Dear Martin,

first, thank-you for your help.

Really I don't know what to do between these possibilities:

1) Using  TrilinosWrappers::PreconditionStokes Class 
(http://www.dealii.org/developer/doxygen/deal.II/classTrilinosWrappers_1_1PreconditionStokes.html).
 
This implies to download, configure and rebuild the subversion of 
DealII. Now I am using the version 6.1.0. Besides, if you tell me that 
perhaps there are some bugs working with NS...

2) Following tutorial step-20 
(http://www.dealii.org/6.1.0/doxygen/tutorial/index.html). Taking into 
account that I am working with NS and that M will be not symmetric. This 
case allows me to use version 6.1.0, that is the present version I have 
installed.

3) Or, as it is shown at the end of the tutorial step-20 to use 
PreconditionLACSolver 
(http://www.dealii.org/6.1.0/doxygen/deal.II/classPreconditionLACSolver.html) 
as InverseMatrix and SchurMatrix 
(http://www.dealii.org/6.1.0/doxygen/deal.II/classSchurMatrix.html) as 
Schur complement. This case allows me to use version 6.1.0, that is the 
present version I have installed.

What option do you think is the best one? Any advice will be welcome.
Best
Isa

Martin Kronbichler wrote:
> Dear Isabel,
>
>> so I mean I would have to use "TrilinosWrappers::PreconditionStokes
>> Class", wouldn't I?
>>
>> And this class is in the developing part of the library DealII, isn't 
>> it?
>
> Yes, you can give it a try (but I'm not sure whether this class will 
> remain in the library in the current form also in the future, since it 
> is very specific, whereas the library is rather meant to be general).
>
> As the author of that piece of code I have to admit that it is not
> really tested (actually I only tested the Stokes part rigorously, but
> not the necessary changes for Navier-Stokes, so I'm pretty sure there
> still are some bugs). You will need to have the matrix in
> TrilinosWrappers::BlockSparseMatrix format and the vector as
> TrilinosWrappers::BlockVector. And then there need to be some more
> matrices around, like a pressure mass, Laplace and convection-diffusion
> matrix (the last one you build just as the Navier-Stokes velocity
> matrix, but on the scalar pressure space)...
>
> So just try it and report me any problems!
>
> Best,
> Martin
>



_______________________________________________

Reply via email to