COMPILING THE CODE
==================

This directory contains the code for calculating the non-relativistic Gaunt
factor. To compile the program, you need an MPI installation based on a recent
g++ version (4.7.0 or later). You also need libgmp and libmpfr installed
(including the devel packages). The code also needs libmpfrc++ written by
Pavel Holoborodko. This library is included in the tarball since it is usually
not available in Linux distros.

Before compiling the program, you need to take a look at the file gauntff_nr.cc.
Near the top are three preprocessor macros that you can define to choose which
form of the Gaunt factor you wish to compute.

#define GENERATE_NONAV 1

When this statement is uncommented, a table of non-averaged Gaunt factors will
be created. This generates the file gauntff_nonav.dat available on
http://data.nublado.org/gauntff

#define GENERATE_THERM_AV 1

When this statement is uncommented, a table of thermally averaged Gaunt
factors will be created. This generates the file gauntff.dat available
on http://data.nublado.org/gauntff

#define GENERATE_TOTAL 1

When this statement is uncommented, a table of total (frequency integrated)
Gaunt factors will be created. This generates the file gauntff_freqint.dat
available on http://data.nublado.org/gauntff

Exactly one of these three statements needs to be uncommented. In the default
setup a table of thermally averaged Gaunt factors will be created.

To compile the code, execute the following commands in some suitable location

tar xvfz gauntff_nr.tar.gz
cd gauntff_nr
make

The code was tested with openMPI, which supplies a compiler wrapper called
mpiCC. Other distros can be used too, but they may give another name to the
wrapper (such as mpicxx). If this is the case with you distro, you can compile
the code as follows

make CXX=mpicxx

You can clean up the files left behind by a compilation with the command

make clean


RUNNING THE CODE
================

The calculation are rather CPU intensive due to the fact that high precision
math routines are being used (with mantissas up to 4096 bits). To keep the
execution time within reasonable bounds, the code needs to be run on a
multi-core machine or cluster using MPI. You can run the code as follows

mpirun -np 32 ./gauntff_nr.exe > gauntff.dat

The number of ranks can be freely chosen (the number 32 shown above is just an
example), except when you are calculating the total Gaunt factor. In the
latter case the number of ranks needs to be a multiple of 4. One rank (in the
case of the total Gaunt factor, 4 ranks) will be reserved as a master rank for
handing out jobs to the other ranks. The master rank will not do any real
calculations itself.


LICENSE
=======

The code is distributed under the GNU GPL v3. The license text is available in
the file license.txt included in this tarball.

When you make use of the code in a scientific publication, please refer to the
following paper

van Hoof P.A.M., Williams R.J.R., Volk K., Chatzikos M., Ferland G.J., Lykins
M., Porter R.L., Wang Y., Accurate determination of the free-free Gaunt factor,
I -- non-relativistic Gaunt factors, 2014, MNRAS, 444, 420


CREDITS
=======

Peter A.M. van Hoof
Royal Observatory of Belgium
Ringlaan 3
B-1180 Brussels
Belgium
p.vanhoof@oma.be