3 replies to this topic

[SamiHuutoniemi]

Short question:

If I am in clip space (unless I've misunderstood, that's where I am after multiplying with the WorldViewProjection matrix), should I not be able to get the z-coordinate of any one pixel from the gl_FragCoord.z value?

[Brother Bob]

Your assumption is correct, you are in clip space after the modelview and the projection matrix, but that does not mean that the fragment coordinate is in clip space. The perspective division stage separates the two. The fragment coordinate is the window coordinate (not sure if it's the normalized or pixel coordinate though) of the fragment, and consequently the depth coordinate is effectively the depth buffer value.

[SamiHuutoniemi]

Your assumption is correct, you are in clip space after the modelview and the projection matrix, but that does not mean that the fragment coordinate is in clip space. The perspective division stage separates the two. The fragment coordinate is the window coordinate (not sure if it's the normalized or pixel coordinate though) of the fragment, and consequently the depth coordinate is effectively the depth buffer value.

So the gl_FragCoord.xy coordinates are screen coordinates? And by depth coordinate, may I assume you mean gl_FragCoord.z?

[Brother Bob]

Correct. But as I said, I am not sure whether the screen coordinates are normalized or not; that is, before or after the viewport transform.