Hello !
As we are moving all our shaders from agal to ogsl, we encountered a little shortcoming.
In AGAL, we use a vector of constants to store some infos (ex: bones matrices) and then in the shader, we would select the needed constant(or matrix) depending on a vertex attribute.
The vertex attribute would be an index in this vector.
Something like this:
ft0 = fc[va1.x]
We noticed 2 issues :
• You do not allow using a variable as an index, the following code is not valid:
var index:float = 3;
constant[index].x; // <= this line would fail in ogsl
• There is no array/vector type with an arbitrary (but fixed on initialisation of course) number of elements, it would be great if you could add a vecXX type.
With theses 2 additions, our needs would be 100% covered.
What do you think ?
Here's a part of the agal (which is actually Easy Agal) prog i'd like to convert to ogsl (OGSL is so cool and powerful) to give you an idea of our troubles:
We store about 16 matrixs from vc40 to vc(40 + 16 * 4)
// matrix average
add(TEMP[7], TEMP[5], CONST[23]); // +40 for matrices
multiply4x4(TEMP[1], TEMP[0], CONST_byIndex(TEMP[7].x)); // vrt * bone matrix 0
multiply4x4(TEMP[2], TEMP[0], CONST_byIndex(TEMP[7].y)); // vrt * bone matrix 1
multiply4x4(TEMP[3], TEMP[0], CONST_byIndex(TEMP[7].z)); // vrt * bone matrix 2
multiply4x4(TEMP[4], TEMP[0], CONST_byIndex(TEMP[7].w)); // vrt * bone matrix 3
multiply(TEMP[1], TEMP[1], TEMP[6].x); // * weight 0
multiply(TEMP[2], TEMP[2], TEMP[6].y); // * weight 1
multiply(TEMP[3], TEMP[3], TEMP[6].z); // * weight 2
multiply(TEMP[4], TEMP[4], TEMP[6].w); // * weight 3
add(TEMP[1], TEMP[1], TEMP[2]); // TEMP[1] = vrt * bone matrix 0 * weight 0 + vrt * bone matrix 1 * weight 1
add(TEMP[1], TEMP[1], TEMP[3]); // + vrt * bone matrix 2 * weight 2
add(TEMP[1], TEMP[1], TEMP[4]); // + vrt * bone matrix 3 * weight 3
Hello !
As we are moving all our shaders from agal to ogsl, we encountered a little shortcoming.
In AGAL, we use a vector of constants to store some infos (ex: bones matrices) and then in the shader, we would select the needed constant(or matrix) depending on a vertex attribute.
The vertex attribute would be an index in this vector.
Something like this:
ft0 = fc[va1.x]
We noticed 2 issues :
• You do not allow using a variable as an index, the following code is not valid:
var index:float = 3;
constant[index].x; // <= this line would fail in ogsl
• There is no array/vector type with an arbitrary (but fixed on initialisation of course) number of elements, it would be great if you could add a vecXX type.
With theses 2 additions, our needs would be 100% covered.
What do you think ?
Here's a part of the agal (which is actually Easy Agal) prog i'd like to convert to ogsl (OGSL is so cool and powerful) to give you an idea of our troubles:
We store about 16 matrixs from vc40 to vc(40 + 16 * 4)
// matrix average
add(TEMP[7], TEMP[5], CONST[23]); // +40 for matrices
multiply4x4(TEMP[1], TEMP[0], CONST_byIndex(TEMP[7].x)); // vrt * bone matrix 0
multiply4x4(TEMP[2], TEMP[0], CONST_byIndex(TEMP[7].y)); // vrt * bone matrix 1
multiply4x4(TEMP[3], TEMP[0], CONST_byIndex(TEMP[7].z)); // vrt * bone matrix 2
multiply4x4(TEMP[4], TEMP[0], CONST_byIndex(TEMP[7].w)); // vrt * bone matrix 3
multiply(TEMP[1], TEMP[1], TEMP[6].x); // * weight 0
multiply(TEMP[2], TEMP[2], TEMP[6].y); // * weight 1
multiply(TEMP[3], TEMP[3], TEMP[6].z); // * weight 2
multiply(TEMP[4], TEMP[4], TEMP[6].w); // * weight 3
add(TEMP[1], TEMP[1], TEMP[2]); // TEMP[1] = vrt * bone matrix 0 * weight 0 + vrt * bone matrix 1 * weight 1
add(TEMP[1], TEMP[1], TEMP[3]); // + vrt * bone matrix 2 * weight 2
add(TEMP[1], TEMP[1], TEMP[4]); // + vrt * bone matrix 3 * weight 3