Skip to content

Conversation

@davebayer
Copy link
Contributor

clang-cuda has always forced us to add __host__ __device__ annotations to deduction guides. However, since clang-22, this is no longer the case + clang emits ton of warnings about this behaviour being deprecated.

This PR introduces _CCCL_DEDUCTION_GUIDE_EXSPACE macro that expands to _host__ __device__ for clang < 22, otherwise is empty.

@davebayer davebayer requested review from a team as code owners December 11, 2025 21:26
@davebayer davebayer requested a review from jrhemstad December 11, 2025 21:26
@github-project-automation github-project-automation bot moved this to Todo in CCCL Dec 11, 2025
@cccl-authenticator-app cccl-authenticator-app bot moved this from Todo to In Review in CCCL Dec 11, 2025
@fbusato
Copy link
Contributor

fbusato commented Dec 11, 2025

I'm a bit hesitant because llvm-22 is still under development. If something change, we need to address the issue. Is the behavior reported on some docs at least?

@github-actions

This comment has been minimized.

@davebayer
Copy link
Contributor Author

I'm a bit hesitant because llvm-22 is still under development. If something change, we need to address the issue. Is the behavior reported on some docs at least?

Aaah, you are right, the change was added to clang last week :D llvm/llvm-project@03e42c9

But it's well documented and I don't expect it to be changed.

@github-actions
Copy link
Contributor

🥳 CI Workflow Results

🟩 Finished in 13h 31m: Pass: 100%/211 | Total: 12d 04h | Max: 5h 13m | Hits: 29%/379225

See results here.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: In Review

Development

Successfully merging this pull request may close these issues.

2 participants