Message-Id:	<200006091512.LAA18302@delorie.com>
From:	"Dieter Buerssner" <buers AT gmx DOT de>
To:	djgpp-workers AT delorie DOT com
Date:	Fri, 9 Jun 2000 17:20:30 +0200
MIME-Version:	1.0
Subject:	Re: rand() comparison
In-reply-to:	<3940EA71.BDD8028F@ecn.nl>
X-mailer:	Pegasus Mail for Win32 (v3.12b)
Reply-To:	djgpp-workers AT delorie DOT com
Errors-To:	nobody AT delorie DOT com
X-Mailing-List:	djgpp-workers AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

Teun Burgers wrote:

> [...] Several generators with good
> sequences are known. I think they are essentially based
> on combining several simple linear congruential generators
> to get a longer sequence.

Some, yes. But not the two you pointed out. 

> Perhaps Dieter can test this one also.

> Perhaps Dieter can test this one also.

From the attachment:

/* This is an implementation of the TGFSR (twisted generalized
 * feedback shift register) random number generator TT800, which was
[...]

If this comment is correct, it will pass all my tests. (I used my own 
implementation of tt800)

Also, Mersenne Twister (by the same researcher), that you pointed out 
in your other mail, will pass all my tests.

They both need about 3 times as long as the MWC32 I suggested for 
DJGPP. (In some situations, probably more, because especially 
Mersenne Twister needs a large table for internal state. When not 
using it in a tight timing loop, this table may not be in the cache 
anymore)

IHMO, anybody who needs such a strong PRNG, will not depend on 
rand().

- Raw text -