3 ms·
Multi-millisecond latency spikes in a process spinning on an atomic update cannot be blamed on TLB flush, which introduces stalls measured in, at worst, microse
by ncmncm 7y ago
Multi-millisecond latency spikes in a process spinning on an atomic update cannot be blamed on TLB flush, which introduces stalls measured in, at worst, microseconds.
- gpderetta 7y ago[very late reply] yes, that sort of spikes are definitely not due to just interrupts. As you say, it is probably a kernel io thread taking over the cpu.