Skip to content

LinearEigensolver: default to eps_target = 0 is conceptually wrong #4634

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

LinearEigensolver: default to eps_target = 0 is conceptually wrong #4634

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 4 additions & 5 deletions firedrake/eigensolver.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,9 +142,10 @@ class LinearEigensolver(OptionsManager):
"eps_largest_real": None
"""

DEFAULT_EPS_PARAMETERS = {"eps_type": "krylovschur",
"eps_tol": 1e-10,
"eps_target": 0.0}
DEFAULT_EPS_PARAMETERS = {
"eps_type": "krylovschur",
"eps_tol": 1e-10,
}

def __init__(self, problem, n_evals, *, options_prefix=None,
solver_parameters=None, ncv=None, mpd=None):
Expand All @@ -158,8 +159,6 @@ def __init__(self, problem, n_evals, *, options_prefix=None,
for key in self.DEFAULT_EPS_PARAMETERS:
value = self.DEFAULT_EPS_PARAMETERS[key]
solver_parameters.setdefault(key, value)
if self._problem.bcs:
solver_parameters.setdefault("st_type", "sinvert")
super().__init__(solver_parameters, options_prefix)
self.set_from_options(self.es)

Expand Down
21 changes: 14 additions & 7 deletions tests/firedrake/regression/test_eigensolver.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,10 +20,13 @@ def evals(n, degree=1, mesh=None, restrict=False):
bc = DirichletBC(V, 0.0, "on_boundary")
eigenprob = LinearEigenproblem(a, bcs=bc, bc_shift=-6666., restrict=restrict)

# Create corresponding eigensolver, looking for n eigenvalues
eigensolver = LinearEigensolver(
eigenprob, n, solver_parameters={"eps_largest_real": None}
)
# Create corresponding eigensolver, looking for n eigenvalues close to 0
# We use shift-and-invert as spectral transform (SLEPc's default is shift)
solver_parameters = {
"eps_target": 0,
"st_type": "sinvert",
}
eigensolver = LinearEigensolver(eigenprob, n, solver_parameters=solver_parameters)
ncov = eigensolver.solve()

# boffi solns
Expand Down Expand Up @@ -67,8 +70,12 @@ def poisson_eigenvalue_2d(i):
ep = LinearEigenproblem(inner(grad(u), grad(v)) * dx,
bcs=bc, bc_shift=666.0)

es = LinearEigensolver(ep, 1, solver_parameters={"eps_gen_hermitian": None,
"eps_largest_real": None})
solver_parameters = {
"eps_gen_hermitian": None,
"eps_target": 0,
"st_type": "sinvert",
}
es = LinearEigensolver(ep, 1, solver_parameters=solver_parameters)

es.solve()
return es.eigenvalue(0)-2.0
Expand All @@ -77,7 +84,7 @@ def poisson_eigenvalue_2d(i):
@pytest.mark.skipslepc
def test_evals_2d():
"""2D Eigenvalue convergence test. As with Boffi, we observe that the
convergence rate convergest to 2 from above."""
convergence rate converges to 2 from above."""
errors = np.array([poisson_eigenvalue_2d(i) for i in range(5)])

convergence = np.log(errors[:-1]/errors[1:])/np.log(2.0)
Expand Down