[petsc-users] Reuse hypre interpolations between successive KSPOnly solves associated with TSStep ?

[Mark Adams]

On Mon, May 4, 2020 at 10:24 AM Sajid Ali <sajidsyed2021 at u.northwestern.edu>
wrote:

> Hi PETSc-developers,
>
> For a linear TS, when solving with -ts_type cn -ksp_type fgmres -pc_type
> gamg, the flag -pc_gamg_reuse_interpolation can be used to re-use the
> GAMG interpolation calculated when solving the first time step. Increasing
> the number of time steps only increases the time spent in application of
> the precondition and the setup time is constant.
>
> Is there an equivalent way to do this with Hypre’s BoomerAMG ?
>

I don't know but you can look at the BoomerAMG documentation and if you
find something that does what you want you can look at the PETSc source
in /src/ksp/pc/impls/hypre/hypre.c and see the idiom for getting a hypre
object and manipulating it. You can then call the Hypre API directly and do
whatever you want.


> I’m using euclid as the smoother but adding the -pc_hypre_euclid_reuse
> flag only seems to make it reuse the euclid ILU smoothing within a
> particular TS time step, i.e. increasing the number of time steps linearly
> increases the time spent in setting up the preconditioner.
>
> Another approach I tried was to use hypre within the PCHMG preconditioner
> (which I hoped would allow me to use BoomerAMG to computer interpolations
> while also reusing the interpolations via the -pc_hmg_reuse_interpolation
> flag.) However, I’m unable to pass the parameter to BoomerAMG when Hypre is
> set as the inner PC for the HMG preconditioner. I didn’t see any options
> for doing so when I searched the -help output and neither does setting
> options via -hmg_inner_pc_hypre_boomeramg_option value work. What am I
> missing here ?
>
> For reference, I’m adding the exact solver options used (and the
> associated log files) for each of the above cases if it helps. (GAMG was
> used on the system of equations arising out of the complex PDE but since I
> was unable to get good convergence rates with GAMG for the real equivalent
> of the same, I switched to Hypre which performed better. Hence I’m looking
> for a way to set the reuse interpolation feature for this as well.)
>
> Thank You,
> Sajid Ali | PhD Candidate
> Applied Physics
> Northwestern University
> s-sajid-ali.github.io
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20200504/0f3b6876/attachment.html>