Problem with matrix in Flash about krafix HOT 7 CLOSED

From @romamik on August 15, 2016 13:46

So I made more simple example and found that the problem is in the agal code.

matrixwtf2.zip

GLSL source:

#version 100

attribute vec3 vertexPosition;
uniform mat4 projectionMatrix;

void kore() {
    gl_Position = projectionMatrix * vec4(vertexPosition, 1.0);
}

Generated ".vert.agal" file:

{
    "varnames": {
        "gl_Position": "vt0",
        "projectionMatrix": "vc0",
        "vertexPosition": "va0"
    },
    "consts": {
        "vc0": [0.5],
        "vc1": [0.5],
        "vc2": [0.5],
        "vc3": [0.5],
        "vc4": [1.0, 1.0, 1.0, 1.0]
    },
    "agalasm": "mov vt0, vc0\nmov vt1, vc1\nmov vt2, vc2\nmov vt3, vc3\nmov vt4, va0\nmov vt5, vt4.xxxx\nmov vt6, vt4.yyyy\nmov vt4, vt4.zzzz\nmov vt5.x, vt5.x\nmov vt5.y, vt6.y\nmov vt5.z, vt4.z\nmov vt5.w, vc4.w\nm44 vt0, vt5, vt0\nmov vt0, vt0\nadd vt1.xxxx, vt0.zzzz, vt0.wwww\nmul vt0.z, vc0.x, vt1.x\nmov op, vt0\n",
} 

And when I change agal to

    "agalasm": "m44 op,va0,vc0\n"

Everything begin to work correctly.

from krafix.

this is the agal generated code from the example:
mov vt0, vc0\n
mov vt1, vc1\n
mov vt2, vc2\n
mov vt3, vc3\n
mov vt4, va0\n
mov vt5, vt4.xxxx\n
mov vt6, vt4.yyyy\n
mov vt4, vt4.zzzz\n
mov vt5.x, vt5.x\n
mov vt5.y, vt6.y\n
mov vt5.z, vt4.z\n
mov vt5.w, vc4.w\n
m44 vt0, vt5, vt0\n
mov vt0, vt0\n
add vt1.xxxx, vt0.zzzz, vt0.wwww\n
mul vt0.z, vc0.x, vt1.x\n
mov op, vt0\n

if you remove the highlighted lines it works as expected. This lines are responsable for adjusting the clip space here . I dont understand why the z is adjusted using the w.
I don't have a 3d test were this line is actually important.

from krafix.

Clip space in z is 0 to w or -w to w depending on the 3D api. In Stage3D it's 0 to w according to the internet, in Kha it's -w to w. That calculation looks correct actually if vc0.x is 0.5, will have to do some tests.

from krafix.

Ok didnt know about the diferent clip spaces. Now I see what the problem is, vc0 is not 0.5, vc0 is the first row of the matrix, a matrix uses 4 registers in this case from vc0 to vc3. Dont worry, I will fix it.

from krafix.

Was this fixed by #40?

from krafix.

Yes, tested with the examples uploaded by @romamik.
PS. You need to delete the old agal generated code, or it wont generate the new one.

from krafix.

Cool Spot

from krafix.