[MPI3 Fortran] Agenda for MPI3 Fortran Working group next week

[Jeff Squyres]

On Jun 5, 2009, at 1:14 PM, Hubert Ritzdorf wrote:

> ! I know that this is not portable and that this is not intened by  
> MPI.
>
> ! But I have seen such usage of derived types in application programs.
> ! Application programmers use this byte stream since it's
> ! too "difficult" to set up a derived MPI datatype for large
> ! Fortran derived tyes and since this works on "all" (they claim)  
> systems.


It will not work in C if the MPI has pointer handles, and it will not  
work in C++.

I do not think that we need to constrain advances in the bindings  
based on erroneous application codes.

Plus, to use the new interface, they'll have to change their codes  
anyway (use mpi -> use mpi3).  So they can choose to keep using the  
old interface where that "works".

-- 
Jeff Squyres
Cisco Systems