Depth buffer

Aren''t you supposed to discard the pixel if it''s outside the view frustum?

(Z - NEAR) / (FAR- NEAR)
This will give a linear depth value in the the range of 0 to 1.0. where 0 = the near plane and 1.0 = the far plane.

However most depth systems use a biased system where 90% of the depth range is used by the near 10% of distance. This is because it is much easier to pick up depth problems in objects close to the eye point than they if they were far away. For example you may have noticed the z tearing of objects far away in games but as you get closer to them the tearing slowly disappears.

oh yeah, for your second problem, all pixels outside of the view frustrum are discarded. You do not write any values to the depth buffer where they are outside of the near/far planes. Also you would want to use an unsigned format for your depth buffer as there is no need for negative values.

Thx! You helped me. I have one more thing. It is slightly something different, but... I'm still unable to corectly simulate OpenGL in coord generation, same triangle in OpenGL seems to be closer to viewer. This is the code I use:

z = 1 / v.z;
mAspect = mHalfWidth / mHalfHeight;
mDist = (mHalfHeight - 0.5f) / tan(mFov * 0.5);
x = ftoi((v.x * z * mDist * mAspect) + mHalfWidth); 
y = ftoi((v.y * z * mDist) + mHalfHeight);   

What's wrong with it? Or is it anything completly differrent?

Edit: Forgot to give computation of z.

[edited by - osh on May 13, 2004 2:08:45 AM]