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

[Aleksandar Donev]

Dan,

> Some on J3 also viewed attaching ASYNCHRONOUS to anything other than
> Fortran I/O as fundamentally the wrong approach.  This is not to say
> that some attribute cannot be added to do something along these lines.
Of course, I said that myself many times. BUT, there was no technical 
discussion in J3 on this, as you know well----the official response (which I 
specifically asked you to write down!) was that it was too much work, no 
time, too late, etc.

Does the MPI group think this is important enough and urgent now? If no, 
cycle.
Does the MPI group not like the existing VOLATILE attribute as a possible 
solution? If yes, end of discussion, write a tutorial note in the MPI-3 
standard. If not, why?
Does the MPI group like the proposed SYNC MEMORY solution (be it ASYCHRONOUS 
or a new ASYNC_EXTERNAL attribute) or wants to propose something else?

Best,
Aleks