Re: [Sbcl-devel] Quality of RANDOM

[Nikodemus S. <nik...@ra...>]

Joubert Nel wrote:

> The post also says that CMUCL uses a Mersenne Twister to generate
> randoms, and hence does not exhibit the ACL issues.
> After some reading I found that there's RANDOM-MT19937, which seems to
> be the RANDOM implementation on SBCL.
> http://sbcl-internals.cliki.net/RANDOM-MT19937

> MT is a very good P-RNG but am wondering whether there are known
> theoretical limitations of SBCL's RANDOM implementation?

As you have discovered, SBCL uses the Mersenne Twister, and there should 
be no limitations that apply to SBCL's implementation but not to MT 
generally.

Or rather, any such limitation would be a bug in SBCL's implementation 
of MT, and no such bugs are known.

Cheers,

  -- Nikodemus