Date: Thu, 25 Jun 1998 15:05:15 -0500 (EST)
From: Jeff Squyres <jsquyres@lsc.nd.edu>
To: mpi-bind@mcs.anl.gov
cc: Michael Holzheu <holzheu@de.ibm.com>
Subject: More from Michael Holzheu
Message-ID: <Pine.SOL.3.96.980625143221.14441A-100000@lsc.nd.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-mpi-core@mcs.anl.gov
Precedence: bulk

Here is Michael's latest questions, and my reply:

> Thanks for your answer! I found the "complex"-class in the standard. 
>  
> I was asking after the "C" MPI_CPP_COMPLEX MPI-Datatype because there
> exist e.g. Fortran Datatypes like MPI::F_COMPLEX in C++ for language
> interoperability, too. If I understood this correctly we can e.g send a
> Fortran-complex from a Fortran program to a C++ program, which must
> interpret the received buffer in an adequate way.  If this is correct,

This is correct.  More specifically, you can:

	Fortran program: CALL MPI_SEND(... MPI_COMPLEX ...)
	C++ prgraom: MPI::COMM_WORLD.Recv(... MPI::F_COMPLEX ...);

and then the C++ program has to interpret the buffer appropriately (or
pass it to a Fortran routine who can handle it natively), because you will
have an instance of a Fortran COMPLEX type.

> so why don't we have this for language interoperability between C and
> C++ ? But perhaps I am totaly wrong in this point... ? 

No, I think that we just didn't put in a corresponding datatype in
Fortran/C for the C++ complex type (since it is different than the Fortran
complex type).  So we have:

			C/Fortran bindings	C++ bindings
  Fortran complex type:	MPI_COMPLEX		MPI::F_COMPLEX
  C++ complex type:	-none-			MPI::COMPLEX

So you are correct in that we should have an "MPI_CPP_COMPLEX" type for
the C/Fortran bindings.

Unless anyone objects, we can put it on Steve's erratta list.

> I have not understood the thing with the depreciated functions. For the
> C++- bindings there do not exist the depreciated functions, don't they?

Ah!  You are correct -- Jeremy and I were wrong in our first reply to
you.

> I assume that "MPI::NULL_COPY_FN" and "MPI::DUP_FN" are of type " 
> MPI::Comm::Copy_attr_function"  and "MPI::NULL_DELETE_FN" is of type
> "MPI::Comm::Delete_attr_function".  So this functions can't be used for
> the depreciated C-call "MPI_Keyval_create()" either.  In the C++ call
> "MPI::Comm::Create_keyval()" we can use the "MPI::Comm::  NULL_COPY_FN"
> etc.  So why do we need the "MPI::NULL_COPY_FN" etc ? 

Actually, it looks like we deleted the typedefs (notice on 346:29 and
346:32 that it says that they are of type MPI::Copy_function and
MPI::Delete_function, respectively, but those typedefs are not declared
anywhere), but we forgot to delete the corresponding constants from
346:29 - 346:33.

These constants do not exist in our C++ bindings implementation.

This should also go on Steve's erratta list.

{+} Jeff Squyres
{+} squyres@cse.nd.edu
{+} Perpetual Obsessive Notre Dame Student Craving Utter Madness
{+} "I came to ND for 4 years and ended up staying for a decade"