Build MCNPX 2.5.F on Power Mac G5 (OS X) with IBM compiler (version 8.1):

(1) Build MPICH 1.2.5.2 using the IBM compiler. If you don't plan to run
    MCNPX in parallel using MPI, skip to step #2.
    (a) Make sure ssh is installed on your system. Then perform the
        following commands:

    ssh-keygen -t rsa
    cp ~/.ssh/identity.pub ~/.ssh/authorized_keys
    chmod go-rxw ~/.ssh/authorized_keys
    ssh-agent $SHELL
    ssh-add

    (b) Build the MPICH libraries:

    cd mpich-1.2.5.2
    ./configure -cc=cc -fc=xlf -f90=xlf90 --prefix=~/mpich-1.2.5.2 -rsh=ssh
    make
    ranlib -cs lib/*.a

(2) Make sure the "xlf90" executable is in your path:
 
    setenv PATH /opt/ibmcmp/xlf/8.1/bin:${PATH}
    setenv F90FLAGS -qsuffix=f=f90

(3) Create an empty file in your $HOME directory called
    libX11.a. This is needed so the configure script can
    find this file and complete execution. We will fix this
    in the future so it will find the libX11.dylib in
    /usr/X11R6 (or in the path specified with --x-libraries).

(4) Make a build directory and run configure. The sequential
    version is built with the following commands:

    mkdir v25f_bld
    cd v25f_bld
    ../v25f/configure --host=rs6000 --with-FOPT="-O2" --with-FC=xlf90
       --with-FFLAGS="-qfixed -qmaxmem=-1"
       --with-CFLAGS="-I/usr/X11R6/include"
       --x-libraries=$HOME --x-includes=/usr/X11R6/include

    And the MPI version with:

    mkdir v25f_bld
    cd v25f_bld
    ../v25f/configure --host=rs6000 --with-FOPT="-O0" --with-FC=xlf90
       --with-FFLAGS="-qfixed -qmaxmem=-1 -I$HOME/mpich-1.2.5.2/include"
       --with-CFLAGS="-I/usr/X11R6/include"
       --with-MPILIB="-L$HOME/mpich-1.2.5.2/lib -lmpich -lpmpich" --with-MPICH
       --x-libraries=$HOME --x-includes=/usr/X11R6/include

    Note for MPICH2 that the --with-MPICH option should be omited
    and the MPILIB option becomes (I know its crazy!):

       --with-MPILIB="-L$HOME/mpich2/lib -lmpich -lpmpich -lmpich"

    A 64-bit integer version can be created by adding the
    "-qintsize=8" compiler flag to the FFLAGS option. The
    following change must be made to the MPICH source to create
    an 8-byte integer version:

    (a) This is a bit strange, but with the XLF compiler you must
        also edit the MPICH include file, called mpif.h, (usually in
        mpich-1.2.5.2/include) and replace INTEGER with INTEGER*4
        (both upper and lower case).

(5) Edit src/mcnpx/Makefile.h and change the XLIBS variable to
    point to the correct location:

        XLIBS = -L/usr/X11R6/lib -lX11

(6) Run "make".

(7) Run the test suite:
    (a) Edit the v25e_bld/src/Test/Makefile.h file and set the TFC
        variable to:

        TFC = xlf

    (b) Run "make".
    (c) The "-O2" optimization level results in tracking differences
        in INP118.The MPI version results in tracking differences in
        INP102 and HTP112 (which are expected). The "-O2" MPI version
        crashes on INP117.