Adopt CUDA stream compatibility accessors

[bdice]

Summary

Use the get() and sync() compatibility aliases added in RMM #2537. These spellings are shared by rmm::cuda_stream_view and cuda::stream_ref.

This preserves existing stream types and public APIs while extracting mechanical accessor updates from the broader stream migration. It is independently buildable without RMM #2372 and leaves the migration PR focused on actual type and signature changes.

This updates raw CUDA, library, kernel-launch, and legacy API boundaries throughout RAFT and uses the CUDA stream umbrella header where stream definitions are required. Existing rmm::cuda_stream_view return types remain unchanged; their migration stays in RAFT #3129.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[bdice]

Validated this preparation branch against RMM upstream/main: the local -j0 build completed and focused select-k tests passed. The dependent migration PR #3129 has been rebased onto this branch.

[coderabbitai]

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: a6d7403b-509b-4b7c-bce7-3d506caaf1c8

📥 Commits

Reviewing files that changed from the base of the PR and between f356e94 and 51a2375.

📒 Files selected for processing (3)

• cpp/include/raft/comms/detail/mpi_comms.hpp
• cpp/include/raft/linalg/strided_reduction.cuh
• cpp/include/raft/solver/linear_assignment.cuh

Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review.

---

📝 Summary

Summary by CodeRabbit

• Bug Fixes

  • Improved CUDA stream interoperability across communication, linear algebra, matrix, random, sparse, spectral, and statistics operations.
  • Ensured asynchronous GPU operations consistently receive valid native stream handles.
  • Updated stream synchronization behavior for more reliable CUDA execution.

• Chores

  • Replaced deprecated CUDA stream header usage with the current stream interface.
  • Updated GPU operation validation coverage without changing public APIs or numerical behavior.

Walkthrough

Changes

CUDA stream handle migration

Layer / File(s)	Summary
Production stream handle conversion cpp/include/raft/**, cpp/src/**	Production code now passes raw cudaStream_t values from stream wrappers through .get(). Deprecated cuda/stream_ref includes and .value() access are replaced.
Test and benchmark stream conversion cpp/tests/**, cpp/bench/**	Tests and benchmarks now pass raw CUDA stream handles to CUDA, RAFT, RMM, and library APIs. Updated synchronization calls use sync().

Estimated code review effort: 3 (Moderate) | ~20 minutes

Merge Risk: ⚪ Minimal · up to 51a23

This updates CUDA stream compatibility accessors while preserving existing stream behavior and public interfaces; no actionable merge risk remains.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 26.73% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 101 functions across 66 files. (2 skipped…	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the main change: adopting CUDA stream compatibility accessors across RAFT.
Description check	✅ Passed	The description directly explains the compatibility accessor updates, preserved APIs, affected boundaries, and migration scope.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

Full details: Docstring Coverage

Explanation

Docstring coverage is 26.73% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 101 functions across 66 files. (2 skipped: 2 unsupported.)

• Fix all pre-merge checks with AI

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[bdice]

CI exposed benchmark targets that the default local build did not compile. The preparation branch now adds the missing raw-stream accessors at those benchmark boundaries (f356e946). Validation:

• build-raft -j0 passes against the current RMM dependency
• all primitive benchmark targets pass with ninja -C cpp/build -j0 after configuring bench-prims against RMM main
• focused pre-commit hooks pass with copyright-only churn excluded