[petsc-users] Fortran interface -- low level access to petsc data
Tabrez Ali
stali at geology.wisc.edu
Sun Oct 27 14:51:01 CDT 2013
On 10/27/2013 12:36 PM, Hossein Talebi wrote:
> Thank you for your answer.
>
> It is not true that F2003 support is that limited: Gfortran, Intel
> Fortran, Cray Fortran, IBM Fortran and PGI Fortran all support nearly
> full F2003. Some other ones support the C portability of F2003
> although not supporting the rest.
You're 100% right but many systems often have older versions of these
compilers that don't support the Fortran C interoperability. There is
absolutely no excuse to not upgrade but then its not always possible.
Unless there is a significant gain in functionality (or ease of support)
by moving to a Fortran 2003 compiler it would be better to be
conservative. One advantage of being able to build PETSc on a old enough
system (e.g., Debian Etch with GCC 4.1; glibc 2.4) is that you can very
easily build a static binary of your code and it will run pretty much on
every Linux machine out there.
T
> * I can contribute on creating this interface. I have experience with
> F2003 to C++ and it works fine while C is a lot easier. My idea is to
> create the F03 interface to petsc as a separate tool. So, anyone
> trying to use petsc with a F03 compiler can compile this interface
> again to avoid the compiler inconsistency of Modules.
>
> We can work on a python (or similar) script to generate the
> interfaces. For example, a project called gtk-fortran uses this
> method. Can any of you guys help a little for planning the interface
> design?
>
>
> Cheers
> Hossein
>
>
>
>
>
>
>
>
>
>
> On Sun, Oct 27, 2013 at 4:42 PM, Satish Balay <balay at mcs.anl.gov
> <mailto:balay at mcs.anl.gov>> wrote:
>
> On Sun, 27 Oct 2013, Matthew Knepley wrote:
>
> > On Sun, Oct 27, 2013 at 6:16 AM, Hossein Talebi
> <talebi.hossein at gmail.com <mailto:talebi.hossein at gmail.com>>wrote:
> >
> > > Hello All,
> > >
> > > I am trying to use petsc with Fortran. The initial mechanism
> to use petsc
> > > with FORTRAN77 that works fine. However, I found the Fortran
> 90 interfaces
> > > difficult to work with. The very first issue is one has to
> recompile petsc
> > > with the Fortran compiler in case different to the one which
> was used to
> > > create the Fortran Mod files.
>
> Yes - one cannot mix fortran code compiled with 2 different compilers
> - or 2 different version of the same compiler wrt modules.
>
> But we generally recommend not mixing compilers for C aswell as
> this
> can cause problems.
>
> Also you don't need to use modules to get f90 interfaces. You can
> include
> the f90 interface include files directly in your code.
>
> #incude "finclude/petsc.h90"
>
> > > Also, we would still need to use C type
> > > #include statements along with Fortran 'use' statement which
> is not so
> > > optimal.
>
> Even if you were to use f90 modules #include is unavoidable as we
> still
> have other stuff in include files.
>
> check include/finclude/petscsysdef.h include/finclude/petscsys.h
>
> > > Since Fortran 2003 offers great interface to C, did anyone try
> to write a
> > > F03 interface to petsc? Using F03 features, we can have very
> low data
> > > access to petsc database as well.
> > >
> >
> > F2003 support is fairly thin in the compiler world, and we have
> not had
> > much call for it. We do gratefully accept contributions.
>
> Currently we have the f90 interfaces [in include/finclude/ftn-auto/]
> automatically generated by bfort tool from the C interface. So to
> support F03 it would be ideal if there is such a tool. [or update
> bfort to do it]
>
> Satish
>
>
>
>
--
No one trusts a model except the one who wrote it; Everyone trusts an observation except the one who made it- Harlow Shapley
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20131027/5f54e702/attachment.html>
More information about the petsc-users
mailing list