Re: random number generation -> PLEASE HELP!

Re: random number generation -> PLEASE HELP!

[Dr. John B. Matthews]

In article <vi8k9uh3r59.fsf@fnal.gov>, Oleg Krivosheev <kriol@fnal.gov> writes:
> ica2ph@alpha1.csv.ica.uni-stuttgart.de (Peter Hermann) writes:
>> Justin Vandenbroucke (rva036@lulu.acns.nwu.edu) wrote:
>> : Does anyone know how to generate "random" numbers in pascal?  I know BASIC
>> : has a standard function that does it, and I know there is some actual code
>> 
>> Ada95, the natural successor of Pascal, has a standard function, of course.
>> see http://lglwww.epfl.ch/Ada/
> 
> C, the natural successor of assembler, has it too... And who cares?
> THere are no words in standard about required rnd genertor quality, so
> it's hardly usefull for anything but (simple) games

Nonsense. GNAT (the GNU Ada compiler from NYU; http://www.gnat.com/)
and MacOS (http://www.apple.com) have very satisfactory psuedo-
random number generators. The former is available in source from
ftp://ftp.cs.nyu.edu/pub/gnat; the latter is shown below.

unit ACMRandom;

interface

   procedure SetACMSeed (seed: LongInt);
   function ACMRandom: LongInt;

implementation

   var
      acmRandSeed: LongInt; {static variable}

   procedure SetACMSeed (seed: LongInt);
   begin
      if seed = 0 then
         seed := 1;
      if seed < 0 then
         seed := -seed;
      acmRandSeed := seed;
   end;

   function ACMRandom: LongInt;
{"minimal standard random number generator" proposed by}
{Parker & Miller in CACM, 1988; it has period (2^31)-1}
      var
         lo, hi, test: LongInt;
   begin
      hi := acmRandSeed div 127773;
      lo := acmRandSeed mod 127773;
      test := 16807 * lo - 2836 * hi;
      if test > 0 then
         acmRandSeed := test
      else
         acmRandSeed := test + maxLongInt;
      ACMRandom := acmRandSeed;
   end;

end.

program RandomTest;

   uses
      QuickDraw, ACMRandom;

   var
      i: Integer;

begin
   randSeed := 123456789;
   SetACMSeed(randSeed);
   for i := 1 to 25 do
      WriteLn(Random : 7, randSeed : 12, ACMRandom : 12);
end.

John
----------------------------------------------------------------
Dr. John B. Matthews
jmatthews@nova.wright.edu; john_matthews@ccmail.dayton.saic.com
"Whom the gods would destroy, they first invite to program in C"

Re: random number generation -> PLEASE HELP!

[Robert I. Eachus]

In article <1996Sep5.165031@nova.wright.edu> jmatthews@nova.wright.edu (Dr. John B. Matthews) writes:

 > Nonsense. GNAT (the GNU Ada compiler from NYU; http://www.gnat.com/)
 > and MacOS (http://www.apple.com) have very satisfactory psuedo-
 > random number generators. The former is available in source from
 > ftp://ftp.cs.nyu.edu/pub/gnat; the latter is shown below...

   Calling the Minimal Standard Random Number Generator good is true
only in comparison to lots of other junk generators out there.  The
generator that comes with GNAT is much better.  If you want the gory
details as to why send me email.  (Or you can just read the published
literature if you want to find out why linear congruential generators
are not very good. Of course, in any case, I really wouldn't trust
implementations which get the credits that badly wrong.  (Among other
things the CACM paper was by Stephen PARK and Keith Miller, but the
generator is actually due to Lewis, Goodman, and Miller. However the
implementation here is apparently from a third source.  The constants
do check out though.)


--

					Robert I. Eachus

with Standard_Disclaimer;
use  Standard_Disclaimer;
function Message (Text: in Clever_Ideas) return Better_Ideas is...

Re: random number generation -> PLEASE HELP!

[Philip Brashear]

>Nonsense. GNAT (the GNU Ada compiler from NYU; http://www.gnat.com/)
>and MacOS (http://www.apple.com) have very satisfactory psuedo-
>random number generators. The former is available in source from
>ftp://ftp.cs.nyu.edu/pub/gnat; the latter is shown below.
>

Version 2.1 of the Ada Compiler Evaluation System has tests for checking the
performance of Ada 95 random value (not just number) generators.  As a side
effect, these tests output some (minimal) statistics relevant to the actual
distribution of the generated values.

The ACES is available at sw-eng.falls-church.va.us, anonymous ftp, directory
public/AdaIC/testing/aces

URL is  http://sw-eng.falls-church.va.us/adaic/testing/aces

Questions?  Contact me.

Phil Brashear
CTA INCORPORATED
(513) 258-0831
brashear.cta@juno.com
brashear@sw-eng.falls-church.va.us

Re: random number generation -> PLEASE HELP!

[Uri Raz]

Look at http://nr.harvard.edu/nr/ - the numerical recipes books contain
  a chapter about random number generators, as well as refernces to other
  books (e.g. Knuth's, which is easily accessable) about the subject.

 Hope this helps,
                  Uri Raz.

 +---------+--------------------+-------+-------+
 | Uri Raz | uraz@iil.intel.com |  Noir |  :-)  |
 |  All opinions are mine. Others may share it. |
 +----------------------------------------------+

Re: random number generation -> PLEASE HELP!

[Richard A. O'Keefe]

Uri Raz <uraz@iil.intel.com> writes:

>Look at http://nr.harvard.edu/nr/ - the numerical recipes books contain
>  a chapter about random number generators, as well as refernces to other
>  books (e.g. Knuth's, which is easily accessable) about the subject.

Urrrr, the Numerical Recipes books have been severely mauled in review.
Their C version in particular is, well, less than idiomatic.
The errata for volumes 1, 2, and 3 of The Art of Computer Programming
are available:  visit Knuth's Web page.  Anyone seriously interested in
random number generation and planning to take advice from Knuth should
*definitely* read the errata to volume 2.  A lot of people have
implemented "Lagged Fibonacci" generators based on that book, but recently
they were found to badly fail a fairly simple statistical test.  This
does mean that the lagged Fibonacci generator in Numerical Recipes should
not be used any longer.  Knuth's errata describe the problem and offer a
choice of two simple fixes.  The code is even in C this time, not MIX!

I offer this definition: "good pseudo-random number generator:
 one whose fatal flaws have not yet been discovered".

-- 
Australian citizen since 14 August 1996.  *Now* I can vote the xxxs out!
Richard A. O'Keefe; http://www.cs.rmit.edu.au/%7Eok; RMIT Comp.Sci.