(j3.2006) (SC22WG5.3677) [ukfortran] N1755: Request for new features from MPI Forum

[N.M. Maclaren]

On Nov 13 2008, Malcolm Cohen wrote:
>
>can argue as to whether it usually is or not).
>> If you call an external I/O routine, like MPI, and it works 
>> asynchronously, there is no difference with Fortran ASYNCHRONOUS I/O
>
>Well ... yes and no.  The optimisation implications are in fact somewhat 
>different (and I'm not sure whether MPI-2's design is consistent with 
>ASYNC).

I think that we need to discuss that!  I was thinking of to MPI-1
non-blocking transfers, which are.

When writing my MPI course, I looked at a range of MPI applications
(PARPACK, CASINO, CETEP/CASTEP/ONETEP, CPMD, FFTW, SPOOLES, ScaLAPACK,
CFX-11 and FLUENT), and the only ones that used MPI-2 one-sided calls
were the last two.  I didn't have the source, so can't tell you how they
used them.  Six of those used MPI-1 non-blocking calls.


Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1 at cam.ac.uk
Tel.:  +44 1223 334761    Fax:  +44 1223 334679