[mpich-discuss] mpifort wrapper compiler?

[Jeff Hammond]

> 
> 
> Will the following tradeoff be acceptable to everyone?
> 
> * We allow users to specify a Fortran compiler using “FC” and “FCFLAGS”.  This will result in a wrapper called “mpifort”.
> 
> * We’ll maintain mpif90 as a symbolic link to mpifort for backward compatibility.
> 
> * If no separate F77 compiler is specified, we’ll use FC as the F77 compiler as well.  In this case, mpif77 will be a symbolic link to mpifort.
> 
> * But users have the ability to provide a separate F77 compiler, in which case a new mpif77 is created.
> 
> This way, we can still support compilers that have a different F77 compiler.  But for most compilers there’ll be a single mpifort (with mpif90 and mpif77 symbolic links for backward compatibility for some time).
> 

Yes. This sounds great. 

Is there any point to having --with-fortran take optional args {header,module,f08mod} to down-select from the default of 'all'? Can one skip the goofy Fortran ABI configure tests if only F08 is requested?

In theory, users could provide all diff Fortran lang compilers or diff flags to indicate language and get wrappers for them but I suppose it is equally valid to say that users/packagers are responsible for doing this by hand. 

Jeff


> 
>> On Dec 12, 2013, at 7:48 AM, Jeff Squyres (jsquyres) <jsquyres at cisco.com> wrote:
>> 
>> On Dec 11, 2013, at 10:29 PM, Rajeev Thakur <thakur at mcs.anl.gov> wrote:
>> 
>>>> I agree that this is right for most cases, eg ifort, ftn and gfortran. I recall PGI has language-named variants.
>> 
>> They also have pgifortran (my PGI license has expired, so I can't say what they do today).
>> 
>>>> 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 Absoft compiler also has different names af77 and af90.
>> 
>> 
>> Correct; it does not look like Absoft has consolidated down to a single Absoft executable.  FWIW: I just asked an Absoft employee -- he said:
>> 
>> -----
>> We have af77, af90, and af95. The af77 driver enables some legacy FORTRAN 77
>> extensions but there is no difference between af90 and af95.
>> 
>> Also, all three drivers define the following symbols if the C pre-processor
>> is invoked by convention ( all cap file extensions .F .F90 .F95 .FOR) or
>> by option ( -cpp ), -D__ABSOFT__  and  -D__ABSVER__=140
>> -----
>> 
>> -- 
>> 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:
>> https://lists.mpich.org/mailman/listinfo/discuss
> 
> --
> Pavan Balaji
> http://www.mcs.anl.gov/~balaji
> 
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss