Firedrake meeting 2020 05 05 - firedrakeproject/firedrake GitHub Wiki

Date and time 2020-05-05 15:00UTC (16:00BST)

Action Items

  1. ALL: (ongoing) triage the open issues and confirm if they are indeed still open (and perhaps provide labels)
  2. KS, (DH, LM): Document describing what we think the mixed domain interface should look like (and hence what is needed in UFL, and whether it matches the existing fenics efforts).
  3. DRS: Open PR for stream plot See PR #1670. Review required
  4. DH: Tackle the remaining complex sprint items
  5. FW: Add the hack to have shape derivatives tests working in complex
  6. SV: Expunge references to COFFEE before tsslac merge
  7. LM: Go through the tsslac PRs
  8. JB: Add mechanism for python3.8
  9. DH: Find time to fix to get final complex sprint test passing.
  10. ALL: Someone to look at TSFC merge needed as part of complex sprint.
  11. DH: Find time to sort out Jenkins to run in real and complex mode.

Minutes

Present:

Apologies:

DRS - Stream Plot PR

Correct metric for various cell geometric quantities discussed. "cell size" vs "cell diameter" where the latter is needed to work on quads. DECISION: "cell size" will return "cell diameter" on quads.

DH Complex Sprint

Now passing in almost all cases - 1 bug (where a test function and trial function get mixed up) needs to be fixed. Noted that real mode could be improved with better implementation of PETSc real matrices.

Merging plan discussed.

PSA: Merging complex will double testing load. Jenkins will stop testing all commits. Only merge commits will be tested and only branches with active pull requests will be tested.

SV: tsslac merge

Work ongoing.

LM: COFFEE in codebase

Reasons to keep coffee in:

  • TSFC used as backend for dolfin;
  • hard work to remove from certain places where need to glue kernels together;
  • loopy doesn't work for everything that users want to do and eats memory for breakfast. Note: In future, prefer to use TSFC kernel interface to decide which code generation to use.

LM: Improving Loopy

Needs someone from firedrake with time to fix things when bugs arrive.

JB: Status of issue #1451

Agenda: In light of dropping COFFEE in the (near?) future, what is the current status of: https://github.com/firedrakeproject/firedrake/issues/1451

Minutes: See LM: COFFEE in codebase

RWH: further discussion needed on #1667

Agenda: Proposed solution from last meeting introduces new problems without obvious solutions. See https://github.com/firedrakeproject/firedrake/pull/1667

Minutes: Will modify make_scalar_element to use mesh topology's ufl_cell. Will make sure that all ufl_cells anywhere else which ultimately are created on a mesh topology will use the mesh topology's ufl cell.

Additional notes

Luke Olson's multigrid lectures start on 2020-05-08 1500UFC (1700UTC):

Date of next meeting

2020-05-12 15:00UTC (16:00BST)