Module affinity 

CPU core affinity helpers for deterministic cross-core latency.

Pinning publisher and subscriber threads to specific CPU cores eliminates OS scheduler jitter and ensures consistent cache-coherence transfer times.

NUMA considerations

On multi-socket systems, pin publisher and subscriber threads to cores on the same socket. Cross-socket communication (QPI/UPI) adds ~100-200 ns of additional latency per cache-line transfer compared to intra-socket L3 snoops (~40-55 ns on Intel Comet Lake).

Example

use photon_ring::affinity;

let cores = affinity::available_cores();
assert!(cores.len() >= 2, "need at least 2 cores");

// Pin to the first core
assert!(affinity::pin_to_core(cores[0]));

Structs

CoreId

This represents a CPU core.

Functions

available_cores

Return the list of CPU cores available to this process.

core_count

Return the number of CPU cores available to this process.

pin_to_core

Pin the current thread to a specific CPU core.

pin_to_core_id

Pin the current thread to a core by its numeric index.