Houston, we have a problem

[Jianwei Li]

Great!

I just found another possible problem within the fortran bindings
when I was looking into the data value written by John's fortran
test. When writting/reading the array data, the fortran interface
just passes the buffer starting address to the C interface. However,
assuming programmers pass fortran memory buffer directly as the argument
to fortran data access functions, the dimension order will get
reversed in the data file of the array (the array data in the file
is always assumed and interpreted in C order).

So I think, apart from passing the fortran arguments to the C
interface, there should be something dealing with fortran order
of dimensions for the memory buffer within the fortran binding
wrapper. Or should we add another argument in the C functions
to indicate Fortran/C order of the memory array? Or should this
dimension order consideration simply be handled by users in their
applications?
// I learned that John's fortran test took care of the dimension
order within the application code :)

In general, for fortran bindings, we need to handle character-string
parameter conventions, convert between column-major-order arrays and
row-major-order arrays, and map between array indices beginning
at one and array indices beginning at zero.
// Also, these 1-based/0-based indices were handled by John's fortran
test code :)

So there is quite some work to do for the fortran bindings ... ?

Jianwei

On Fri, 1 Aug 2003, Robert Latham wrote:

> On Fri, Aug 01, 2003 at 11:19:23AM -0700, John Tannahill wrote:
> > Jianwei,
> >
> > At first glance, it looks good to me:  diff, delmin, & delmax all 0 and
> > ncdump is showing reasonable values for tt.  Great work!  Let me know
> > when I can grab a new release and I will check it out further.
>
> ok, i've put a 0.8.9 release out (release anouncement to follow
> shortly).  This has Jianwei and Bill's changes to the Fortran
> bindings.  Please try it out and let us know how it works for you.
>
> ==rob
>
> --
> Rob Latham
> Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
> Argonne National Labs, IL USA                B29D F333 664A 4280 315B
>