-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 (Repost)
- -------- Original Message -------- Subject: Re: [dealii] parameterGUI Date: Sat, 15 Jan 2011 13:31:38 -0600 From: Carsten Burstedde <[email protected]> To: Wolfgang Bangerth <[email protected]> CC: Martin Steigemann <[email protected]>, [email protected] > I have one concern: p4est does some asynchronous communication. If there were > two triangulations that both communicate on MPI_COMM_WORLD, I could imagine > situations where a message that one triangulation sends gets picked up by the > other triangulation. We pass an MPI communicator object to the > parallel::distributed::Triangulation when we create it. I think that if we > start using multiple triangulation, we should clone this communicator (there > is a function in Utilities::System::MPI for this) and make sure that each > triangulation has its own, unique communicator that has the same processors > but is distinct from the one passed to the class. In copy_triangulation, we > would then clone it again. Can you take a look at what this would take? p4est communication is confined within the major API functions. All communications initiated in such a function will be terminated before it returns. Problems with duplicate message IDs can only occur in multithreaded programs that call such functions concurrently. So in the current non-threaded implementation it is ok to copy a p4est, as long as the communicator passed into p4est_new stays alive (it is copied, not MPI_Comm_dup'ed). Carsten -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAk0x9n0ACgkQRClUBgYbo/mSagCdEp665hFqnDySaLVj1/fnd+EQ j1oAoKG0EUAkI5sFfNDkN0RLZGVy4I8p =bP/B -----END PGP SIGNATURE----- _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
