Struct crossbeam::mem::CachePadded
[−]
[src]
pub struct CachePadded<T> { // some fields omitted }
Pad T
to the length of a cacheline.
Sometimes concurrent programming requires a piece of data to be padded out
to the size of a cacheline to avoid "false sharing": cachelines being
invalidated due to unrelated concurrent activity. Use the CachePadded
type
when you want to avoid cache locality.
At the moment, cache lines are assumed to be 32 * sizeof(usize) on all architectures.
Warning: the wrapped data is never dropped; move out using ptr::read
if you need to run dtors.
Methods
impl<T: ZerosValid> CachePadded<T>
fn zeroed() -> CachePadded<T>
[−]
A const fn equivalent to mem::zeroed().
impl<T> CachePadded<T>
fn new(t: T) -> CachePadded<T>
[−]
Wrap t
with cacheline padding.
Warning: the wrapped data is never dropped; move out using
ptr:read
if you need to run dtors.