OT: Help with Testing Random Number Generators

[Namir]

Hi All,

I am interested in learning about testing for high performance random number generators that can be used in Monte Carlo methods. Such methods require "Rolls Royce" versions of RNG and not you regular "Ford" versions (apologies to car makers for comparison).

Any good links to such tests?

Thanks!

Namir

Edited: 1 Aug 2011, 12:06 p.m.

[spb]

The standard starting point for serious randomness testing is probably Marsaglia's diehard tests. It is over a decade old, and it doesn't include some modern approaches (phase space visualisations, for example), but it's probably a pretty good starting point.

What's the application? It isn't really meaningful to talk about any measure of randomness in the abstract; you can only meaningfully talk about whether some measure is suitable for a particular application, threat model, or whatever.

[Allen]

Chapter 3 of TAOCP is devoted to RNG entirely. Although I'm sure other people have done more recent work, it's still worth the 175 pg read.

[Namir]

Thanks guys for the tips!

Namir

[htom trites jr]

It's a joy to see someone asking about testing PRNG, rather than how to make fast ones. A good PRNG is a hard problem to solve.

[Paul Dale]

The GNU Scientific Library has some links to generators and test suites.

- Pauli