Skip to content

Firedrake meeting 2026 05 12

Pablo Brubeck edited this page May 12, 2026 · 16 revisions

Date and time 2026-05-12 1600 UTC+1

Action Items

  1. Pick Chair and Minuter (LC to pick -> PB)
  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. CW: More testing configurations (minutes)
  7. JHC: New manual release

Agenda

Present: PB, DH, CW, JHC, LC, IM

Apologies:

Team catch up

PB: Should we merge SubmeshHierarchy #5083 into release?

The main substance of the PR is SubmeshHierarchy, which is a very minor feature that does not break API.

We also added the first Submesh demo, demonstrating how to solve a very simple PDE with volume-surface coupling. The demo will be an excellent point of reference on how to deal with Measure intersection.

At least two users have asked about using multigrid + Submesh since the April release. Do we want to wait until the next release?

Conclusion: A change in merge policy, as we are rapidly evolving, we will start accepting new feature additions to be merged into release, as long as they are strictly additive.

Simple fix needed in order for Irksome + Submesh work out of the box.

Merged into release.

DH: we can set do_remove_component_tensors conditionally in release, and undonditionally in main. In that way we make sure that we don't cause different behaviour until the next relase.

CW: Testing configurations

We now have:

  • Linux x86, Linux ARM, macOS
  • real, complex
  • single, double
  • no GPU, CUDA
  • int32, int64 (ideally)

I want to make sure that we can test them all thoroughly but without destroying our CI runners.

I propose:

  • On every PR and push to release or main we run the full test suite for the 'default' configuration (Linux x86, real, double, no GPU)
  • If a PR is marked ci:complex, ci:gpu, ci:macos, ci:single etc then we run the full test suite for that configuration, not just firedrake-check
  • When we push to release or main we build every possible configuration and run a subset of the test suite. This subset will be more than firedrake-check and include any tests specific to that configuration (e.g. GPU+OffloadPC).

DH: too many combinations

CW: we will explicitly enumerate them and claim that we support them, but we only test a subset of these. We will allow for commuinity configuations, but not necessarily test them.

CW: we will only run certain subset of the test suite per configuation.

CW: Related PR

https://github.com/firedrakeproject/firedrake/pull/5099

CW: New manual release

Someone who isn't yet getting citations please do this :)

JHC: I will do it.

Looking for feedback based on the demo.

DH: do things in the AppCtx die at the right time?

JHC: yes, we don't keep a handle of those things

Merge PRs

Date of next meeting

1600 UTC+0100 2026-05-19

Clone this wiki locally