Normalization cube maps


I am trying to implement the Example 8-1 and 8-2 (C8E1v_bumpWall and C8E2f_bumpSurf shader programs) from the Cg Tutorial book and I fail at initializing the normalization cube map.

The fragment shader calls for a “uniform samplerCUBE” and I try to pass the because the book said it is a normalization map I can use. But all I get is white color all over the rectangle I try to texture - without the normalization cube map (i.e. just with the normal map) I do see the normal map and everything is fine.

How am I to load a normalization cube map into OpenGL so that Cg can use it? I am using DevIL and do it like this:

static void myInit (void) {
   glClearColor(0.0f, 0.0f, 0.0f, 0.0f);

   // initialize devIL

   textureIdx0 = ilutGLLoadImage("");
   textureIdx1 = ilutGLLoadImage("");



static void initCG() {
    cgNormalMap = cgGetNamedParameter(cgFragmentProgram1, "normalMap");
    cgNormalizeCube = cgGetNamedParameter(cgFragmentProgram1, "normalizeCube");
    cgGLSetTextureParameter(cgNormalMap, textureIdx0);

    cgGLSetTextureParameter(cgNormalizeCube, textureIdx1);

Do I have to load the cube map differently? It probably isn’t a texture coordinates problem, because when I load just the normal map everything seems fine.

I know that this isn’t just an OpenGL problem, but it lies directly on the line between Cg and OpenGL so I didn’t know where else to turn. I posted this also on the Cg developer forum and hope that someone will be able to help me. Any help will be appreciated.

It’s difficult to say what is going on because of the library calls you’re making, all those il* calls that really don’t have anything to do with OpenGL but obviously include a lot of OpenGL calls.

Same with the cg stuff really, it’s obviously attribute/resource setup for a shader. The only OpenGL code apparent is set clear color to black, enable depth test and disable texture.

Ok, let me rephrase … I want to use a normalization cube map saved in a single .dds file in Cg. For that I need to load it in OpenGL and then pass it on to Cg - OpenGL will have its texturing disabled because Cg takes care of that. I am fairly certain that the passing of the texture to Cg works. Now if I wasn’t to use the DevIL stuff to load the .dds as a normalization cube map, how would I do it in OpenGL only?

Keep in mind that I do not have 6 different images for each side of the cube and therefore cannot create a cube map “the usual way”.

Any help will be appreciated.

One question springs to mind, is DevIL smart enough to recognize the cubemap and load it to the cubemap face texture targets? I doubt it but I could be wrong.

You do know what’s in a normalization cubemap right? Each point just represents the color of a unit normal in that direction. As such you don’t really need to load it, you can generate the cubemap procedurally quite easily. Treat each image like it’s +1.0 to -1,0 on the face of a unit cube and the missing (major) axis is fixed at 1.0, normalize the vector and convert the range -1->+1 to 0->1.0, then store that pixel as an rgb triplet in your chosen format, next load the image array to the appropriate texture face. Rinse & repeat for each face, transposing axes & switching signs accordingly.

You shouldn’t need to load the images from disk.

From the source code of DevIL, it seems it supports loading of cubemap textures, so this can not be the problem…

But I agree with dorbie, you shouldn’t be loading a texture from file that can easily be generated on the fly.

Ok, thx for the tipps. I will try to create the cube map procedurally.