Move optimization passes from opt_level=0 to opt_level=1 (#18206)

Differential Revision: D96766073

Pull Request resolved: #18206

Author: mcremon-meta

backends/cadence/aot/BUCK

@@ -300,6 +300,7 @@ fbcode_target(_kind = runtime.python_library,
     ],
     typing = True,
     deps = [
+        ":fuse_ops",
         ":ops_registrations",
         "//caffe2:torch",
         "//executorch/backends/cadence/aot:pass_utils",

backends/cadence/aot/decompose_ops.py

@@ -23,10 +23,12 @@
 from torch.fx.node import Argument
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
+@register_cadence_pass(CadencePassAttribute(opt_level=1))
 class DecomposeAtenApproxGeluPass(ExportPass):
     """
-    Decompose the aten gelu op with an approximate arg to a series of simpler ops
+    Decompose the aten gelu op with an approximate arg to a series of simpler ops.
+    This is an optimization - gelu has a portable kernel fallback, but decomposing
+    may be more efficient on some backends.
     """
 
     def call_operator(

backends/cadence/aot/functions.yaml

@@ -309,10 +309,15 @@
     - arg_meta: null
       kernel_name: impl::generic::quantized_relu_asym8u_asym8u_per_tensor_out
 
-- func: cadence::quantized_max_pool2d.out(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode, *, Tensor(a!) out) -> Tensor(a!)
+- func: cadence::quantized_max_pool2d_nchw.out(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode, *, Tensor(a!) out) -> Tensor(a!)
   kernels:
     - arg_meta: null
-      kernel_name: impl::generic::quantized_max_pool2d_out
+      kernel_name: impl::generic::quantized_max_pool2d_nchw_out
+
+- func: cadence::quantized_max_pool2d_nhwc.out(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::generic::quantized_max_pool2d_nhwc_out
 
 - func: cadence::quantized_add.per_tensor_out(Tensor X, float X_scale, int X_zero_point, Tensor Y, float Y_scale, int Y_zero_point, float out_scale, int out_zero_point, *, Tensor(a!) out) -> Tensor(a!)
   kernels:

backends/cadence/aot/fuse_ops.py

@@ -1170,7 +1170,10 @@ def can_fuse_for_chain(
             return False
 
         # checking that permut2(permut1(identity)) == identity, modulo unitary dimensions
-        input_shape = cast(torch.fx.Node, producer.args[0]).meta["val"].shape
+        producer_input = cast(torch.fx.Node, producer.args[0])
+        if "val" not in producer_input.meta:
+            return False
+        input_shape = producer_input.meta["val"].shape
         ident_dims = list(range(len(input_shape)))
         # this mapping helps to handle both transpose and permutations
         f: dict[Any, Callable] = {

backends/cadence/aot/ops_registrations.py

@@ -214,10 +214,16 @@ def register_fake(
 )
 
 lib.define(
-    "quantized_max_pool2d(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode) -> Tensor"
+    "quantized_max_pool2d_nchw(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode) -> Tensor"
 )
 lib.define(
-    "quantized_max_pool2d.out(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode, *, Tensor(a!) out) -> Tensor(a!)"
+    "quantized_max_pool2d_nchw.out(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode, *, Tensor(a!) out) -> Tensor(a!)"
+)
+lib.define(
+    "quantized_max_pool2d_nhwc(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode) -> Tensor"
+)
+lib.define(
+    "quantized_max_pool2d_nhwc.out(Tensor input, int[] kernel_size, int[] stride, int[] padding, int[] dilation, bool ceil_mode, *, Tensor(a!) out) -> Tensor(a!)"
 )
 
 lib.define(
@@ -2277,8 +2283,8 @@ def quantized_relu_asym8u_asym8u_per_tensor_meta(
     return input.new_empty(input.size(), dtype=input.dtype)
 
 
-@register_fake("cadence::quantized_max_pool2d")
-def quantized_max_pool2d_meta(
+@register_fake("cadence::quantized_max_pool2d_nchw")
+def quantized_max_pool2d_nchw_meta(
     input: torch.Tensor,
     kernel_size: list[int],
     stride: list[int],
@@ -2318,6 +2324,47 @@ def quantized_max_pool2d_meta(
     return input.new_empty([batch, channels, height_out, width_out], dtype=input.dtype)
 
 
+@register_fake("cadence::quantized_max_pool2d_nhwc")
+def quantized_max_pool2d_nhwc_meta(
+    input: torch.Tensor,
+    kernel_size: list[int],
+    stride: list[int],
+    padding: list[int],
+    dilation: list[int],
+    ceil_mode: bool,
+) -> torch.Tensor:
+    assert (
+        len(kernel_size) == 2
+    ), f"kernel_size must have 2 elements, got {len(kernel_size)}"
+    assert len(stride) == 2, f"stride must have 2 elements, got {len(stride)}"
+    assert len(padding) == 2, f"padding must have 2 elements, got {len(padding)}"
+    assert len(dilation) == 2, f"dilation must have 2 elements, got {len(dilation)}"
+    assert (
+        len(input.size()) == 4
+    ), f"input must be 4D (N, H, W, C), got {len(input.size())}D"
+
+    batch = input.size(0)
+    height_in = input.size(1)
+    width_in = input.size(2)
+    channels = input.size(3)
+
+    height_out_raw = (
+        height_in + 2 * padding[0] - dilation[0] * (kernel_size[0] - 1) - 1
+    ) / stride[0] + 1
+    width_out_raw = (
+        width_in + 2 * padding[1] - dilation[1] * (kernel_size[1] - 1) - 1
+    ) / stride[1] + 1
+
+    if ceil_mode:
+        height_out = ceil(height_out_raw)
+        width_out = ceil(width_out_raw)
+    else:
+        height_out = int(height_out_raw)
+        width_out = int(width_out_raw)
+
+    return input.new_empty([batch, height_out, width_out, channels], dtype=input.dtype)
+
+
 @register_fake("cadence::fully_connected")
 def fully_connected_meta(
     src: torch.Tensor,

backends/cadence/aot/passes.py

@@ -25,6 +25,7 @@
 from executorch.backends.cadence.aot.remove_ops import (
     CadenceRemoveNops,
     RemoveNopSliceOrViewOpPass,
+    RemovePermutesAroundElementwiseOps,
     RemoveRedundantOps,
 )
 from executorch.backends.cadence.aot.reorder_ops import CadenceReorderOpsInGraph
@@ -89,6 +90,7 @@ def get_passes_in_default_order() -> list[Type[ExportPass]]:
         CadenceSimplifyOpsInGraph.passes,
         FinalizePipeline,
         FuseFullThenReshapePass,
+        RemovePermutesAroundElementwiseOps,
         FuseTransposeOrPermuteOpPairsPass,
         RemoveNopSliceOrViewOpPass,
         CompileTimeTypeDispatchPass,

backends/cadence/aot/quantizer/patterns.py

@@ -459,7 +459,7 @@ def get_anchors(
         )
 
     def replacement_op(self) -> OpOverload:
-        return torch.ops.cadence.quantized_max_pool2d.default
+        return torch.ops.cadence.quantized_max_pool2d_nchw.default
 
 
 class MaxPool2dWithoutIndicesPattern(QuantizationPattern):
@@ -498,7 +498,10 @@ def get_anchors(
         )
 
     def replacement_op(self) -> OpOverload:
-        return torch.ops.cadence.quantized_max_pool2d.default
+        return torch.ops.cadence.quantized_max_pool2d_nchw.default
+
+
+# This is a base class for ReLU
 
 
 # This is a base class for ReLU, since it can be used with two different aten ops

backends/cadence/aot/ref_implementations.py

@@ -1868,8 +1868,8 @@ def rms_norm(
     return W * nn.RMSNorm(list(normalized_shape), eps=eps, dtype=X.dtype)(X)
 
 
-@impl_tracked(m, "quantized_max_pool2d")
-def quantized_max_pool2d(
+@impl_tracked(m, "quantized_max_pool2d_nchw")
+def quantized_max_pool2d_nchw(
     input: torch.Tensor,
     kernel_size: list[int],
     stride: list[int],
@@ -1897,6 +1897,37 @@ def quantized_max_pool2d(
     )
 
 
+@impl_tracked(m, "quantized_max_pool2d_nhwc")
+def quantized_max_pool2d_nhwc(
+    input: torch.Tensor,
+    kernel_size: list[int],
+    stride: list[int],
+    padding: list[int],
+    dilation: list[int],
+    ceil_mode: bool,
+) -> torch.Tensor:
+    """
+    Quantized max pooling in NHWC layout.
+
+    Converts NHWC→NCHW, performs max pooling, then converts back NCHW→NHWC.
+    """
+    # Convert NHWC [N, H, W, C] to NCHW [N, C, H, W]
+    input_nchw = input.permute(0, 3, 1, 2).contiguous()
+
+    # Call the NCHW version
+    output_nchw = quantized_max_pool2d_nchw(
+        input_nchw,
+        kernel_size=kernel_size,
+        stride=stride,
+        padding=padding,
+        dilation=dilation,
+        ceil_mode=ceil_mode,
+    )
+
+    # Convert NCHW [N, C, H_out, W_out] back to NHWC [N, H_out, W_out, C]
+    return output_nchw.permute(0, 2, 3, 1).contiguous()
+
+
 @impl_tracked(m, "where_Scalar")
 def where_Scalar(
     condition: torch.Tensor,

backends/cadence/aot/remove_ops.py

@@ -14,14 +14,15 @@
 
 import torch
 import torch.fx
+
+from executorch.backends.cadence.aot.fuse_ops import FuseTransposeOrPermuteOpPairsPass
 from executorch.backends.cadence.aot.pass_utils import (
     CadencePassAttribute,
     get_arg,
     register_cadence_pass,
     RemoveOrReplacePassInterface,
     set_arg,
 )
-
 from executorch.backends.cadence.aot.simplify_ops import SimplifySliceOpPass
 from executorch.backends.cadence.aot.utils import get_edge_overload_packet
 from executorch.backends.transforms.remove_clone_ops import RemoveCloneOpsTransform
@@ -33,7 +34,7 @@
 from torch.fx.node import Node
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
+@register_cadence_pass(CadencePassAttribute(opt_level=1))
 class RemoveCloneOpsTransformImported(ExportPass):
     def call(self, graph_module: torch.fx.GraphModule) -> PassResult:
         finalize_passes: List[PassType] = [
@@ -44,7 +45,7 @@ def call(self, graph_module: torch.fx.GraphModule) -> PassResult:
         return result
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
+@register_cadence_pass(CadencePassAttribute(opt_level=1))
 class RemoveDetachCopyPass(RemoveOrReplacePassInterface):
     @property
     def targets(self) -> list[EdgeOpOverload]:
@@ -66,7 +67,7 @@ class RemoveRedundantOps:
     ]
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
+@register_cadence_pass(CadencePassAttribute(opt_level=1))
 class RemoveZeroSizedCatArgsPass(RemoveOrReplacePassInterface):
     @property
     def targets(self) -> list[EdgeOpOverload]:
@@ -120,11 +121,11 @@ def maybe_remove_or_replace(self, node: Node) -> bool:
         return False
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
+@register_cadence_pass(CadencePassAttribute(opt_level=1))
 class RemoveNopExpandOpPass(RemoveOrReplacePassInterface):
     """
     For an expand op, if the operator shape matches the expand shape, then the
-    expand is a nop.
+    expand is a nop. This is an optimization that removes unnecessary ops.
     """
 
     @property
@@ -143,9 +144,9 @@ def maybe_remove_or_replace(self, node: Node) -> bool:
         return False
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
+@register_cadence_pass(CadencePassAttribute(opt_level=1))
 class RemoveToOpsPass(RemoveOrReplacePassInterface):
-    # aten.to.* as of now are all nops
+    # aten.to.* ops are no-ops in inference - this is an optimization
     @property
     def targets(self) -> list[EdgeOpOverload]:
         return [
@@ -264,11 +265,11 @@ def maybe_remove_or_replace(self, node: Node) -> bool:
         return True
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
+@register_cadence_pass(CadencePassAttribute(opt_level=1))
 class RemoveAliasCopyOpPass(RemoveOrReplacePassInterface):
     """
-
     alias_copy is a no-op and can be removed.
+    This is an optimization that removes unnecessary ops.
     """
 
     @property
@@ -412,6 +413,9 @@ class Subgraph:
         exir_ops.edge.quantized_decomposed.dequantize_per_tensor.default,
         exir_ops.edge.cadence.quantize_per_tensor.default,
         exir_ops.edge.cadence.dequantize_per_tensor.default,
+        exir_ops.edge.cadence.quantized_relu.per_tensor,
+        exir_ops.edge.cadence.requantize.per_tensor,
+        exir_ops.edge.cadence.quantized_add.per_tensor,
         # Ops that require special handling.
         exir_ops.edge.aten.cat.default,
         exir_ops.edge.aten.mean.dim,
@@ -804,6 +808,7 @@ class CommonRemovePasses:
         RemoveToOpsPass,
         RemoveZeroSizedCatArgsPass,
         RemovePermutesAroundElementwiseOps,
+        FuseTransposeOrPermuteOpPairsPass,
         RemoveSqueezeViewBeforeElementwiseOps,
         RemoveCatFromSliceCopyPass,
         RemoveCloneOpsTransformImported,