Dear Bruno,
Thanks a lot for the positive feedback! I’m really glad to know that those
changes worked well for you. Its always great to hear that the improvements
that we make to deal.II (be them small or large) are impactful and help our
users, and it certainly encourages us to keep at it.
Dear Jean-Paul,
Thank you, everything works. Right now for simple test cases I have managed
to get interesting results:
- Going from ILU to AMG leads to speed-up of 4X on the same number of
processor.
- More importantly, my number of GMRES iteration remains extremely low
(below
Dear Jean-Paul,
Thank you very much. I will look into this and the documentation.
I will post you the final results.
I greatly appreciate your support and these very nice changes to the
wrapper :)
Best
Bruno
On Tuesday, 2 April 2019 05:55:26 UTC-4, Jean-Paul Pelteret wrote:
>
> Dear Bruno,
>
>
Dear Bruno,
These are the basic steps that you need to take:
1. Create and initialise a PreconditionAMG::AdditionalData object
(“additional_data”) as per usual. Here you set up the constant modes as is
needed using DoFTools::extract_constant_modes().
2. Create a Teuchos::ParameterList
Dear Jean-Paul,
I've compiled the new DEALII version with your changes, but I am unsure of
how to use the changes correctly.
If I understand correctly, you can now set Teuchos parameters after you
have initialized the TrilinosWrappers::PreconditionAMG.
This is achieved by creating a
Dear Jean-Paul.
That's good, I will test the AMG with the stabilized approach and see what
I can get.
If the results are interesting (well they already are right now for ILU +
GMRES) i'll open up an issue on the github for the dicussion on that more
specialized topic.
Thank you for everything
On 3/12/19 3:06 AM, Jean-Paul Pelteret wrote:
>
> Thanks for being willing to share your findings wrt. AMG parameters. I think
> that would be really useful. I also think that encapsulating a parameter
> study
> for NS + AMG/GMG within a tutorial could be interesting and valuable. Maybe
> you
Dear Bruno,
I believe that my proposed changes are working now, so you could always pull
that branch from my repository and use it immediately. That might be the most
convenient way for you to move forward.
Thanks for being willing to share your findings wrt. AMG parameters. I think
that
Dear Jean-Paul and Wolfgang,
Is it better if I try to go with the Teuchos way or should I wait until
Jean-Paul finishes the merge on his branch ? I think the latter might be a
good option.
In all cases, I would be more than glad to share my experience in
parameters tuning for the AMG solver
On 3/8/19 9:26 AM, Bruno Blais wrote:
>
> Clearly, parameters like fill-level, rel. treshold and abs. threshold could
> be
> modified. Is there any way to access them and modify them? I have not been
> able to as of now.
Bruno -- Jean-Paul already outlined what's necessary. But I wanted to
Dear Bruno,
Indeed, it appears that one isn't able to change the settings passed to the AMG
preconditioner other than those set by the AdditionalData object. Your only
current alternative is to create the Teuchos::ParameterList yourself, which you
can send in to the call to initialize()
Hello everyone,
Just a small question.
Looking at the documentation and at the .h of the
TrilinosWrappers::PreconditionAMG I have not found a way to access the
smoother and coarse solver to play with their parameters. Say if you use an
ILU or ILUT smoother.
When you use the verbose output of
12 matches
Mail list logo
