Skip to main content

run_core_latency_bench

Function run_core_latency_bench 

Source
fn run_core_latency_bench(nr_cpus: usize) -> Vec<Vec<f64>>
Expand description

Run a core-to-core latency benchmark using atomic ping-pong. Hot loop uses only wrapping_add(1) — no multiply or checked add — so debug builds don’t inflate measurements with overflow checks. Runs 3 attempts per pair with warmup, takes the minimum.