normalizing cube map

The idea of it to turn a vector into the normalized version of itself. Each pixel in the cube map has a vector stored in it. rgb = 0.5*xyz+0.5 which is then converted into a signed vertex. It is used mainly in the pixel shader because it is a lot cheaper than a full normalization function and more accurate than estimated ways. The most useful thing about them is I believe they only use one instruction. A full blown algorithm would take up a good deal more.