[cortex-m] initial commit

backends/cortex_m/README.md

@@ -0,0 +1,3 @@
+# Cortex-M Backend
+
+WIP. This is a temporary/placeholder backend for Cortex-M CPUs. It is not intended to be used in production, but rather as a proof of concept. Things will change without notice.

backends/cortex_m/ops/TARGETS

@@ -0,0 +1,21 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+load("@fbcode_macros//build_defs:python_library.bzl", "python_library")
+load("@fbcode_macros//build_defs:export_files.bzl", "export_file")
+load("@fbsource//xplat/executorch/codegen:codegen.bzl", "executorch_generated_lib")
+
+oncall("executorch")
+
+python_library(
+    name = "ops",
+    srcs = [
+        "operators.py",
+    ],
+    deps = [
+        "fbcode//caffe2:torch",
+    ]
+) 

backends/cortex_m/ops/operators.py

@@ -0,0 +1,98 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+import torch
+from executorch.exir.dialects._ops import (
+    ops as exir_ops,
+)  # To provide the implementation of the operators
+from torch.library import impl, Library, register_fake
+
+# New operator library with a custom namespace to allow fusion etc.
+lib = Library("cortex_m", "DEF")
+
+###
+# dequantize_per_tensor
+###
+
+lib.define(
+    "quantize_per_tensor(Tensor input, float scale, int zero_point, int quant_min, int quant_max, ScalarType dtype) -> (Tensor Z)"
+)
+
+lib.define(
+    "quantize_per_tensor.out(Tensor input, float scale, int zero_point, int quant_min, int quant_max, ScalarType dtype, *, Tensor(a!) out) -> Tensor(a!)"
+)
+
+
+@register_fake("cortex_m::quantize_per_tensor")
+def quantize_per_tensor_meta(
+    input: torch.Tensor,
+    scale: float,
+    zero_point: int,
+    quant_min: int,
+    quant_max: int,
+    dtype: torch.dtype,
+) -> torch.Tensor:
+    return torch.empty_like(input, dtype=dtype)
+
+
+@impl(lib, "quantize_per_tensor", "CompositeExplicitAutograd")
+def quantize_per_tensor_impl(
+    input: torch.Tensor,
+    scale: float,
+    zero_point: int,
+    quant_min: int,
+    quant_max: int,
+    dtype: torch.dtype,
+) -> torch.Tensor:
+    """
+    The implementation of the quantize_per_tensor operator is the same as the
+    quantize_per_tensor operator in the edge dialect.
+    """
+    return exir_ops.edge.quantized_decomposed.quantize_per_tensor.default(
+        input, scale, zero_point, quant_min, quant_max, dtype
+    )
+
+
+###
+# dequantize_per_tensor
+###
+
+lib.define(
+    "dequantize_per_tensor(Tensor input, float scale, int zero_point, int quant_min, int quant_max, ScalarType dtype) -> (Tensor Z)"
+)
+lib.define(
+    "dequantize_per_tensor.out(Tensor input, float scale, int zero_point, int quant_min, int quant_max, ScalarType dtype, *, Tensor(a!) out) -> Tensor(a!)"
+)
+
+
+@register_fake("cortex_m::dequantize_per_tensor")
+def dequantize_per_tensor_meta(
+    input: torch.Tensor,
+    scale: float,
+    zero_point: int,
+    quant_min: int,
+    quant_max: int,
+    dtype: torch.dtype,
+) -> torch.Tensor:
+    return torch.empty_like(input, dtype=torch.float)
+
+
+@impl(lib, "dequantize_per_tensor", "CompositeExplicitAutograd")
+def dequantize_per_tensor_impl(
+    input: torch.Tensor,
+    scale: float,
+    zero_point: int,
+    quant_min: int,
+    quant_max: int,
+    dtype: torch.dtype,
+) -> torch.Tensor:
+    """
+    The implementation of the dequantize_per_tensor operator is the same as the
+    dequantize_per_tensor operator in the edge dialect.
+    """
+    return exir_ops.edge.quantized_decomposed.dequantize_per_tensor.default(
+        input, scale, zero_point, quant_min, quant_max, dtype
+    )

backends/cortex_m/passes/TARGETS

@@ -0,0 +1,21 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+load("@fbcode_macros//build_defs:python_library.bzl", "python_library")
+
+oncall("executorch")
+
+python_library(
+    name = "replace_quant_nodes_pass",
+    srcs = ["replace_quant_nodes_pass.py"],
+    deps = [
+        "//caffe2:torch",
+        "//executorch/exir:lib",
+        "//executorch/exir:pass_base",
+        "//executorch/exir/dialects:lib",
+        "//executorch/backends/cortex_m/ops:ops",
+    ]
+)

backends/cortex_m/passes/replace_quant_nodes_pass.py

@@ -0,0 +1,62 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+from typing import Callable, Dict, Tuple
+
+import executorch.backends.cortex_m.ops.operators  # noqa
+import torch
+
+from executorch.exir.dialects._ops import ops as exir_ops
+from executorch.exir.dialects.edge._ops import EdgeOpOverload
+from executorch.exir.pass_base import ExportPass, NodeMetadata, ProxyValue
+
+
+class ReplaceQuantNodesPass(ExportPass):
+    """
+    Replace quantize and dequantize nodes with the corresponding
+    cortex_m.quantize_per_tensor and cortex_m.dequantize_per_tensor nodes.
+    """
+
+    @staticmethod
+    def _is_qualified_int8_node(args) -> bool:
+        return (
+            args[3] >= torch.iinfo(torch.int8).min  # qmin
+            and args[4] <= torch.iinfo(torch.int8).max  # qmax
+            and args[5] == torch.int8  # dtype
+        )
+
+    def __init__(self):
+        super().__init__()
+        self.op_replacements = {
+            exir_ops.edge.quantized_decomposed.quantize_per_tensor.default: {
+                "new_target": exir_ops.edge.cortex_m.quantize_per_tensor.default,
+                "qualifier": self._is_qualified_int8_node,
+            },
+            exir_ops.edge.quantized_decomposed.dequantize_per_tensor.default: {
+                "new_target": exir_ops.edge.cortex_m.dequantize_per_tensor.default,
+                "qualifier": self._is_qualified_int8_node,
+            },
+        }
+
+    def call_operator(
+        self,
+        op: Callable[..., object],
+        args: Tuple[object, ...],
+        kwargs: Dict[str, object],
+        meta: NodeMetadata,
+    ) -> ProxyValue:
+        assert isinstance(
+            op, EdgeOpOverload
+        ), "Op must be an EdgeOpOverload. Run this pass after to_edge()."
+
+        if op in self.op_replacements and self.op_replacements[op]["qualifier"](args):
+            return super().call_operator(
+                self.op_replacements[op]["new_target"],
+                args,
+                kwargs,
+                meta,
+            )
+        return super().call_operator(op, args, kwargs, meta)

backends/cortex_m/test/TARGETS

@@ -0,0 +1,18 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+load("@fbcode_macros//build_defs:python_unittest.bzl", "python_unittest")
+
+python_unittest(
+    name = "test_replace_quant_nodes",
+    srcs = ["test_replace_quant_nodes.py"],
+    deps = [
+        "//pytorch/ao:torchao",  # @manual
+        "//caffe2:torch", 
+        "//executorch/backends/cortex_m/passes:replace_quant_nodes_pass",
+        "//executorch/backends/cortex_m/ops:ops",
+    ],
+)