Allow for solving multi-domain problems involving codim-0 submeshes by ksagiyam · Pull Request #3478 · firedrakeproject/firedrake

ksagiyam · 2024-03-27T15:44:33Z

Depends on:
~~UFL: FEniCS/ufl#264~~ Merged.

Allow for solving multi-domain problems involving codim-0 submeshes.

tests/submesh/test_submesh_assemble.py

firedrake/mesh.py

github-actions · 2024-11-15T11:25:10Z

	Tests	Passed ✅	Skipped ⏭️	Failed ❌
Firedrake complex	8354 ran	6710 passed	1644 skipped	0 failed

github-actions · 2024-11-15T11:32:59Z

	Tests	Passed ✅	Skipped ⏭️	Failed ❌
Firedrake real	8315 ran	7601 passed	714 skipped	0 failed

pbrubeck · 2024-12-22T22:36:16Z

tsfc/fem.py

+    # This way of caching is fragile.
+    # Should Implement _hash_key_() for ModifiedTerminal and use the entire mt as key.
+    return (ufl_element, mt.local_derivatives, ctx.point_set, ctx.integration_dim, entity_id, coordinate_element, mt.restriction, domain._ufl_hash_data_())


I believe this key should not depend on the mesh. We just need to call finat once, independently of mt.terminal (and its domain).

CoordinateMapping uses mt.terminal:

@serial_cache(hashkey=make_basis_evaluation_key) def basis_evaluation(self, finat_element, mt, entity_id): return finat_element.basis_evaluation(mt.local_derivatives, self.point_set, (self.integration_dim, entity_id), coordinate_mapping=CoordinateMapping(mt, self))

But internally it only cares about the element

We should probably make CoordinateMapping only depend on the UFL element

In concept that might be true, but we probably need refactoring. CoordinateMapping also has:

def cell_size(self): return self.interface.cell_size(self.mt.restriction)

which implicitly depends on domain.

Either way, as long as we pass mt to basis_evaluation, we need to use the entire mt in the cache key to ensure that we can safely use mt.something_new in the relevant parts of the code.

github-actions · 2025-03-07T18:26:43Z

	Tests	Passed ✅	Skipped ⏭️	Failed ❌
Firedrake default	8267 ran	7568 passed	699 skipped	0 failed

ksagiyam · 2025-06-18T10:14:44Z

tests/firedrake/submesh/test_submesh_assemble.py

+    print(mesh, mesh_l, mesh_r, mesh_rl)
+    dS = Measure(
+        "dS", domain=mesh,
+        additional_domain_integral_type_map={


Suggested change

additional_domain_integral_type_map={

additional_domain_integral_type_map={

Suggested change

additional_domain_integral_type_map={

extra_measures={

ksagiyam · 2025-06-25T14:00:41Z

firedrake/assemble.py

+    def iter_active_coordinates(form, kinfo):
+        """Yield the form coordinates referenced in ``kinfo``."""
+        all_meshes = extract_domains(form)
+        for i in kinfo.active_domain_numbers.coordinates:
+            yield all_meshes[i].coordinates
+


Note: Make GlocalKernel args and Parloop args only for the active coordinates (and so on).

ksagiyam · 2025-06-25T14:06:04Z

tests/firedrake/submesh/test_submesh_assemble.py

+    dS = Measure(
+        "dS", domain=mesh,
+        extra_measures={
+            mesh_l: "dS",
+            mesh_r: "ds",
+            mesh_rl: "ds",
+        }
+    )


Note: Define Measures like this.

ksagiyam · 2025-06-25T14:09:37Z

tsfc/kernel_interface/common.py

+    def entity_number(self, domain, restriction):
        """Facet or vertex number as a GEM index."""
-        # Assume self._entity_number dict is set up at this point.
-        return self._entity_number[restriction]
+        if not hasattr(self, "_entity_numbers"):
+            raise RuntimeError("Haven't called set_entity_numbers")
+        return self._entity_numbers[domain][restriction]


Note: Define these gem.Variable(s) per domain. When we grab these from tsfc/fem.py, we need to use the domain as key.

ksagiyam · 2025-06-25T14:10:19Z

tsfc/kernel_interface/firedrake_loopy.py

+ActiveDomainNumbers = namedtuple('ActiveDomainNumbers', ['coordinates',
+                                                         'cell_orientations',
+                                                         'cell_sizes',
+                                                         'exterior_facets',
+                                                         'interior_facets',
+                                                         'exterior_facet_orientations',
+                                                         'interior_facet_orientations'])
+ActiveDomainNumbers.__doc__ = """
+    Active domain numbers collected for each key.
+
+    """


Note: ActiveDomainNumbers.

ksagiyam · 2025-06-25T14:11:10Z

tsfc/kernel_interface/firedrake_loopy.py

+    def set_coordinates(self, domains):
+        """Set coordinates for each domain.
+
+        Parameters
+        ----------
+        domains : list or tuple
+            All domains in the form.

-        :arg domain: :class:`ufl.Domain`
        """
        # Create a fake coordinate coefficient for a domain.
-        f = Coefficient(FunctionSpace(domain, domain.ufl_coordinate_element()))
-        self.domain_coordinate[domain] = f
-        self._coefficient(f, "coords")
+        for i, domain in enumerate(domains):
+            if isinstance(domain, MeshSequence):
+                raise RuntimeError("Found a MeshSequence")
+            f = Coefficient(FunctionSpace(domain, domain.ufl_coordinate_element()))
+            self.domain_coordinate[domain] = f
+            self._coefficient(f, f"coords_{i}")


Note: We do these things for each domain.

ksagiyam · 2025-06-25T14:14:15Z

tsfc/kernel_interface/firedrake_loopy.py

+        active_domain_numbers_coordinates, args_ = self.make_active_domain_numbers({d: self.coefficient_map[c] for d, c in self.domain_coordinate.items()},
+                                                                                   active_variables,
+                                                                                   kernel_args.CoordinatesKernelArg)
+        args.extend(args_)


Note: After calling self.compile_gem(), we use self.make_active_domain_numbers() to get indices of the active domains.

dham · 2025-06-25T15:49:26Z

tests/firedrake/submesh/test_submesh_assemble.py

+    dS = Measure(
+        "dS", domain=mesh,
+        extra_measures={
+            mesh_l: "dS",


I think the value here should be a Measure

ksagiyam · 2025-10-15T19:38:06Z

Reviewed multiple times. Merging.

ksagiyam force-pushed the ksagiyam/submesh_core branch 3 times, most recently from 9b8f70c to 1e7c035 Compare April 4, 2024 01:07

ksagiyam force-pushed the ksagiyam/submesh_core branch 10 times, most recently from 219240f to d555615 Compare April 11, 2024 12:19

ksagiyam force-pushed the ksagiyam/submesh_core branch 2 times, most recently from 9b1d48b to 23caf2e Compare April 17, 2024 21:14

ksagiyam force-pushed the ksagiyam/submesh_core branch 7 times, most recently from f5afedb to 824f2a5 Compare May 7, 2024 14:57

ksagiyam force-pushed the ksagiyam/submesh_core branch 2 times, most recently from 76e89ef to 7e3c248 Compare May 15, 2024 22:16

ksagiyam force-pushed the ksagiyam/submesh_core branch from b6a41e7 to 9c839e0 Compare May 27, 2024 13:20

ksagiyam force-pushed the ksagiyam/submesh_core branch 2 times, most recently from 1167181 to fc6545c Compare June 4, 2024 19:41

ksagiyam force-pushed the ksagiyam/submesh_core branch from a76cca5 to 630bfb4 Compare June 10, 2024 21:49

ksagiyam force-pushed the ksagiyam/submesh_core branch from 2fe02f9 to 8ba65d6 Compare June 26, 2024 21:36

ksagiyam force-pushed the ksagiyam/submesh_core branch from e204069 to b8cc6f6 Compare July 12, 2024 09:49

ksagiyam force-pushed the ksagiyam/submesh_core branch 4 times, most recently from 9608d6e to 654d56a Compare August 20, 2024 15:04

ksagiyam commented Aug 21, 2024

View reviewed changes

tests/submesh/test_submesh_assemble.py Outdated Show resolved Hide resolved

ksagiyam commented Aug 21, 2024

View reviewed changes

firedrake/mesh.py Show resolved Hide resolved

ksagiyam force-pushed the ksagiyam/submesh_core branch 5 times, most recently from 9d39ea2 to 6972622 Compare August 22, 2024 14:42

ksagiyam mentioned this pull request Aug 22, 2024

allow for ghost only mesh #3743

Merged

ksagiyam force-pushed the ksagiyam/submesh_core branch from 6972622 to a6335d4 Compare August 22, 2024 21:23

ksagiyam force-pushed the ksagiyam/submesh_core branch 2 times, most recently from c9497b3 to 435fc60 Compare September 12, 2024 02:32

ksagiyam force-pushed the ksagiyam/submesh_core branch from 435fc60 to a9ce114 Compare November 15, 2024 10:19

ksagiyam force-pushed the ksagiyam/submesh_core branch 2 times, most recently from a683850 to ed0e3b1 Compare November 16, 2024 12:01

pbrubeck reviewed Dec 22, 2024

View reviewed changes

ksagiyam commented Jun 18, 2025

View reviewed changes

ksagiyam commented Jun 25, 2025

View reviewed changes

dham reviewed Jun 25, 2025

View reviewed changes

This was referenced Sep 2, 2025

adapt for MeshSequence g-adopt/g-adopt#287

Merged

adapt for MeshSequence firedrakeproject/gusto#655

Merged

leo-collins mentioned this pull request Sep 29, 2025

MixedInterpolator #4596

Merged

Enable solving multi-domain problems involving codim-0 submeshes

aebe6cd

	additional_domain_integral_type_map={
	additional_domain_integral_type_map={

Conversation

ksagiyam commented Mar 27, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

github-actions bot commented Nov 15, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Nov 15, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Mar 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ksagiyam Jun 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ksagiyam commented Oct 15, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

ksagiyam commented Mar 27, 2024 •

edited

Loading

github-actions bot commented Nov 15, 2024 •

edited

Loading

github-actions bot commented Nov 15, 2024 •

edited

Loading

github-actions bot commented Mar 7, 2025 •

edited

Loading

ksagiyam Jun 25, 2025 •

edited

Loading