When using a perspective matrix in a vertex shader am I supposed to write code to divide by w or is it done automatically in a later stage?
The reason for my question is that I have seen lots of vertex shaders using:
gl_Position = matrix * pos;
which makes sense if there is a later stage that divides the vector with its w component.
However I never got it to work until I used the following in my vertex shader:
gl_Position = matrix * pos;
gl_Position = gl_Position / gl_Position.w;
Is the second example the correct one or could some other setting be missing?
Put it another way: which of the steps shown in the OpenGL Vertex transformation(first image) do I have to write in the vertex shader?
I know for sure that ModelView and Projection matrix belongs there(or a merge of the two). The viewport transform is not a part of the vertex shader, but what about the divide by w?
My setup is some simple triangles having coordinates within [-1 1] for x/y/z.
The Perspective matrix is supposed to project coordinates from z=-1 to -10 onto z=-1, x=[-1,1], y=[-1,1].
-1.0 0.0 0.0 0.0
0.0 -1.0 0.0 0.0
0.0 0.0 -1.2 -2.2
0.0 0.0 1.0 0.0
It was generated by:
x = 2.0f * zNear / (xMax - xMin);
y = 2.0f * zNear / (yMax - yMin);
a = -(xMax + xMin) / (xMax - xMin);
b = -(yMax + yMin) / (yMax - yMin);
c = (zFar + zNear) / (zNear - zFar);
d = -(2.0f * zFar * zNear) / (zNear - zFar);
To make the matrix P:
x, 0, a, 0
0, y, b, 0
0, 0, c, d
0, 0, 1, 0;
Finally I generate the final matrix by matrix = P * T where T is a translation (0,0,-2)
I have tried to do the math on the CPU and it appears to work generating expected results, however there I also do the divide by w manually.
Update: Solved but need understanding
I negated all components in the matrix (multiply by -1) and now it works. The example above also had an issue with projecting both positive and negative z-coordinates onto the projection plane which also got solved by this change.
Any references or explanation why it got solved by this change is welcome.