[MPI3 Fortran] Request for a straw vote.
donev1 at llnl.gov
Fri Jun 12 14:29:55 CDT 2009
On Friday 12 June 2009 10:11, Craig Rasmussen wrote:
> Drat, I thought copy-in/copy-out was solved. If the compiler does
> copy-in/copy-out what does the dope vector look like? This will be
> in a BIND(C) interface so I think the compiler should know it can't
> control optimizations and shouldn't do copy-in/copy-out, it should
> just pass the original dope vector.
This is why it is important that the committee first really finalize the
assumed type and rank stuff. We could do what you ask for, but I do not
recall it being discussed. The latest unofficial draft from J3 contains
some comments, such as:
[CER - add to last sentence]
and copy-in/copy-out of the actual argument is prohibited.
(who is CER?), which are not "official".
> Since the C semantics don't allow for non-contiguous buffers, the MPI
> standard could decide to say this is an error and have the MPI vendor
> return an error condition.
OK, got it.
Yes, the MPI vendor should be allowed to flag an error, but only for the
*non-blocking* versions. For the blocking ones, a copy can trivially be
made, and should be done (it's the "Fortran" thing to do).
More information about the mpiwg-fortran