You will need to link the code with the International Mathematical 
and Statistical libraries (IMSL) to access the subroutine that finds 
the roots of the polynomial and with the GQOPT libraries to perform 
the numerical maximization of the likelihood function.  If you have 
other routines to perform these numerical procedures it should be fairly
easy to rewrite the code.  I follow the notation in my paper ``Maximum 
Likelihood Estimation of Stationary Univariate Fractionally Integrated 
Time Series Models'' in the Journal of Econometrics, 1992.

Once the coded is running this data set should produce the output files.  
The results are from my paper  ``Modeling Long Run Behavior with the 
Fractional ARIMA Model,'' in the Journal of Monetary Economics, 1992.