Re: Randomness tests

[Gautier write-only <gautier_niouzes@hotmail.com>]

Nicholas Paul Collin Gloucester:

> Please let us know your conclusions.

So - the mentioned package is indeed the "universal" random generator
from Dr. George Marsaglia, which seems to have appeared in the Ada
letters in 1988, which circulates under the name U_Rand.
According to the machine and compiler, I get a speedup from 5x to 6.8x
(net of overhead).
No more advanced tests yet...
There are a couple of others generators I've to try, e.g. the Mersenne
Twister:
http://adrianhoe.com/adrianhoe/projects/adamt19937/
or a very simple one from Peter:
http://groups.google.ch/group/comp.lang.ada/msg/cf300f84f8782702

I've just "repimped" U_Rand with generics, no more global variables
but a type Generator, and some Ada 95 A.N.Float_Random-style
"renames", which facilitates the switch between random packages, like
in this example:

  -- *** Choice of the floating-point type used for the whole
Portfolio Model:
  subtype Real is Long_Float;

  package Real_U_Rand is new U_Rand(Real);

  -- *** Choice of a random generator: A.N.F_R, or U_Rand (faster),
or...:
  package RRand renames
    -- Ada.Numerics.Float_Random;
    Real_U_Rand;

The funny thing is that RRand can be plugged in its turn into another
generic package

  package GRA is new Ada.Numerics.Generic_Real_Arrays(Real);

  package RCopulas is new Copulas(
    Real,
    RRand.Uniformly_Distributed, RRand.Generator, RRand.Random,
    GRA
  );

In case someone is interested, mail me.
Anytime soon I'll open a SourceForge project with various random
variable goodies.
_________________________________________________________
Gautier's Ada programming -- http://sf.net/users/gdemont/
NB: For a direct answer, e-mail address on the Web site!