[petsc-dev] mpifort versus mpif90

Barry Smith bsmith at mcs.anl.gov
Sat Jun 3 20:34:05 CDT 2017

> On Jun 3, 2017, at 5:56 PM, Jed Brown <jed at jedbrown.org> wrote:
> Satish Balay <balay at mcs.anl.gov> writes:
>> On Sat, 3 Jun 2017, Jed Brown wrote:
>>> We only need one FC now and the preferred executable name is mpifort
>>> (other variants like mpif90 are deprecated in both MPICH and Open MPI).
>>> I would just add this everywhere we currently have mpif90 but the code
>>> is already super convoluted and I'd rather simplify if possible.  Does
>>> anyone use --with-gnu-compilers versus --with-vendor-compilers?  Is it
>>> tested anywhere?
>> Its not tested in the nightlybuilds - but I've seen these options (esp
>> with-gnu-compilers) used in configure.log sent by users.
>> I'm not a fan of this interface - and won't miss it if removed.
>> [its a bit ambiguous when one does stuff like: '--with-gnu-compilers=1 --with-mpi-dir=.']
>> A related issue is - how does one specify compatible compilers [for
>> c/c++/fortran]
> --with-mpi-dir or specify each directly CC=cc FC=fort

   Yes, but they may pass a FC that is not compatible with the CC. Configure will likely error but it is unlikely it will provide a "very useful error message" about why it failed.

More information about the petsc-dev mailing list