[petsc-users] DIVERGED_NANORINF when using HYPRE/BoomerAMG

[Mark Adams]

We don't know the details of hypre, we just provide an interface. For
algorithmic questions you would need to ask the hypre tea.

Tell them your equations and discretizations and what you have observed. I
imagine they can help.

Mark

On Thu, Jul 9, 2020 at 8:59 AM Andrea Iob <andrea_iob at hotmail.com> wrote:

> Sorry, I've clicked the send button by mistake.
>
> The setup I'm using comes from the following guide:
>
> https://mooseframework.inl.gov/application_development/hypre.html
>
> Since I'm new to BoomerAMG I've been using that guide as a reference, but
> I'm not sure if those setting are appropriate from my system.
>
> Best regards,
> Andrea
>
>
> ------------------------------
> *From:* Andrea Iob <andrea_iob at hotmail.com>
> *Sent:* Thursday, July 9, 2020 2:55 PM
> *To:* Mark Adams <mfadams at lbl.gov>
> *Cc:* Barry Smith <bsmith at petsc.dev>; Matthew Knepley <knepley at gmail.com>;
> petsc-users at mcs.anl.gov <petsc-users at mcs.anl.gov>
> *Subject:* Re: [petsc-users] DIVERGED_NANORINF when using HYPRE/BoomerAMG
>
> Yes, I'm using AIJ matrices.
>
> These are the options I'm using to set up BoomerAMG:
>
>         PCHYPRESetType(preconditioner, "boomeramg");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_smooth_type",
> "Pilut");
>         PetscOptionsSetValue(nullptr,
> "-pc_hypre_boomeramg_relax_type_all", "sequential-Gauss-Seidel");
>         PetscOptionsSetValue(nullptr,
> "-pc_hypre_boomeramg_strong_threshold", "0.7");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_coarsen_type",
> "HMIS");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_interp_type",
> "classical");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_interp_type",
> "ext+i");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_P_max", "2");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_truncfactor",
> "0.3");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_agg_nl", "4");
>         PetscOptionsSetValue(nullptr, "-pc_hypre_boomeramg_agg_num_paths",
> "5");
>         PetscOptionsSetValue(nullptr, "-ksp_error_if_not_converged", "");
>
> This setup come from the following
>
> https://mooseframework.inl.gov/application_development/hypre.html
>
> ------------------------------
> *From:* Mark Adams <mfadams at lbl.gov>
> *Sent:* Friday, July 3, 2020 2:58 PM
> *To:* Andrea Iob <andrea_iob at hotmail.com>
> *Cc:* Barry Smith <bsmith at petsc.dev>; Matthew Knepley <knepley at gmail.com>;
> petsc-users at mcs.anl.gov <petsc-users at mcs.anl.gov>
> *Subject:* Re: [petsc-users] DIVERGED_NANORINF when using HYPRE/BoomerAMG
>
>
>
>
> 3) Schwarz-smoothers
>
> #0  0x00007ffff50897f4 in dtrsm_ () from
> /opt/lapack/3.8.0-gcc.7.4.0/lib64/libblas.so.3
> #1  0x00007ffff550924b in dpotrs_ () from
> /opt/lapack/3.8.0-gcc.7.4.0/lib64/liblapack.so.3
>
>
> This seems to be Cholesky ... which would not work for you if you are not
> symmetric (Navier-Stokes).
>
> If GAMG works then and pilut going crazy but quite there yet, I would
> think there is a hypre interface problem of some sort.
>
> Hypre has been used a lot so this is a surprise. Do you use AIJ matrices?
> Not BAIJ or anything else.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20200709/d2e2ace3/attachment.html>