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

[MPI Forum]

#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                      
----------------------------------------------------------+-----------------
 Multiple threads attach to the same endpoint. One thread makes the
 communication call while the others make a call to some 'helper' routine,
 or in some other way "lend" themselves to the MPI implementation. The
 communication is then free to use these additional threads as needed.
 However, this does tend to imply larger numbers of threads, or more re-
 associating of threads to endpoints (and the program must somehow know to
 do that). It is also unclear if that would allow the implementation to
 fully parallelize the communication, if endpoints are intended to be the
 unit of parallelism and possibly represent actual hardware.

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