GLhalf Should Be a Floating Point Type

[TheFloatingBrain]

Hello, great library, I have been enjoying using it with strong typing guarantees and enum classes :)

I have been implementing some conversion code to associate opengl types with associated enum values for glVertexAttribPointer/Format

According too this and to the enum value (I would get you a line permalink but my browser is viewing glbinding/gl/enum.h as raw, but it is line 6999).

However it is defined as a short integer type.

Thanks - TFB :)

[scheibel]

This is an interesting request.
It reminded me of an earlier issue here in glbinding: #236

However, I want to refer to other official, normative information on the GLhalf data type.
In the official C headers from Khronos, GLhalf is defined as follows:

typedef khronos_uint16_t GLhalf;

And in the gl.xml, it is defined as such:

<type requires="khrplatform">typedef khronos_uint16_t <name>GLhalf</name>;</type>

Which is of little surprise, as this XML mirrors the C header.

Coming from this, I think most platforms expect a GLhalf to be just storage from a CPU and RAM perspective. Only the GPU performs meaningful operations on them.
From what I see, I think that we just need some kind of placeholder type and not actual FP semantics.
I had a look into glm, a common dependency of our downstream projects that also use glbinding, and they even removed their half data type in the past.

Can you be a little more specific what you expect a GLhalf to behave?
In which scenarios do you need FP semantics?

Best,
Willy