PROM Circuit Extraction #326

phdum-a · 2025-01-10T17:33:24Z

Obtain circuit from PROM operator.

~~Draft: Pending #302, #309 merges and expecting follow up changes from those.~~

hughcars

Took a quick look whilst waiting on your fixes on the normalization.

palace/models/romoperator.cpp

hughcars · 2025-10-02T16:26:51Z

palace/drivers/drivensolver.cpp

+    }
+    const auto &orth_R = prom_op.GetRomOrthogonalityMatrix();
+    MFEM_VERIFY(orth_R.isDiagonal(), "Lumped port modes should have exactly zero overlap");
+    // TODO(phdum): Test this.


Debris, I assume wip given our inner product discussions.

palace/models/romoperator.hpp

palace/models/romoperator.cpp

hughcars · 2025-10-02T19:34:47Z

palace/models/romoperator.cpp

+  // nullptr check. Kr, Mr, Cr would require a numerical check on imag elements.
+
+  auto unit_henry = units.GetScaleFactor<Units::ValueType::INDUCTANCE>();
+  auto m_Linv = ((1.0 / unit_henry) * v_d) * Kr * v_d;


m_Linv isn't a great unit name, is this the Linv for a mode?

hughcars · 2025-10-02T19:35:31Z

palace/models/spaceoperator.cpp

+  RHS.SetSize(GetNDSpace().GetTrueVSize());
+  RHS.UseDevice(true);
+  RHS = 0.0;
+
+  MFEM_VERIFY(RHS.Size() == GetNDSpace().GetTrueVSize(),
+              "Invalid T-vector size for AddExcitationVector1Internal!");
+  SumVectorCoefficient fb(GetMesh().SpaceDimension());
+
+  const auto &data = lumped_port_op.GetPort(port_idx);
+
+  mfem::Array<int> attr_list;
+  mfem::Array<int> attr_marker;
+
+  for (const auto &elem : data.elems)
+  {
+    attr_list.Append(elem->GetAttrList());
+    fb.AddCoefficient(
+        elem->GetModeCoefficient(1.0 / (elem->GetGeometryWidth() * data.elems.size())));
+  }
+  auto &mesh = GetNDSpace().GetParMesh();
+  int bdr_attr_max = mesh.bdr_attributes.Size() ? mesh.bdr_attributes.Max() : 0;
+  mesh::AttrToMarker(bdr_attr_max, attr_list, attr_marker);
+
+  mfem::LinearForm rhs1(&GetNDSpace().Get());
+  rhs1.AddBoundaryIntegrator(new VectorFEBoundaryLFIntegrator(fb), attr_marker);
+  rhs1.UseFastAssembly(false);
+  rhs1.UseDevice(false);
+  rhs1.Assemble();
+  rhs1.UseDevice(true);
+  GetNDSpace().GetProlongationMatrix()->AddMultTranspose(rhs1, RHS.Real());
+
+  if (zero_metal)
+  {
+    linalg::SetSubVector(RHS.Real(), nd_dbc_tdof_lists.back(), 0.0);
+  }
+  return true;


I'm not looking at this rn based on our discussions, I believe you're fixing this up based on your new understanding.

hughcars · 2025-10-02T19:35:43Z

palace/models/spaceoperator.hpp

  bool GetExcitationVector(int excitation_idx, Vector &RHS);
  bool GetExcitationVector(int excitation_idx, double omega, ComplexVector &RHS);

+  bool GetLumpedPortExcitationVector(int port_idx, ComplexVector &RHS, bool zero_metal);


Doc string.

palace/drivers/drivensolver.cpp

- Replace manual vector search and insert with stl functions

- Don't fix max basis size by using vector reserve - De-duplicate code - Fix PROM printing bug

- Improve comments - Add verifications and small code cleaning

- Add unit tests for both weighted and existing orthogonalization

- Move logic to add port modes into prom_op. port mode adding into rom operator - Small comments fixup

- Add orthogonality tolerance in UpdatePROM - Always use refined Gram-Schmidt in RomOp with circuit synthesis

- Works on 3x2x1 domain - Test lumped port normalizations, power, voltage etc - Includes single and multi-element port - Includes cases with proximate PEC to port

- Weight matrix of hybrid domain-port mass term

- Merge weighted and unweighted inner product

- Make a private function to help testing

- Bind && to local object in ctor

phdum-a force-pushed the phdum/multi_excitation branch 3 times, most recently from 0bbd871 to 04a6a38 Compare March 28, 2025 21:56

phdum-a force-pushed the phdum/multi_excitation branch 2 times, most recently from 576db7f to 9a8b33a Compare April 1, 2025 01:00

phdum-a force-pushed the phdum/prom_synthesis branch from 1345ffc to 99faed2 Compare April 2, 2025 18:48

phdum-a force-pushed the phdum/prom_synthesis branch 3 times, most recently from 4f9dcd1 to 1de618d Compare April 16, 2025 20:36

Base automatically changed from phdum/multi_excitation to main April 16, 2025 21:50

hughcars force-pushed the phdum/prom_synthesis branch from 1de618d to 777dbba Compare April 16, 2025 21:52

phdum-a force-pushed the phdum/prom_synthesis branch from 777dbba to 8410b5b Compare May 20, 2025 18:21

phdum-a force-pushed the phdum/prom_synthesis branch from 8410b5b to 645e19f Compare July 14, 2025 22:39

hughcars force-pushed the phdum/prom_synthesis branch 2 times, most recently from 00c89d6 to 6dea781 Compare August 26, 2025 16:40

phdum-a force-pushed the phdum/prom_synthesis branch from 6dea781 to 06c680d Compare August 30, 2025 00:12

phdum-a force-pushed the phdum/prom_synthesis branch from 06c680d to 70f588c Compare September 17, 2025 16:28

phdum-a marked this pull request as ready for review September 17, 2025 21:42

phdum-a requested a review from hughcars September 17, 2025 21:43

phdum-a force-pushed the phdum/prom_synthesis branch 5 times, most recently from 1e98f37 to 10a0443 Compare September 25, 2025 18:28

phdum-a force-pushed the phdum/prom_synthesis branch from 10a0443 to f9691b4 Compare September 29, 2025 21:47

hughcars reviewed Oct 2, 2025

View reviewed changes

phdum-a force-pushed the phdum/prom_synthesis branch from f9691b4 to eef61b9 Compare October 3, 2025 00:48

phdum-a force-pushed the phdum/prom_synthesis branch 3 times, most recently from 1b98be6 to f94f27f Compare November 20, 2025 00:27

phdum-a added 29 commits December 2, 2025 14:02

Add basic test for Minimal Rational Interpolation

a79b091

Quick-fix MRI multiple max errors

ec8b984

Better MRI multiple max error

f12ea81

- Replace manual vector search and insert with stl functions

Update docs and changelog

21ba6f3

Cleanup UpdatePROM

176f683

- Don't fix max basis size by using vector reserve - De-duplicate code - Fix PROM printing bug

Add new test labels to romoperator unit test

91d7b82

Address PR feedback (Round 1)

f078649

- Improve comments - Add verifications and small code cleaning

Add driven solver port orthogonality test for circuits

a2964dd

Explicit instantiation of BuildParSumOperator for single operator

b0acf9d

Add weighted orthogonalization

a91e246

- Add unit tests for both weighted and existing orthogonalization

Cleanup for prom synthesis

3948b63

- Move logic to add port modes into prom_op. port mode adding into rom operator - Small comments fixup

Revert rom matrix normalization to ports

6be71b8

Comment test due to MPI bug

0621cd7

Add both dual and primary vector to PROM

ff6795b

- Add orthogonality tolerance in UpdatePROM - Always use refined Gram-Schmidt in RomOp with circuit synthesis

Add LumpedPort basic test

f7b1505

- Works on 3x2x1 domain - Test lumped port normalizations, power, voltage etc - Includes single and multi-element port - Includes cases with proximate PEC to port

Fix LumpedPort documentation

a3d1415

First pass of new PROM orthogonalization

9ce7579

- Weight matrix of hybrid domain-port mass term

Add basic port meshes

22ac49c

Update orthog routines

91f0c56

- Merge weighted and unweighted inner product

Minor cleanup

683a4b9

Add IoData ctor to take parsed json class

8a9efee

Rework PROM inner product using hybrid matrix approach

016cbad

Small clang-tidy adjustments

41dec5c

Separate out ROM matrix normalization from printing

e1ec633

- Make a private function to help testing

Finalize ROM synthesis inner product test

f251216

Minor clang-tidy fixes

98a554b

Mark single argument ctors of IoData explicit

6b54351

- Bind && to local object in ctor

Prefer std::size_t over plain size_t

6bba335

Quick-fix Stratton-Chu test

6fc8c82

phdum-a force-pushed the phdum/prom_synthesis branch from 240d841 to 6fc8c82 Compare December 2, 2025 22:05

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

PROM Circuit Extraction #326

PROM Circuit Extraction #326

phdum-a commented Jan 10, 2025 •

edited

Loading

Uh oh!

hughcars left a comment

Uh oh!

Uh oh!

Uh oh!

hughcars Oct 2, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

hughcars Oct 2, 2025

Uh oh!

hughcars Oct 2, 2025

Uh oh!

hughcars Oct 2, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

PROM Circuit Extraction #326

Are you sure you want to change the base?

PROM Circuit Extraction #326

Conversation

phdum-a commented Jan 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hughcars left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

hughcars Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

hughcars Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

hughcars Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

hughcars Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

phdum-a commented Jan 10, 2025 •

edited

Loading