[infra] Add `clang-22` CUDA build job to nightly/weekly CI by davebayer · Pull Request #8888 · NVIDIA/cccl

davebayer · 2026-05-09T07:04:37Z

LLVM 22.1 is out, so I thought we could add it to our nightly/weekly CI runs as clang-cuda-next run to catch potential issues early. It brings CUDA 12.9 support.

Waiting for llvm22 + cuda-ext devcontainers: rapidsai/devcontainers#695

coderabbitai · 2026-05-13T17:33:37Z

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: bc0b52c2-7dad-40d0-bd88-e1f34c731b8c

📥 Commits

Reviewing files that changed from the base of the PR and between 48998ef and 46769aa.

📒 Files selected for processing (5)

cub/cub/agent/agent_histogram.cuh
cub/cub/device/dispatch/dispatch_histogram.cuh
cub/cub/util_device.cuh
cub/test/CMakeLists.txt
libcudacxx/test/libcudacxx/std/utilities/memory/smartptr/unique.ptr/unique.ptr.create/make_unique_for_overwrite.pass.cpp

✅ Files skipped from review due to trivial changes (2)

cub/test/CMakeLists.txt
cub/cub/agent/agent_histogram.cuh

📝 Walkthrough

Summary by CodeRabbit

New Features
- Added devcontainer configurations for CUDA 12.9, 12.9ext, and 13.2 with LLVM 22 to standardize development environments.
Chores
- Extended CI matrix to include clang-cuda-next and LLVM 22 toolchain combinations and updated devcontainer mappings.
Bug Fixes
- Improved std::span deduction attributes, adjusted CUDA architecture dispatch, added runtime-check wrapping, enforced a histogram vector-size constraint, and suppressed a Clang CUDA warning in tests.
Tests
- Reworked make_unique_for_overwrite detection to a SFINAE-based trait for more robust test coverage.

important:

Walkthrough

Adds three LLVM‑22 Dev Container definitions (CUDA 12.9, 12.9ext, 13.2), extends CI matrix/devcontainer aliasing for clang-cuda-next and clang‑22, and applies small libcudacxx test and CUB compile-time/verification adjustments.

important:

Changes

LLVM 22 Devcontainers and clang-cuda-next CI Support

Layer / File(s)	Summary
LLVM 22 devcontainers for CUDA 12.9, 12.9ext, and 13.2 `.devcontainer/cuda12.9-llvm22/devcontainer.json`, `.devcontainer/cuda12.9ext-llvm22/devcontainer.json`, `.devcontainer/cuda13.2-llvm22/devcontainer.json`	Three new devcontainer definitions add container lifecycle/runArgs, initialization and post-attach commands (WSL/Codespaces conditionals), SCCache/AWS/toolchain/container env, workspace mounts and persistent build/wheelhouse volumes, and VS Code extension/editor/clangd/clang-format settings.
CI matrix clang-cuda-next wiring and toolchain aliases `ci/matrix.yaml`	Adds clang-cuda-next entries to override/nightly/weekly workflows with a fixed SM list; adds `dc_ext` entry for CTK 12.X with `clang22`; expands CTK 12.9 aliases to include `clang-cuda-next`; adds `clang` 22 host_compiler alias `cuda-next`.
libcudacxx span deduction-guide attributes `libcudacxx/include/cuda/std/span`	Two `std::array`-based `span` deduction guides updated to use `_CCCL_DEDUCTION_GUIDE_ATTRIBUTES` instead of `_CCCL_HOST_DEVICE`.
CUB compile-time arch dispatch tweak `cub/cub/util_device.cuh`	`ChainedPolicy::Invoke` now prepends `0` to the instantiated architecture integer pack for the Clang CUDA >=22 `__CUDA_ARCH_LIST__` branch.
CUB histogram safety `cub/cub/agent/agent_histogram.cuh`, `cub/cub/device/dispatch/dispatch_histogram.cuh`	Enforce `VEC_SIZE` ∈ {1,2,4} via `static_assert` and check `MaxPolicy::Invoke` return with `_CCCL_VERIFY(... == cudaSuccess, ...)`.
CUB test compile flags `cub/test/CMakeLists.txt`	Conditionally add `-Wno-pass-failed` for tests when building with Clang CUDA compiler >= 22.0.0.
libcudacxx test trait modernization `libcudacxx/test/.../make_unique_for_overwrite.pass.cpp`	Replace concept-based `HasMakeUniqueForOverwrite` with a SFINAE `void_t`/`decltype` trait and add `type_traits` include.

Suggested reviewers:

bdice
bernhardmgruber

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

coderabbitai

Actionable comments posted: 2

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 07e1774d-884b-47d6-b073-04cffb238887

📥 Commits

Reviewing files that changed from the base of the PR and between e49bdfa and cfac25f.

📒 Files selected for processing (4)

.devcontainer/cuda12.9-llvm22/devcontainer.json
.devcontainer/cuda12.9ext-llvm22/devcontainer.json
.devcontainer/cuda13.2-llvm22/devcontainer.json
ci/matrix.yaml

coderabbitai

Actionable comments posted: 1

🧹 Nitpick comments (1)

.devcontainer/cuda12.9-llvm22/devcontainer.json (1)

41-42: ⚡ Quick win

suggestion: Make timezone mounting host-portable.

Line 41 binds /etc/timezone, which is absent on some hosts (notably macOS), and can block container startup. Prefer relying on /etc/localtime only, or gate /etc/timezone by host type in generation logic.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: ba867447-8cfa-4c48-b786-93c75c6c110b

📥 Commits

Reviewing files that changed from the base of the PR and between cfac25f and df91fd6.

📒 Files selected for processing (5)

.devcontainer/cuda12.9-llvm22/devcontainer.json
.devcontainer/cuda12.9ext-llvm22/devcontainer.json
.devcontainer/cuda13.2-llvm22/devcontainer.json
ci/matrix.yaml
libcudacxx/include/cuda/std/span

✅ Files skipped from review due to trivial changes (1)

libcudacxx/include/cuda/std/span

🚧 Files skipped from review as they are similar to previous changes (3)

.devcontainer/cuda13.2-llvm22/devcontainer.json
.devcontainer/cuda12.9ext-llvm22/devcontainer.json
ci/matrix.yaml

coderabbitai

🧹 Nitpick comments (1)

cub/cub/util_device.cuh (1)

783-785: 💤 Low value

suggestion: The 0 sentinel in the variadic pack guards against empty macro expansion and ensures the fold expression is non-empty. While functionally harmless, a comment explaining this defensive measure would clarify the intent for maintainers.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: f67611ff-f0b1-45f0-a428-47047edfaab3

📥 Commits

Reviewing files that changed from the base of the PR and between df91fd6 and 48998ef.

📒 Files selected for processing (1)

cub/cub/util_device.cuh

github-actions · 2026-05-14T08:10:50Z

😬 CI Workflow Results

🟥 Finished in 29m 17s: Pass: 66%/12 | Total: 3h 20m | Max: 29m 12s | Hits: 97%/17693

See results here.

Jacobfaib · 2026-05-14T14:17:25Z

 #if _CCCL_HAS_HOST_STD_LIB()
 template <class _Tp, size_t _Sz>
-_CCCL_HOST_DEVICE span(::std::array<_Tp, _Sz>&) -> span<_Tp, _Sz>;
+_CCCL_DEDUCTION_GUIDE_ATTRIBUTES span(::std::array<_Tp, _Sz>&) -> span<_Tp, _Sz>;


There are a lot of other places where this is just _CCCL_API (for example the ranges). Is that a problem?

I've changed all deduction guide attributes in #6964, so none should use _CCCL_HOST_DEVICE anymore :)

Jacobfaib · 2026-05-14T14:19:32Z

-_CCCL_CONCEPT HasMakeUniqueForOverwrite = _CCCL_REQUIRES_EXPR((T, variadic Args), T t, Args&&... args)(
-  (cuda::std::make_unique_for_overwrite<T>(cuda::std::forward<Args>(args)...)));


What was the issue with the original concept macro formulation?

clang-22 in CUDA mode failed to evaluate it correctly in C++17 for some reason

davebayer requested a review from a team as a code owner May 9, 2026 07:04

github-project-automation Bot added this to CCCL May 9, 2026

davebayer requested a review from jrhemstad May 9, 2026 07:04

github-project-automation Bot moved this to Todo in CCCL May 9, 2026

cccl-authenticator-app Bot moved this from Todo to In Review in CCCL May 9, 2026

This comment has been minimized.

Sign in to view

davebayer force-pushed the add_clang_cuda22 branch from fa2cde8 to 1d4d5b2 Compare May 9, 2026 07:20

This comment has been minimized.

Sign in to view

bernhardmgruber approved these changes May 11, 2026

View reviewed changes

davebayer changed the title ~~[infra] Add clang-22 build job to nightly/weekly CI~~ [infra] Add clang-22 CUDA build job to nightly/weekly CI May 11, 2026

This comment has been minimized.

Sign in to view

coderabbitai Bot reviewed May 13, 2026

View reviewed changes

Comment thread .devcontainer/cuda12.9ext-llvm22/devcontainer.json

Comment thread ci/matrix.yaml

This comment has been minimized.

Sign in to view

alliepiper approved these changes May 13, 2026

View reviewed changes

davebayer added 2 commits May 13, 2026 21:23

[infra] Add clang-22 build job to nightly/weekly CI

de6c93f

fix span

df91fd6

davebayer force-pushed the add_clang_cuda22 branch from badebe2 to df91fd6 Compare May 13, 2026 19:24

davebayer requested a review from a team as a code owner May 13, 2026 19:24

coderabbitai Bot reviewed May 13, 2026

View reviewed changes

Comment thread .devcontainer/cuda12.9-llvm22/devcontainer.json

This comment has been minimized.

Sign in to view

davebayer requested a review from a team as a code owner May 13, 2026 21:09

coderabbitai Bot reviewed May 13, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

fixes

46769aa

davebayer force-pushed the add_clang_cuda22 branch from 48998ef to 46769aa Compare May 14, 2026 07:39

davebayer requested a review from a team as a code owner May 14, 2026 07:39

Jacobfaib reviewed May 14, 2026

View reviewed changes

		_CCCL_CONCEPT HasMakeUniqueForOverwrite = _CCCL_REQUIRES_EXPR((T, variadic Args), T t, Args&&... args)(
		(cuda::std::make_unique_for_overwrite<T>(cuda::std::forward<Args>(args)...)));

Conversation

davebayer commented May 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

coderabbitai Bot commented May 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Walkthrough

Changes

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

This comment has been minimized.

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

github-actions Bot commented May 14, 2026

😬 CI Workflow Results

🟥 Finished in 29m 17s: Pass: 66%/12 | Total: 3h 20m | Max: 29m 12s | Hits: 97%/17693

Uh oh!

Jacobfaib May 14, 2026

Choose a reason for hiding this comment

Uh oh!

davebayer May 14, 2026

Choose a reason for hiding this comment

Uh oh!

Jacobfaib May 14, 2026

Choose a reason for hiding this comment

Uh oh!

davebayer May 14, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

davebayer commented May 9, 2026 •

edited

Loading

coderabbitai Bot commented May 13, 2026 •

edited

Loading