I have a 3D simplex noise implementation with a weird bug: There seems to be a singularity around (0,0,0), which I can't quite explain, and searching the net found no mention of it, so I must assume it is just something I screwed up in my implementation.

It is modelled after Gustavson's paper.

And this is the code: http://www.pasteit4me.com/26002

The important bit is the -noise3dWithVector: method, starting on line 100. for some reason, if the input of that is close to (0,0,0), it doesn't do the right thing, but I can't quite see why.

If anyone wants to take a stab it, it'd be appreciated, I've been trying to figure this out for two days now, and I am tired of it.

I suspect that the code chooses the "wrong" simplex at (0,0,0), but I don't quite see which would be the right one, or how to remedy the problem.

Yay, I of course noticed an error five minutes after posting. And it's looking ok now, so hopefully it was THE error. Fixed code for posterity: http://www.pasteit4me.com/25008