[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [mpi-21] Missing MPI::CONVERSION_FN_NULL
- To: mpi-21@xxxxxxxxxxxxx
- Subject: Re: [mpi-21] Missing MPI::CONVERSION_FN_NULL
- From: Jeff Squyres <jsquyres@xxxxxxxxx>
- Date: Fri, 7 Dec 2007 08:34:38 -0500
- Authentication-results: sj-dkim-1; header.From=jsquyres@cisco.com; dkim=pass ( sig from cisco.com/sjdkim1004 verified; );
- Delivered-to: mpifrm-mpi-21-outgoing@mailbouncer.mcs.anl.gov
- Delivered-to: mpifrm-mpi-21@mailbouncer.mcs.anl.gov
- Dkim-signature: v=0.5; a=rsa-sha256; q=dns/txt; l=1166; t=1197034481; x=1197898481; c=relaxed/simple; s=sjdkim1004; h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version; d=cisco.com; i=jsquyres@cisco.com; z=From:=20Jeff=20Squyres=20<jsquyres@cisco.com> |Subject:=20Re=3A=20[mpi-21]=20Missing=20MPI=3A=3ACONVERSION_FN_NULL |Sender:=20; bh=Fw59bdo0XBMTEugUlIb/7PXjWG45G4gWDDSVRtg00Tw=; b=S5TOaPXR00RZAfNdBMYHtEUGrlC9WbYe0fHIjhLwHwJVqtWLeNiSiCUUPJN+hFofsvMzU2YT cnGyOFVtCnc+P96fqn5ANXtw1M7bq1m8y+M4ox9CdwXKzGT/TKpREinVERPQ93EjPKN2cp1UPE kLvo7fZyAciS12p3gAeCYBH/k=;
- In-reply-to: <5F7D4031-9087-4E51-80EF-5E5921C2BD65@cisco.com>
- References: <5F7D4031-9087-4E51-80EF-5E5921C2BD65@cisco.com>
- Reply-to: mpi-21@xxxxxxxxxxxxx
- Sender: owner-mpi-21@xxxxxxxxxxxxx
On Dec 6, 2007, at 6:00 PM, Jeff Squyres wrote:
Here's another item for MPI-2.1: I see MPI_CONVERSION_FN_NULL
defined in MPI-2 (see MPI-2:9.5.3).
It would be convenient for implementations (and consistent for all
other constants) if MPI::CONVERSION_FN_NULL was also defined, and
was the only constant you could use in the C++ bindings (i.e., you
couldn't use MPI_CONVERSION_FN_NULL).
The same should also be true for all the pre-defined callback function
pointer values for attributes: MPI_*_NULL_DELETE_FN (add
MPI::*_NULL_DELETE_FN, or perhaps even MPI::*::DELETE_FN?),
MPI_*_NULL_COPY_FN, and MPI_*_DUP_FN (and the deprecated MPI-1
equivalents).
It would ease implementations by not forcing extra overloaded
functions and translating back and forth between the C and C++
function pointers (e.g., it doesn't make sense to create a keyval with
the C++ binding with a C callback because the callback function
signatures are different).
It would be simpler/cleaner to say that the C constants should be used
for C, and the C++ constants should be used for C++.
--
Jeff Squyres
Cisco Systems