[mpich-discuss] mpifort wrapper compiler?
jeff.science at gmail.com
Wed Dec 11 21:56:37 CST 2013
> On Dec 11, 2013, at 6:39 PM, "Jeff Squyres (jsquyres)" <jsquyres at cisco.com> wrote:
> I deleted all the stuff that JeffH and I argued about in person. It wasn't useful to keep arguing about those things here on the list. :-)
> On Dec 7, 2013, at 3:50 PM, Jeff Hammond <jeff.science at gmail.com> wrote:
>>> So other than the backwards compatibility issue, a single "mpifort" wrapper that makes all 3 interfaces available makes perfect sense.
>>> I don't think there can be any expectation that "mpif77" provides a true, honest, and only Fortran 77 compiler (because MPI never worked with Fortran 77).
>> How about we have mpifisffc (MPI FIxed Source Form Fortran Compiler)
>> and mpifrsffc (MPI FRee Source Form Fortran Compiler) then?
> When I asked earlier, everyone wanted "mpifort", and it's very much in conjunction with what several Fortran vendors are doing (i.e., a single executable name, and optional CLI options and/or .suffix examination if you want specific version syntax enforcement).
> Open MPI has already released mpifort into the wild, and I hope MPICH does the same, because it will really make users' lives easier if they can assume that "mpifort" exists everywhere.
> The point is that MPI shouldn't try to subvert what the underlying Fortran compilers are doing. If they're consolidating down to a single name, so should we. And if you have platforms where they're not, great -- don't consolidate to a single name (e.g., mpifort) on those platforms, and/or rely on downstream vendors (e.g., IBM) to do the right things for their platforms/compiler suites (e.g., the mpif* scripts that IBM is apparently generating).
I agree that this is right for most cases, eg ifort, ftn and gfortran. I recall PGI has language-named variants. Not sure what you think should be done there. That is a case that actually matters to MPICH and OpenMPI builds from source by users (as opposed to IBM on BG).
>> The only natural conclusion at this point is that it is critical to
>> deprecate both mpif.h and the 'mpi' module, since the 'mpi_f08' module
>> is clearly the only sane way to use MPI from Fortran.
> Yes, this was the long-term plan. The Fortran WG should probably meet again to take steps towards deprecating mpif.h... and possibly the mpi module. :-)
> Jeff Squyres
> jsquyres at cisco.com
> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
> discuss mailing list discuss at mpich.org
> To manage subscription options or unsubscribe:
More information about the discuss