Dependent texture reads and nVidia

[endrelidal]

Hi, I have a newbe question, I think. I’m trying to do dependent texturing, like reading a value from a 3Dtexture and use this value as texture coordinate for a 1Dtexture. Like a color look-up table.

The program is simple:

uniform sampler1D Colormap;
uniform sampler3D IndexTexture;

void main(void)
{
vec4 index = texture3D(IndexTexture, gl_TexCoord[0]);
gl_FragColor = texture1D(Colormap, index[0]);
}

The problem is that it doesn’t work on my nVidia 5700 card. If I program it using GL_ARB_FRAGMENT_PROGRAM or CG it works.

Is it true that the nVidia driver compiles the GLSlang program to CG before compiling it to GPU-code?

If so, how does the driver know which CG profile to use?

With the last version of CG the cgc (command-line compiler) can compile GLSlang programs. If I do not specify a CG profile I get an error message, if I use profile supporting dependent texture reads, like fp30 or fp40 the programs compile into correct GPU asm code.

Any idea, anyone?

General question about GLSlang: Is there a notion of profiles, like in CG, to distinguish between different GPU hardware?

I’m running on Linux and have the latest driver (6629).

[jra101]

Have you set the samplers to different texture units using glUniform1iARB? Initially all samples are bound to texture unit 0 so you’ll need to bind them to the appropriate texture units before rendering anything that uses that shader.