Stencil Shadows & Carmack reverse.

[Antorian]

I would like your opinion on Stencil Shadows (ok i know it have already posted somewhere) about carmack's reverse.
My question is: When using carmack's reverse it avoid all problems and bugs with shadows?
Is still remain some cases to manage after that?
Sorry I'm new with shadows.
Thank you.

PS: I look the NVidia news about their last GForce Fx 5600... but!! I saw on their site the Ti4800 doing 136Millions Vertices/s
--> http://www.nvidia.com/view.asp?PAGE=geforce4ti
and the FX56000 doing only 88 millions..?
--> http://www.nvidia.com/view.asp?PAGE=geforcefx
Did they loose some figures? http://www.opengl.org/discussion_boards/ubb/biggrin.gif or is it true?

May be I don't know reading http://www.opengl.org/discussion_boards/ubb/tongue.gif

[JustHanging]

No, not really. The remaining problems, however, are quite easy to solve, so it's a good technique if you are doing stencil shadows.

The biggest problem comes with smooth shading, when certain polygons start popping in and out of the shadow. This recent thread covers the problem and solutions quite well: http://www.opengl.org/discussion_boards/ubb/Forum3/HTML/008969.html

The zfail technique, which I think means the same as the "Carmacs reverse" also requires the volumes to be closed. This can be a problem especially if your model isn't closed. This has also been discussed here recently, not very hard to fix. The volumes aren't allowed to be clipped by the far clipping plane. This can be solved by a projection matrix trick, see the nvidia paper called "Infinite shadow volumes".

-Ilkka