Arm backend: Add MAX_POOL2D tosa dialect op

[AdrianLundell]

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18970

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure, 6 Cancelled Jobs, 7 Pending

As of commit 81cb360 with merge base 9d72936 ():

NEW FAILURE - The following job has failed:

• pull / android / run-emulator (gh)
  The process '/usr/bin/sh' failed with exit code 1

CANCELLED JOBS - The following jobs were cancelled. Please retry:

• periodic / test-models-linux (buck2, mv3, portable, linux.2xlarge, 90) / linux-job (gh)
  ##[error]The operation was canceled.
• periodic / test-models-linux (buck2, mv3, xnnpack-quantization-delegation, linux.2xlarge, 90) / linux-job (gh)
  ##[error]The operation was canceled.
• periodic / test-models-linux (cmake, mv3, portable, linux.2xlarge, 90) / linux-job (gh)
  ##[error]The operation was canceled.
• periodic / test-models-linux (cmake, mv3, xnnpack-quantization-delegation, linux.2xlarge, 90) / linux-job (gh)
  ##[error]The operation was canceled.
• periodic / test-models-linux (cmake, vit, portable, linux.2xlarge, 90) / linux-job (gh)
  ##[error]The operation was canceled.
• periodic / test-models-linux (cmake, vit, xnnpack-quantization-delegation, linux.2xlarge, 90) / linux-job (gh)
  ##[error]The operation was canceled.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[Copilot]

Pull request overview

Adds support for lowering 2D max pooling to a TOSA dialect MAX_POOL2D op in the Arm backend, including a rewrite pass and corresponding operator visitor updates.

Changes:

• Introduce a new TOSA dialect fake op MAX_POOL2D with shape/dtype validation and output meta computation.
• Add RewriteMaxPool2dPass and integrate it into the Arm pass pipelines (including SizeAdjustInputPass dependencies).
• Update the Arm operator visitor to lower tosa.MAX_POOL2D.default, and add a new unit test covering the rewrite.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
backends/arm/tosa/dialect/ops/max_pool2d.py	Adds fake TOSA op schema + output meta for MAX_POOL2D.
backends/arm/tosa/dialect/__init__.py	Ensures the new dialect op module is imported/registered.
backends/arm/_passes/rewrite_max_pool2d_pass.py	New pass rewriting edge max_pool2d into backend TOSA MAX_POOL2D.
backends/arm/_passes/arm_pass_manager.py	Inserts RewriteMaxPool2dPass into the TOSA pipeline.
backends/arm/_passes/size_adjust_input_pass.py	Adds RewriteMaxPool2dPass to required-after passes.
backends/arm/_passes/__init__.py	Exposes the new rewrite pass.
backends/arm/operators/op_tosa_max_pool2d.py	Switches visitor to handle tosa.MAX_POOL2D.default and uses provided attrs directly.
backends/arm/operators/__init__.py	Registers the new op_tosa_max_pool2d visitor (removes old import).
backends/arm/test/passes/test_rewrite_max_pool2d_pass.py	Adds rewrite-pass test coverage.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

args[2] (stride) can be an int (e.g., max_pool2d(..., stride=2)), but this code calls len(args[2]), which will raise TypeError: object of type 'int' has no len(). Normalize stride via type checks (int/Sequence/None) before using len() and before passing it to _to_2tuple so the pass works for both scalar and tuple/list strides.

Suggested change

	if len(args) > 2 and args[2] is not None and len(args[2]) > 0:
	stride = _to_2tuple(args[2])
	else:
	stride = kernel
	raw_stride = args[2] if len(args) > 2 else None
	if raw_stride is None:
	stride = kernel
	else:
	stride = _to_2tuple(raw_stride)

[Copilot]

Typo in comment: "Cannnot" should be "Cannot".

Suggested change

	) # Cannnot run aten graph with tosa dialect ops
	) # Cannot run aten graph with tosa dialect ops

[oscarandersson8218]

Doesn't seem to need any Buck-changes. LGTM!

[AdrianLundell]

Fail unrelated