need size for uniforms

[RemiArnaud]

split from #100

need to add count to uniforms when it is a multiple of the size defined in the type.

for example

"jointMat": {
                    "semantic": "JOINT_MATRIX",
                    "type": 35676
                },

will become

"jointMat": {
                    "semantic": "JOINT_MATRIX",
                    "type": 35676,
                    "count": 7
                },

[fabrobinet]

It is not strictly needed but developers will appreciate that, especially to allocate the memory needed upfront if they want to do it while reading techniques.
That count isn't explicitly present in the WebGL API because as it is contained in the array:

uniformMatrix2fv(WebGLUniformLocation? location, GLboolean transpose, 
                          Float32Array value);

But (of course) it is there on GL ES 2.0...

void glUniform4iv(  GLint location,
    GLsizei count,
    const GLint *value)

[pjcozzi]

This is OK with me. Have we thought about uniform arrays in general, e.g., is the value array for a parameter still a 1D array where every sizeof(type) elements is one value in the uniform array?

[fabrobinet]

the count will be only for arrays, and thus will actually indicate we deal with an array. mat4 joint[1] has to be differentiated from just mat4 joint;

[pjcozzi]

So lack of count indicates a non-array and the presence of count (with count always >= 1) indicates an array? If so, I can update the 0.7 spec.

[fabrobinet]

exactly

[pjcozzi]

Spec updated in d10b1d0.

This impacts both uniforms and vertex attributes since both can reference parameters. This also keeps values as a 1D array so, for example, if a parameter is an array of two vec3s, the values array would have six elements, where the first three are one vec3 and the last three are the other.

A material parameter needs to override the entire array, not just individual elements. Is this OK for 1.0? We could add more fine-grained control later without a breaking change by adding offset and count properties.

Is there anything else we need to consider for this change?

[fabrobinet]

That sounds fine to me, thanks @pjcozzi closing now.