[Mpi3-hybridpm] [MPI Forum] #210: MPI3 Hybrid Programming: multiple endpoints per collective, option 3

MPI Forum mpi-22 at lists.mpi-forum.org
Thu Jan 7 07:35:47 CST 2010


#210: MPI3 Hybrid Programming: multiple endpoints per collective, option 3
----------------------------------------------------------+-----------------
                    Reporter:  dougmill@…                 |                       Owner:  dougmill@…             
                        Type:  Enhancements to standard   |                      Status:  new                    
                    Priority:  Not ready / author rework  |                   Milestone:  2010/01/19 Atlanta, USA
                     Version:  MPI 3.0                    |                    Keywords:                         
              Implementation:  Unnecessary                |           Author_bill_gropp:  0                      
          Author_rich_graham:  0                          |           Author_adam_moody:  0                      
      Author_torsten_hoefler:  0                          |        Author_dick_treumann:  0                      
Author_jesper_larsson_traeff:  0                          |       Author_george_bosilca:  0                      
           Author_david_solt:  0                          |   Author_bronis_de_supinski:  0                      
        Author_rajeev_thakur:  0                          |         Author_jeff_squyres:  0                      
    Author_alexander_supalov:  0                          |    Author_rolf_rabenseifner:  0                      
----------------------------------------------------------+-----------------

Comment(by dougmill@…):

 In order to make this work, I assume that the endpoint must be set
 MPI_THREAD_MULTIPLE, although internally it may not be required. Since the
 communications calls are not independent communications but all part of
 the same communication, perhaps MPI_THREAD_MULTIPLE is not required. But
 there is still confusion with multiple, independent, communications by
 threads attached to the same endpoint - unless we declare that to be an
 invalid usage.

 So, this would probably require some additional indicator of how the
 multiple calls should relate to each other, which requires changes to the
 communications calls API (e.g. ticket:209)

-- 
Ticket URL: <https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/210#comment:2>
MPI Forum <https://svn.mpi-forum.org/>
MPI Forum




More information about the mpiwg-hybridpm mailing list