Arm(R) Ethos(TM)-U NPU Pooling operators support (#9384)

This commit adds support for the max and average pooling primitive operators for the Arm(R) Ethos(TM)-U NPU and includes a few minor rewording changes.

python/tvm/relay/backend/contrib/ethosu/legalize.py

@@ -16,7 +16,8 @@
 # under the License.
 # pylint: disable=invalid-name, unused-argument, import-outside-toplevel, no-value-for-parameter
 """A set of passes to legalize some of operations for the NPU"""
-from typing import List
+from typing import List, Type
+
 import numpy as np  # type: ignore
 
 import tvm  # type: ignore
@@ -26,6 +27,7 @@
 from tvm.relay.dataflow_pattern import wildcard
 from tvm.relay.dataflow_pattern import is_op
 from tvm.relay.dataflow_pattern import rewrite
+from tvm.relay.dataflow_pattern import CallPattern
 from tvm.relay.backend.contrib.ethosu import op as ethosu_ops  # type: ignore
 from tvm.relay.backend.contrib.ethosu.errors import UnsupportedLayout  # type: ignore
 from tvm.relay.backend.contrib.ethosu import vela_api
@@ -121,7 +123,7 @@ def __call__(self, *args, **kwargs):
         pass
 
 
-class EthosUConv2DRewriter(DFPatternCallback):
+class Conv2DRewriter(DFPatternCallback):
     """Convert conv2d related composite functions into ethosu_conv2d operators"""
 
     def __init__(self):
@@ -193,22 +195,22 @@ def callback(
 
 
 @ir.transform.module_pass(opt_level=1)
-class LegalizeEthosUConv2D:
-    """This is the pass that wraps the EthosUConv2DRewriter"""
+class LegalizeConv2D:
+    """This is the pass that wraps the Conv2DRewriter"""
 
     def transform_module(
         self, mod: tvm.ir.IRModule, ctx: tvm.ir.transform.PassContext
     ) -> tvm.ir.IRModule:
         for global_var, func in mod.functions.items():
-            func = rewrite(EthosUConv2DRewriter(), func)
+            func = rewrite(Conv2DRewriter(), func)
             mod.update_func(global_var, func)
         return mod
 
     def __call__(self, *args, **kwargs):
         pass
 
 
-class EthosuDepthwiseConv2DRewriter(DFPatternCallback):
+class DepthwiseConv2DRewriter(DFPatternCallback):
     """Convert ethosu.qnn_depthwise_conv2d composite functions to ethosu_depthwise_conv2d
     operators"""
 
@@ -286,14 +288,124 @@ def callback(
 
 
 @ir.transform.module_pass(opt_level=1)
-class LegalizeEthosUDepthwiseConv2D:
-    """This is the pass that wraps the EthosUDepthwiseConv2DRewriter"""
+class LegalizeDepthwiseConv2D:
+    """This is the pass that wraps the DepthwiseConv2DRewriter"""
+
+    def transform_module(
+        self, mod: tvm.ir.IRModule, ctx: tvm.ir.transform.PassContext
+    ) -> tvm.ir.IRModule:
+        for global_var, func in mod.functions.items():
+            func = rewrite(DepthwiseConv2DRewriter(), func)
+            mod.update_func(global_var, func)
+        return mod
+
+    def __call__(self, *args, **kwargs):
+        pass
+
+
+class PoolingRewriter(DFPatternCallback):
+    """Convert ethosu.avgpool2d and ethosu.maxpool2d composite functions to
+    ethosu_pooling operators"""
+
+    def __init__(
+        self,
+        params_class: Type,
+        pattern: CallPattern,
+    ):
+        super().__init__(require_type=True)
+        self.params_class = params_class
+        self.pattern = pattern
+
+    def callback(
+        self, pre: tvm.relay.Expr, post: tvm.relay.Expr, node_map: tvm.ir.container.Map
+    ) -> tvm.relay.Expr:
+        params = self.params_class(post.op.body)
+        params.ifm.tensor = post.args[0]
+        channels_map = {
+            "NHWC": 3,
+        }
+        if str(params.ofm.layout) not in channels_map.keys():
+            raise UnsupportedLayout(str(params.ofm.layout))
+
+        activation_map = {"clip": "CLIP"}
+        if params.activation:
+            activation = activation_map[params.activation.op.name]
+            clip_min = int(params.activation.attrs.a_min)
+            clip_max = int(params.activation.attrs.a_max)
+        else:
+            activation = "NONE"
+            clip_min = 0
+            clip_max = 0
+
+        # Activations requiring LUT is currently not supported, so setting it to an empty list
+        lut = relay.const([], dtype="int8")
+
+        return ethosu_ops.ethosu_pooling(
+            ifm=post.args[0],
+            lut=lut,
+            pooling_type=params.pooling_type,
+            ifm_scale=params.ifm.q_params.scale_f32,
+            ifm_zero_point=params.ifm.q_params.zero_point,
+            ofm_scale=params.ofm.q_params.scale_f32,
+            ofm_zero_point=params.ofm.q_params.zero_point,
+            pool_shape=params.pool_shape,
+            ofm_channels=params.ofm.shape[channels_map[str(params.ofm.layout)]],
+            strides=params.strides,
+            padding=params.padding,
+            activation=activation,
+            clip_min=clip_min,
+            clip_max=clip_max,
+            upscale="NONE",
+            ifm_layout=str(params.ifm.layout),
+            ofm_layout=str(params.ofm.layout),
+        )
+
+
+class MaxPoolingRewriter(PoolingRewriter):
+    def __init__(self):
+        super().__init__(
+            params_class=ethosu_patterns.MaxPool2DParams,
+            pattern=(
+                wildcard().has_attr({"Composite": ethosu_patterns.MaxPool2DParams.composite_name})
+            )(wildcard()),
+        )
+
+
+@ir.transform.module_pass(opt_level=1)
+class LegalizeMaxPooling:
+    """This is the pass that wraps the MaxPoolingRewriter"""
+
+    def transform_module(
+        self, mod: tvm.ir.IRModule, ctx: tvm.ir.transform.PassContext
+    ) -> tvm.ir.IRModule:
+        for global_var, func in mod.functions.items():
+            func = rewrite(MaxPoolingRewriter(), func)
+            mod.update_func(global_var, func)
+        return mod
+
+    def __call__(self, *args, **kwargs):
+        pass
+
+
+class AvgPoolingRewriter(PoolingRewriter):
+    def __init__(self):
+        super().__init__(
+            params_class=ethosu_patterns.AvgPool2DParams,
+            pattern=(
+                wildcard().has_attr({"Composite": ethosu_patterns.AvgPool2DParams.composite_name})
+            )(wildcard()),
+        )
+
+
+@ir.transform.module_pass(opt_level=1)
+class LegalizeAvgPooling:
+    """This is the pass that wraps the AvgPoolingRewriter"""
 
     def transform_module(
         self, mod: tvm.ir.IRModule, ctx: tvm.ir.transform.PassContext
     ) -> tvm.ir.IRModule:
         for global_var, func in mod.functions.items():
-            func = rewrite(EthosuDepthwiseConv2DRewriter(), func)
+            func = rewrite(AvgPoolingRewriter(), func)
             mod.update_func(global_var, func)
         return mod
 
@@ -312,8 +424,10 @@ def transform_module(
         self, mod: tvm.ir.IRModule, ctx: tvm.ir.transform.PassContext
     ) -> tvm.ir.IRModule:
         mod = LegalizeSplit()(mod)
-        mod = LegalizeEthosUConv2D()(mod)
-        mod = LegalizeEthosUDepthwiseConv2D()(mod)
+        mod = LegalizeConv2D()(mod)
+        mod = LegalizeDepthwiseConv2D()(mod)
+        mod = LegalizeMaxPooling()(mod)
+        mod = LegalizeAvgPooling()(mod)
         return mod
 
     def __call__(self, *args, **kwargs):

python/tvm/relay/backend/contrib/ethosu/op/__init__.py

@@ -18,3 +18,4 @@
 
 from .convolution import ethosu_conv2d
 from .depthwise import ethosu_depthwise_conv2d
+from .pooling import ethosu_pooling

python/tvm/relay/backend/contrib/ethosu/op/convolution.py

@@ -113,7 +113,7 @@ def ethosu_conv2d(
     ofm_layout: str = "NHWC",
 ) -> tvm.relay.Call:
     """This is a quantized 2D convolution operation as supported by the
-    the NPU. It accepts either NHWC or NHCWB16 format
+    Ethos(TM)-U NPU. It accepts either NHWC or NHCWB16 format
     for the input data and OHWI format for the kernel weights.
 
     Reference: https://developer.arm.com/documentation/102420/0200/
@@ -132,7 +132,7 @@ def ethosu_conv2d(
     scale_bias : tvm.relay.Expr
         The packed per-channel weight scale and bias tensor.
     lut : tvm.relay.Expr
-        The look-up table values to use if activation = "LUT".
+        The look-up table of values to use if activation = "LUT".
     ifm_scale : float
         The quantization scale for the Input Feature Map tensor.
     ifm_zero_point : int
@@ -146,7 +146,7 @@ def ethosu_conv2d(
     kernel_shape : tuple of int
         The 2 dimensional kernel shape as (kernel_height, kernel_width).
     ofm_channels : int
-        The number of OFM channels.
+        The number of the Output Feature Map channels.
     strides : tuple of int, optional
         The 2 dimensional strides as (stride_height, stride_width).
     padding : tuple of int, optional

python/tvm/relay/backend/contrib/ethosu/op/depthwise.py

@@ -112,8 +112,8 @@ def ethosu_depthwise_conv2d(
     ifm_layout: str = "NHWC",
     ofm_layout: str = "NHWC",
 ) -> tvm.relay.Call:
-    """This is a quantized 2D depthwise convolution operation as supported
-    by the NPU. It accepts either NHWC or NHCWB16 format
+    """This is a quantized 2D depthwise convolution operation as supported by the
+    Ethos(TM)-U NPU. It accepts either NHWC or NHCWB16 format
     for the input data and OHWI format for the kernel weights.
 
     Reference: https://developer.arm.com/documentation/102420/0200/
@@ -132,7 +132,7 @@ def ethosu_depthwise_conv2d(
     scale_bias : tvm.relay.Expr
         The packed per-channel weight scale and bias tensor.
     lut : tvm.relay.Expr
-        The look-up table values to use if activation = "LUT"
+        The look-up table of values to use if activation = "LUT"
     ifm_scale : float
         The quantization scale for the Input Feature Map tensor.
     ifm_zero_point : int
@@ -146,7 +146,7 @@ def ethosu_depthwise_conv2d(
     kernel_shape : tuple of int
         The 2 dimensional kernel shape as (kernel_height, kernel_width).
     ofm_channels : int
-        The number of OFM channels.
+        The number of the Output Feature Map channels.
     strides : tuple of int, optional
         The 2 dimensional strides as (stride_height, stride_width).
     padding : tuple of int, optional