Struct PeekPermits 

pub struct PeekPermits {
    semaphore: Arc<Semaphore>,
    bound: Mutex<Bound>,
    workers: usize,
}

The bound on how many offloaded peek walks run at once.

The resource it protects is CPU rather than any one worker’s thread, so one instance covers every worker that shares it.

Fields

semaphore: Arc<Semaphore>

bound: Mutex<Bound>

Under one lock, so a resize computing its delta and a release deciding whether to return its permit see one state.

workers: usize

The workers the configured fraction is a fraction of.

Implementations

impl PeekPermits

pub fn new(workers: usize) -> Self

Creates a bound over the workers a process runs, admitting one walk per worker until it is configured otherwise.

async fn acquire(self: &Arc<Self>) -> WalkPermit

Waits for a permit.

fn absorb_release(&self) -> bool

Whether the permit about to be released is owed to a shrink, in which case the caller forgets it instead of returning it.

fn permits_for(workers: usize, fraction: f64) -> usize

The permits fraction asks for over workers, at least one and at most what a semaphore can hold.

A fraction that is negative or NaN lands on the floor of one rather than being rejected, because a misconfigured bound should pace the offload rather than stop it.

fn resize(&self, fraction: f64)

Resizes the bound to what fraction asks for.

Raising it takes effect at once. Lowering it interrupts no walk: the permits free right now go, and the rest go as the walks holding them finish, see WalkPermit.