Question by Tudor · Sep 22, 2015 at 04:22 PM · unity 5 shader cg opengl matrix4x4

Difference between pos.z and pos.w?

I'd like some clarification on what's stored in the builtin float4's at different stages.

What value does the .w parameter of a vertex Position hold in the stages prior to the Projection stage? Is it 1? (is it at all useful in the vertex program of a unity CG shader?)

From theory I know that if you want to do a perspective divide on your view frustum in your clip / projection stage, you need to divide the current position by the inverse of its distance from camera. Which in (viewspace and) clipspace should be 1/ the .z component (and same as 1/ depth), no?

But in opengl books and articles I see it's always divided by .w, not by .z.

Is .w actually == 1/.z and is it only there for homogenous space coordinates reasons (and also to save you having to do an extra 1/ division?)? I'm asking because you have to turn .z into 1/.z (and then back) anyway in order to keep it linear during interpolation (and OpenGL does that automatically between vert and frag), so I'm a bit confused.

As well I see in unity's shader variables that _ProjectionParams.w == 1/FarPlane and ProjectionParams.z == FarPlane.

Does ProjectionParamscontain the camera's near and far clip plane positions in worldspace if you access it in the vert program, vs it being in normalized device coordinates in frag?

0 Show 1

_Gkxd · Sep 22, 2015 at 07:37 PM 1

3D points are stored as 4D vectors so that homogeneous transforms can be used. The 4-vector (x,y,z,1) represents the point (x,y,z).

Points are passed into the vertex shaders as (x,y,z,1), which is then multiplied at some point by the perspective projection matrix:

alt text

The resulting point then goes through the perspective divide, and so on.

ProjectionParams is just a structure that allows you to obtain certain properties about the current view. Think of it as four separate variables that are stored together instead of a point in 3D space.

projection-matrix.png (10.9 kB)