Accept integral Double select scalars (#21387)

[JCNTH]

Summary:

Loads existing Vulkan-serialized select graphs whose integer arguments alias Double constants.

Problem: Vulkan serialization deduplicates equal scalar values, so an integer select dimension or index can reference a previously emitted Double. WebGPU rejected those graphs even though Vulkan accepts the scalar representation.

Solution:

• Before: aten.select_copy.int accepted only serialized Int values.
• After: it also accepts exactly integral Double values in the int64_t range and rejects fractional, non-finite, out-of-range, and symbolic values.

Implementation:

• Select.cpp — validates Double range and integrality before conversion.
• test_webgpu_native.cpp — adds direct VK00 success, boundary, and rejection coverage.
• Mirrors Vulkan ComputeGraph::extract_scalar<int64_t> scalar acceptance while retaining stricter checked conversion.

Constraints: The shader, dispatch, resize behavior, and SymInt policy are unchanged.

Co-authored-with: Claude Code.

Differential Revision: D113588820

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21387

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[meta-codesync]

@JCNTH has exported this pull request. If you are a Meta employee, you can view the originating Diff in D113588820.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

[Copilot]

Pull request overview

This PR improves WebGPU runtime compatibility with Vulkan-serialized graphs by allowing aten.select_copy.int to consume Double scalar values when they are exactly integral and within int64_t range, matching Vulkan’s scalar deduplication behavior while still rejecting fractional/non-finite/out-of-range and symbolic values.

Changes:

• Extend WebGPU select scalar parsing to accept Double values only when they are integral and safely convertible to int64_t.
• Add WebGPU native regression tests covering successful execution for integral Double dim/index, boundary cases, and expected build-time rejections (fractional, NaN/Inf, out-of-range, SymInt).

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
backends/webgpu/runtime/ops/select/Select.cpp	Accept integral Double scalars for select dim/index with validated conversion; keep SymInt/dynamic rejection behavior.
backends/webgpu/test/test_webgpu_native.cpp	Add regression + negative tests for Vulkan-serialized select graphs where integer args alias cached Double scalars.

---

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

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

[SS-JIA]

Review automatically exported from Phabricator review in Meta.