How do I get screen space co-ordinate of a vertex in vertex shader ?

Questions & Answers
1

I rendered a scene with several test objects to a texture, in flat color (purple) test.
Now I am trying to render the same objects again… mapping that texture directly onto them… seems like it would be easy, if I can just get the screen co-ordinate of each vertex in vertex shader, and map it as UV onto my texture from step 1 un 0-1 space. This is where the trouble begins…

using MVP transform… this does not appear to be in -1 to +1 screen space, but some other space called clip space, which seems local to each object. My texture from step 1 gets mapped wholly onto to each object… no good. almost like its mapping to to the bound of each object instead of the entire camera space.

I also tried manually calculating offset of vertex world space from camera space… transforming it to view space, then dividing by Z or by Range, for my own perspective calculation… This gets pretty close to mapping step1 texture over entire screen, but comes with some kind of distortion/warping in a saddle shape…

Anyone know the correct way to get arbitrary vertices on any/all objects… screen space position -1 tp 1 or 0-1 space ? help appreciated. thanks.

example code…
void vert (inout appdata_full v, out Input o)
{
UNITY_INITIALIZE_OUTPUT(Input,o);

o.worldPos = mul (_Object2World, v.vertex);
float3 relPos = o.worldPos - _WorldSpaceCameraPos;
float dist = length (relPos);
o.locPos = mul (UNITY_MATRIX_V, o.worldPos.xyz);  	//  just V
o.locPos = o.locPos /dist;	//true space perspective,   instead of /z    try this if no P  on the MVP transform
o.locPos *= _MetaScalar;		//scalar shader param, so I can experiment with it

}

2

If you’re targeting shader model 3, you get this automatically using a variable with the VPOS semantic. See the Unlit/Screen Position example at Unity - Manual: Shader semantics.

Otherwise, use the ComputeScreenPos macro. First define a new float4 member in the structure passed from your vertex to fragment shader, say, scrPos. Assign it an unused TEXCOORD semantic.

Populate it in the vertex program using the ComputeScreenPos macro:

v2f vert(appdata_full v)
 {
     o.scrPos = ComputeScreenPos(v.pos);
     ...

And then use it as follows in the fragment shader:

 fixed4 frag(v2f i) : COLOR
 {
     float2 screenPosition = (i.scrPos.xy/i.scrPos.w);