Struct Placement 

pub struct Placement {
    pub core: usize,
    pub wait: WaitStrategy,
}

Where one pipeline thread runs and how it waits there.

The two are one value because they constrain each other: a thread that busy-spins needs a core to itself, and a thread with no core of its own (core == 0, the unpinned sentinel) can only ever yield — a spinner the scheduler is free to place is the co-location bug with the victim chosen at random. PipelineCores::validate enforces the cross-thread half of that rule; the constructors here enforce the per-thread half.

Fields

core: usize

Logical CPU to pin to. 0 = unpinned (OS scheduled).

wait: WaitStrategy

How the thread waits when it has nothing to do.

Implementations

impl Placement

pub const fn spinning(core: usize) -> Self

Pinned to core and busy-spinning: the thread owns that core. A core of 0 cannot spin — see the type docs — so this yields the unpinned placement instead.

pub const fn yielding(core: usize) -> Self

Pinned to core, spinning briefly then yielding: the thread may share that core with other yielding threads.

pub const fn unpinned() -> Self

No core of its own: the OS scheduler places it within the CPU set the process was started with — and therefore it yields.

pub fn is_pinned(&self) -> bool

Whether this thread is pinned to a core (core != 0).

Trait Implementations

impl Clone for Placement

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for Placement

impl Debug for Placement

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for Placement

impl PartialEq for Placement

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for Placement