WebNN: Define XNNPACK Node for conv2d MLOperator

This CL implements the DefineXnnNodeForConv2d() method for WebNN conv2d
MLOperator that defines XNNPACK conv2d or depthwise conv2d Node
according to MLConv2dOptions.

This CL only supports the input layout “nhwc”, filter layout “ohwi” for
regular conv2d and filter layout “ihow” for depthwise conv2d. Other
input and filter layouts could be supported by inserting transpose
operators later. There is also another proposal [1] that suggests
simplifying the layouts support. The implementation will be updated
according to the WG’s consensus.

For unit tests, this CL implements Conv2dTest of MLGraphXnnpackTest that
tests against conv2d, depthwise conv2d with and without fused bias
operand and relu activation.

[1]: webmachinelearning/webnn#324


Bug: 1273291
Change-Id: I8a70ee7bb053b386e12ff46e67d139683b044383
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/4031876
Reviewed-by: Jiewei Qian <[email protected]>
Commit-Queue: ningxin hu <[email protected]>
Cr-Commit-Position: refs/heads/main@{#1093328}

Author: huningxin

third_party/blink/renderer/modules/ml/webnn/ml_graph_builder.cc

@@ -7,7 +7,6 @@
 #include <algorithm>
 
 #include "base/numerics/checked_math.h"
-#include "third_party/abseil-cpp/absl/types/optional.h"
 #include "third_party/blink/renderer/bindings/core/v8/script_promise_resolver.h"
 #include "third_party/blink/renderer/bindings/modules/v8/v8_ml_clamp_options.h"
 #include "third_party/blink/renderer/bindings/modules/v8/v8_ml_conv_2d_options.h"
@@ -133,53 +132,6 @@ MLOperand* BuildElementWiseBinary(MLGraphBuilder* builder,
   return output;
 }
 
-struct PaddingSizes {
-  uint32_t begin;
-  uint32_t end;
-};
-
-// Calculate the padding given auto pad, input size, filter size, stride and
-// dilation. Return the calculated padding sizes if no error.
-absl::optional<PaddingSizes> CalculatePaddingForAutoPad(
-    V8MLAutoPad::Enum auto_pad,
-    const uint32_t input_size,
-    const uint32_t filter_size,
-    const uint32_t stride,
-    const uint32_t dilation) {
-  auto checked_output_size =
-      (base::MakeCheckedNum<uint32_t>(input_size) + stride - 1) / stride;
-  auto checked_dilated_filter_size =
-      (base::MakeCheckedNum<uint32_t>(filter_size) - 1) * dilation + 1;
-  auto checked_needed_input_size =
-      (checked_output_size - 1) * stride + checked_dilated_filter_size;
-  if (!checked_needed_input_size.IsValid()) {
-    return absl::nullopt;
-  }
-  auto checked_total_padding =
-      checked_needed_input_size.ValueOrDie() > input_size
-          ? checked_needed_input_size - input_size
-          : base::MakeCheckedNum<uint32_t>(0);
-  base::CheckedNumeric<uint32_t> checked_padding_begin, checked_padding_end;
-  switch (auto_pad) {
-    case V8MLAutoPad::Enum::kSameUpper:
-      checked_padding_begin = checked_total_padding / 2;
-      checked_padding_end = (checked_total_padding + 1) / 2;
-      break;
-    case V8MLAutoPad::Enum::kSameLower:
-      checked_padding_begin = (checked_total_padding + 1) / 2;
-      checked_padding_end = checked_total_padding / 2;
-      break;
-    default:
-      NOTREACHED();
-  }
-  uint32_t padding_begin, padding_end;
-  if (!checked_padding_begin.AssignIfValid(&padding_begin) ||
-      !checked_padding_end.AssignIfValid(&padding_end)) {
-    return absl::nullopt;
-  }
-  return PaddingSizes({.begin = padding_begin, .end = padding_end});
-}
-
 // Calculate the output size for conv2d based on WebNN spec:
 // https://www.w3.org/TR/webnn/#api-mlgraphbuilder-conv2d
 // Return the calculated output size if no error.
@@ -315,7 +267,7 @@ absl::optional<FloatSize2D> ValidateAndCalculateConv2dOutputSizes(
   // options.padding array are ignored and the explicit padding values need to
   // be calculated.
   if (auto_pad != V8MLAutoPad::Enum::kExplicit) {
-    auto padding_sizes_height = CalculatePaddingForAutoPad(
+    auto padding_sizes_height = MLGraphBuilder::CalculatePaddingForAutoPad(
         auto_pad.AsEnum(), input_height, filter_height, stride_height,
         dilation_height);
     if (!padding_sizes_height) {
@@ -327,9 +279,9 @@ absl::optional<FloatSize2D> ValidateAndCalculateConv2dOutputSizes(
     }
     padding_beginning_height = padding_sizes_height.value().begin;
     padding_ending_height = padding_sizes_height.value().end;
-    auto padding_sizes_width =
-        CalculatePaddingForAutoPad(auto_pad.AsEnum(), input_width, filter_width,
-                                   stride_width, dilation_width);
+    auto padding_sizes_width = MLGraphBuilder::CalculatePaddingForAutoPad(
+        auto_pad.AsEnum(), input_width, filter_width, stride_width,
+        dilation_width);
     if (!padding_sizes_width) {
       exception_state.ThrowDOMException(
           DOMExceptionCode::kDataError,
@@ -561,6 +513,47 @@ MLContext* MLGraphBuilder::GetContext() const {
   return ml_context_;
 }
 
+// static
+absl::optional<MLGraphBuilder::PaddingSizes>
+MLGraphBuilder::CalculatePaddingForAutoPad(V8MLAutoPad::Enum auto_pad,
+                                           const uint32_t input_size,
+                                           const uint32_t filter_size,
+                                           const uint32_t stride,
+                                           const uint32_t dilation) {
+  auto checked_output_size =
+      (base::MakeCheckedNum<uint32_t>(input_size) + stride - 1) / stride;
+  auto checked_dilated_filter_size =
+      (base::MakeCheckedNum<uint32_t>(filter_size) - 1) * dilation + 1;
+  auto checked_needed_input_size =
+      (checked_output_size - 1) * stride + checked_dilated_filter_size;
+  if (!checked_needed_input_size.IsValid()) {
+    return absl::nullopt;
+  }
+  auto checked_total_padding =
+      checked_needed_input_size.ValueOrDie() > input_size
+          ? checked_needed_input_size - input_size
+          : base::MakeCheckedNum<uint32_t>(0);
+  base::CheckedNumeric<uint32_t> checked_padding_begin, checked_padding_end;
+  switch (auto_pad) {
+    case V8MLAutoPad::Enum::kSameUpper:
+      checked_padding_begin = checked_total_padding / 2;
+      checked_padding_end = (checked_total_padding + 1) / 2;
+      break;
+    case V8MLAutoPad::Enum::kSameLower:
+      checked_padding_begin = (checked_total_padding + 1) / 2;
+      checked_padding_end = checked_total_padding / 2;
+      break;
+    default:
+      NOTREACHED();
+  }
+  uint32_t padding_begin, padding_end;
+  if (!checked_padding_begin.AssignIfValid(&padding_begin) ||
+      !checked_padding_end.AssignIfValid(&padding_end)) {
+    return absl::nullopt;
+  }
+  return PaddingSizes({.begin = padding_begin, .end = padding_end});
+}
+
 MLOperand* MLGraphBuilder::input(String name,
                                  const MLOperandDescriptor* desc,
                                  ExceptionState& exception_state) {

third_party/blink/renderer/modules/ml/webnn/ml_graph_builder.h

@@ -5,7 +5,9 @@
 #ifndef THIRD_PARTY_BLINK_RENDERER_MODULES_ML_WEBNN_ML_GRAPH_BUILDER_H_
 #define THIRD_PARTY_BLINK_RENDERER_MODULES_ML_WEBNN_ML_GRAPH_BUILDER_H_
 
+#include "third_party/abseil-cpp/absl/types/optional.h"
 #include "third_party/blink/renderer/bindings/core/v8/script_promise.h"
+#include "third_party/blink/renderer/bindings/modules/v8/v8_ml_auto_pad.h"
 #include "third_party/blink/renderer/core/typed_arrays/array_buffer_view_helpers.h"
 #include "third_party/blink/renderer/core/typed_arrays/dom_array_buffer_view.h"
 #include "third_party/blink/renderer/modules/ml/webnn/ml_operator.h"
@@ -48,6 +50,22 @@ class MODULES_EXPORT MLGraphBuilder final : public ScriptWrappable {
 
   MLContext* GetContext() const;
 
+  struct PaddingSizes {
+    uint32_t begin;
+    uint32_t end;
+  };
+
+  // Calculate the effective padding based on WebNN auto padding rules.
+  //
+  // TODO(crbug.com/1273291): Add the link to WebNN spec's algorithm once it is
+  // defined, tracked by: https://github.com/webmachinelearning/webnn/issues/326
+  static absl::optional<PaddingSizes> CalculatePaddingForAutoPad(
+      V8MLAutoPad::Enum auto_pad,
+      const uint32_t input_size,
+      const uint32_t filter_size,
+      const uint32_t stride,
+      const uint32_t dilation);
+
   // ml_graph_builder.idl
   MLOperand* input(String name,
                    const MLOperandDescriptor* desc,

third_party/blink/renderer/modules/ml/webnn/ml_graph_builder_test.cc

@@ -105,17 +105,20 @@ NotShared<DOMArrayBufferView> CreateDOMArrayBufferView(
   return buffer_view;
 }
 
-MLOperand* BuildConstant(V8TestingScope& scope,
-                         MLGraphBuilder* builder,
-                         const Vector<uint32_t>& dimensions,
-                         V8MLOperandType::Enum type) {
+MLOperand* BuildConstant(
+    V8TestingScope& scope,
+    MLGraphBuilder* builder,
+    const Vector<uint32_t>& dimensions,
+    V8MLOperandType::Enum type,
+    absl::optional<NotShared<DOMArrayBufferView>> user_buffer_view) {
   auto* desc = MLOperandDescriptor::Create();
   desc->setDimensions(dimensions);
   desc->setType(type);
   size_t size = std::accumulate(dimensions.begin(), dimensions.end(), size_t(1),
                                 std::multiplies<uint32_t>());
   NotShared<DOMArrayBufferView> buffer_view =
-      CreateDOMArrayBufferView(size, type);
+      user_buffer_view ? std::move(user_buffer_view.value())
+                       : CreateDOMArrayBufferView(size, type);
   auto* constant =
       builder->constant(desc, buffer_view, scope.GetExceptionState());
   EXPECT_NE(constant, nullptr);

third_party/blink/renderer/modules/ml/webnn/ml_graph_builder_test.h

@@ -6,6 +6,7 @@
 #define THIRD_PARTY_BLINK_RENDERER_MODULES_ML_WEBNN_ML_GRAPH_BUILDER_TEST_H_
 
 #include "testing/gtest/include/gtest/gtest.h"
+#include "third_party/abseil-cpp/absl/types/optional.h"
 #include "third_party/blink/renderer/bindings/modules/v8/v8_ml_clamp_options.h"
 #include "third_party/blink/renderer/bindings/modules/v8/v8_ml_context_options.h"
 #include "third_party/blink/renderer/bindings/modules/v8/v8_ml_conv_2d_options.h"
@@ -40,10 +41,12 @@ NotShared<DOMArrayBufferView> CreateDOMArrayBufferView(
     size_t size,
     V8MLOperandType::Enum type);
 
-MLOperand* BuildConstant(V8TestingScope& scope,
-                         MLGraphBuilder* builder,
-                         const Vector<uint32_t>& dimensions,
-                         V8MLOperandType::Enum type);
+MLOperand* BuildConstant(
+    V8TestingScope& scope,
+    MLGraphBuilder* builder,
+    const Vector<uint32_t>& dimensions,
+    V8MLOperandType::Enum type,
+    absl::optional<NotShared<DOMArrayBufferView>> buffer_view = absl::nullopt);
 
 NotShared<DOMArrayBufferView> CreateArrayBufferViewForOperand(
     const MLOperand* operand);