Enum AffinityIntent 

pub enum AffinityIntent {
    Inherit,
    RandomSubset {
        from: BTreeSet<usize>,
        count: usize,
    },
    LlcAligned,
    CrossCgroup,
    SingleCpu,
    Exact(BTreeSet<usize>),
    SmtSiblingPair,
}

Scenario-level affinity intent for a group of workers.

Resolved to a concrete ResolvedAffinity at runtime based on the cgroup’s effective cpuset and the VM’s topology. When attached to a WorkSpec, determines per-worker sched_setaffinity masks.

Resolution uses resolve_affinity_for_cgroup().

Naming pattern (Intent vs Resolved)

AffinityIntent and ResolvedAffinity form a pre/post-resolution pair. Variant names line up where the same shape exists on both sides; payload differences encode the intent → concrete-CPU-set distinction:

AffinityIntent	ResolvedAffinity
Inherit (no payload)	None
Exact(BTreeSet<usize>)	Fixed(BTreeSet<usize>)
RandomSubset { from, count }	Random { from, count }
SingleCpu (no payload)	SingleCpu(usize)
LlcAligned / CrossCgroup	Fixed(...) (resolver expands)
SmtSiblingPair (no payload)	Fixed({sibling_a, sibling_b})

Constructor helpers: AffinityIntent::exact takes any IntoIterator<Item = usize> for the Exact set; AffinityIntent::random_subset takes the same iterator shape for the RandomSubset pool plus a sample-count argument.

The SingleCpu pair specifically: AffinityIntent::SingleCpu expresses “pin to one CPU; resolver picks which based on cgroup state and worker index”, and ResolvedAffinity::SingleCpu records the concrete CPU id chosen. Reusing the variant name keeps the pre/post mapping lexically obvious — payload presence distinguishes intent from resolution without renaming the variant.

AffinityIntent::RandomSubset carries the resolved pool (from) and sample size (count) — sampling itself is deferred to spawn time so each worker gets an independent draw. The scenario engine’s resolve_affinity_for_cgroup materialises the pool from cgroup cpuset / topology before constructing this variant; spawn-time resolve_affinity samples per-worker. Construct directly via AffinityIntent::random_subset.

Variants

Inherit

No affinity constraint – inherit from parent cgroup.

RandomSubset

Pin each worker to a random subset of from, sampling count CPUs per worker. Sampling is deferred to spawn time so each worker gets an independent draw — mirrors ResolvedAffinity::Random semantics. Construct with the resolved pool already materialised; the scenario engine pre- resolves topology-aware “pick from cgroup state” intent before building this variant.

Fields

from: BTreeSet<usize>

count: usize

LlcAligned

Pin to the CPUs in the worker’s LLC.

CrossCgroup

Pin to all CPUs (crosses cgroup boundaries).

SingleCpu

Pin to a single CPU.

Exact(BTreeSet<usize>)

Pin to an exact set of CPUs.

SmtSiblingPair

Pin all workers in the group to the two SMT siblings of one physical core. Tests how the scheduler handles two compute-bound tasks placed on SMT siblings — both threads contend for the core’s shared front-end / execution resources, exposing scheduler decisions about co-running vs. spreading compute load across cores.

Designed for WorkType::SmtSiblingSpin and other worker_group_size = 2 variants (WorkType::FutexPingPong, WorkType::AsymmetricWaker, WorkType::SignalStorm, etc.) where both workers in a group are intended to run on a sibling pair. The variant has no payload — the resolver picks an SMT-sibling pair from the cgroup’s effective cpuset (or the full topology when no cpuset is active).

Resolution is performed by the scenario engine’s resolve_affinity_for_cgroup (topology-aware, not available at the bare WorkloadHandle::spawn gate). The resolver searches the cpuset for a physical core with at least two thread siblings present and resolves to ResolvedAffinity::Fixed containing those two CPU IDs. All workers in the group get pinned to that 2-CPU set; when num_workers == 2 the kernel runs one worker on each sibling, which is the contention pattern this intent targets.

Returns an error from the resolver — NOT a silent fallback — when no SMT-sibling pair is available (threads_per_core == 1, or the cpuset isolates each sibling onto a different CPU set). Callers must handle the error; running WorkType::SmtSiblingSpin without SMT siblings would produce a misleading result.

Implementations

impl AffinityIntent

pub fn exact(cpus: impl IntoIterator<Item = usize>) -> Self

Construct an Exact affinity from any iterator of CPU indices.

Accepts arrays, ranges, Vec, BTreeSet, or any IntoIterator<Item = usize>.

pub fn random_subset( from: impl IntoIterator<Item = usize>, count: usize, ) -> Self

Construct a RandomSubset from a pool iterator and a sample size. Mirrors the Self::exact constructor’s iterator flexibility — accepts arrays, Vec, BTreeSet, ranges, or any IntoIterator<Item = usize> for the pool.

Sampling is deferred to spawn time; each worker gets an independent count-sized draw from from. count > from.len() is clamped to from.len() at sample time (topology fact, not caller error). count == 0 and empty from are rejected at the spawn-time affinity gate with an actionable diagnostic — use AffinityIntent::Inherit for no affinity constraint.

Trait Implementations

impl Clone for AffinityIntent

fn clone(&self) -> AffinityIntent

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AffinityIntent

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

Formats the value using the given formatter.

impl Default for AffinityIntent

fn default() -> AffinityIntent

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for AffinityIntent

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Hash for AffinityIntent

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for AffinityIntent

fn eq(&self, other: &AffinityIntent) -> 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 Serialize for AffinityIntent

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Eq for AffinityIntent

impl StructuralPartialEq for AffinityIntent