<div dir="ltr">On Wed, Apr 17, 2013 at 12:26 PM, Jozsef Bakosi <span dir="ltr"><<a href="mailto:jbakosi@lanl.gov" target="_blank">jbakosi@lanl.gov</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On 04.05.2013 08:34, Jozsef Bakosi wrote:<br>
> Hi folks,<br>
><br>
> In switching from 3.1-p8 to 3.3-p6, keeping the same ML ml-6.2.tar.gz, I get<br>
> indefinite preconditioner with the newer PETSc version. Has there been anything<br>
> substantial changed around how PCs are handled, e.g. in the defaults?<br>
><br>
> I know this request is pretty general, I would just like to know where to start<br>
> looking, where changes in PETSc might be clobbering the (supposedly same)<br>
> behavior of ML.<br>
><br>
<br>
Alright, here is a little more information about what we see. Running the same<br>
setup/solve using ML (using the same ML and application source code) and<br>
switching from PETSc 3.1-p8 to 3.3-p6 appears to work differently, in some<br>
cases, resulting in divergence compared to the old version.<br>
<br>
I attach the output from KSPView() called after KSPSetup() for the 3.1-p8<br>
(old.out) and for the 3.3-p6 (new.out), both running on 4 MPI ranks.<br>
<br>
A diff reveals some notable differences:<br>
<br>
* using (PRECONDITIONED -> NONE) norm type for convergence test<br>
<br>
* (using -> not using) I-node routines<br>
<br>
* tolerance for zero pivot (1e-12 -> 2.22045e-14) for PPE_mg_levels_[12]_sub_<br>
(stayed the same for PPE_mg_coarse_redundant_)<br>
<br>
So we are wondering what might have changed in the PETSc defaults around how<br>
PCs, in particular ML, is used.<br>
<br>
Thanks, and please let me know if I can give you more information,<br></blockquote><div><br></div><div style>1) Please do not give error reports without the full error message</div><div style><br></div><div style>2) If you get "Indefinite PC" (I am guessing from using CG) it is because the preconditioner</div>
<div style> really is indefinite (or possible non-symmetric). We improved the checking for this in one</div><div style> of those releases.</div><div style><br></div><div style>AMG does not guarantee an SPD preconditioner so why persist in trying to use CG?</div>
<div style><br></div><div style> Matt</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Thanks,<br>
Jozsef<br>
</blockquote></div><br><br clear="all"><div><br></div>-- <br>What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>
-- Norbert Wiener
</div></div>