[WebGPU EP] Reduce forward declaration boilerplate in kernel registration

[edgchen1]

Description

Update WebGPU EP kernel registration to remove redundant forward declarations.

E.g., change from this:

class ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kMSDomain, 1, Attention);

  static const BuildKernelCreateInfoFn function_table[] = {
      ...
      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kMSDomain, 1, Attention)>,

to this:

static const BuildKernelCreateInfoFn function_table[] = {
    ...
    BuildKernelCreateInfo<class ONNX_OPERATOR_KERNEL_CLASS_NAME(kWebGpuExecutionProvider, kMSDomain, 1, Attention)>,

One subtlety is that the function table needs to be moved outside of the function now, since function-local (block scope) forward declarations would introduce names with no linkage. Those names should have external linkage.

We already do something similar here:

onnxruntime/onnxruntime/test/autoep/library/example_plugin_ep_kernel_registry/ep_kernel_registration.cc

Line 12 in aaa4944

BuildKernelCreateInfo<class ONNX_OPERATOR_KERNEL_CLASS_NAME(kOnnxDomain, 14, Mul)>,

Motivation and Context

Reduce boilerplate.

[github-actions]

You can commit the suggested changes from lintrunner.

[Copilot]

Pull request overview

Refactors WebGPU EP kernel registration to eliminate large blocks of redundant forward declarations by using elaborated type specifiers (class ...) directly in BuildKernelCreateInfo<...> entries, and moving the function pointer tables to namespace scope to ensure the introduced declarations have external linkage.

Changes:

• Replace explicit forward-declaration lists with BuildKernelCreateInfo<class ONNX_OPERATOR_...> usage.
• Move kernel create-info function tables out of function scope and reuse them in registration loops.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
onnxruntime/core/providers/webgpu/webgpu_execution_provider.cc	Moves WebGPU kernel create-info function table to namespace scope and removes forward-declaration boilerplate via class in template arguments.
onnxruntime/contrib_ops/webgpu/webgpu_contrib_kernels.cc	Applies the same pattern for WebGPU contrib kernel registration by hoisting the function table and removing forward declarations.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.