Build custom portable ops

[lucylq]

Summary

Build custom portable ops, allow for easy import in python. This is mostly for testing but it does allow us to run memory planning tests (and potentially others) in OSS.

• Build portable_custom_ops_aot_lib shared library (registers allclose.out and allclose.Tensor custom ops into PyTorch) in CMake, following the same pattern as quantized_ops_aot_lib
• Make test_memory_planning.py importable outside Buck by replacing the Buck-only load_library("//executorch/...") with import executorch.kernels.portable
• Add kernels/portable/init.py that discovers and loads the built shared library, mirroring kernels/quantized/init.py

Test plan

Confirm memory planning can be tested in python

python -m unittest exir.tests.test_memory_planning

Confirm buck test still works

buck test @fbcode//mode/dev fbcode//executorch/exir/tests:memory_planning 

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/17642

• 📄 Preview Python docs built from this PR

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

❌ 4 New Failures

As of commit 0d07e8e with merge base 19de115 ():

NEW FAILURES - The following jobs have failed:

• Copilot code review / Agent (gh)
  Process completed with exit code 1.
• Test Metal Backend / export-model-metal-artifact (mistralai, Voxtral-Mini-3B-2507, non-quantized) / macos-job (gh)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 1
• Test Metal Backend / export-model-metal-artifact (mistralai, Voxtral-Mini-4B-Realtime-2602, quantized-int4-metal) / macos-job (gh)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 1
• Test Metal Backend / export-model-metal-artifact (nvidia, parakeet-tdt, quantized-int4-metal) / macos-job (gh)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 1

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

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

[meta-codesync]

@lucylq has imported this pull request. If you are a Meta employee, you can view this in D94555555.

[Copilot]

Pull request overview

Adds build + Python import support for portable custom ops (AOT) so OSS Python tests (e.g., memory planning) can load the shared library without Buck-specific mechanisms.

Changes:

• Introduces EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT CMake option and Torch discovery at root scope.
• Builds portable_custom_ops_aot_lib in kernels/portable for registering custom ops into PyTorch.
• Updates test_memory_planning.py to import executorch.kernels.portable and adds a loader kernels/portable/__init__.py.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
tools/cmake/preset/pybind.cmake	Enables a new overridable build option in the pybind preset.
tools/cmake/preset/default.cmake	Defines the new EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT option.
src/executorch/kernels/portable	Adds a symlink-like entry pointing to kernels/portable.
kernels/quantized/CMakeLists.txt	Fixes/rewraps a formatting instruction comment.
kernels/portable/init.py	Adds runtime discovery + torch.ops.load_library for the custom ops AOT library.
kernels/portable/CMakeLists.txt	Builds portable_custom_ops_aot_lib and sets RPATH for runtime loading.
exir/tests/test_memory_planning.py	Replaces Buck-only load with importing executorch.kernels.portable.
CMakeLists.txt	Finds Torch at root scope when custom AOT is enabled.
.lintrunner.toml	Excludes kernels/portable and kernels/quantized from lintrunner.

Comments suppressed due to low confidence (1)

src/executorch/kernels/portable:1

• This appears to be a symlink-style entry. Git symlinks can degrade into plain text files on some Windows configurations/checkout modes, which can break imports/packaging. If this is part of the Python package layout, consider replacing it with a real package directory (e.g., src/executorch/kernels/portable/__init__.py that re-exports) or adjust packaging configuration to include kernels/portable without relying on symlinks.

---

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

[Copilot]

This adds the portable_custom_ops_aot_lib target but does not add any install(TARGETS ...) rule (or otherwise ensure the built shared library ends up inside the Python package/wheel). Since kernels/portable/__init__.py tries to discover and load the library at runtime, missing installation/packaging will prevent import executorch.kernels.portable from actually registering ops in an installed environment. Add an install rule consistent with the quantized AOT library pattern and align it with the loader’s expected search location.

Suggested change

	)
	)
	install(
	TARGETS portable_custom_ops_aot_lib
	EXPORT ExecuTorchTargets
	DESTINATION ${CMAKE_INSTALL_LIBDIR}/executorch/extensions/pybindings
	)

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

---

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

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

---

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

[Copilot]

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated no new comments.

---

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