depends on how many objects are moving, so there may be only one model matrix and just in the next frame there may be 30 of them).

You need to define in shader the static length of the uniform array of matricies, and since that, the array of floats you pass to uniform setting function cannot be shorter than that (but can be longer since some updates). There is also inablitity to specify offset in the array, so it will always fill the uniform from the begining of array (this is so silly its hard to believe, but I just cannot find the offset accepting functions in specs!)

But there are more ways to set matricies to shaders, you can simply create a vec4 array, and set that, and use it to construct the matricies or just use them as rows the raw way in shader, doting 4d vectors with them.

@JohnnyCode

I like the second method, but I'm not sure what would be more computational expensive - creating matrices from an array of vec4 on GPU or sending on each draw call filled Float32Array with let's say 32 matrices? Also I don't need to offset anything since I found a clever way of managing model matrices, so it is not a problem for now.

I don't remember if I succeeded in sending Matrix4 arrays.

My engine is sending Vec4 arrays for now.

The reason behind this is that Transformation matrix can be 3x4 which spare you 1 Vec4 per matrix.

Considering you have a limit of uniforms in the vertex shader, this is mostly welcome!

Sending an array of floats and saving uniform vectors slots sounds more promising to me than sending ready matrices and binding every one of them separately, so I think I'd rather go that way.

BTW Is there a way to define size of an array in the vertex shader basing on MAX_VERTEX_UNIFORM_VECTORS value, so it would be more 'cross-hardware'?

BTW Is there a way to define size of an array in the vertex shader basing on MAX_VERTEX_UNIFORM_VECTORS value, so it would be more 'cross-hardware'?

I would like to correct you here that you should establish as many vectors in a compiled shader as you need, not a single bit more.

If amount you expect shrinks or grows it will bogus up your API calls. (no count or offset parameters in webgl specs uniform setters)

So establish as many vec4a as you need, not more, and check with gl.getParameter(gl.MAX_VERTEX_UNIFORM_VECTORS)) to see if the current setup actualy even supports you least needed amount. If the MVUV provided in outer API of yours is smaller than your needs, you need to fallback on your own responsibility and pick lesser shaders - if even.

(because of the fact that the amount of needed floats would dynamically change at the runtime).

Ah I see what you ment. You want to maximize the array , as much as it can get.

Well if you want to declare an array size of unifrom declaration of a shader to be maximal on a system, you may call gl.getParameter(gl.MAX_VERTEX_UNIFORM_VECTORS)) and put this value to your shader code string, before you compile it (and also knowinng this value for your very aplication) .

Wheather this constant (MAX_VERTEX_UNIFORM_VECTORS) is available in GLSL code compiler I do not know .

But bare in mind that once shader gets compiled, the uniform array keeps its size, so change at the runtime will not be possible unless you recompile, but I do not see why would someone need to change an unifrom array length through runtime? (just fill the unsed rest always, even bogus, you cannot specify count in webgl so always pass conforming array of floats)

You can index vec4 array with an expresion in vertex shader I am sure.

like :

uniform vec4 u_vecs[256];

....

int ind = 6 ; // sixth matrix

mat4 themat=mat4(u_vecs[ind*4],u_vecs[ind*4+1],u_vecs[ind*4+2],u_vecs[ind*4+3]);