In the last week, I did some experiments with lighting models (not the illumination models, which is alway Lambertian). Right now, I feel I don't need this experiment anymore. I got the goal and I simply don't want to put any more effort on it. Since there's always the need of tutorials (this is not a tutorial mind you), I may make it public (most of my experiments simply fade away in the corners of my HD but this is somewhat more involved). Not that I feel there's the need for it: after all, those parallax mapping things everyone keeps speaking about are really easy but maybe you can use the framework for additional experiments, who knows: the more, the better. You may remember NVSDK included a program to test the various fog modes. This program is its "ideological follower" and does the same for the various illumination shaders you fetch. Mind you, it's ugly. Anyway, it does per-vertex, per-pixel, dot3, parallax mapping (with and without offset limiting), steep parallax mapping and its interpolated version (that is, parallax occlusion mapping as I realize it). Another thing I wanted to test was to render a great amount of lights. Although the performances on my system were too low to get excited about, I'm still confident I could get something out of it. The bottom line is, as always, very trivial: don't use interpolators to pass lighting info, which is passed in two LUTs. If you think this could somehow turn out useful, I'll keep an eye on this thread.