The tag also factors in here. What I said in the blog entry was: "The MPI specification doesn’t define which message arrives first. It defines which message is matched first at the receiver: the first one (which happens to be the long one). Specifically, between a pair of peers, MPI defines that messages sent on the same communicator and tag will be matched at the receiver in the same relative order."
On Jan 25, 2012, at 1:20 AM, George Bosilca wrote: > Mateus, > > MPI guarantee message ordering per communicator per peer. In other words any > message going from peer A to peer B in the same communicator will be > __matched__ on the receiver in the exact same order as they were sent (this > remains true even for multi-threaded libraries). MPI does not mandate any > other type of ordering, such as between communicators or between different > pairs of processes. > > Now, what I previously said is only true for the matching logic. Completion > of message reception is a totally different thing. > > George. > > > > On Jan 24, 2012, at 23:53, Mateus Augusto <teleomin...@yahoo.com.br> wrote: > >> After a read: http://blogs.cisco.com/performance/more_traffic/ >> I understood that if a large message is sent and then a short message is >> sent, then the short message can reach before. But what if the messages have >> the same size, and are small enough so that no fragmentation occurs, the >> ordering in delivery will be guaranteed? >> >> >> _______________________________________________ >> users mailing list >> us...@open-mpi.org >> http://www.open-mpi.org/mailman/listinfo.cgi/users > _______________________________________________ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users -- Jeff Squyres jsquy...@cisco.com For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
