[petsc-dev] so

[William Gropp]

The current choices were not mine, but there is a rationale - GNU autoconf has decided that FC is the environment variable for the Fortran compiler.  Sort of like Automake telling you that C code is compiled with the C++ compiler (by using “CC file.c”) instead of giving you some useful information, such as the name of the C compiler).  But almost no fortran programer uses fc (and not a single one uses FC), and the MPI Forum has made a big point of saying that there is a single Fortran, and that the three Fortran bindings (with mpif.h, use mpi, and use mpi_f08) are not the Fortran 77, Fortran 9x, and Fortran 2008+TS29xxx versions, but just different bindings for Fortran.  What feedback we got suggested that Fortran programmers would prefer mpifort for a new name for the Fortran wrapper.

My preference would have been for MPICH to support backward compatibility, but that wasn’t the decision.

Bill

William Gropp
Director, Parallel Computing Institute
Thomas M. Siebel Chair in Computer Science
University of Illinois Urbana-Champaign







On Sep 4, 2014, at 11:23 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:

> 
>  MPICH configure rejects options like F90=fortran compiler as an error requiring FC=fortran compiler instead but what is the MPI Fortran compiler wrapper called?  Hint, it is not mpifc
> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20140905/b12be9b8/attachment.html>