guile-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Nearly finished (re)integrating GMP for bignums.


From: Kevin Ryde
Subject: Re: Nearly finished (re)integrating GMP for bignums.
Date: Thu, 13 Feb 2003 09:00:47 +1000
User-agent: Gnus/5.090013 (Oort Gnus v0.13) Emacs/21.2 (i386-pc-linux-gnu)

Mikael Djurfeldt <address@hidden> writes:
>
> 2. What do you mean by a "good" RNG?  I'd say it is good if it is
>    fast.

In gmp we're looking at a Blum-Blum-Shub algorithm for possible
inclusion, it's based on cryptographic principles, and so supposedly
suits such applications (which are rather specialized of course).

>    Our generator is fast.  Can you name any single other
>    criterion which is not "religious" and which our generator doesn't
>    fulfil?  (I'm talking about something measurable here.)

For reference, the next gmp will be have a Mersenne Twister generator
contributed by Pedro Gimeno, and it'll be the default too
(gmp_randinit_default).  It's fast, and very random according to those
who pay attention to such things (I'm no expert).

Rob Browning <address@hidden> writes:
>
> I wonder how fast the GMP RNGs are (I currently have no idea).

The current linear congruential generator isn't as good as it could
be, it uses bignum operations even for the common case of a modulus
that fits in one or two machine words.

Mersenne Twister ought to end up being as fast or faster than L-C, and
the lack of randomness of L-C is well-known, so there oughtn't be any
direct call for the latter any more, not except maybe for special
purposes like experimenting with very un-random sequences.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]