原子累加器及其原理-JUC-并发编程(Java)

1、原子整数累加和原子累加器性能比较

import java.util.ArrayList;
import java.util.List;
import java.util.concurrent.atomic.AtomicLong;
import java.util.concurrent.atomic.LongAdder;
import java.util.function.Consumer;
import java.util.function.Supplier;

public class Test01 {
    public static void main(String[] args) {
        for(int i = 0; i < 10; i++) {
            demo(() -> new AtomicLong(0), AtomicLong::incrementAndGet);
        }
        for (int i = 0; i < 10; i++) {
            demo(LongAdder::new, LongAdder::increment);
        }
    }

    private static <T> void demo(Supplier<T> adderSupplier, Consumer<T> action) {
        T adder = adderSupplier.get();
        List<Thread> ts = new ArrayList<>();
        // 4个线程，没个线程累加500000次
        for (int i = 0; i < 4; i++) {
            ts.add(new Thread(() -> {
                for (int j = 0; j < 500000; j++) {
                    action.accept(adder);
                }
            }));
        }
        long start = System.nanoTime();
        ts.forEach(Thread::start);
        ts.forEach(t -> {
            try {
                t.join();
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        });
        long end = System.nanoTime();
        System.out.println(adder + " cost: "+ (end - start)/1000_000);
    }
}

// 多次执行测试结果
2000000 cost: 53
2000000 cost: 50
2000000 cost: 45
2000000 cost: 46
2000000 cost: 46
2000000 cost: 46
2000000 cost: 46
2000000 cost: 46
2000000 cost: 46
2000000 cost: 46
2000000 cost: 12
2000000 cost: 9
2000000 cost: 8
2000000 cost: 6
2000000 cost: 6
2000000 cost: 6
2000000 cost: 5
2000000 cost: 6
2000000 cost: 5
2000000 cost: 5

可以发现原子累加器性能明显提升，为什么呢？

在发生竞争时，原子累加器设置了多个累加单元，Thread-0累加Cell[0],Thread-1累加Cell[1],…,最后将结果汇总。这样它们在累加的时操作不同的累加变量，因此减少了CAS失败重试，从而提高了性能。

2、LongAddr原理（源码）分析

2.1、LondAddr关键字段

要分析原理呢，我们先来看下源码。

LongAddr类几个关键的字段

/**
     * Table of cells. When non-null, size is a power of 2.
     */
    transient volatile Cell[] cells;

    /**
     * Base value, used mainly when there is no contention, but also as
     * a fallback during table initialization races. Updated via CAS.
     */
    transient volatile long base;

    /**
     * Spinlock (locked via CAS) used when resizing and/or creating Cells.
     */
    transient volatile int cellsBusy;

• Cell[] cells：累加单元数组，懒惰初始化
• base：基础值，如果没有竞争，则用cas累加这个字段
• cellsBusy：在cells创建或者扩容时，置为1，表示加锁；默认0表示未加锁

2.2、cas锁原理

import lombok.extern.slf4j.Slf4j;

import java.util.concurrent.TimeUnit;
import java.util.concurrent.atomic.AtomicInteger;

@Slf4j(topic = "c.LockCas")
public class LockCas {
    private AtomicInteger state = new AtomicInteger(0);

    public void lock() {
        while (true) {
            if (state.compareAndSet(0, 1)) {
                break;
            }
        }
    }

    public void unlock() {
        log.debug("unlock...");
        state.set(0);
    }
    public static void main(String[] args) {
        LockCas lock = new LockCas();
        new Thread(() -> {
            log.debug("begin...");
            lock.lock();
            try {
                log.debug("lock...");
                TimeUnit.SECONDS.sleep(1);
            } catch (InterruptedException e) {
                e.printStackTrace();
            } finally {
                lock.unlock();
            }
        }, "t1").start();

        new Thread(() -> {
            log.debug("begin...");
            lock.lock();
            try {
                log.debug("lock...");
                TimeUnit.SECONDS.sleep(1);
            } catch (InterruptedException e) {
                e.printStackTrace();
            } finally {
                lock.unlock();
            }
        }, "t2").start();
    }
}

// 测试结果
2021-09-29 10:28:45.960 DEBUG [t2] c.LockCas - begin...
2021-09-29 10:28:45.960 DEBUG [t1] c.LockCas - begin...
2021-09-29 10:28:45.962 DEBUG [t1] c.LockCas - lock...
2021-09-29 10:28:46.962 DEBUG [t1] c.LockCas - unlock...
2021-09-29 10:28:46.962 DEBUG [t2] c.LockCas - lock...
2021-09-29 10:28:47.962 DEBUG [t2] c.LockCas - unlock...

一旦某个线程加锁，其他线程就会一直循环等待，直到持有锁的线程释放锁。

2.3、缓存行伪共享

Cell类

@sun.misc.Contended 
static final class Cell {
        volatile long value;
        Cell(long x) { value = x; }
        final boolean cas(long cmp, long val) {
            return UNSAFE.compareAndSwapLong(this, valueOffset, cmp, val);
        }

        // Unsafe mechanics
        private static final sun.misc.Unsafe UNSAFE;
        private static final long valueOffset;
        static {
            try {
                UNSAFE = sun.misc.Unsafe.getUnsafe();
                Class<?> ak = Cell.class;
                valueOffset = UNSAFE.objectFieldOffset
                    (ak.getDeclaredField("value"));
            } catch (Exception e) {
                throw new Error(e);
            }
        }
    }

• 注解@sun.misc.Contended 用于防止缓存行伪共享

CUP与主存速度比较

从CUP到	需要的时钟周期
寄存器	1 cycle
L1	3~4 cycle
L2	10~20 cycle
L3	40~45 cycle
内存	120~240 cycle

因为CUP与内存速度的差异很大，需要靠预读取数据至缓存来提示效率

而缓存以缓存行为单位，没个缓存行对应一块内存，一般是64 byte

缓存的加入会造成数据副本的产生，即内存的同一份数据会缓存在不同核心（多核CUP）的缓存行中

CPU要保证数据的一致性，如果某个CPU核心更改了数据，其他CPU核心对应的缓存行会失效

因为cells是数组类型，在内存中是练习存储的，一个Cell为24字节（16字节对象头和8字节的value)，因此缓存行可以存下 2个Cell对象。这样问题来了

• Core-0要修改Cell[0]
• Core-1要修改Cell[1]

无论谁修改成功，都会导致对方Core缓存行失效

@sun.misc.Contended注解用来解决这个问题，它的原理就是在使用了此注解的对象或者字段前后各家128字节大小的padding,从而让CPU将对象预读至缓存时占用不同的缓存行，这样，不会对方缓存行的失效

2.4、成员方法

• add方法

public void add(long x) {
        Cell[] as; long b, v; int m; Cell a;
        if ((as = cells) != null || !casBase(b = base, b + x)) {
            boolean uncontended = true;
            if (as == null || (m = as.length - 1) < 0 ||
                (a = as[getProbe() & m]) == null ||
                !(uncontended = a.cas(v = a.value, v + x)))
                longAccumulate(x, null, uncontended);
        }
    }

• x此处为1

• longAccumulate方法

 final void longAccumulate(long x, LongBinaryOperator fn,
                              boolean wasUncontended) {
        int h;
        if ((h = getProbe()) == 0) {
            ThreadLocalRandom.current(); // force initialization
            h = getProbe();
            wasUncontended = true;
        }
        boolean collide = false;                // True if last slot nonempty
        for (;;) {
            Cell[] as; Cell a; int n; long v;
            if ((as = cells) != null && (n = as.length) > 0) {
                if ((a = as[(n - 1) & h]) == null) {
                    if (cellsBusy == 0) {       // Try to attach new Cell
                        Cell r = new Cell(x);   // Optimistically create
                        if (cellsBusy == 0 && casCellsBusy()) {
                            boolean created = false;
                            try {               // Recheck under lock
                                Cell[] rs; int m, j;
                                if ((rs = cells) != null &&
                                    (m = rs.length) > 0 &&
                                    rs[j = (m - 1) & h] == null) {
                                    rs[j] = r;
                                    created = true;
                                }
                            } finally {
                                cellsBusy = 0;
                            }
                            if (created)
                                break;
                            continue;           // Slot is now non-empty
                        }
                    }
                    collide = false;
                }
                else if (!wasUncontended)       // CAS already known to fail
                    wasUncontended = true;      // Continue after rehash
                else if (a.cas(v = a.value, ((fn == null) ? v + x :
                                             fn.applyAsLong(v, x))))
                    break;
                else if (n >= NCPU || cells != as)
                    collide = false;            // At max size or stale
                else if (!collide)
                    collide = true;
                else if (cellsBusy == 0 && casCellsBusy()) {
                    try {
                        if (cells == as) {      // Expand table unless stale
                            Cell[] rs = new Cell[n << 1];
                            for (int i = 0; i < n; ++i)
                                rs[i] = as[i];
                            cells = rs;
                        }
                    } finally {
                        cellsBusy = 0;
                    }
                    collide = false;
                    continue;                   // Retry with expanded table
                }
                h = advanceProbe(h);
            }
            else if (cellsBusy == 0 && cells == as && casCellsBusy()) {
                boolean init = false;
                try {                           // Initialize table
                    if (cells == as) {
                        Cell[] rs = new Cell[2];
                        rs[h & 1] = new Cell(x);
                        cells = rs;
                        init = true;
                    }
                } finally {
                    cellsBusy = 0;
                }
                if (init)
                    break;
            }
            else if (casBase(v = base, ((fn == null) ? v + x :
                                        fn.applyAsLong(v, x))))
                break;                          // Fall back on using base
        }
    }

cells数组创建且不为空

cells未创建或者为空

符合上述条件

• sum方法获取最终结果

  public long sum() {
          Cell[] as = cells; Cell a;
          long sum = base;
          if (as != null) {
              for (int i = 0; i < as.length; ++i) {
                  if ((a = as[i]) != null)
                      sum += a.value;
              }
          }
          return sum;
      }
  

  • 初始值base
  • 变量cells数组
  • 如果不为空，累加对应cell的值

QQ:806797785

仓库地址：https://gitee.com/gaogzhen/concurrent