[Mpi3-hybridpm] Reminder for telecon: 03/30

Douglas Miller dougmill at us.ibm.com
Mon Mar 29 09:41:04 CDT 2010


It is the leave that blocks, that is where the threads actually perform
work. Between join and leave, if a thread makes calls to MPI_Wait (et al.)
- presumably for it's own communications - then it will also perform any
other work assigned to it at that time.


_______________________________________________
Douglas Miller                  BlueGene Messaging Development
IBM Corp., Rochester, MN USA                     Bldg 030-2 A410
dougmill at us.ibm.com               Douglas Miller/Rochester/IBM


                                                                           
             Pavan Balaji                                                  
             <balaji at mcs.anl.g                                             
             ov>                                                        To 
             Sent by:                  mpi3-hybridpm at lists.mpi-forum.org   
             mpi3-hybridpm-bou                                          cc 
             nces at lists.mpi-fo                                             
             rum.org                                               Subject 
                                       Re: [Mpi3-hybridpm] Reminder for    
                                       telecon: 03/30                      
             03/29/2010 09:35                                              
             AM                                                            
                                                                           
                                                                           
             Please respond to                                             
             mpi3-hybridpm at lis                                             
             ts.mpi-forum.org                                              
                                                                           
                                                                           




Hi Doug,

Douglas Miller wrote:
> The example is as I intended. The usage I've been promoting is that all
> team members do the join and leave. This is the only way to get all
threads
> participating.

Sorry if I'm forgetting something we already discussed -- in your
example, are the t != 0 threads blocking inside MPI_Endpoint_join()? My
guess is no, since the t == 0 thread which needs to make the Allgather
call needs to come out.

Then how will MPI use these threads (they need to be under MPI's control
to be used).

  -- Pavan

--
Pavan Balaji
http://www.mcs.anl.gov/~balaji
_______________________________________________
Mpi3-hybridpm mailing list
Mpi3-hybridpm at lists.mpi-forum.org
http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-hybridpm





More information about the mpiwg-hybridpm mailing list