4 depth samples, perform 4 R comparisons then use the
| results of those 4 tests to determine if the fragment is shadowed
| or not.  I'm also curious how mipmapping will work out.

If I remember correctly, SGIX_shadow is written so that the depth
comparisons happen before the mipmap filtering.  So in theory you get
a shadow value ranging from 0.0 to 1.0, rather than a hard-edged
0.0-or-1.0, and you avoid misclassifying geometry near edges.

It's still possible to get artifacts from depth-value quantization and
undersampling, of course.

Allen