home *** CD-ROM | disk | FTP | other *** search
- Path: news.ioffe.rssi.ru!usenet
- From: "Eugene I Levin" <Levin@lh.ioffe.rssi.ru>
- Newsgroups: comp.lang.c,comp.lang.c++
- Subject: Re: RANDOM NUMBER GENERATOR
- Date: Fri, 5 Jan 1996 16:18:24 +0300 (MSK)
- Organization: Ioffe Institute
- Distribution: world
- Message-ID: <ABWKIxmi6N@lh.ioffe.rssi.ru>
- NNTP-Posting-Host: ccpti.ioffe.rssi.ru
- Mime-Version: 1.0
- Content-Transfer-Encoding: 8bit
- X-Mailer: dMail [Demos Mail for DOS v1.23]
-
- Bill,
-
- > The recent paper:
- > Vattulainen et al. (1994) Physical Review Letters, 73, 2513-2516
- > finds problems with your generator which they call R250.
- >
- > A linear congruential generator GGL passes the tests.
-
- Many thanks :-( ... I really trusted the generator: a lot of Monte-
- Carlo simulation of Coulomb-interacting particles has been done with
- it, and from our experience the Coulomb energy is VERY sensitive to
- bad generators: that's how we sorted our some bad congruential
- generators (including infamous IBM "RANDU"). Also I ran just \chi^2
- test of 3D distribution of consequent numbers of r250, and the
- results were EXCELLENT.
-
- I'll try to find the paper (it's not that easy nowadays in Russia:
- central St.Petersburg library has no money to subscribe Phys.Rev. :-().
-
- Again: many thanks for the warning.
-
- Regards,
- --
- Eugene I Levin, St.Petersburg, Russia (Levin@lh.ioffe.rssi.ru).
-
-