I am making something that requires a very good randomizer but cannot find one that doesn't have some sort of bias in one output or another. Typically the main thing I see is that in liquid/gas randomizers there is always a bias to the left or right because TPT actually renders from left to right. (and up to down)

I was thinking that perhaps if it hasn't been done already you could combine multiple random data streams with each other and switch around their most significant bits? This could possibly negate the left/right bias and make the output closer resemble the random numbers the TPT code is generating.

When I said "render" I guess I meant "update". Anyway I am fairly certain this causes some sort of bias when it comes to moving particles. I went through with my idea and combined the outputs of multiple randomizers and the results are pretty good.


Each 8-bit randomizer has its bit significance shifted like so:

001 002 004 008 016 032 064 128
128 001 002 004 008 016 032 064
064 128 001 002 004 008 016 032
032 064 128 001 002 004 008 016
016 032 064 128 001 002 004 008
008 016 032 064 128 001 002 004
004 008 016 032 064 128 001 002
002 004 008 016 032 064 128 001

All the outputs are then XORred together into one 8-bit value.

My hope is that this will offset any kind of left/right bias in the movement of the particles that generate the output; and thus get closer to that glorious system - rand() function embedded in the C++.

Next I want to be able to generate a random number between a lower and upper limit while still utilizing the 0-255 output. Does anyone know of an efficient algorithm to do this?