Enum Placement 

pub enum Placement {
    Auto {
        max_partitions: Option<usize>,
    },
    Explicit(Vec<CpuSet>),
    Unpinned {
        partitions: usize,
    },
}

How partitions map onto the machine.

Variants

Auto

One partition per NUMA node in the affinity mask.

The partition count is the node count rounded down to a power of two (at least one); when that is fewer than the node count, adjacent nodes are merged into a partition rather than dropped, so every allowed CPU stays in play. Each partition takes as many threads as it has CPUs.

Fields

max_partitions: Option<usize>

Upper bound on the partition count, itself rounded down to a power of two.

Explicit(Vec<CpuSet>)

One partition per listed CPU set, exactly as given.

Sets may overlap (the resolve logs a warning and proceeds), but each must be non-empty and a subset of the process’s affinity mask.

Tuple Fields

0: Vec<CpuSet>

The CPU sets, one per partition; the count must be a power of two.

Unpinned

Partitions with no pinning at all, for tests and CI, each taking an even share of available_parallelism (at least one thread).

Fields

partitions: usize

Number of partitions; must be a power of two.

Trait Implementations

impl Clone for Placement

fn clone(&self) -> Placement

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Placement

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

Formats the value using the given formatter.

impl PartialEq for Placement

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

Tests for self and other values to be equal, and is used by ==.

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

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Eq for Placement

impl StructuralPartialEq for Placement