Good "deteriministic" RNG algorithm?

I'm trying to find something that will allow me to initialize it with a seed, and for every set of 3-space coordinates I pass it, it will output a random integer - the integer needs to always be the same for each coordinate set and should vary greatly according to the seed. I've played around with a few algorithms that work ok - mostly of my own creation, but I lack the statistics knowledge to know how "random" my "muss it all up - but logically" algorithms are.

Is there a well known creme de la creme method for doing this?

 

If I understand you correctly you're wanting a unique random integer for each vector3 coordinate, and that random number should always be the same for that particular vector3 coordinate. The standard way that's done is as follows:

Code (JavaScript):

1. Random.seed = x * 100 + y * 50 + z * 10;

The number each axis is multiplied by can be changed to whatever you want and so long as you always multiply by the same numbers you'll end up with a unique random integer for every unique vector3 coordinate.

I too have tried a lot of algorithms I've created myself, some of them pretty complex, and none have ever given me results as good as this simple formula.

 

I was trying to avoid reinitializing a "Random" class each time... Although I haven't tested the speed at which Random is initialized. It's for a 3-d Perlin algorithm, so there's going to be a lot of coordinates and it needs to be as fast as possible. I may do some Diagnostics.Stopwatch tries on your method.

Here's the source code for System.Random() in .NET - as you can see the initialization sequence is a whole bunch of words: http://www.dotnetframework.org/defa...lr/src/BCL/System/Random@cs/1305376/Random@cs