An optimized version of openQCD-1.2 for BlueGene/Q and other machines
=====================================================================

DESCRIPTION

   This software is a modification of the original openQCD code 
   from http://luscher.web.cern.ch/luscher/openQCD/openQCD-1.2.tar.gz
   which was written by Martin Luscher and Stefan Schaefer 

   The additional optimizations include:

   1) use of intrinsics for QPX instructions for BlueGene/Q

      These instructions are only available on BlueGene/Q and
      with the IBM xlc compiler. They are enabled by default
      (by including "-DQPX" in CFLAGS in main/Makefile).

   2) implementation of global communications by different MPI functions

      Wrapper functions for global summation and broadcast allow
      to implement them by different MPI cunctions (e.g. using 
      Reduce + Bcast or Allreduce for global sums). 

      This can be used also on other machines, e.g. SuperMUC, and 
      may help to significantly reduce the time spent in global 
      communications or to work around unpleasant features of some 
      MPI implementations.

      The MPI functions used are selected in modules/utils/utils.c 
      through the CPP macros USE_MPI_BCAST and USE_MPI_ALLREDUCE.
      The default setting has been found benificial on BlueGene/Q
      and SuperMUC.

AUTHORS

   The original openQCD code has been written by Martin Luscher
   and Stefan Schaefer with contributions from others (see
   http://luscher.web.cern.ch/luscher/openQCD/)

   The optimizations and modifications with respect to the
   original openQCD code have been implemented and tested by
   Dalibor Djukanovic, Mauro Papinutto, and Hubert Simma.

LICENSE

   The code may be used under the terms of the GNU General Public License (GPL)
   http://www.fsf.org/licensing/licenses/gpl.html