Thanks a lot for your comments Jeff..
I will try some of your advices.. and further I will let you know... in the mean time we can try at least to convince the old school of MPIers to include the Pascal interface... :)
2007/10/22, Jeff Squyres <firstname.lastname@example.org>:
On Oct 22, 2007, at 6:44 PM, Lourival Mendes wrote:
> Hy everybody, I'm interested in use the MPI on the Pascal
> environment. I tryed the MPICH2 list but no success. On the Free
> Pascal Compiler list, DaniŽl invited me to subscribe this list and
> open a discussion on the interface of OpenMPI for Pascal.
> Probably as DaniŽl knows there is almost no reference on the MPI
> for Pascal interface, only some very few tryes, one of them in
> I would like to know if there is someone working on the interface
> of OpenMPI for Pascal?
There was a mail or two about it a while ago; you might want to dig
through the OMPI list archives. The short version is that none of
the current Open MPI members have a desire to add Pascal bindings to
MPI. It also might be somewhat of an uphill battle to convince the
old-school MPI'ers to include a Pascal interface in Open MPI, even if
it was developed by a 3rd party and contributed to the project.
However, that should not deter you from pursuing a Pascal interface
if you want one. Traditionally, extensions to MPI have been
implemented in an MPI-neutral fashion and released into the wild as
3rd party libraries (such as the C++ bindings for MPI several years
ago). The Pascal bindings likely don't need to know anything about
the internals of an MPI implementation -- they can just call the C
bindings. So it's possible/likely that you would write up a Pascal
interface that would work with both Open MPI and MPICH (and any other
MPI's out there).
As I typed out the above, I have a dim recollection of the Pascal
interface needing to obtain the values of the C constants during its
setup/compilation phase (note that these values are going to be
different between different MPI implementations). You have a few
options here; you could write a parser for mpi.h to extract the
values you need (
e.g., in perl or somesuch) or write a short C
program to extract them and printf the values that you capture into a
Pascal header file or something (I'm not sure if you need the literal
or symbolic values -- I remember very little of Pascal). Either way,
with a little diligence and creativity, it could probably be done.
> Also I'm putting a very newbie question: What is the main
> difference between the OpenMPI and MPICH ?
Short version: we're different projects implementing the same API
users mailing list
Lourival J. Mendes Neto