question

[William Gropp]

At 11:00 PM 7/22/2003 -0500, Robert Latham wrote:
>On Tue, Jul 22, 2003 at 01:34:11PM -0700, John Tannahill wrote:
> > I am steadily making progress.  I have installed parallel-netcdf on
> > an IBM-SP at NERSC (seaborg).  I have put together a C test code and
> > gotten it to work.
>
>fantastic. thanks for the success report!
>
> > I now have created a Fortran90 version of the
> > test code, but have not gotten it to link and run yet.  Does the
> > default install put together a parallel-netcdf interface for C only?
>
>The default install *should* try to build the fortran interface, but
>it is not nearly as well tested as the c interface.
>
> > Do I need to change something and rebuild it to get a Fortran90
> > interface?
>
>i don't know f90 very well... if we have to do anything special for a
>f90 interface that we wouldn't have to do for an f77 interface, then
>maybe that's the problem?

We haven't built a specific f90 version; we should build a pnetcdf module 
to at least provide the interface definitions.  But the F77 version should 
work, as long as the name mapping from the C to Fortran is correct.  We 
should also document the override for that in the case where the automatic 
code fails.


> > Then do I just change the ncmpi_ calls to nfmpi_, and add
> > an ierror argument at the end?
>
>that's how it works in theory :>
>
> > Anywhere I could have looked to get answers to these questions
> > myself?
>
>nope :> you and troy baer are the only end users who have tried out
>our fortran interface.   I'll see what i can find out tomorrow, and
>i'll be sure to let you know.
>
>Thanks for exercising the fortran side of things for us.  Sorry you
>are running into so much difficulty, but we'll try to get you
>straight as soon as we can.

We'll add Fortran-specific documentation as well.  Thanks for being on the 
bleeding edge.

Bill