24

在Object.hashCode()的 Javadoc 中，它指出

在合理可行的情况下，类定义的 hashCode 方法Object确实为不同的对象返回不同的整数。（这通常通过将对象的内部地址转换为整数来实现，但 Java™ 编程语言不需要这种实现技术。）

这是一个常见的误解，这与内存地址有关，但它不会在没有通知的情况下更改，并且 hashCode() 不会也不能更改对象。

@Neet 提供了一个很好的答案链接https://stackoverflow.com/a/565416/57695但我正在寻找更多细节。

这是一个例子来说明我的担忧

Field theUnsafe = Unsafe.class.getDeclaredField("theUnsafe");
theUnsafe.setAccessible(true);
Unsafe unsafe = (Unsafe) theUnsafe.get(null);

for (int t = 0; t < 10; t++) {
    System.gc();
    Object[] objects = new Object[10];
    for (int i = 0; i < objects.length; i++)
        objects[i] = new Object();

    for (int i = 0; i < objects.length; i++) {
        if (i > 0) System.out.print(", ");
        int location = unsafe.getInt(objects, Unsafe.ARRAY_OBJECT_BASE_OFFSET + Unsafe.ARRAY_OBJECT_INDEX_SCALE * i);
        System.out.printf("%08x: hc= %08x", location, objects[i].hashCode());
    }
    System.out.println();
}

印刷

eac00038: hc= 4f47e0ba, eac00048: hc= 2342d884, eac00058: hc= 7994d431, eac00068: hc= 19f71b53, eac00078: hc= 2e22f376, eac00088: hc= 789ddfa3, eac00098: hc= 44c58432, eac000a8: hc= 036a11e4, eac000b8: hc= 28bc917c, eac000c8: hc= 73f378c8
eac00038: hc= 30813486, eac00048: hc= 729f624a, eac00058: hc= 3dee2310, eac00068: hc= 5d400f33, eac00078: hc= 18a60d19, eac00088: hc= 3da5f0f3, eac00098: hc= 596e0123, eac000a8: hc= 450cceb3, eac000b8: hc= 4bd66d2f, eac000c8: hc= 6a9a4f8e
eac00038: hc= 711dc088, eac00048: hc= 584b5abc, eac00058: hc= 3b3219ed, eac00068: hc= 564434f7, eac00078: hc= 17f17060, eac00088: hc= 6c08bae7, eac00098: hc= 3126cb1a, eac000a8: hc= 69e0312b, eac000b8: hc= 7dbc345a, eac000c8: hc= 4f114133
eac00038: hc= 50c8c3b8, eac00048: hc= 2ca98e77, eac00058: hc= 2fc83d89, eac00068: hc= 034005e1, eac00078: hc= 6041f871, eac00088: hc= 0b1df416, eac00098: hc= 5b83d60d, eac000a8: hc= 2c5a1e6b, eac000b8: hc= 5083198c, eac000c8: hc= 4f025f9f
eac00038: hc= 00c5eb8a, eac00048: hc= 41eab16b, eac00058: hc= 1726099c, eac00068: hc= 4240eca3, eac00078: hc= 346fe350, eac00088: hc= 1db4b415, eac00098: hc= 429addef, eac000a8: hc= 45609812, eac000b8: hc= 489fe953, eac000c8: hc= 7a8f6d64
eac00038: hc= 7e628e42, eac00048: hc= 7869cfe0, eac00058: hc= 6aceb8e2, eac00068: hc= 29cc3436, eac00078: hc= 1d77daaa, eac00088: hc= 27b4de03, eac00098: hc= 535bab52, eac000a8: hc= 274cbf3f, eac000b8: hc= 1f9fd541, eac000c8: hc= 3669ae9f
eac00038: hc= 772a3766, eac00048: hc= 749b46a8, eac00058: hc= 7e3bfb66, eac00068: hc= 13f62649, eac00078: hc= 054b8cdc, eac00088: hc= 230cc23b, eac00098: hc= 1aa3c177, eac000a8: hc= 74f2794a, eac000b8: hc= 5af92541, eac000c8: hc= 1afcfd10
eac00038: hc= 396e1dd8, eac00048: hc= 6c696d5c, eac00058: hc= 7d8aea9e, eac00068: hc= 2b316b76, eac00078: hc= 39862621, eac00088: hc= 16315e08, eac00098: hc= 03146a9a, eac000a8: hc= 3162a60a, eac000b8: hc= 4382f3da, eac000c8: hc= 4a578fd6
eac00038: hc= 225765b0, eac00048: hc= 17d5176d, eac00058: hc= 26f50154, eac00068: hc= 1f2a45c7, eac00078: hc= 104b1bcd, eac00088: hc= 330e3816, eac00098: hc= 6a844689, eac000a8: hc= 12330301, eac000b8: hc= 530a3ffc, eac000c8: hc= 45eee3fb
eac00038: hc= 3f9432e0, eac00048: hc= 1a9830bc, eac00058: hc= 7da79447, eac00068: hc= 04f801c4, eac00078: hc= 363bed68, eac00088: hc= 185f62a9, eac00098: hc= 1e4651bf, eac000a8: hc= 1aa0e220, eac000b8: hc= 385db088, eac000c8: hc= 0ef0cda1

作为旁注；如果你看这段代码

if (value == 0) value = 0xBAD ;

看起来 0xBAD 的可能性是任何 hashCode 的两倍，因为 0 映射到该值。如果你运行足够长的时间，你会看到

long count = 0, countBAD = 0;
while (true) {
    for (int i = 0; i < 200000000; i++) {
        int hc = new Object().hashCode();
        if (hc == 0xBAD)
            countBAD++;
        count++;
    }
    System.out.println("0xBAD ratio is " + (double) (countBAD << 32) / count + " times expected.");
}

印刷

0xBAD ratio is 2.0183116992481205 times expected.
4

4 回答 4

23

这显然是特定于实现的。

下面我包括Object.hashCode()OpenJDK 7 中使用的实现。

该函数支持六种不同的计算方法，其中只有两种注意到对象的地址（“地址”是 C++oop转换为intptr_t）。这两种方法中的一种按原样使用地址，而另一种则进行一些操作，然后将结果与不经常更新的随机数混合。

其余方法中，一种返回常量（可能用于测试），一种返回序列号，其余的基于伪随机序列。

似乎可以在运行时选择该方法，并且默认似乎是方法 0，即os::random(). 后者是一个线性同余生成器，其中包含所谓的竞争条件。:-) 竞争条件是可以接受的，因为在最坏的情况下它会导致两个对象共享相同的哈希码；这不会破坏任何不变量。

第一次需要哈希码时执行计算。为了保持一致性，结果随后存储在对象的标头中，并在后续调用时返回hashCode()。缓存在此函数之外完成。

总而言之，Object.hashCode()基于对象地址的概念在很大程度上是一种历史产物，已被现代垃圾收集器的特性所淘汰。

// hotspot/src/share/vm/runtime/synchronizer.hpp

// hashCode() generation :
//
// Possibilities:
// * MD5Digest of {obj,stwRandom}
// * CRC32 of {obj,stwRandom} or any linear-feedback shift register function.
// * A DES- or AES-style SBox[] mechanism
// * One of the Phi-based schemes, such as:
//   2654435761 = 2^32 * Phi (golden ratio)
//   HashCodeValue = ((uintptr_t(obj) >> 3) * 2654435761) ^ GVars.stwRandom ;
// * A variation of Marsaglia's shift-xor RNG scheme.
// * (obj ^ stwRandom) is appealing, but can result
//   in undesirable regularity in the hashCode values of adjacent objects
//   (objects allocated back-to-back, in particular).  This could potentially
//   result in hashtable collisions and reduced hashtable efficiency.
//   There are simple ways to "diffuse" the middle address bits over the
//   generated hashCode values:
//

static inline intptr_t get_next_hash(Thread * Self, oop obj) {
  intptr_t value = 0 ;
  if (hashCode == 0) {
     // This form uses an unguarded global Park-Miller RNG,
     // so it's possible for two threads to race and generate the same RNG.
     // On MP system we'll have lots of RW access to a global, so the
     // mechanism induces lots of coherency traffic.
     value = os::random() ;
  } else
  if (hashCode == 1) {
     // This variation has the property of being stable (idempotent)
     // between STW operations.  This can be useful in some of the 1-0
     // synchronization schemes.
     intptr_t addrBits = intptr_t(obj) >> 3 ;
     value = addrBits ^ (addrBits >> 5) ^ GVars.stwRandom ;
  } else
  if (hashCode == 2) {
     value = 1 ;            // for sensitivity testing
  } else
  if (hashCode == 3) {
     value = ++GVars.hcSequence ;
  } else
  if (hashCode == 4) {
     value = intptr_t(obj) ;
  } else {
     // Marsaglia's xor-shift scheme with thread-specific state
     // This is probably the best overall implementation -- we'll
     // likely make this the default in future releases.
     unsigned t = Self->_hashStateX ;
     t ^= (t << 11) ;
     Self->_hashStateX = Self->_hashStateY ;
     Self->_hashStateY = Self->_hashStateZ ;
     Self->_hashStateZ = Self->_hashStateW ;
     unsigned v = Self->_hashStateW ;
     v = (v ^ (v >> 19)) ^ (t ^ (t >> 8)) ;
     Self->_hashStateW = v ;
     value = v ;
  }

  value &= markOopDesc::hash_mask;
  if (value == 0) value = 0xBAD ;
  assert (value != markOopDesc::no_hash, "invariant") ;
  TEVENT (hashCode: GENERATE) ;
  return value;
}
于 2012-12-13T13:05:59.213 回答
5

它通常是对象的内存地址。但是，第一次在hashcode对象上调用该方法时，整数存储在该对象的标头中，以便下一次调用将返回相同的值（如您所说，压缩垃圾收集可以更改地址）。据我所知，这就是它在 Oracle JVM 中的实现方式。

编辑：深入研究 JVM 源代码，这就是显示的内容（synchronizer.cpp）：

// hashCode() generation :
//
// Possibilities:
// * MD5Digest of {obj,stwRandom}
// * CRC32 of {obj,stwRandom} or any linear-feedback shift register function.
// * A DES- or AES-style SBox[] mechanism
// * One of the Phi-based schemes, such as:
//   2654435761 = 2^32 * Phi (golden ratio)
//   HashCodeValue = ((uintptr_t(obj) >> 3) * 2654435761) ^ GVars.stwRandom ;
// * A variation of Marsaglia's shift-xor RNG scheme.
// * (obj ^ stwRandom) is appealing, but can result
//   in undesirable regularity in the hashCode values of adjacent objects
//   (objects allocated back-to-back, in particular).  This could potentially
//   result in hashtable collisions and reduced hashtable efficiency.
//   There are simple ways to "diffuse" the middle address bits over the
//   generated hashCode values:
//

static inline intptr_t get_next_hash(Thread * Self, oop obj) {
  intptr_t value = 0 ;
  if (hashCode == 0) {
     // This form uses an unguarded global Park-Miller RNG,
     // so it's possible for two threads to race and generate the same RNG.
     // On MP system we'll have lots of RW access to a global, so the
     // mechanism induces lots of coherency traffic.
     value = os::random() ;
  } else
  if (hashCode == 1) {
     // This variation has the property of being stable (idempotent)
     // between STW operations.  This can be useful in some of the 1-0
     // synchronization schemes.
     intptr_t addrBits = intptr_t(obj) >> 3 ;
     value = addrBits ^ (addrBits >> 5) ^ GVars.stwRandom ;
  } else
  if (hashCode == 2) {
     value = 1 ;            // for sensitivity testing
  } else
  if (hashCode == 3) {
     value = ++GVars.hcSequence ;
  } else
  if (hashCode == 4) {
     value = intptr_t(obj) ;
  } else {
     // Marsaglia's xor-shift scheme with thread-specific state
     // This is probably the best overall implementation -- we'll
     // likely make this the default in future releases.
     unsigned t = Self->_hashStateX ;
     t ^= (t << 11) ;
     Self->_hashStateX = Self->_hashStateY ;
     Self->_hashStateY = Self->_hashStateZ ;
     Self->_hashStateZ = Self->_hashStateW ;
     unsigned v = Self->_hashStateW ;
     v = (v ^ (v >> 19)) ^ (t ^ (t >> 8)) ;
     Self->_hashStateW = v ;
     value = v ;
  }

  value &= markOopDesc::hash_mask;
  if (value == 0) value = 0xBAD ;
  assert (value != markOopDesc::no_hash, "invariant") ;
  TEVENT (hashCode: GENERATE) ;
  return value;
}

所以在Oracle JVM中有6种不同的方法，其中一种相当于我说的……值通过方法调用存储在对象的头中get_next_hash（那个被调用FastHashCode并从native版本调用）的Object.hashCode()。

于 2012-12-13T12:52:35.383 回答
2

恕我直言，虽然依赖于实现，但 JVM 中的对象引用永远不是实际的内存地址，而是指向实际地址的内部 JVM 引用。这些内部引用最初是根据内存地址生成的，但它们仍然与对象相关联，直到它被丢弃。

我这样说是因为 Java HotSpot GC 正在复制某种形式的收集器，并且它们通过遍历并将它们从一个堆复制到另一个堆来工作，随后live objects销毁。old heap所以当 GC 发生时，JVM 不必更新所有对象中的引用，而只是改变映射到内部引用的实际内存地址。

于 2012-12-13T13:16:25.667 回答
0

Javadoc 对 Object.hashCode() 的建议是，它是通过从内存地址派生哈希码来实现的，这已经过时了。

可能没有人打扰（或注意到）复制垃圾收集器不再可能实现此实现路径（因为当对象被复制到另一个内存位置时它会更改哈希码）。

在复制垃圾收集器出现之前以这种方式实现哈希码是很有意义的，因为它可以节省堆空间。今天，非复制 GC (CMS) 可能仍以这种方式实现哈希码。

于 2012-12-13T15:44:12.597 回答