[Mpi3-rma] Updated MPI-3 RMA proposal 1

Jeff Hammond jeff.science at gmail.com
Wed Jun 16 20:53:48 CDT 2010


On Wed, Jun 16, 2010 at 8:37 PM, Underwood, Keith D
<keith.d.underwood at intel.com> wrote:
> Adding all_flush_all to passive target is adding an active target-like call, which is ugly from an API design perspective.

This ship has sailed.  You're the only person who has voiced an
objection to all_flush_all and numerous others have demonstrated at
length why it is necessary.  I do not find the obsession with
pure-passive target semantics to be a compelling reason to reject
all_flush_all.

> Allowing multiple locks invites deadlock from the user.

So do blocking send/recv.  The potential for deadlock is not a valid
counterargument.  What don't you provide a specific example of how an
expert user could generate a deadlock in the course of following the
standard?

> Do we have the use cases for both of those and some attempt at quantification of the performance advantage in some implementation?

Yes.  I will not repeat the use case for all_flush_all since we have
discussed that topic far too long already.  Somewhat else should
present the multiple locks issue generically before I give the Global
Arrays use case.

Jeff

-- 
Jeff Hammond
Argonne Leadership Computing Facility
jhammond at mcs.anl.gov / (630) 252-5381
http://www.linkedin.com/in/jeffhammond



More information about the mpiwg-rma mailing list