<div dir="ltr">Thank you very much for this preprint, Lawrence. I have also planned to use the pressure mass matrix for the A11 block.<div><br></div><div>Unfortunately, at this time, I have no time for implementing things. What I would like to do is to get the best out of the built-in methods of fieldsplit/PETSc.</div><div><br></div><div>Any hint is welcome!</div><div><br></div><div>Nicolas</div></div><div class="gmail_extra"><br><div class="gmail_quote">2016-12-13 19:02 GMT+01:00 Lawrence Mitchell <span dir="ltr"><<a href="mailto:lawrence.mitchell@imperial.ac.uk" target="_blank">lawrence.mitchell@imperial.ac.uk</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class=""><br>
<br>
On 13/12/16 16:50, Karin&NiKo wrote:<br>
> Dear Petsc-gurus,<br>
><br>
> I am solving Biot's poroelasticity problem :<br>
</span><span class="">> Images intégrées 1<br>
><br>
> I am using a mixed P2-P1 finite element discretization.<br>
><br>
> I am using the fieldsplit framework to solve the linear systems. Here<br>
> are the options I am using :<br>
> -pc_type fieldsplit<br>
> -pc_field_split_type schur<br>
> -fieldsplit_0_pc_type gamg<br>
> -fieldsplit_0_pc_gamg_<wbr>threshold -1.0<br>
> -fieldsplit_0_ksp_type gmres<br>
> -fieldsplit_0_ksp_monitor<br>
> -fieldsplit_1_pc_type sor<br>
> -fieldsplit_1_ksp_type gmres<br>
> -pc_fieldsplit_schur_<wbr>factorization_type upper<br>
><br>
><br>
> By increasing the mesh size, I get increasing numbers of outer<br>
> iterations.<br>
><br>
> According to your own experience, among all the features of<br>
> fieldsplit, was is the "best" set of preconditioners for this rather<br>
> classical problem in order to get an extensible solver (I would like<br>
> to solve this problem on some tens millions of unknowns of some<br>
> hundreds of procs)?<br>
<br>
</span>Here's a recent preprint that develops a three-field formulation of<br>
the problem that gets reasonably mesh and parameter-independent<br>
iteration counts using block-diagonal preconditioning.<br>
<br>
<a href="https://arxiv.org/abs/1507.03199" rel="noreferrer" target="_blank">https://arxiv.org/abs/1507.<wbr>03199</a><br>
<br>
(No need for schur complements)<br>
<br>
If you can create the relevant blocks it should be implementable with<br>
-pc_fieldsplit_type additive<br>
<span class="HOEnZb"><font color="#888888"><br>
<br>
Lawrence<br>
<br>
</font></span></blockquote></div><br></div>