[MPI3 Fortran] [Fwd: Library-based ASYNCHRONOUS I/O and SYNC MEMORY]

[Dan Nagle]

Hi,

On Sep 9, 2008, at 1:06 PM, Aleksandar Donev wrote:

> On Tuesday 09 September 2008 05:19, Dan Nagle wrote:
>> Look, MPI works with Fortran, and it has for years.
> What a false statement, unless your Fortran does *not* mean "The ISO  
> Fortran
> Standard".

I can write perfectly portable Fortran using perfectly portable MPI
that works perfectly well.  It's been done for years.  If you want
to stand on a technicality, calling *any* C procedure is a violation
of the (pre-f03) Fortran standard.

> MPI works with a hacked Fortran, thanks to hacks done by vendors,
> users, and the MPI Forum itself.

The vendors take care that their compilers work with standard MPI.
Making things work is not always a hack.

> It does NOT work with standard Fortran.

If you combine the Fortran standard with the restrictions in the MPI  
standard,
your program works, and it has worked for years.

> Period. There is an emergency,

It's an "emergency" that's lasted for years, without ill effect
beyond some ugly code.

Even if there is an "emergency", there's no reason to rush
a hack solution when we can think a bit and do things right.

> if you want the standard to have any
> significance.

The standards involved have the same significance they always have:
If you follow them, your code is portable.

-- 
Cheers!

Dan Nagle