<div dir="ltr"><div class="gmail_quote">On Thu, Mar 31, 2011 at 14:57, Ormiston, Scott J. <span dir="ltr"><<a href="mailto:SJ_Ormiston@umanitoba.ca">SJ_Ormiston@umanitoba.ca</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
If there is no parallel direct solver in PETSc, what should I use? I have seen some people recommend SuperLU. This seems to be a totally separate package from PETSc with its own syntax and setup. Therefore, I must start from scratch with the SuperLU setup and I cannot take advantage of the parallel matrix setup that I have already done to solve my problem with a ksp solver with PETSc.</blockquote>
</div><div><br></div><div>Hey now, we're not sadists. ;-)</div><br><div>First configure PETSc with --download-superlu_dist or --download-mumps (or install these packages yourself and use --with-superlu_dist-dir=, see the output from ./configure --help for details).</div>
<div>Then just run your program with -pc_type lu -pc_factor_mat_solver_type superlu_dist, no need to modify any code.</div></div>