opengl - Can I pack both floats and ints into the same array buffer?

Question

...because the floats seem to be coming out fine, but there's something wrong with the ints.

Essentially I have a struct called "BlockInstance" which holds a vec3 and an int. I've got an array of these BlockInstances which I buffer like so (translating from C# to C for clarity):

glBindBuffer(GL_ARRAY_BUFFER, bufferHandle);
glBufferData(GL_ARRAY_BUFFER, sizeof(BlockInstance)*numBlocks, blockData, GL_DYNAMIC_DRAW);
glVertexAttribPointer(3,3,GL_FLOAT,false,16,0);
glVertexAttribPointer(4,1,GL_INT,false,16,12);
glVertexAttribDivisor(3,1);
glVertexAttribDivisor(4,1);

And my vertex shader looks like this:

#version 330

layout (location = 0) in vec3 Position;
layout (location = 1) in vec2 TexCoord;
layout (location = 2) in vec3 Normal;
layout (location = 3) in vec3 Translation;
layout (location = 4) in int TexIndex;

uniform mat4 ProjectionMatrix;

out vec2 TexCoord0;

void main()
{
    mat4 trans = mat4(
        1,0,0,0,
        0,1,0,0,
        0,0,1,0,
        Translation.x,Translation.y,Translation.z,1);
    gl_Position = ProjectionMatrix * trans * vec4(Position, 1.0);
    TexCoord0 = vec2(TexCoord.x+TexIndex,TexCoord.y)/16;
}

When I replace TexIndex on the last line of my GLSL shader with a constant like 0, 1, or 2, my textures come out fine, but if I leave it like it is, they come out all mangled, so there must be something wrong with the number, right? But I don't know what it's coming out as so it's hard to debug.

I've looked at my array of BlockInstances, and they're all set to 1,2, or 19 so I don't think my input is wrong...

What else could it be?

---

Note that I'm using a sprite map texture where each of the tiles is 16x16 px but my TexCoords are in the range 0-1, so I add a whole number to it to choose which tile, and then divide it by 16 (the map is also 16x16 tiles) to put it back into the proper range. The idea is I'll replace that last line with

TexCoord0 = vec2(TexCoord.x+(TexIndex%16),TexCoord.y+(TexIndex/16))/16;

-- GLSL does integer math, right? An int divided by an int will come out as whole number?

---

If I try this:

TexCoord0 = vec2(TexCoord.x+(TexIndex%16),TexCoord.y)/16;

The texture looks fine, but it's not using the right sprite. (Looks to be using the first sprite)

If I do this:

TexCoord0 = vec2(TexCoord.x+(TexIndex%16),TexCoord.y+(TexIndex/16))/16;

It comes out all white. This leads me to believe that TexIndex is coming out to be a very large number (bigger than 256 anyway) and that it's probably a multiple of 16.

Answer 1

layout (location = 4) in int TexIndex;

There's your problem.

glVertexAttribPointer is used to send data that will be converted to floating-point values. It's used to feed floating-point attributes. Passing integers is possible, but those integers are converted to floats, because that's what glVertexAttribPointer is for.

What you need is glVertexAttribIPointer (notice the I). This is used for providing signed and unsigned integer data.

So if you declare a vertex shader input as a float or some non-prefixed vec, you use glVertexAttribPointer to feed it. If you declare the input as int, uint, ivec or uvec, then you use glVertexAttribIPointer.