BUG: interpolating to VOM produces two blocks on the adjoint tape

[colinjcotter]

Describe the bug

The attached code does one interpolation to a vertex only mesh function. The resulting tape has two interpolate blocks.
I obtained this whilst trying to debug a more substantial issue where there is only one interpolate block but it is unconnected
to my functional on the tape graph. I couldn't find an MFE for it so I'm hoping this is related and exposes the same issue.

Steps to Reproduce
Steps to reproduce the behavior:

run the code below and view the tape graph output as pdf.

from firedrake import *
from firedrake_adjoint import *

mesh = UnitSquareMesh(10,10)

V = FunctionSpace(mesh, "CG", 1)

W = V * V

x, y = SpatialCoordinate(mesh)

w0 = Function(W).assign(1.0)

w1 = Function(W)

v0, v1 = w1.subfunctions
w1.project(w0)

VOM = VertexOnlyMesh(mesh, [[0.51,0.51]])
VVOM = FunctionSpace(VOM, "DG", 0)
y = Function(VVOM)

y.interpolate(v0)

J = assemble(y*y*dx)
Jhat = ReducedFunctional(J, (Control(w0)))

tape = pyadjoint.get_working_tape()
tape.visualise_pdf("testtape.pdf")

Expected behavior

I expect to only see one interpolate block on the grpah.

Error message
N/A

Environment:

• OS: Linux
• Python version: Python 3.10.6
• Output of firedrake-status
  Firedrake Configuration:
  package_manager: True
  minimal_petsc: False
  mpicc: None
  mpicxx: None
  mpif90: None
  mpiexec: None
  disable_ssh: False
  honour_petsc_dir: False
  with_parmetis: False
  slepc: False
  packages: []
  honour_pythonpath: False
  opencascade: False
  tinyasm: False
  torch: False
  petsc_int_type: int32
  cache_dir: /home/cjc1/firedrake/.cache
  complex: False
  remove_build_files: False
  with_blas: None
  netgen: False
  Additions:
  None
  Environment:
  PYTHONPATH: None
  PETSC_ARCH: None
  PETSC_DIR: None
  Status of components:

---

|Package |Branch |Revision |Modified |

|FInAT |master |47f6c37 |False |
|PyOP2 |master |edae2884 |False |
|fiat |master |8c66270 |False |
|firedrake |master |dd124559f |False |
|h5py |firedrake |6cc4c912 |False |
|libspatialindex |master |4768bf3 |True |
|libsupermesh |master |69012e5 |False |
|loopy |main |cd388f74 |False |
|petsc |firedrake |9364cb008b|False |
|pyadjoint |reduced_functional_parameters |05f3e18 |False |
|pytest-mpi |main |a478bc8 |False |
|tsfc |master |ef39f72 |False |
|ufl |ufl2unicode_vom |f4d7d915 |False |

• Any relevant environment variables or modifications [eg: PYOP2_DEBUG=1]
  N/A

[stephankramer]

I suspect this is completely harmless. It's just recording the interpolate it does inside the creation of the VertexOnlyMesh: firedrake/mesh.py:3156
Should probably be excluded from the tape, but doubt it does any harm at the moment.

[colinjcotter]

Ah ok. One strange thing though is that the extra block disappears when you replace subfunctions with split (). On 1 Jun 2023, at 17:54, Stephan Kramer ***@***.***> wrote:  I suspect this is completely harmless. It's just recording the interpolate it does inside the creation of the VertexOnlyMesh: firedrake/mesh.py:3156<https://github.com/firedrakeproject/firedrake/blob/b612fdf04a5155f0d5c3c3c703cac9fe2d5ddd82/firedrake/mesh.py#L3156> Should probably be excluded from the tape, but doubt it does any harm at the moment. — Reply to this email directly, view it on GitHub<#2954 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABOSV4T7XC6WL6WX2KF7JY3XJDCKLANCNFSM6AAAAAAYXFWWHQ>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[dham]

I am unable to reproduce this difference between subfunctions and split(). If I replace:

v0, v1 = w1.subfunctions

with

v0, v1 = w1.split()

Then I see the deprecation warning, but I produce exactly the same tape visualisation.