[xkevio]

Related to #131513.

Allows annotating nested functions with #[offload_kernel], similar to autodiff.

Also fixes the error message when #[offload_kernel] is used on anything other than functions.

r? @ZuseZ4

(@Sa4dUs, ping)

[rustbot]

Error: Failed to set assignee to Sa4dUs: invalid assignee

Note: Only org members with at least the repository "read" role, users with write permissions, or people who have commented on the PR may be assigned.

Please file an issue on GitHub at triagebot if there's a problem with this bot, or reach out on #triagebot on Zulip.

[xkevio]

I think the test runner just took too long? In any case, I finally had a bit more time for thorough testing and can confirm that it works without issues.

My main goal was to actually enable nested offload functions in main but I only realized after implementing that due to conditional compilation via cfg(target_os), that that might not work as I had hoped ^^

[ZuseZ4]

Our CI timeout is after 6hrs iirc, some of our runners regulary take over 3hrs. Just force-push to trigger a rerun, this might be a spurious failure. github CI is often a bit flaky.

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[xkevio]

PR is rebased, checks are green and the new unimplemented! expansion is reflected in the tests!

[ZuseZ4]

Just for some code cleanup, can you deduplicate the two by first extracting the inner item from both branches?
Then you can handle the ItemKind::Fn case once. Also if let could be a bit nicer.

View changes since the review

[ZuseZ4]

lgtm, minor some code feedback. Can you squash it to one or two commits terwards so the wip/fix comments are gone? There's also a bors command for it if you want.