Programs contained in this directory contain no array operations (only
using FORALL or parallel loops).

Therefore the flag -F77 or -F90 have the same meaning (F77 loops will
not translated to F90 array operations).

Both programs check their results by comparing the results with the 
original values.

a) FFT
======

fft.f: Fast Fourier Transformation (indirect addressing)

  Algorithm applies to a random sequence the fft and its inverse. It
  compares the original sequence with the new computed sequence.

  Output (should be near to zero):

      Max =   2.2942879E-07

  On Alliant FX/2800 there are problems with optimizations switched on
  (use -Og instead of -Oguv)

  With Adaptor 1.0 the old version runs in some problems due to the array
  normalization (lower bounds of arrays must be 1).

b) LU and MATMUL
================

lumm.f: LU Decomposition and Matrix Multiplication

  The Algorithm takes a n x n matrix with random values and makes an
  LU Decomposition. Then it computes the inverse to a.

  Finally a and its inverse are multiplied. The result must be the
  unity matrix.

  Output (maximal eps should be near to zero):

     Arrays are allocated
     PROGRAM STARTS
     LU ready
     now finds Y with L * Y = E
     now finds X with R * X = E
     b is the inverse
     lu = a * b
     matmul correct, now check = E
     maximal eps =   2.4130940E-04

