Fix more _CCCL_DEVICE static -> _CCCL_DEVICE_API

[Jacobfaib]

Description

More LLVM23 -Wunused-template warnings. After this round of fixes, however, we should just disable the warning entirely.

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[coderabbitai]

Important

Review skipped

Review was skipped as selected files did not have any reviewable changes.

💤 Files selected but had no reviewable changes (1)

• libcudacxx/include/cuda/std/__atomic/functions/cuda_ptx_generated.h

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 63278e33-303a-414c-acf0-a7051180f863

📥 Commits

Reviewing files that changed from the base of the PR and between cb688b8 and e4dc70c.

📒 Files selected for processing (1)

• libcudacxx/include/cuda/std/__atomic/functions/cuda_ptx_generated.h

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: b241ebe1-18d3-4315-827b-b091e19c955d

📥 Commits

Reviewing files that changed from the base of the PR and between 6864b6a and cb688b8.

📒 Files selected for processing (5)

• libcudacxx/codegen/generators/compare_and_swap.h
• libcudacxx/codegen/generators/exchange.h
• libcudacxx/codegen/generators/fence.h
• libcudacxx/codegen/generators/fetch_ops.h
• libcudacxx/include/cuda/std/__atomic/functions/cuda_ptx_generated.h

Included review availability: Your plan provides up to 12 included reviews per hour; 8 remain after this review.

---

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes
  • Improved compatibility for CUDA dispatch, atomic, load/store, and synchronization helpers.
  • Enabled broader use of these helpers across compilation units and supported toolchains.
  • Preserved existing interfaces, runtime behavior, memory-order handling, and dispatch logic.
  • Standardized API exposure for device and runtime helper functions, including top-k, histogram, reduction, run-length encoding, and transform operations.

Walkthrough

Changes

CUB dispatch declarations and CUDA atomic helper declarations no longer use internal linkage. Function parameters, dispatch logic, atomic operations, and fence behavior remain unchanged.

CUB and CUDA helper linkage updates

Layer / File(s)	Summary
CUB dispatch linkage cub/cub/detail/env_dispatch.cuh, cub/cub/device/...	Environment, TopK, histogram, reduce-by-key, RLE, and transform dispatch functions no longer use static.
Generated atomic load and store linkage libcudacxx/codegen/generators/ld_st.h	Generated atomic load and store helpers, dispatchers, functors, and wrappers now use _CCCL_DEVICE_API.
Generated atomic operation linkage libcudacxx/codegen/generators/compare_and_swap.h, libcudacxx/codegen/generators/exchange.h, libcudacxx/codegen/generators/fetch_ops.h	Generated compare-exchange, exchange, and fetch-operation helpers now use _CCCL_DEVICE_API.
CUDA atomic helper linkage libcudacxx/include/cuda/std/__atomic/functions/cuda_ptx_derived.h, libcudacxx/include/cuda/std/__atomic/platform/msvc_to_builtins.h	Atomic operations and fence helpers no longer use internal linkage. Their implementations remain unchanged.
Generated fence linkage libcudacxx/codegen/generators/fence.h	Generated membar, fence, and thread-fence functions now use _CCCL_DEVICE_API.

Suggested reviewers: miscco, wmaxey, griwes

Merge Risk: ⚪ Minimal · up to cb688

This PR makes localized annotation changes to address compiler warnings, and no actionable merge-blocking risk remains beyond normal checks and review.

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

🥳 CI Workflow Results

🟩 Finished in 3h 13m: Pass: 100%/405 | Total: 16d 21h | Max: 2h 52m | Hits: 31%/2206446

See results here.