[Trilinos-Users] How best to solve Poisson eq. in cylindrical coordinates

[Travis Austin]

We solved our problem by catching an error in how we were defining the solver.  
Now ML performs great and we only need about 15 its for our tougher problem.  

I'd suggest playing around with the threshold value in ML.  Does performance
slowly degrade as you increase the aspect ratios?  How bad are they?  I may
be able to easily test out performance for a similar problem to get you a 
comparison.

Travis

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Travis Austin, Ph.D.
VP of Computational Mathematics
Tech-X Corporation
5621 Arapahoe Ave, Suite A
Boulder, CO 80303
austin at txcorp.com
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^



On Dec 27, 2011, at 6:06 PM, chris.jackson at mayahtt.com wrote:

> I'd also be interested to hear the reply.  We use ML as the preconditioner to BiCGStab for the Poisson problem in Cartesian coordinates, but we have similar slow convergence - presumably for the same reason - when the finite volume cells have large aspect ratios.
> 
> Thanks
> 
> Chris Jackson
> 
> Sent from my BlackBerry device on the Rogers Wireless Network
> 
> -----Original Message-----
> From: John R Cary <cary at colorado.edu>
> Sender: "trilinos-users-bounces at software.sandia.gov"
> 	<trilinos-users-bounces at software.sandia.gov>
> Date: Tue, 27 Dec 2011 10:20:45 
> To: Trilinos Users<Trilinos-Users at software.sandia.gov>
> Subject: [Trilinos-Users] How best to solve Poisson eq. in cylindrical
> coordinates
> 
> In this case, the coefficients of the operator
> vary dramatically, as the operator in the radial
> direction is
> 
> (1/r)(d/dr)(r d/dr) = d^2/dr^2 + (1/r)(d/dr)
> 
> We try plain ol' ML and convergence is slow,
> like 80 iterations.
> 
> Are there suggestions for this?
> 
> Thx....John Cary
> 
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users
> 
> 
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://software.sandia.gov/pipermail/trilinos-users/attachments/20111227/a1c4a726/attachment.html