Is there is a replacement for this? Something that is in core GL perhaps. I'm using shaders so I can do some manipulation but it would cost instructions. I could use floating point format but it would cause a slowdown. I think I will go for floating point for the moment. I have the following from a demo code glTexImage3D(GL_TEXTURE_3D, 0, GL_SIGNED_RGBA_NV, size, size, size, 0, GL_RGBA, GL_BYTE, data);

Quote:Original post by V-man
Is there is a replacement for this? Something that is in core GL perhaps.

I don't think. As you write, half-floats are obviously going to have a different footprint but they may give a few unpredicted advantages. Anyway, this is old FFP notions, it's good they didn't survive.

Quote:Original post by V-man
I'm using shaders so I can do some manipulation but it would cost instructions.

Are you really expecting a MAD to slowdown? Any card supporting decent floating point formats won't have any issue in managing a bunch of MADs, let alone one.

If I use float, then I don't need to add additional instructions in the shader. It is just that the texture slows down things because of the format. Also, problem with floating point is that I can't have mipmaps on some hw.