Hi Brian, When we state that accumulates with the same operation are non-conflicting, should I be interpreting this as CAS is only non-conflicting with other CAS operations? This is a bit of a stretch since CAS doesn't take an op argument, but I suppose it's doable. Some text to help the reader make this connection for CAS would definitely help. ~Jim. On 03/02/2011 10:30 AM, Barrett, Brian W wrote:

I agree with most of the edits, but not the move from "accumulate operations" to "atomic operations". I understand the reluctance to call compare-and-swap an accumulate operation, but we used "accumulate operations" so that we wouldn't run into any semantic changes to the greater document. This was the same reason we didn't make rational overlap semantics for put/get. If we're going to change our minds and make the change to accumulate operations, but not change put/get semantics, that seems rather odd.

Brian

On 3/1/11 1:54 PM, "James Dinan"<[email protected]> wrote:

...

Hi All,

I did a read through and attached my edits. I also scanned in the edited document and typed up the edits so you don't have to interpret my scribbled notes. I *think* there might have been one or two pages that were flipped over when I scanned it, so keep an eye out.

A couple of those edits are non-trivial, so feel free to move anything that should be discussed to the mailing list.

For example, I suggested merging MPI_Win_query into a window attribute - we've remove the datatype and op from the call, so now it should work cleanly as an attribute.

Thanks, ~Jim.

On 02/28/2011 03:43 PM, Torsten Hoefler wrote:

...

Hello all,

I updated the version again (hopefully for a last time) with Brian's cleanup of the examples (Brian, I edited the style a bit). See

https://svn.mpi-forum.org/trac/mpi-forum-web/attachment/wiki/mpi3-rma-pro posal1/

All the Best, Torsten