VBO colour data

OpenGL OpenGL: Basic Coding
1

Hi folks,

I think I’m not quite understanding the bits and bytes stuff when it comes to storing colour data. I’m trying to store raw data from an image file in a vbo. I’m avoiding using a texture object.

I’m collecting the image data via a std::vector< unsigned char > container. The colour data is good, I can print random values, which are correctly returned in their 0-255 colour range. I’m sending the container with:


// note: container in this case is a reference, as the data is collated elsewhere and passed to my "make buffer function" via a reference
// ..glGenBuffers() etc here..
glBufferData(GL_ARRAY_BUFFER,container->size() * sizeof(unsigned char),container->data(),GL_STATIC_DRAW);

My vertex attribute pointer in the draw function looks like this:


// id is good
glEnableVertexAttribArray(id);
glVertexAttribPointer(id,3,GL_UNSIGNED_BYTE,GL_FALSE,sizeof(unsigned char)*3,0);

I send the info to the shader using an ivec3:


// in my shader
in ivec3 ColourIn_int;
out vec4 colour;
...

void main()
{
...
colour.x = ColourIn_int.x / 255.0;
...etc...
}

The issue I’m getting is that the colour drawn with my shader is not correct. I have to fiddle around with large numbers (e.g. 4+billion) before I start to get anything that remotely resembles the given image. Dividing by 255.0 in this case doesn’t give the desired results.

I don’t appear to be getting any errors with glGetError() anywhere. Is there anything else I can provide?
Where/what am I tripping up with here?

WP.

2

I’ve just noted that glVertexAttribPointer should probably be glVertexAttribIPointer. Hadn’t realised there was more than one of them!

Part way there…

WP.

EDIT: this has made a significant difference in being able to see the image straight up, but the colours still aren’t correct.

3

OK, solved. After many days being unable to see the forest through the trees, apparently talking to myself here helped me see the light! My apologies.

The issue was, I needed glVertexAttribIPointer, and I needed to use GL_UNSIGNED_BYTE as the flag. I had accidently used GL_BYTE which had come about through testing other flags in the original glVertexAttribPointer, and was why I was getting incorrect colours as noted in the second post above.

For others on the learning boat, there’s two types of pointer attributes:
glVertexAttribPointer <-- for data with floats
glVertexAttribIPointer <-- for data with chars/ints/longs (my interpretation, correct me if I’m wrong someone!)

Maybe of some help to someone else in the future.

WP.

4

There’s actually three types:

  • glVertexAttribPointer() is for attributes which are accessed via a single-precision floating-point variable (scalar, vector or matrix) in the shader. The source data can be floating point, integers which are normalised (converted to floats in the range 0…1 or -1…1) or integers which are converted directly to floats.

  • glVertexAttribIPointer() (OpenGL 3.0 and later) is for attributes which are accessed via a (signed or unsigned) integer variable in the shader. The source data must be integers.

  • glVertexAttribLPointer() (OpenGL 4.1 and later) is for attributes which are accessed via a double-precision floating-point variable in the shader. The source data must be double-precision floating point.

5

Well there you go, there’s three! Thanks GClements!

WP.

6

Actually for your data set you can use plain old glVertexAttribPointer but set the normalized parameter to GL_TRUE: this will result in the source 0…255 range data being converted to a 0…1 floating point range when it is accessed, and your shader will then declare the colour attrib as vec3.

This is effectively how the old fixed-pipeline glColorPointer worked.

The only thing to be aware of is that GPUs really don’t like 3-component colours and using 3 components may even drop you to software emulation on some platforms. You’d be strongly advised to burn an extra byte in exchange for the performance. Further reading: Common Mistakes: Deprecated - OpenGL Wiki