[ExecuTorch][WebGPU] Fix CI: link shared wgpu-native to avoid LTO bitcode clash#20054
Merged
Conversation
…code clash Pull Request resolved: #20036 The "Test WebGPU Backend" CI job fails on every PR while building the editable executorch wheel. The pybind _portable_lib thin-LTO link aborts with `LLVM gold plugin has failed to create LTO module: Invalid value (Producer: 'LLVM21.1.3-rust-1.92.0-stable' Reader: 'LLVM 12.0.1')`. Root cause: with `-DEXECUTORCH_BUILD_WEBGPU=ON` the backend static-links the prebuilt Rust `libwgpu_native.a`, whose objects embed LLVM-21 `.llvmbc` bitcode (377 of 487 archive members), into the thin-LTO'd pybind extension that pybind11 builds with the CI image's clang-12. LLVM-12's gold plugin cannot parse LLVM-21 bitcode, so the link fails before any test runs. Fix: link the prebuilt shared `libwgpu_native.so` (already downloaded by `setup-wgpu-native.sh`) instead of the static `.a`. The `.so` carries no `.llvmbc`, so no foreign bitcode enters the LTO link. This mirrors how the QNN/CoreML/OpenVINO backends consume prebuilt native deps (shared, runtime-loaded) rather than static-linking foreign objects. Also flips the setup-script idempotency check to the `.so`. Authored with Claude Code. ghstack-source-id: 390126987 @exported-using-ghexport Differential Revision: [D107539304](https://our.internmc.facebook.com/intern/diff/D107539304/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20054
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-cla
Bot
added
the
CLA Signed
This PR needs a
|
JulianCloudNTH
approved these changes
Jun 5, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #20036 by @JulianCloudNTH
^ Please use this as the source of truth for the PR details, comments, and reviews
ghstack PR base: https://github.com/pytorch/executorch/tree/gh/JulianCloudNTH/10/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/JulianCloudNTH/10/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/main
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/JulianCloudNTH/10/orig
@diff-train-skip-merge