Skip to content

Generalise linear solvers #648

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 33 commits into
base: main
Choose a base branch
from
Open

Generalise linear solvers #648

wants to merge 33 commits into from

Conversation

@JHopeCollins
Copy link
Member

@JHopeCollins JHopeCollins commented Aug 7, 2025

No description provided.

@tommbendall tommbendall added enhancement Involves adding a new capability design discussion Issues relating to some code design. May be better as a discussion tidying Involves tidying up code and removed design discussion Issues relating to some code design. May be better as a discussion labels Aug 8, 2025
JHopeCollins and others added 23 commits August 8, 2025 10:18
@JHopeCollins
remove aP from ConvectiveSWESolver
99ca85b
@tommbendall
update linearisation map and default solver
9cfe8b1
@atb1995
Saving progress before merge in of linearisation map
807c7c0
@atb1995
Merge remote-tracking branch 'origin/TBendall/SolverEqns' into auxili…
5e51f2c 
…ary-equation-pc
@JHopeCollins
PC for calculating a schur complement using slate
e508449
@JHopeCollins
remove MoistConvectiveSWSolver
593f7ba
@atb1995
Revert "Merge remote-tracking branch 'origin/TBendall/SolverEqns' int…
32ce053 
…o auxiliary-equation-pc"

This reverts commit 5e51f2c, reversing
changes made to 807c7c0.
@atb1995
Monolithic solver for hard coded linear thermal eqns added
6ae2e4b
@atb1995
added monolithic solver for thermal shallow water eqns
55f7cc0
@atb1995
Remove declaration..
1f636e4
@JHopeCollins
refactor monolithic thermal swe linear solver
a326eda
@JHopeCollins
Merge branch 'main' into auxiliary-equation-pc
f036239
@JHopeCollins
expunge shallow water linear solvers
fd6c755
@JHopeCollins
boussinesq schur complement?
86da3dc
@atb1995
Merge to main
caf2b02
@atb1995
Solver set up for compressible euler - still requires testing
bbcfe93
@atb1995
Minor changes
177c06f
@atb1995
Solver changes
44d0037
@atb1995
Fixing of avg terms
6a2e8e8
@atb1995
First stab at custom PC for gusto solver, almost guaranteed to fail..
766cd25
@atb1995
Code runs! But the solver is unstable...
1ef983a
@atb1995
New linear preconditioner appears to be working!
017a875
@atb1995
all still works - reference profile is properly updated + dy is only …
adf3b3a 
…set to 0 with a flag
JHopeCollins
JHopeCollins commented Dec 3, 2025
View reviewed changes
gusto/timestepping/semi_implicit_quasi_newton.py Outdated
Comment on lines 983 to 984
if self.enforce_pc_on_rhs:
self.dy.assign(Constant(0.0))
Copy link
Member Author

@JHopeCollins JHopeCollins Dec 3, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think you always want to do this for a linear solve. Are there cases when you've found you need enforce_pc_on_rhs=True to match previous results?

Copy link
Collaborator

@atb1995 atb1995 Dec 3, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes - in which case I can just force it to be 0

Copy link
Member Author

@JHopeCollins JHopeCollins Dec 3, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry, I meant is there anywhere you need enforce_pc_on_rhs=False!

@atb1995
Large tidy up and connecting of solver options. Still to do: incompre…
07ce4e8 
…ssible boussinesq, ThermalSW equiv b, moist convective SW, double check compressible Euler reproducibility, edit TR-BDF2
JHopeCollins
JHopeCollins commented Dec 4, 2025
View reviewed changes
Copy link
Member Author

@JHopeCollins JHopeCollins left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If you add the pc.getOptionsPrefix() to the compressible pc subsolvers then you can pass options at the expected level in the options dictionary.

JHopeCollins
JHopeCollins commented Dec 4, 2025
View reviewed changes
JHopeCollins
JHopeCollins commented Dec 4, 2025
View reviewed changes
gusto/solvers/solver_presets.py

__all__ = ["HybridisedSolverParameters"]

def HybridisedSolverParameters(equation, alpha=0.5, tau_values=None, nonlinear=False):
Copy link
Member Author

@JHopeCollins JHopeCollins Dec 4, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seeing as all the defaults use some form of hybridisation, it could be useful to have a hybridisation_params optional argument so that people can change the solver options for the trace system (either "hybridization_" or "condensed_field_") when they use this function.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@atb1995 atb1995 atb1995 left review comments

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

enhancement Involves adding a new capability tidying Involves tidying up code

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@JHopeCollins @atb1995 @tommbendall