Skip to content

[libc++] Increase atomic_ref's required alignment for small types #99654

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 7 commits into from
Aug 1, 2024
Merged

[libc++] Increase atomic_ref's required alignment for small types #99654

Changes from 4 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
8d1fc92
[libc++] Increase atomic_ref's required alignment for small types
dalg24 Jul 19, 2024
8545aeb
fixup! [libc++] Increase atomic_ref's required alignment for small types
dalg24 Jul 19, 2024
6027a62
Fix typo and follow the suggestion from the review
dalg24 Jul 19, 2024
056d4af
Fixup bool -> size_t
dalg24 Jul 23, 2024
e3b95d4
Merge branch 'main' of https://github.com/llvm/llvm-project into incr…
dalg24 Jul 26, 2024
edeef46
Follow James' suggestion and add aligned attribute to the pointer dat…
dalg24 Jul 26, 2024
976dbd9
Use alignas to make objects satisfy atomic_ref's alignement requirement
dalg24 Aug 1, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion libcxx/include/__atomic/atomic_ref.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,10 +95,14 @@ struct __atomic_ref_base {

friend struct __atomic_waitable_traits<__atomic_ref_base<_Tp>>;

// require types that are 1, 2, 4, 8, or 16 bytes in length to be aligned to at least their size to be potentially
// used lock-free
static constexpr size_t __min_alignment = (sizeof(_Tp) & (sizeof(_Tp) - 1)) || (sizeof(_Tp) > 16) ? 0 : sizeof(_Tp);

public:
using value_type = _Tp;

static constexpr size_t required_alignment = alignof(_Tp);
static constexpr size_t required_alignment = alignof(_Tp) > __min_alignment ? alignof(_Tp) : __min_alignment;
Copy link
Member

@EricWF EricWF Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we should calculate the alignment using something like this?

template <class T>
constexpr size_t calc_alignment() {
    size_t align = alignof(T);
    while (align <= alignof(max_align_t)) {
        if (__atomic_always_lock_free(sizeof(T), align))
            return align;
        align *= 2;
    }
    return alignof(T);
}

Copy link
Member

@ldionne ldionne Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I really like this, I think this directly translates to what we are actually trying to achieve (which is to calculate the smallest alignment that allows us to be lockfree).

Copy link
Member

@jyknight jyknight Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That is not a valid implementation, because it causes the ABI of the type to change depending on the microarchitecture selection.

Copy link
Member

@ldionne ldionne Jul 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

TIL


// The __atomic_always_lock_free builtin takes into account the alignment of the pointer if provided,
// so we create a fake pointer with a suitable alignment when querying it. Note that we are guaranteed
Expand Down