[MPIWG Fortran] [mpiwg-tools] Proposal: MPI_SIZEOF not profiled
longb at cray.com
Mon May 19 11:00:54 CDT 2014
On May 19, 2014, at 10:45 AM, Jeff Squyres (jsquyres) <jsquyres at cisco.com> wrote:
> On May 19, 2014, at 11:34 AM, Bill Long <longb at cray.com> wrote:
>> I was curious why the PMPI versions had been exempted in the text quoted in the ticket. It seems to contradict the text quoted above. I’m inclined to require that the C library PMPI versions be supplied even if the MPI versions are implemented as non-interceptable. The PMPI version could be as simple as a wrapper around a macro. But it ensures that the C library PMPI version exists for cross-language calling (where macros don’t work).
> p555 says that the PMPI_ versions must be supplied. But it does not say anything about forcing them to be in a library (e.g., being able to call the C version from Fortran). That would be a new requirement.
Ah, OK. I then assume “supplied” can be satisfied by providing a macro for PMPI_Wtick, for example. In that case, the Fortran side can supply a non-MPI-named C wrapper that uses that macro. It seems reasonable to expect that the Fortran and C versions of routines like MPI_WTICK return the same values, which would be ensured if the Fortran version always calls the C equivalent. It’s useful to realize that some level of trickery might be needed to make that happen.
> Jeff Squyres
> jsquyres at cisco.com
> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
> mpiwg-fortran mailing list
> mpiwg-fortran at lists.mpi-forum.org
Bill Long longb at cray.com
Fortran Technical Suport & voice: 651-605-9024
Bioinformatics Software Development fax: 651-605-9142
Cray Inc./ Cray Plaza, Suite 210/ 380 Jackson St./ St. Paul, MN 55101
More information about the mpiwg-fortran