[Mpi3-rma] RMA telecon, 1/24, noon Central

Barrett, Brian W bwbarre at sandia.gov
Mon Jan 24 11:11:45 CST 2011


On Jan 23, 2011, at 11:22 AM, Pavan Balaji wrote:

> On 01/23/2011 12:20 PM, Barrett, Brian W wrote:
>> Before you commit, it would be good to get some favorable reviews.
>> 
>> I'm having some trouble figuring out what you merged in from
>> Torsten's document - I can see the purple removes, but given that
>> there was already a bunch of red and dark red, it's a bit difficult
>> to pick out the additions. Would it be possible to either pick a
>> different color for your additions, merge into Torsten's clean
>> document, or list all the changes?
> 
> I only merged the MPI_RMA_PUT/GET/ACCUMULATE/GET_ACCUMULATE functions 
> that we agreed on in the telecon last time. I can put it in a different 
> color if you really want, but I don't see a point of repeating the 
> entire exercise again.

First, I'm not entirely sure the working group reached consensus on the addition of RMA_* functions.  I know there has been significant discussion, but I don't believe discussion was reached.  I think this merits more discussion at today's telecon; the semantics still seem rather poorly defined and there has been discussion about a presentation to the entire forum in a couple weeks which would require having all the details sorted out.

Second, the diffs make it much easier to verify what the total sum of changes were.  A number of people have done extremely close readings of the RMA document (what Torsten sent out a couple days ago), and fine language details are now being discussed.  Once that level of discussion has been reached, making such large changes (with diffs that don't make sense - such as purple marking the addition of text that was never in the main document and shouldn't be in the main document) doesn't inspire confidence.  Being able to see the exactly what was brought in would make the process much smoother.

Brian

-- 
  Brian W. Barrett
  Dept. 1423: Scalable System Software
  Sandia National Laboratories






More information about the mpiwg-rma mailing list