Documentation for MPI tranpose routines:

This directory contains in-place transpose routines for MPI, used by
fftwnd_mpi.  These routines can actually be used completely
independently of FFTW, but no documentation is available for them yet.
(Email the authors at fftw@theory.lcs.mit.edu if you have any
questions.)

Note that these routines are specialized for transposing arrays
whose elements are of type double.  If you wish to have different
element types (e.g. if you change fftw_real to float), you must
change the transpose element type definitions in transpose_mpi.h.

You will need to compile these files before using fftwnd_mpi.  To do
this, simply type make in this directory.  The Makefile will probably
have to be modified to reflect whatever incantation is necessary on
your machine to compile MPI programs.

The makefile will additionally create a test program, test_transpose_mpi.

This program is run as follows:

test_transpose_mpi nx ny [el_size]

This will test the transpose on an nx x ny matrix.  The third,
optional parameter, el_size, allows each element of the test matrix to
consist of more than one number.  You should probably run the test
program with various numbers of processes and a few values of each of
these parameters to make sure everything is working correctly on your
system.

	  << DOCUMENTATION FOR SUBROUTINES TO BE ADDED... >>
