Developer's notes:

(1) Download the tar balls and retrieve the source code, utility code, libraries;
(2) study Makefile (cf. end of beta_notes for more info);
(3) build graph partitioning library ParMetis - follow instruction inside;
(4) make executible;
(5) selfe_verification_tests/ has a number of tests, most of which are up to date; 
(6) useful info about code and input format changes can be found in beta_notes.

About the source code: 
(1) all serial code options (and more) are now available in MPI code;
(2) Input files: the main differences between serial and MPI version are in param.in, hotstart.in 
    and nudging inputs for S,T. Note that param.in now uses a free format; see user manual for details.

How to run:

Remember to make a subdirectory inside the run directory called 'outputs/'
where all process-specific outputs are stored.

On Intel, 

mpiexec -n 24 pelfe

Ask your system manager if your system is different.

Postprocessing:

Currently each process will spit out results
 within its own subdomain, and the latest code combine_output*.f90 inside Utility/
 can be used to gather these files
 into one global output. You may use the simple perl script autocombine_MPI_elfe.pl to do the combining automatically. 

Most outputs are process-specific also. This is the case for hotstart outputs, 
and so in order to hotstart, you need to combine outputs into one hotstart.in
 using the latest combine_hotstart*.f90 in Utility/.

