[petsc-users] Insufficient memory when using GMRES + Boomeramg

[TAY wee-beng]

On 27/3/2014 12:14 AM, Matthew Knepley wrote:
> On Wed, Mar 26, 2014 at 10:42 AM, TAY wee-beng <zonexo at gmail.com 
> <mailto:zonexo at gmail.com>> wrote:
>
>     On 26/3/2014 11:22 PM, Matthew Knepley wrote:
>>     On Wed, Mar 26, 2014 at 9:59 AM, TAY wee-beng <zonexo at gmail.com
>>     <mailto:zonexo at gmail.com>> wrote:
>>
>>         Hi,
>>
>>         I am running a CFD solver. The Poisson eqn was originally
>>         solved using HYPRE's geometric multigrid.
>>
>>
>>     Is this on a structured grid?
>
>     Yes.
>
>
> Then you can replicate the Hypre structured MG with the PCMG, and it 
> can be lighter memory than GAMG. You
> will need to code the problem in the style of SNES ex5, which is a 
> Poisson for which geometric MG works from
> the command line.
>
>   Thanks,
>
>      Matt
Hi Matt,

First of all, is there an easier way out? Is it Boomeramg or GMRES which 
has a large memory requirement? Will changing e.g. GMRES to FGMRES or 
other ksp solvers solve the problem?

Also, I switched from geometric multigrid (GMG) to Boomeramg because the 
latter is faster. If I use PCMG, am I going back to the GMG path, which 
was slower?

Thanks!
>
>>       Matt
>>
>>         Recently, I tested it with Boomeramg as the preconditioner
>>         and GMRES as the ksp solver. There's a 20% increase in speed.
>>
>>         However, when I increased the grid resolution, I got the out
>>         of memory error. Changing the solver back to HYPRE solved the
>>         problem.
>>
>>         So does GMRES  + Boomeramg used more memory than other
>>         solvers? Are there alternatives?
>>
>>         Thank you.
>>
>>         -- 
>>         Yours sincerely,
>>
>>         TAY wee-beng
>>
>>
>>
>>
>>     -- 
>>     What most experimenters take for granted before they begin their
>>     experiments is infinitely more interesting than any results to
>>     which their experiments lead.
>>     -- Norbert Wiener
>
>
>
>
> -- 
> What most experimenters take for granted before they begin their 
> experiments is infinitely more interesting than any results to which 
> their experiments lead.
> -- Norbert Wiener

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20140327/b71a1a65/attachment.html>