source: pypar/contrib/schiotz @ 622

Dear Ole Nielsen,

I saw your announcement of Pypar on comp.lang.python.announce.  In our
programs, we are using Konrad Hinsen's Scientific Python module.  I
have a small technical question.  You write that one advantage of your
approach is, that one does not need a special version of the Python
executable, just the pypar module.  In Scientific Python it is
necessary to modify Python because MPI requires that MPI_Init is
called before the program looks at its command line arguments, and
Python needs to look at its command line arguments to figure out which
script to run.  How do you bypass this problem?  I know there are MPI
implementations where this is not a problem as MPI only adds extra
stuff to the command-line on the non-master nodes, but there are some
implementation (MPICH, for example) where the whole command line is
replaced with MPI stuff and the command line is only restored when
MPI_Init is called.

I am very interested to hear if you have found an elegant way to solve
this, as it is slightly annoying to have to have an extra version of
the Python executable.


Best regards

Jakob Schiotz


-- 
Jakob Schiotz, Ph.D.             !  Email: schiotz@fysik.dtu.dk
CAMP, Department of Physics      !  WWW: http://www.fysik.dtu.dk/~schiotz/
Technical University of Denmark  !  Phone: (+45) 45 25 32 28
DK-2800 Lyngby, Denmark          !  Fax:   (+45) 45 93 23 99