My parents took the same lazy approach as the MPI Forum regarding asynchronous progress in RMA, so I can only work on the RMA chapter when I am communicating with Forum members via email or telecon. This makes me totally useless but I still tell my friends that I am a high-quality implementation. I fully support #2. I'm sure that #1 makes sense but since PSCW makes my brain hurt and I stopped using point-wise Lock/Unlock as soon as MPI-3 dropped, I haven't thought about it much. Best, Jeff On Mon, Jan 19, 2015 at 6:42 AM, Jim Dinan <[email protected]> wrote:
Bumping this thread for progress. Please take a look a the two errata suggestions that I captured from the December meeting. If folks feel that this is a good direction, I can convert the informal text below into proper errata proposals.
~Jim.
On Wed, Dec 17, 2014 at 6:39 PM, Jim Dinan <[email protected]> wrote:
Hi All,
Reviewing my notes from the WG meeting, it appears that the following errata would clarify many misunderstandings of the MPI-3 RMA spec:
1) Clarify the target process rank for load/store operations on shared memory windows. This is needed so that synchronization operations (e.g. lock/unlock and PSCW) are well defined on shared memory windows.
2) Clarify that that while there is one copy of the window data in memory in the unified memory model, there are still public and private views that must be synchronized according to the RMA semantics.
Is there enough support for these that they should be captured for future discussion?
Cheers, ~Jim.
_______________________________________________ mpiwg-rma mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma
-- Jeff Hammond [email protected] http://jeffhammond.github.io/