Firedrake meeting 2025 06 25 - firedrakeproject/firedrake GitHub Wiki

Date and time 2025-06-25 1600 UTC

Action Items

  1. Pick Chair and Minuter (CW to pick)
  2. ALL: (ongoing) triage the open issues and confirm if they are indeed still open (and perhaps provide labels)
  3. ALL: do things with SV's branches
  4. DH: Email to Andreas to have 2 (+ others!!!) loopy PRs merged TODO: FIND OUT WHICH PRS THESE ARE
  5. DH: Talk to GregVernon about PR#2116.
  6. JB: Enable merge queues (minutes)
  7. LC: Try to merge RNH' PR: Movable VOM
  8. PB: Profile and speed up some tests (minutes, minutes)
  9. CW: More testing configurations (minutes)
  10. All: Do post merge works (interpolate, abstract reduced functionals, etc.)
  11. CW: Preliminary prep for PETSc User Meeting 2026 (minutes)

Agenda

Present: DH, KS, IM, DRS, JHC, CW, LC

Apologies: PB

CW: petsc_raises

https://github.com/firedrakeproject/firedrake/pull/4384

DH: Upgrade Python on firedrake CI?

CW: We're currently using Ubuntu latest. We'd need to use Pyenv, which slows down CI.

DH: Don't upgrade Python, this is only a problem for 8 months until new Ubuntu.

DRS: hard to debug error

https://github.com/danshapero/mantle-convection-test

DRS: mantle convection example not working. Also failing on RK. Was working on install from April. Failing on most recent docker.

DRS: Failing in line search. Can fix sometimes by increasing max number of iterations.

JHC: With line search off, does it still fail?

DH: Residual is exploding. Is the problem ill-conditioned?

JHC: Is the problem non-dimensionalised? Mantle problems have large dimensions

DRS: It is. Rayleigh number is 10^6. Everything else is O(1).

JHC: Is the nonlinearity causing the ill-conditioning?

DRS: Was working on older installs.

JHC: Look at MUMPS version. If on Ubuntu you could be on an older MUMPS.

DH: Try passing --no-package-manager to firedrake-configure. This will configure PETSc closer to how we did it a few months ago.

KS: Submesh PRs

Small dependency:

UFL: https://github.com/FEniCS/ufl/pull/382

KS: small PR to fix UFL indexing

DH: Looks fine, need to ask fenics

Core implementation:

UFL: https://github.com/FEniCS/ufl/pull/264

DH: renamed MixedMesh to MeshSequence in PR title

DH: Needs more documentation for extra_measures parameter. It is publicly-facing

DH: Once we have submeshes, we can do integrals of quantities defined on different meshes. For facets, an exterior facet of a subdomain may be an interior facet of another subdomain. We integrate over intersections of primary measures and extra_measures.

KS: updated form signature

DH: correct variable names to hashdata

Looks good, just add documentation and also run by fenics

Firedrake: https://github.com/firedrakeproject/firedrake/pull/3478

KS: large file is output from benchmark

DH: probably not the right way to do this

DH: extra_measures dict should contain Measure and not strings. This is nearly there, just need to pester fenics people (mainly Jørgen).

Merge PRs

Note that PRs put in this section should either be trivial or already have been reviewed. Discussion-worthy PRs should be separate agenda items.

CW: Is the API change fine?

DH: API change is necessary. PR is approved.

DH: Pytest doesn't respect namespaces

DH: Lots of special-casing just for a single PETSc mat type. Is it necessary?

DH: Also need more tests.

KS: we should grab lgmap and underlying data in cython and change that.

DH: Australians noticed that if you use single memory output you get the wrong answer.

DH: checkpoints were being discarded using marked_in_path.

DH: fix was to split flags in BlockVariable.

DH: I need to add their MFE in as a test.

Date of next meeting

1600 UTC 2025-07-02