Cortex-M (CMSIS-NN) Backend

Architecture

Not a delegate backend — no partitioner. Custom ops and graph passes replace ATen quantized ops with CMSIS-NN equivalents at the graph level.

Pipeline

Uses standard PT2E quantization (prepare_pt2e / convert_pt2e), then CortexMPassManager rewrites quantized ops to cortex_m:: equivalents.

python
1from executorch.backends.cortex_m.quantizer.quantizer import CortexMQuantizer
2from executorch.backends.cortex_m.passes.cortex_m_pass_manager import CortexMPassManager
3from torch.export import export
4from torchao.quantization.pt2e.quantize_pt2e import convert_pt2e, prepare_pt2e
5from executorch.exir import to_edge_transform_and_lower, EdgeCompileConfig
6
7quantizer = CortexMQuantizer()
8captured = export(model, example_inputs).module()
9prepared = prepare_pt2e(captured, quantizer)
10prepared(*example_inputs)  # calibration
11quantized = convert_pt2e(prepared)
12
13exported = export(quantized, example_inputs)
14edge = to_edge_transform_and_lower(
15    exported,
16    compile_config=EdgeCompileConfig(_check_ir_validity=False),
17)
18edge._edge_programs["forward"] = CortexMPassManager(
19    edge.exported_program(), CortexMPassManager.pass_list
20).transform()
21et_program = edge.to_executorch()

In tests, CortexMTester wraps this pipeline:

python
1from executorch.backends.cortex_m.test.tester import CortexMTester
2
3tester = CortexMTester(model, example_inputs)
4tester.quantize().export().to_edge().run_passes().to_executorch()

Key Files

C++ kernels calling CMSIS-NN APIs live under backends/cortex_m/ops/.

Testing

Toolchain setup (required for test_implementation tests):

bash
1./examples/arm/setup.sh --i-agree-to-the-contained-eula
2source ./examples/arm/arm-scratch/setup_path.sh

Run all tests:

bash
1source ./examples/arm/arm-scratch/setup_path.sh
2pytest backends/cortex_m/test/

test_dialect_* tests verify graph correctness (pure Python, no toolchain needed). test_implementation_* tests verify numerical accuracy on the Corstone-300 FVP (requires toolchain on PATH).

Baremetal build:

bash
1backends/cortex_m/test/build_test_runner.sh

Adding a New Op

1. Define the op schema, meta function, and reference implementation in operators.py
2. Write the C++ kernel in backends/cortex_m/ops/ calling CMSIS-NN APIs
3. Register the .out kernel in operators.yaml
4. Add a pass to rewrite the ATen op → cortex_m:: op
5. Test with CortexMTester.test_dialect() (graph correctness) and test_implementation() (numerical accuracy on FVP)