Cortex-M backend: Refactor ConvertToCortexMPass#20070
Merged
Merged
Conversation
Change to use the general AtenToDialectPass structure. Removes the 1-1 ensure check as it was to restrictive, cortex-m backend needs to both insert scratch nodes and for BMM an additional transpose at the time of dialect replacement. Bonus small fix: Moves yolo import into test to avoid download at test collection time Signed-off-by: Adrian Lundell <adrian.lundell@arm.com> Change-Id: Ic62964522704d31f75c653bc265bcce02ce375c6
AdrianLundell
added
the
partner: arm
AdrianLundell
added
ciflow/trunk
release notes: none
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20070
Note: Links to docs will display an error until the docs builds have been completed. ❌ 3 New FailuresAs of commit d46edf4 with merge base e56be3e ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
Contributor
There was a problem hiding this comment.
Pull request overview
Refactors the Cortex-M backend’s ATen→dialect lowering to use the shared AtenToDialectPass substitution/decorator structure (replacing ConvertToCortexMPass), and relaxes the prior 1:1 “call_function count preserved” postcondition to allow scratch/aux node insertion (needed for CMSIS-NN scratch buffers and BMM transpose handling). Also adjusts the YOLO test to avoid triggering weights download during test collection.
Changes:
- Remove the
requires()/ensures()“call_function count preserved” check fromAtenToDialectPassand delete the corresponding test. - Introduce
AtenToCortexMPassand wire it into the Cortex-M pass manager in place ofConvertToCortexMPass. - Move YOLO model instantiation/weights loading into the test body to avoid collection-time side effects.
Reviewed changes
Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| backends/transforms/test/test_aten_to_dialect_pass.py | Removes the test that asserted the old 1:1 call_function count invariant. |
| backends/transforms/aten_to_dialect_pass.py | Removes the 1:1 substitution invariant enforcement (requires/ensures). |
| backends/cortex_m/test/models/test_yolo11.py | Defers YOLO weights/model creation to test runtime to avoid collection-time downloads. |
| backends/cortex_m/passes/cortex_m_pass_manager.py | Switches the pass pipeline to use AtenToCortexMPass. |
| backends/cortex_m/passes/aten_to_cortex_m_pass.py | New pass implementing Cortex-M dialect substitutions + scratch alloc sizing logic. |
| backends/cortex_m/passes/init.py | Updates public pass imports to expose AtenToCortexMPass. |
Comments suppressed due to low confidence (1)
backends/cortex_m/test/models/test_yolo11.py:28
McuTestCase.modelis typed astorch.nn.Module, but this test passesNonewith a type-ignore. This can hide real issues and could break if future test helpers start usingtest_case.model. Use a lightweight placeholder module instead so the dataclass contract stays valid.
"yolo11n": McuTestCase(
model=None, # type: ignore[arg-type]
example_inputs=lambda: (
torch.randn(1, 3, 640, 640).to(memory_format=torch.channels_last),
),
),
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
rascani
approved these changes
Jun 5, 2026
rascani
left a comment
rascani
left a comment
Contributor
There was a problem hiding this comment.
Can you also update backends/cortex_m/passes/BUCK?
Signed-off-by: Adrian Lundell <adrian.lundell@arm.com> Change-Id: Ida9d298a0bf57c18957fb9117244adf260aadc12
Collaborator
Author
Done, can you verify that it looks correct? |
Contributor
LGTM, thanks for doing that. |
rascani
added a commit
to rascani/executorch
that referenced
this pull request
Jun 9, 2026
Adapt _get_avg_pool2d_replacement from the old ConvertToCortexMPass class to a @register_dialect_substitution function on AtenToCortexMPass, matching the pattern established by pytorch#20070. Add casts to satisfy mypy's node.args union type. Add a structural assertion to test_dialect_avg_pool2d that verifies the scratch buffer size matches cmsis_nn.avgpool_buffer_size() for the given input/output dimensions. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Change to use the general AtenToDialectPass structure, quantized_op_fusion_pass
to be changed similarly in an upcoming PR.
Removes the 1-1 ensure check as it was to restrictive, cortex-m backend needs to both insert scratch nodes and for BMM an additional transpose at the time of dialect replacement.
Bonus small fix: Moves yolo import into test to avoid download at test collection time
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @rascani