Floating point FBO with VTF

Hello I am trying to use an floating point FBO attachment for vertex texture fetch, this doc (http://download.nvidia.com/developer/Papers/2005/OpenGL_2.0/NVIDIA_OpenGL_2.0_Support.pdf)

says

҉ۢ While 1D and 2D texture targets for vertex textures are supported, the 3D, cube
map, and rectangle texture targets are not hardware accelerated for vertex
textures.”

which I take to mean I must use power of 2… and then says the exact opposite…

“These features are supported for hardware-accelerated vertex textures:
• All wrap modes for S and T including all the clamp modes, mirrored repeat, and
the three “mirror once then clamp” modes. Any legal border color is allowed.
• Level-of-detail (LOD) bias and min/max clamping.
• Non-power-of-two sizes for the texture image width and height.”

Is it possible to use a power of 2 floating point texture with FBO or if that fails at none power of 2 floating point texture with VTF without it falling into software?

Originally posted by Fitz:
which I take to mean I must use power of 2… and then says the exact opposite…

Where did you read that?

It tells clearly: supported are 1D and 2D textures. Has nothing to do with POT or NPOT textures. If tehy tell they support NPOT, then they probably will. As far as i know nvidia only accelerates floating-point formats for vertex texture fetch, NPOT textures very probable.
Hovewer, why don’t you just try it out?

You are not required to use power of 2 textures for vertex texturing, the GL_ARB_texture_non_power_of_two extension allows you to load NPOT textures for all texture targets (1D, 2D, 3D, etc).

Of the texture targets, 1D and 2D are the only ones that support hardware accelerated vertex texturing.