天下事有难易乎?为之,则难者亦易矣;不为,则易者亦难矣。

HashCode() 的返回值到底是不是对象内存地址?

往事如烟 781次浏览 0个评论

点击上方“终码一生”,选择“加为星标”

第一时间获取技术干货!

1、基于OpenJDK 8

一直以为Java Object.hashCode()的结果就是通过对象的内存地址做相关运算得到的,但是无意在网上看到有相应的意见争论,故抽时间从源码层面验证了剖析了hashCode的默认计算方法。

先说结论:OpenJDK8 默认hashCode的计算方法是通过和当前线程有关的一个随机数+三个确定值,运用Marsaglia’s xorshift scheme随机数算法得到的一个随机数。和对象内存地址无关。

下面通过查找和分析OpenJDK8源码实现来一步步分析。

1. 查找java.lang.Object.hashCode()源码

public native int hashCode();

2. 导出Object的JNI头文件

切换到Object.class文件所在目录,执行 javah -jni java.lang.Object,得到java_lang_Object.h文件,文件内容如下:

/* DO NOT EDIT THIS FILE - it is machine generated */
#include <jni.h>
/* Header for class java_lang_Object */
#ifndef _Included_java_lang_Object
#define _Included_java_lang_Object
#ifdef __cplusplus
extern "C" {
#endif
/*
 * Class: java_lang_Object
 * Method: registerNatives
 * Signature: ()V
 */
JNIEXPORT void JNICALL Java_java_lang_Object_registerNatives
  (JNIEnv *, jclass);

/*
 * Class: java_lang_Object
 * Method: getClass
 * Signature: ()Ljava/lang/Class;
 */
JNIEXPORT jclass JNICALL Java_java_lang_Object_getClass
  (JNIEnv *, jobject);

/*
 * Class: java_lang_Object
 * Method: hashCode
 * Signature: ()I
 */
JNIEXPORT jint JNICALL Java_java_lang_Object_hashCode
  (JNIEnv *, jobject);

/*
 * Class: java_lang_Object
 * Method: clone
 * Signature: ()Ljava/lang/Object;
 */
JNIEXPORT jobject JNICALL Java_java_lang_Object_clone
  (JNIEnv *, jobject);

/*
 * Class: java_lang_Object
 * Method: notify
 * Signature: ()V
 */
JNIEXPORT void JNICALL Java_java_lang_Object_notify
  (JNIEnv *, jobject);

/*
 * Class: java_lang_Object
 * Method: notifyAll
 * Signature: ()V
 */
JNIEXPORT void JNICALL Java_java_lang_Object_notifyAll
  (JNIEnv *, jobject);

/*
 * Class: java_lang_Object
 * Method: wait
 * Signature: (J)V
 */
JNIEXPORT void JNICALL Java_java_lang_Object_wait
  (JNIEnv *, jobject, jlong);

#ifdef __cplusplus
}
#endif
#endif

3 . 查看Object的native方法实现

OpenJDK源码链接:

http://hg.openjdk.java.net/jdk8u/jdk8u/jdk/file/3462d04401ba/src/share/native/java/lang/Object.c

查看Object.c文件,可以看到hashCode()的方法被注册成由JVM_IHashCode方法指针来处理。

static JNINativeMethod methods[] = {
    {"hashCode", "()I", (void *)&JVM_IHashCode},//hashcode的方法指针JVM_IHashCode
    {"wait", "(J)V", (void *)&JVM_MonitorWait},
    {"notify", "()V", (void *)&JVM_MonitorNotify},
    {"notifyAll", "()V", (void *)&JVM_MonitorNotifyAll},
    {"clone", "()Ljava/lang/Object;", (void *)&JVM_Clone},
};

而JVM_IHashCode方法指针在 openjdk\hotspot\src\share\vm\prims\jvm.cpp中定义为:

JVM_ENTRY(jint, JVM_IHashCode(JNIEnv* env, jobject handle))
  JVMWrapper("JVM_IHashCode");
  // as implemented in the classic virtual machine; return 0 if object is NULL
  return handle == NULL ? 0 : ObjectSynchronizer::FastHashCode (THREAD, JNIHandles::resolve_non_null(handle)) ;
JVM_END

从而得知,真正计算获得hashCode的值是ObjectSynchronizer::FastHashCode

4 . ObjectSynchronizer::fashHashCode方法的实现

openjdk\hotspot\src\share\vm\runtime\synchronizer.cpp 找到其实现方法。

intptr_t ObjectSynchronizer::FastHashCode (Thread * Self, oop obj) {
  if (UseBiasedLocking) {
    if (obj->mark()->has_bias_pattern()) {
      // Box and unbox the raw reference just in case we cause a STW safepoint.
      Handle hobj (Self, obj) ;
      // Relaxing assertion for bug 6320749.
      assert (Universe::verify_in_progress() ||
              !SafepointSynchronize::is_at_safepoint(),
             "biases should not be seen by VM thread here");
      BiasedLocking::revoke_and_rebias(hobj, false, JavaThread::current());
      obj = hobj() ;
      assert(!obj->mark()->has_bias_pattern(), "biases should be revoked by now");
    }
  }

  // hashCode() is a heap mutator ...
  // Relaxing assertion for bug 6320749.
  assert (Universe::verify_in_progress() ||
          !SafepointSynchronize::is_at_safepoint(), "invariant") ;
  assert (Universe::verify_in_progress() ||
          Self->is_Java_thread() , "invariant") ;
  assert (Universe::verify_in_progress() ||
         ((JavaThread *)Self)->thread_state() != _thread_blocked, "invariant") ;

  ObjectMonitor* monitor = NULL;
  markOop temp, test;
  intptr_t hash;
  markOop mark = ReadStableMark (obj);

  // object should remain ineligible for biased locking
  assert (!mark->has_bias_pattern(), "invariant") ;

  if (mark->is_neutral()) {
    hash = mark->hash(); // this is a normal header
    if (hash) { // if it has hash, just return it
      return hash;
    }

    hash = get_next_hash(Self, obj); // allocate a new hash code
    temp = mark->copy_set_hash(hash); // merge the hash code into header
    // use (machine word version) atomic operation to install the hash
    test = (markOop) Atomic::cmpxchg_ptr(temp, obj->mark_addr(), mark);
    if (test == mark) {
      return hash;
    }

  } else if (mark->has_monitor()) {
    monitor = mark->monitor();
    temp = monitor->header();
    assert (temp->is_neutral(), "invariant") ;
    hash = temp->hash();
    if (hash) {
      return hash;
    }

    // Skip to the following code to reduce code size
  } else if (Self->is_lock_owned((address)mark->locker())) {
    temp = mark->displaced_mark_helper(); // this is a lightweight monitor owned
    assert (temp->is_neutral(), "invariant") ;
    hash = temp->hash(); // by current thread, check if the displaced
    if (hash) { // header contains hash code
      return hash;
    }
  }

  // Inflate the monitor to set hash code
  monitor = ObjectSynchronizer::inflate(Self, obj);
  // Load displaced header and check it has hash code
  mark = monitor->header();
  assert (mark->is_neutral(), "invariant") ;
  hash = mark->hash();
  if (hash == 0) {
    hash = get_next_hash(Self, obj);
    temp = mark->copy_set_hash(hash); // merge hash code into header
    assert (temp->is_neutral(), "invariant") ;
    test = (markOop) Atomic::cmpxchg_ptr(temp, monitor, mark);
    if (test != mark) {
      // The only update to the header in the monitor (outside GC)
      // is install the hash code. If someone add new usage of
      // displaced header, please update this code
      hash = test->hash();
      assert (test->is_neutral(), "invariant") ;
      assert (hash != 0, "Trivial unexpected object/monitor header usage.");
    }
  }

  return hash;
}

该方法中

// Load displaced header and check it has hash code
  mark = monitor->header();
  assert (mark->is_neutral(), "invariant") ;
  hash = mark->hash();
  if (hash == 0) {
    hash = get_next_hash(Self, obj);
...
}

对hash值真正进行了计算,查看get_next_hash方法源码

http://hg.openjdk.java.net/jdk8u/jdk8u/hotspot/file/87ee5ee27509/src/share/vm/runtime/synchronizer.cpp#l555

static inline intptr_t get_next_hash(Thread * Self, oop obj) {
  intptr_t value = 0 ;
  if (hashCode == 0) {
     value = os::random() ;
  } else
  if (hashCode == 1) {
     intptr_t addrBits = cast_from_oop<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 = cast_from_oop<intptr_t>(obj) ;
  } else {
     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;
}

对于OpenJDK8版本,其默认配置为:

product(intx, hashCode, 5, \

          "(Unstable) select hashCode generation algorithm") \

其对应的hashCode计算方案为:

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 ;

其中Thread->_hashStateX, Thread->_hashStateY, Thread->_hashStateZ, Thread->_hashStateW在

http://hg.openjdk.java.net/jdk8u/jdk8u/hotspot/file/87ee5ee27509/src/share/vm/runtime/thread.cpp#I263

有定义:

// thread-specific hashCode stream generator state - Marsaglia shift-xor form
_hashStateX = os::random() ;
_hashStateY = 842502087 ;
_hashStateZ = 0x8767 ; // (int)(3579807591LL & 0xffff) ;
_hashStateW = 273326509 ;

所以,JDK8 的默认hashCode的计算方法是通过和当前线程有关的一个随机数+三个确定值,运用Marsaglia’s xorshift scheme随机数算法得到的一个随机数。欢迎关注我们,公号终码一生。

xorshift是由George Marsaglia发现的一类伪随机数生成器,其通过移位和与或计算,能够在计算机上以极快的速度生成伪随机数序列。其算法的基本实现如下:

unsigned long xor128(){
static unsigned long x=123456789,y=362436069,z=521288629,w=88675123;
unsigned long t;
t=(xˆ(x<<11));x=y;y=z;z=w; return( w=(wˆ(w>>19))ˆ(tˆ(t>>8)) );

这就和上面计算hashCode的OpenJDK代码对应了起来。

5 . 其他几类hashCode计算方案:

  • hashCode == 0

此类方案返回一个Park-Miller伪随机数生成器生成的随机数

OpenJdk 6 &7的默认实现。

http://hg.openjdk.java.net/jdk7u/jdk7u/hotspot/file/5b9a416a5632/src/share/vm/runtime/globals.hpp#l1100

http://hg.openjdk.java.net/jdk6/jdk6/hotspot/file/5cec449cc409/src/share/vm/runtime/globals.hpp#l1128

if (hashCode == 0) {
     value = os::random() ;
}
  • hashCode == 1

此类方案将对象的内存地址,做移位运算后与一个随机数进行异或得到结果

if (hashCode == 1) {
     intptr_t addrBits = cast_from_oop<intptr_t>(obj) >> 3 ;
     value = addrBits ^ (addrBits >> 5) ^ GVars.stwRandom ;
}
  • hashCode == 2

此类方案返回固定的1

if (hashCode == 2) {
     value = 1 ;
}
  • hashCode == 3

此类方案返回一个自增序列的当前值

if (hashCode == 3) {
     value = ++GVars.hcSequence ;
}
  • hashCode == 4

此类方案返回当前对象的内存地址

if (hashCode == 4) {
     value = cast_from_oop<intptr_t>(obj) ;
}

可以通过在JVM启动参数中添加-XX:hashCode=4,改变默认的hashCode计算方式。

PS:防止找不到本篇文章,可以收藏点赞,方便翻阅查找哦。

 

转自公号:终码一生


ITZOO版权所有丨如未注明 , 均为原创丨转载请注明来自IT乐园 ->HashCode() 的返回值到底是不是对象内存地址?
发表我的评论
取消评论
表情 贴图 加粗 删除线 居中 斜体 签到

Hi,您需要填写昵称和邮箱!

  • 昵称 (必填)
  • 邮箱 (必填)
  • 网址