On Fri, Jun 15, 2012 at 3:35 AM, Nakib Haider Protik <span dir="ltr"><<a href="mailto:nprot048@uottawa.ca" target="_blank">nprot048@uottawa.ca</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">
I am running 3.0.0-p12. Upgrading to 3.3 has proven difficult for me.<br></blockquote><div><br></div><div>This is 5 years old. You will notice that we have an email list to help people</div><div>with upgrade problems. We have not received any mails about problems.</div>
<div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
In my version I used -mg_levels_ksp_type chebychev and the error is<br>
greater than both ordinary gmres with lu, and ml with gmres and lu on each<br></blockquote><div><br></div><div>The solver has nothing to do with the error. This is a matter for the iterative</div><div>tolerance and condition number of the problem.</div>
<div><br></div><div> Matt</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
level. Also, it takes a very long time to converge for even a 200 x 200<br>
grid.<br>
<br>
Thanks.<br>
<br>
> On Thu, Jun 14, 2012 at 1:09 PM, Nakib Haider Protik<br>
> <<a href="mailto:nprot048@uottawa.ca">nprot048@uottawa.ca</a>>wrote:<br>
><br>
>> I have a 2 dimensional Poisson problem and the Laplacian is defined on a<br>
>> mesh that has both uniform and non-uniform regimes. I am using the<br>
>> following command:<br>
>><br>
>> -pc_type ml -mg_levels_ksp_type gmres -mg_levels_pc_type lu<br>
>><br>
><br>
> This does a direct solve on every level, of course it's going to be slow.<br>
><br>
><br>
>><br>
>> This gives a good result. (However, if I don't specify the<br>
>> -mg_level_ksp_type and -mg_levels_pc_type to gmres and lu respectively,<br>
>> the default are used (richardson and sor respectively). These lead to<br>
>> wrong results and the solver is extremely slow.)<br>
>><br>
><br>
> Are you using petsc-3.3?<br>
><br>
> Run with -mg_levels_ksp_type chebyshev (and only that). Does that converge<br>
> better?<br>
><br>
><br>
>><br>
>> Now, the following method:<br>
>><br>
>> -ksp_type gmres -pc_type lu<br>
>><br>
>> gives a slightly better result and is faster. The speed difference is<br>
>> quite conspicuous for a 501 x 1001 grid size.<br>
>><br>
>> According to my limited knowledge, asymmetric matrices are better dealt<br>
>> with by gmres and lu. However, multigrid methods are supposed to have<br>
>> O(n)<br>
>> complexity. Why is the non-multigrid method doing better in both speed<br>
>> and<br>
>> accuracy?<br>
>><br>
><br>
<span class="HOEnZb"><font color="#888888"><br>
<br>
--<br>
Nakib :)<br>
</font></span></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<br>