Although using parallel mesh, a MPI broadcast(mesh) is called after reading mesh on a single processor, then each processor has a copy of the global mesh, isn't it?
The first partition (either Metis or ParMetis) requires a redistribute() whereby the send/receive are called among all processors. I think until "delete_remote_elements" is called, each processor has a global mesh rather than only its own elements, even for parallel mesh. > -----原始邮件----- > 发件人: "Junchao Zhang" <junchao.zh...@gmail.com> > 发送时间: 2015年7月30日 星期四 > 收件人: "Roy Stogner" <royst...@ices.utexas.edu> > 抄送: libmesh-users <libmesh-users@lists.sourceforge.net> > 主题: Re: [Libmesh-users] redistribute() in a parallel mesh > > On Thu, Jul 30, 2015 at 8:31 AM, Roy Stogner <royst...@ices.utexas.edu> > wrote: > > > > > On Thu, 30 Jul 2015, li....@siat.ac.cn wrote: > > > > > For a parallel mesh, is that every processor has a global mesh? > > > > For SerialMesh, for every properly initialized mesh every processor > > has a full copy of the mesh. > > > > I guess after refinement, processors need an all-to-all communication to > get a full picture of the mesh (This is an expensive operation) > I also guess even the mesh is duplicated, the work is not. libmesh need to > call METIS to partition the mesh/work. In the parallel mesh case, ParMETIS > should be called instead. > Am I correct? > > > > For ParallelMesh, each processor only has a copy of its own elements > > and a layer of "ghost" elements. > > > > Repartitioning thus requires a redistribute() call on the latter but > > not the former. > > --- > > Roy > > > > > > ------------------------------------------------------------------------------ > > _______________________________________________ > > Libmesh-users mailing list > > Libmesh-users@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/libmesh-users > > > ------------------------------------------------------------------------------ > _______________________________________________ > Libmesh-users mailing list > Libmesh-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/libmesh-users -- Li Luo https://sites.google.com/site/rolyliluo/ Shenzhen Institutes of Advanced Technology Address: 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, P.R.China Tel: +86-15899753087 li....@siat.ac.cn ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
