The ABINIT Users Mailing List ( CLOSED )

[James Johns <jej7x@xenon.cchem.berkeley.edu>]

Hi,
I'm the relatively inexperienced system admin for a few computers in my 
experimental lab.  I'm  using a SMP 4 intel xenon processor SUSE10.1 
machine, and have compiled and run abinit quite sucessfully, but only in a 
serial capacity.  I'd like to use the larger power of my machine by using 
the parallel version of the code.  To that end, I installed the mpich 
package.  That, however, did not install a header file mpi.h which the 
tutorial on abinip said to look for.  I then installed the mpich 
development package/kit/thingy from SUSE.  This has the mpich header, as 
well as a several libraries for the fortran compiler.  First question, it 
is unclear from the tutorial if I should still use the configuration flag 
FC=ifort (I had used the intel fortran compiler to make the serial 
version).  Also, using the default configuration script, it does not 
detect a native mpi support.    If I use the command configure 
--with-mpi-prefix=/opt/mpich/include (this is where mpi.h is) it also 
can't find anything.  If I use the directory /opt/mpich/ch-p4 it can at 
least find that there is a general parallel environment, but does not 
recognize it as mpi.  This also does not put a copy of abinip in the 
~abinit/src/main directory.  Attempting to run make and then make install 
or make multi and make install both give errors.  In the tutorial, it says 
to replace a directory in the line with a directory you have already 
written down, but early in the same paragraph it says to take note of the 
location of the mpi.h file, the name and location of the MPI library file 
ending in .a (which of the many library files I'm not sure) and the 
dependencies of the MPI Library.  The tutorial I'm reading from is
http://www.abinit.org/Infos_v5.5/tutorial/lesson_parallelism.html

I know discussions of compilation of abinit, and specifically how to 
compile and use the parallel version have come up before, but I have 4 
years of this forum saved on my hard drive, and searching for parallel or 
compilation problems yields far to many results.  If anyone has any 
clarification, that would be greatly appreciated. Thanks,
James Johns
UC Berkeley