Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uwm.edu!bionet!hayes.fai.alaska.edu!acad3.fai.alaska.edu!ftpam1
From: ftpam1@acad3.fai.alaska.edu (MUNTS PHILLIP A)
Newsgroups: comp.lang.pascal
Subject: Re: help on random number generator & tpu display requested
Message-ID: <1990Jul21.211554.13179@hayes.fai.alaska.edu>
Date: 21 Jul 90 21:15:54 GMT
References: <23899@adm.BRL.MIL> <1990Jul19.180723.8770@hayes.fai.alaska.edu> <1990Jul21.115203.2435@maytag.waterloo.edu>
Sender: usenet@hayes.fai.alaska.edu (J Random USENET)
Reply-To: ftpam1@acad3.fai.alaska.edu
Organization: University of Alaska Fairbanks
Lines: 66
News-Software: VAX/VMS VNEWS 1.3-4

In article <1990Jul21.115203.2435@maytag.waterloo.edu>, dmurdoch@watstat.uwaterloo.ca (Duncan Murdoch) writes...
>In article <1990Jul19.180723.8770@hayes.fai.alaska.edu> ftpam1@acad3.fai.alaska.edu writes:
>>In article <23899@adm.BRL.MIL>, NELIS%NLR.nl@cunyvm.cuny.edu (W.J.M. Nelis, NLR-NOP) writes...
>>>-1- Which pseudo random number is used by TP?  It is a multiplicative
>>>    linear congruential generator, but I can't determine its parameters.
>>
>>     See the article in the October 1988 issue of "Communications of the ACM"
>>for stuff on random number generators.  Don't use the built-in version.
> 
>What problems does the built-in one have?
> 
>Duncan Murdoch

     Most presently used random number generators have arithmetic overflow
in the internal calculations, which causes strange things to the sequences.
Hence the large body of lore concerning which seeds to use (or not use!).  A
generator without overflow (the article referenced above evades it with 
various clever techniques) will always generate the entire sequence, no matter
what seed you use.  The article is actually pre-5.5, so it is possible that
they have fixed it.

     Last year in a DSP class we did some simulations.  Generators used by
other students included those intrinsic to MS Basic and MS Fortran and generated
noticeably spiky results.  I used the "Minimal Standard Generator" described
in the article and my results were absolutely flat.

     I will append my random number generator unit.

Philip Munts N7AHL
NRA Extremist, etc.
University of Alaska, Fairbanks

UNIT MSRandom;

INTERFACE

  FUNCTION random(VAR seed : longint) : real;

IMPLEMENTATION

(* From CACM Oct 1988 *)

  FUNCTION random(VAR seed : longint) : real;

    CONST
      a = 16807;
      m = 2147483647;

      q = 127773;
      r = 2836;

    VAR
      lo, hi, test : longint;

  BEGIN
    hi := seed DIV q;
    lo := seed MOD q;
    test := a*lo - r*hi;
    IF test > 0 THEN
      seed := test
    ELSE
      seed := test + m;
    random := seed/m;
  END;

END.