# Three programs:  scilab (v3.1.1), matlab (v7sp3), comsol script (v3.2)
# Both the proprietary ones are just single-threaded for everything!
# Scilab has had "atlas" lapack and other libraries compiled for 4 P4 CPUs
# but this is not ideal:  there are only 2CPUs (each hyperthreaded) and
# they are Xeons (any difference apart from cache?).
# So, let's assume scilab's got about the equiv. of 2 CPUs...
#
# 2005-09-27
#
__________________________
INVERSE
  clear a b
  a = rand(3000,3000);
  tic
  b = inv(a);
  clear a b
  toc

Scilab:   9.27  9.26  9.36
Matlab:  15.01 15.02 15.28
Comsol:  19.39 19.23 19.32

_________________________
LU FACTORIZATION
  clear a b
  a = rand(3000,3000);
  tic
  [bl, bu] = lu(a); 
  toc

Scilab:  4.85 4.51 4.51
Matlab:  5.36 5.37 5.39
Comsol:  7.56 7.23 7.46 


_________________________
FFT
  clear a b
  a = rand(1,2^24);
  tic
  b = fft(a);
  toc

Scilab: 36.34 36.22 36.27  # the FFT routine is _not_ SMP
Matlab:  6.84  6.87  6.85
Comsol:  6.99  7.01  7.02


__________________________