[MPI3 Fortran] Agenda for MPI3 Fortran Working group next week
Bill Long
longb at cray.com
Wed Jun 3 08:46:44 CDT 2009
As another of the "Fortran guys" I've proposed going even farther. But I
see Jeff's suggestions as a reasonable compromise.
On a full disclosure note, most of the "Fortran guys" participating in
this discussion are active in J3 - the body that writes the Fortran
standard, and is currently finishing Fortran 2008. As such, we tend to
have a future-focused view of Fortran. There are certainly many users
out there who are firmly planted in the (sometimes distant) past and
might have a different view. Given that the task is to design an
interface for use in the future, I think meshing with the language in
the same time frame is the right approach. As the "Fortran guys", we
can suggest what is possible and recommend. But first, the MPI group
needs to decide which direction it's looking.
Cheers,
Bill
Lionel, Steve wrote:
> Jeff Squyres wrote:
>
>
>> Before the Fortran guys storm in with their replies ;-), I'll mention
>> that I think that using integer handles is a horrid idea for Fortran.
>> It was done with good reason back in the 90s, but the language has
>> evolved such that we can give much better type safety these days.
>>
>
> As a "Fortran guy", I'll agree 100% with Jeff. If the object is truly a handle, it should be declared as an opaque type in Fortran, which is done as an interoperable derived type with PRIVATE components, similar to C_PTR. The major objection that would be raised to this is that it means changing existing code that was written to use integer handles. In most cases, it may just be changing a declaration here and there, but it is a barrier to adopting the new interface.
>
> On the other hand, if we do it right, there will be many advantages to the new bindings and adoption should go well.
>
--
Bill Long longb at cray.com
Fortran Technical Support & voice: 651-605-9024
Bioinformatics Software Development fax: 651-605-9142
Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120
More information about the mpiwg-fortran
mailing list