bilinear interpolation type in opengl

Hi all,

Context :
I’m programming in Cg in an Opengl program (using NV30emulate).

Problem :
I try to store textures coordinates (u,v) in a texture (used for indirection), but the bilinear interpolation of theses (u,v) is not fine enough. As a result, in the final image, the texture seems to be rendered as a GL_NEAREST mapping.
Notice that when bilinear interpolation is used for final color, no visual artifacts are presents. The problem is that the indirection map requires more precision.

Question :
Do someone know what is the type (GL_RGBA,GL_RGBA16,float, etc.) used by the hardware to compute the bilinear interpolation ?

I’ve supposed that it was the texture storage format, but using GL_RGBA16 inside of GL_RGBA8 do not change anything…

Please help me…