On Tue, Oct 23, 2012 at 5:21 PM, Jinquan Zhong <span dir="ltr"><<a href="mailto:jzhong@scsolutions.com" target="_blank">jzhong@scsolutions.com</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
That is new for me. What would you suggest, Matt?</blockquote></div><br><div>Were you using LU or Cholesky before? That is the difference between -pc_type lu and -pc_type cholesky. Use -pc_factor_mat_solver_package mumps to choose MUMPS. You can access the MUMPS options with -mat_mumps_icntl_opaquenumber.</div>
<div><br></div><div>It looks like PETSc's Clique interface does not work in parallel. (A student was working on it recently and it seems to work in serial.) When it is fixed to work in parallel, that is almost certainly what you should use. Alternatively, there may well be a Fast method, depending on the structure of the system and that fat dense block.</div>