Coming next: simplex noise

I'm also developing a different approach to noise,
in which the cubic lattice is replaced by a simplicial lattice.
This becomes a big win in higher dimensions, starting at
noise over R^{4}, and then just keeps getting better.
If you evaluate a point based on the influence of vertices of a surrounding
simplex, then you only need to do n+1 operations, one per vertex.
In contrast, the grid method gets twice as slow with each
added dimension, which becomes untenable beyond four dimensions.
Since at each vertex of the simplex you need to do a dot product between items of length n, the total cost of the simplex algorithm is O(n^{2}) with respect to dimension. In contrast, the traditional noise implementation is at least O(2^{n}). This allows you to do things like procedurally textured BRDFs, which are four dimensional, and become five dimensional when you vary them through time. My current plan is to explain the simplex approach in detail in a paper to be submitted for SIGGRAPH 2001. 