An additional work item is that I am going to discuss offline with Hubert and hopefully Tim Mattson about shared-memory windows on non-coherent systems to see if this is viable. If it is, I will provide all the details to the WG for further discussion. I wanted this to be stated for all to see in case others are interested in talking about non-coherent systems in detail. Jeff On Wed, Dec 5, 2012 at 2:14 PM, Pavan Balaji <[email protected]> wrote:

We talked about different items that need to be addressed for (1) MPI-3 Errata (should only be bug fixes, not all small changes), and (2) MPI-3.1.

We still need to discuss the below items in a telecon; this is just a basic separation we did based on the WG discussion.

MPI-3 Errata ============ 1. Ticket #348 -- one of the examples has a bug in it, which should be fixed.

2. Ticket #350 -- we should state that for MPI_ACCUMULATE with MPI_NO_OP, the MPI implementation is not allowed to throw an error when the origin buffer is NULL.

MPI-3.1 ======= 1. Ticket #344 -- Shared memory attribute query.

2. Ticket #346 -- Info keys are not consistent across window creation routines.

3. Ticket #347 -- Info keys are not listed in the appendix.

4. Ticket #349 -- Addition of MPI_Disp_to_aint and MPI_Aint_to_disp.

5. Addition of the MPI_MODE_NOLOAD assertion. RMA slides have info on this.

6. Info argument addition for req_mem_model. RMA slides have info on this.

-- Pavan

-- Pavan Balaji http://www.mcs.anl.gov/~balaji _______________________________________________ mpi3-rma mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-rma

-- Jeff Hammond Argonne Leadership Computing Facility University of Chicago Computation Institute [email protected] / (630) 252-5381 http://www.linkedin.com/in/jeffhammond https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond