On Tue, Aug 11, 2009 at 1:50 PM, Vijay S. Mahadevan<vija...@gmail.com> wrote: > Hi guys, > > I have an ImplicitSystem to which I add an additional matrix. The > allocation happens fine or so it seems but when I try to assemble the > matrix, each add_matrix call takes quite a while which in turn > avalanches to a huge hit in performance. I do not see the same problem > with the System matrix when I assemble it. > > I tried two different ways: 1) add a matrix to the system and then > call init() on system. 2) create the matrix myself and attach the > dof_map and call init on PetscMatrix to have the nonzero sparsity > preallotted correctly. But both of these code paths behave similarly > in terms of the long assembly time. Am I missing something ?
Could you post the output of PETSc's -log_summary or -log_info or whatever it is that prints out the number of mallocs done during matrix assembly? If that's not zero then something's up with the sparsity pattern computation, and it can make the code run really slow. -- John ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
