如何使用OpenGL ES2，GLSL，C ++将模型矩阵数据缓冲到纹理中？

Question

According to this answer , OpenGL allows you to store arbitrary data within a texture buffer that can be accessed in the vertex shader.

I created a Float32Array of size 4096 * 256 * 4 which contains the world matrix for every model (enough for ~256k models). Each model has a modelIndex attribute which is used to read its matrix from the texture. Then at each frame, gl.texSubImage2D the entire texture and draw as many as possible in each draw call.

An example scenario:
I have a bunch of unique models who share a common shader, their vertex positions packed into a single VBO. I'd like to update each model with a unique model matrix fetched from the texture buffer, then draw these in a single glDrawElements() call.

Some very rough pseudo code:

// In C++
struct Vertex {
    GLFloat x, y;
    GLuint matrixID;
}

// Create an array of floats that represent a unique model matrix
// Assume pre-calculation of values
GLfloat data[32] { /* 32 floats, two 4x4 matrices */ }
glTexImage2D(..., data);

// Then in the vertex shader
attribute vec4 in_Position; // x, y, 1.0, matrixID;
uniform mat4 uf_Projection;
void main() {
    // I need help implementing this magical function...
    mat4 model = getMatrixFromTextureBuffer(in_Position.w);

    // Apply unique model matrix
    gl_Position = uf_Projection * model * vec4(in_Position.xy, 1.0, 1.0);
}

// Draw everything!
glDrawElements(...);

Could someone please share an example of how to implement this? I'm struggling to find relevant information online.

Answer 1

OpenGL allows you to store arbitrary data within a texture buffer that can be accessed in the vertex shader.

It's possible in some implementations of OpenGL ES 2.0, but it's not required in the specification (max vertex texturing units is allowed to be 0), and so many OpenGL ES 2.0 only GPUs don't support it.

Similarly, OpenGL ES 2.0 doesn't support floating point texturing (except as an extension) so that isn't guaranteed to work either.

Finally, even if you could get it working, loading a 16 element FP32 matrix from a texture for every vertex is going to be insanely slow, so just don't do it unless this is just a toy project for a desktop GPU ... (ie if your target is shipping commercial mobile content using OpenGL ES 2.x then this is probably not a good idea).