pub(crate) fn read_irq_ticks(nr_cpus: usize) -> Option<Vec<u64>>Expand description
Per-CPU time spent in interrupt handlers, in kernel ticks, from
/proc/stat (irq + softirq columns, CONFIG_IRQ_TIME_ACCOUNTING).
Handler TIME is the harm itself, not a proxy: a wake suffers exactly when it lands on a CPU whose handler is running, and the time share IS that probability. Counting interrupt lines mis-priced both directions – a 1000 Hz mouse with a ~1 µs handler steals no more time than an unflagged CPU, while a slow line with a heavy handler could never reach a count bar (§G33). Softirq time is included because NAPI network processing shadows wakes the same way and never appeared as line counts.