Add several datatypes to Fortran and C++ bindings

[ompiteam]

MPI-2.1 states:

MPI_LONG_LONG_INT, MPI_LONG_LONG (as synonym), MPI_UNSIGNED_LONG_LONG, MPI_SIGNED_CHAR, and MPI_WCHAR are moved from optional to official and they are therefore defined for all three language bindings.

We have all of these types in mpi.h, but a quick glance shows that we don't have them in mpif.h and some are missing from the C++ bindings.

[ompiteam]

Imported from trac issue 1368. Created by jsquyres on 2008-07-01T09:14:10, last modified: 2014-03-26T19:47:42

[ompiteam]

Trac comment by jsquyres on 2008-07-11 18:26:00:

I'm now quite confused by this ticket. We argued about C/Fortran datatypes in Menlo Park (w.r.t. MPI-2.1). Specifically, as noted in the initial ticket, it says that these names are to be defined in all 3 language bindings. But:

• Neither OMPI nor MPICH2 define C datatypes in mpif.h (e.g., MPI_INT, MPI_FLOAT, etc.)
• Both OMPI and MPICH2 define Fortran datatypes in mpi.h (e.g., MPI_INTEGER, MPI_DOUBLE, etc.)
• Ditto for C++ (e.g., MPI::INTEGER, MPI::DOUBLE, etc.)

But the wording of of the item says "defined for all 3 language bindings". Gah! Are both OMPI and MPICH2 out of spec because we don't have Fortran bindings for all the existing C types, and this just highlights the problem? Or is the wording in the changelog wrong -- they ''shouldn't'' be defined in Fortran?

I argued fervently on the side that they ''should'' all be defined in Fortran, but now I'm confounded by the fact that neither OMPI nor MPICH2 do for any of the other types. Is there some other statement elsewhere in MPI that says that the C/C++ types don't have to be defined in Fortran?

[ompiteam]

Trac comment by jsquyres on 2008-07-15 11:37:44:

I will:

• Check what some other MPI's do
• If an ambiguity exists, I'll check with the Forum

[ompiteam]

Trac comment by jsquyres on 2008-07-15 21:42:08:

I notice that Intel MPI and HP MPI do the same thing (C MPI datatypes are not available in mpif.h). So I think that this is an issue for MPI-2.2. Pushing this off to "Future".

[ompiteam]

Trac comment by jsquyres on 2008-10-14 15:26:46:

Changing the title of this ticket to reflect that this is now an MPI-2.2 issue.

[ompiteam]

Trac comment by jsquyres on 2008-10-23 16:23:25:

Filed a ticket about this with MPI-2.2:

https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/56

[ompiteam]

Trac comment by jsquyres on 2009-01-29 10:22:03:

I was reminded today by Dave Goodell (and previously by Hubert) that MPI-2.1 Section 16.3.6 MPI Opaque Objects, page 483, lines 46-47:

"All predefined datatypes can be used in datatype constructors in any language."

(this sentence has existed since MPI 2.0)

So while none of us are currently compliant, this doesn't seem to be a huge issue for MPI applications (i.e., no one has complained about it). Hence, we should ensure that all predefined datatypes exist in all language bindings, but it might not be a huge priority.

[ompiteam]

Trac comment by hjelmn on 2014-03-26 19:47:42:

Jeff, from what I can tell all of these types are defined for fortran and C++ on the trunk. Can you verify?