Struct CachePadding

#[repr(align(128))]
pub struct CachePadding {
    atomic: AtomicUsize,
}

Intentionally force alignment to 128 bytes to make a best effort attempt to place each atomic on its own cache line. This reduces contention and improves performance for common CPU caching protocols such as MESI.

Fields

atomic: AtomicUsize

Implementations

impl CachePadding

Convenience wrapper methods around atomic operations. Both start and end indices are packed into a single atomic so that we can use the fastest and easiest to reason about Relaxed ordering.

fn new(n: usize) -> Self

fn increment(&self) -> usize

fn load(&self) -> usize

fn store(&self, n: usize)

fn compare_exchange(&self, current: usize, new: usize) -> bool