So basically i store 6 shadowmap mvp matrices ( each representing a different view direction (FRONT,BACK,LEFT,RIGHT,BOTTOM,TOP))

I basically use the formula like this:(model *view) *projection and pass this to shader

However i end up with these matrices

-1  0  0  135.957  0  1  0  -255.867  0  0  -1  331.663  0  0  0  1
  1  0  0  -135.957  0  1  0  -255.867  0  0  1  -331.663  0  0  0  1
  0  0  -1  331.663  0  1  0  -255.867  1  0  0  -135.957  0  0  0  1
  0  0  1  -331.663  0  1  0  -255.867  -1  0  0  135.957  0  0  0  1
  -1  0  0  135.957  0  0  -1  331.663  0  -1  0  255.867  0  0  0  1
  -1  0  0  135.957  0  0  1  -331.663  0  1  0  -255.867  0  0  0  1

So i have 0 0 0 1 at last row everytime

Having this code in fragment shader

	shadowvclip.x = -dp43(SMVP[i*4 + 0], vertex_pos);
		shadowvclip.y = dp43(SMVP[i*4 + 1], vertex_pos);
		shadowvclip.z = dp43(SMVP[i*4 + 2], vertex_pos);
		shadowvclip.w = dp43(SMVP[i*4 + 3], vertex_pos);
		
		svc = shadowvclip.xyz / shadowvclip.w;
		svc = svc * 0.5 + 0.5;

It happens to w to be always 1 no matter what thus i always get value beyound 0..1

rendering shadow compare useless.

float dp43(vec4 matrow, vec3 p)
{
return ( (matrow.x*p.x) + (matrow.y*p.y) + (matrow.z*p.z) + matrow.w );
}

But however when i use the samr techinque to render a 3d scene ( its not related to shadowmapping - i just render the scene everythings renders where it should be i mean)

Imagine i pass one of these shadow mvp matrices to a different vertex shader

precision highp float;
attribute vec3 Vpos;

uniform vec4 MVP1;
uniform vec4 MVP2;
uniform vec4 MVP3;
uniform vec4 MVP4;

vec4 vertexClip;

float dp43(vec4 matrow, vec3 p)
{
return ( (matrow.x*p.x) + (matrow.y*p.y) + (matrow.z*p.z) + matrow.w );
}

void main()
{
vertexClip.x = -dp43(MVP1, Vpos);
vertexClip.y = dp43(MVP2, Vpos);
vertexClip.z = dp43(MVP3, Vpos);
vertexClip.w = dp43(MVP4, Vpos);

gl_Position = vertexClip;

}

GPU somehow magically converts that vertexClip.w to something usefull however it stays last row is 0 0 0 1 all the time…

and dont bother with vclip.x = -dp43()

It just changes wicked view of opengl to something usefull

So how is that in vertex shader gpu somehow manages to do perspective division with this w component and i get ndc coords, but whenever i try to manually calculate it, it doesnt?

It took me 5 days to realize that i was passing only view matrix to shader not shadow mvp…