mateu / Algorithm-Simplex

Tucker Tableau version of Simplex Algorithm.

The benchmark_random_LPs.pl script shows how to 
use the Models of which there are three:

* Float - Normal Perl Numbers
* Rational - Provided by Math::Cephes::Fraction.=
* PDL - Use piddles for execution speed in larger sized problems

Also see t/example_LPs.t for how to implement a solver subroutine.

The normal Perl types model has OK performance and is actually the fastest 
up to around 200x200 matrices.

The Fractions (Rationals) model uses the module Math::Fractions::Cephes.
It is the slowest, but it is exact when starting with a tableau (matrix) of 
rationals. 

PDL Model uses the Perl Data Language.  It is intended for larger sized 
problems as it's the fastest choice when the tableau is > 200x200.  It has
been tested on matrices of 2000x2000 in size.  Don't be fooled - PDL is fast.
Hey it's C and fortran doing the arithmetic operations.

The algorithm implementation was strongly influenced by Nering and Tucker's
book "Linear Programs and Related Problems", along with the guidance of
George McRae of the University of Montana.  Any errors are solely mine  ;)