multithreading · advanced

Spinlock with std::atomic_flag

Build a spinlock from std::atomic_flag: test_and_set spins until it grabs the flag, clear releases it - busy-waiting instead of sleeping.

The code

class SpinLock {
    std::atomic_flag flag = ATOMIC_FLAG_INIT;

public:
    void lock() {
        while (flag.test_and_set(std::memory_order_acquire))
            ; // spin: flag was already set -> keep trying
    }
    void unlock() {
        flag.clear(std::memory_order_release);
    }
};

What this lesson walks through

1. 01A spinlock = one atomic flag
2. 02T1 lock(): test_and_set -> was clear
3. 03T2 lock(): old = true -> spin
4. 04T2 spins while T1 works
5. 05T1 unlock(): flag.clear() (release)
6. 06T2's spin succeeds -> acquires
7. 07The whole spinlock

A spinlock is just a single atomic_flag. lock() repeatedly tries to set it; unlock() clears it. No OS sleep - the waiting thread busy-waits.