面试二十七、 CAS和Atomic

本文主要是介绍面试二十七、 CAS和Atomic，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

CAS锁机制（无锁、自旋锁、乐观锁、轻量级锁）-CSDN博客

1. ABA问题

在C++中，可以使用std::atomic和版本号来解决ABA问题。C++标准库没有直接提供类似Java的AtomicStampedReference，但可以通过将版本号和指针组合在一起实现类似的效果。

#include <atomic>
#include <iostream>
#include <thread>template <typename T>
class AtomicStampedReference {
public:AtomicStampedReference(T initial_value, int initial_stamp): value_(initial_value), stamp_(initial_stamp) {}bool compare_and_set(T expected_value, T new_value, int expected_stamp, int new_stamp) {auto current_value = value_.load();auto current_stamp = stamp_.load();return (current_value == expected_value && current_stamp == expected_stamp) &&value_.compare_exchange_strong(current_value, new_value) &&stamp_.compare_exchange_strong(current_stamp, new_stamp);}T get_reference() const {return value_.load();}int get_stamp() const {return stamp_.load();}private:std::atomic<T> value_;std::atomic<int> stamp_;
};void aba_example() {AtomicStampedReference<int> atomicStampedRef(100, 0);auto thread1 = [&atomicStampedRef]() {int stamp = atomicStampedRef.get_stamp();int reference = atomicStampedRef.get_reference();std::cout << "Thread 1 initial stamp: " << stamp << "\n";std::cout << "Thread 1 initial value: " << reference << "\n";if (atomicStampedRef.compare_and_set(reference, reference + 1, stamp, stamp + 1)) {std::cout << "Thread 1 new stamp: " << atomicStampedRef.get_stamp() << "\n";std::cout << "Thread 1 new value: " << atomicStampedRef.get_reference() << "\n";}};auto thread2 = [&atomicStampedRef]() {int stamp = atomicStampedRef.get_stamp();int reference = atomicStampedRef.get_reference();atomicStampedRef.compare_and_set(reference, reference + 1, stamp, stamp + 1);atomicStampedRef.compare_and_set(reference + 1, reference, stamp + 1, stamp + 2);std::cout << "Thread 2 stamp after ABA: " << atomicStampedRef.get_stamp() << "\n";std::cout << "Thread 2 value after ABA: " << atomicStampedRef.get_reference() << "\n";};std::thread t1(thread1);std::thread t2(thread2);t1.join();t2.join();
}int main() {aba_example();return 0;
}

2. 自旋时间长开销大

循环时间长开销大

描述：当多个线程竞争同一个CAS操作时，如果一直失败，线程会不断自旋重试，造成CPU资源的浪费。尤其在高并发情况下，频繁的CAS重试会导致较高的CPU开销。
解决办法：可以使用自适应自旋锁或其他锁机制来减轻自旋的开销。

自旋锁在高竞争情况下会导致大量CPU资源浪费。可以使用自适应自旋锁或结合其他锁机制来减轻这种开销。以下是一个简单的自适应自旋锁实现示例。

#include <atomic>
#include <thread>
#include <iostream>
#include <chrono>class AdaptiveSpinLock {
public:AdaptiveSpinLock() : flag(ATOMIC_FLAG_INIT) {}void lock() {int spin_count = 0;while (flag.test_and_set(std::memory_order_acquire)) {++spin_count;if (spin_count > max_spin_count) {std::this_thread::yield(); // 出让CPUspin_count = 0;}}}void unlock() {flag.clear(std::memory_order_release);}private:static const int max_spin_count = 1000; // 自适应阈值std::atomic_flag flag;
};void spinlock_example() {AdaptiveSpinLock lock;int counter = 0;auto increment = [&lock, &counter]() {for (int i = 0; i < 1000; ++i) {lock.lock();++counter;lock.unlock();}};std::thread t1(increment);std::thread t2(increment);t1.join();t2.join();std::cout << "Final counter value: " << counter << "\n";
}int main() {spinlock_example();return 0;
}

3. 多变量原子操作

CAS只能保证单个变量的原子操作，要保证多个变量的原子操作，可以使用锁机制。以下是使用std::mutex来保证多个变量的原子操作的示例。

#include <mutex>
#include <thread>
#include <iostream>class MultiVariable {
public:void update(int a, int b) {std::lock_guard<std::mutex> lock(mtx);var1 = a;var2 = b;}void get(int& a, int& b) {std::lock_guard<std::mutex> lock(mtx);a = var1;b = var2;}private:int var1 = 0;int var2 = 0;std::mutex mtx;
};void multivariable_example() {MultiVariable mv;auto writer = [&mv]() {for (int i = 0; i < 1000; ++i) {mv.update(i, i * 2);}};auto reader = [&mv]() {for (int i = 0; i < 1000; ++i) {int a, b;mv.get(a, b);std::cout << "Read values: " << a << ", " << b << "\n";}};std::thread t1(writer);std::thread t2(reader);t1.join();t2.join();
}int main() {multivariable_example();return 0;
}