Re: MPI code error raises a question

On 2008-05-08 04:02:00 -0300, Janne Blomqvist <foo@[EMAIL PROTECTED]
> said:

> On 2008-05-08, Richard Maine <nospam@[EMAIL PROTECTED]
> wrote:
>> If, perhaps, you are talking about problems relating to TKR-cheating,
> 
> Yes, this is essentially it. From
> 
>
http://www.open-mpi.org/papers/euro-pvmmpi-2005-fortran/euro-pvm-mpi-2005-fortran.pdf

"""
Not 
> 
> only must interfaces be defined for arrays of each intrinsic data
> type, but for each array dimension as well. Depending on the compiler,
> there may be approximately 15 type / size combinations.1 Each of these
> combinations can be paired with up to a maximum of seven array
> dimensions. With approximately 50 MPI functions that have one choice
> buffer, this means that 5,250 interface declarations must be specified
> (i.e., 15 types × 7 dimensions × 50 functions).  Note that this does
> not include the approximately 25 MPI functions with two choice
> buffers. This leads to an additional 6.8M interface declarations
> (i.e., (15× 7 × 25)2).
> """

Most arguments of that type are followed by a comment "and then I
implemented a trivial text processing program to produce all of them".
Sounds like a marvelous project for a student in need of a practical
problem. A bit of clever use of generic forms may be called for.

> In practice, implementations either resort to implicit interfaces in
> the MPI module, or provide a subset of explicit interfaces and expect
> the programmer to change to the old
> 
> INCLUDE 'mpif.h'
> 
> thing in case they do something outside the subset defined in the mpi
> module. Or something like that.
> 
> For MPI 3, there is an effort underway to implement the kind of
> bindings with ISO_C_BINDING as is described in the paper. See
> 
> http://svn.mpi-forum.org/trac/mpi-forum-web/wiki/FtnWikiPage