Thomas, I have checked into this before (for MPI based solvers), as I also have worked on problems where I create a 3x3 unsymmetric block matrix. You may want to check out MUMPS:
http://mumps.enseeiht.fr/ I think that both PETSc and Trilinos have support to include MUMPS, which can then be utilized through the normal interfaces in the respective packages. Honestly, I never got it working, but I never gave it a really serious try either. Please let me know if you do get it going and how well the solver scales. Regards, Jon Pitt On Mon, Jan 25, 2010 at 4:29 AM, Thomas Wick < [email protected]> wrote: > Dear all! > > I've got a question regarding to direct solvers within deal.II. > My system matrix is a 3x3 block system or 4x4 block system. From > theoretical and practical point of view there is (up to know) almost no > iterative solver which can deal with my configuration. So, I solve my > (2-dimensional) system with our UMFPACK solver. But, all of you know that, > this direct solvers become slow when computing large linear system. My > question and hope: does anybody has experiences with parallel direct solvers > and their use within deal.II? > > Does it make sense to have a parallel direct solver in deal.II? > > Kind regards, > > Thomas > > > --------------------------------------->> >>> >> Thomas Wick > Universität Heidelberg > AG Numerik, Raum 213 > Im Neuenheimer Feld 293 > 69120 Heidelberg > > Tel.: +49-(0)6221 / 54-5449 > E-Mail: [email protected] > Web: numerik.iwr.uni-heidelberg.de/ > >> --------------------------------------->> >>> >> > _______________________________________________ > dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii > >
_______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
