[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Fortran 2003 compliant replacements for Netcdf FORTRAN interfaces

Hi Rick,

>   The attached gzipped tar file contains my implementation of new
> FORTRAN interfaces that use the FORTRAN 2003 C interoperability facility
> to replace your current C based wrapper routines with standard
> conforming FORTRAN code. I don't know if anyone in the netCDF community
> has done something similar. If they have, then you have another
> implementation to compare against. I've tested the new interfaces with
> the two compilers I have access to (CRAY X1 and g95) that support the
> F03 C interop. facility. However, I would like to release it to the
> netCDF community for further testing with compilers I don't have access
> to (such as version 10 of IBM's xlf compiler for AIX and Linux). There
> is a PDF in the tar file that describes the new interfaces and my
> motivations for developing them. 
> 
> Let me know what you think and if this work is something UCAR is
> interested in developing further.

I'm *very* interested in this, as our cfortran.h approach is starting
to become unmaintainable.  I knew the Fortran standards group was
proposing a standard way to call C funtions from Fortran, but I didn't
know g95 and possibly other commercial compilers already implemented
it.  

I'm embarrassed to admit I didn't even know Fortran 2003 was an
approved standard.  But now that I look at the Draft Standard at

  http://j3-fortran.org/doc/standing/2003/007.pdf

I see it includes an "ISO C BINDING module" that portably solves the
problem of invoking C from Fortran and vice-versa, and even provides
portable mappings between C types and Fortran types.  This is a great
addition to standard Fortran!

As far as I know, no one in the netCDF community has done anything
like this, so you're the first.  I encourage you to let people in the
community know about this for testing.  I would be very interested in
finding out what compilers support the Fortran 2003 interoperability
with C.  We probably won't be able to abandon the cfortran.h approach
until most of the compilers people use have this capability.

We are about to make a 3.6.2 release.  I see that the compressed tar
you sent is a complete replacement for 3.6.1.  [Ed: it's in
/scratch/russ/netcdf-3.6.1-f03/].  Do you think it would be practical
to maintain a set of patches so this could be applied to later
releases than 3.6.1?  Alternatively, maybe our autotools approach
could detect whether the Fortran compiler supports C interoperability
and use it in that case.

This has the potential to eliminate a lot of the support questions we
get about the Fortran-C interface with various compilers and
development platforms.

On behalf of the netCDF community, thanks for making this valuable
development work available!

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden                     http://www.unidata.ucar.edu