It's nonsense to do this when you can be preempted. But you can run one thread per core and avoid preemption. You can tune the linux kernel to avoid almost all

It's nonsense to do this when you can be preempted. But you can run one thread per core and avoid preemption. You can tune the linux kernel to avoid almost all preemption, due to TLB shootdowns etc: https://rigtorp.se/low-latency-guide/ https://rigtorp.se/low-latency-guide/