Java中volatile关键字实现原理

/** * @author 五月的仓颉http://www.cnblogs.com/xrq730/p/7048693.html
 */
public class LazySingleton {
  private static volatile LazySingleton instance = null;
  public static LazySingleton getInstance() {
    if (instance == null) {
      instance = new LazySingleton();
    }
    return instance;
  }
  public static void main(String[] args) {
    LazySingleton.getInstance();
  }
}

-server -Xcomp -XX:+UnlockDiagnosticVMOptions -XX:+PrintAssembly -XX:CompileCommand=compileonly,*LazySingleton.getInstance

Java HotSpot(TM) 64-Bit Server VM warning: PrintAssembly is enabled; turning on DebugNonSafepoints to gain additional output
CompilerOracle: compileonly *LazySingleton.getInstance
Loaded disassembler from D:\JDK\jre\bin\server\hsdis-amd64.dll
Decoding compiled method 0x0000000002931150:
Code:
Argument 0 is unknown.RIP: 0x29312a0 Code size: 0x00000108
[Disassembling for mach='amd64']
[Entry Point]
[Verified Entry Point]
[Constants]
 # {method} 'getInstance' '()Lorg/xrq/test/design/singleton/LazySingleton;' in 'org/xrq/test/design/singleton/LazySingleton'
 #      [sp+0x20] (sp of caller)
 0x00000000029312a0: mov   dword ptr [rsp+0ffffffffffffa000h],eax
 0x00000000029312a7: push  rbp
 0x00000000029312a8: sub   rsp,10h      ;*synchronization entry
                        ; - org.xrq.test.design.singleton.LazySingleton::getInstance@-1 (line 13)
 0x00000000029312ac: mov   r10,7ada9e428h  ;  {oop(a 'java/lang/Class' = 'org/xrq/test/design/singleton/LazySingleton')}
 0x00000000029312b6: mov   r11d,dword ptr [r10+58h]
                        ;*getstatic instance
                        ; - org.xrq.test.design.singleton.LazySingleton::getInstance@0 (line 13)
 0x00000000029312ba: test  r11d,r11d
 0x00000000029312bd: je   29312e0h
 0x00000000029312bf: mov   r10,7ada9e428h  ;  {oop(a 'java/lang/Class' = 'org/xrq/test/design/singleton/LazySingleton')}
 0x00000000029312c9: mov   r11d,dword ptr [r10+58h]
 0x00000000029312cd: mov   rax,r11
 0x00000000029312d0: shl   rax,3h      ;*getstatic instance
                        ; - org.xrq.test.design.singleton.LazySingleton::getInstance@16 (line 17)
 0x00000000029312d4: add   rsp,10h
 0x00000000029312d8: pop   rbp
 0x00000000029312d9: test  dword ptr [330000h],eax ;  {poll_return}
 0x00000000029312df: ret
 0x00000000029312e0: mov   rax,qword ptr [r15+60h]
 0x00000000029312e4: mov   r10,rax
 0x00000000029312e7: add   r10,10h
 0x00000000029312eb: cmp   r10,qword ptr [r15+70h]
 0x00000000029312ef: jnb   293135bh
 0x00000000029312f1: mov   qword ptr [r15+60h],r10
 0x00000000029312f5: prefetchnta byte ptr [r10+0c0h]
 0x00000000029312fd: mov   r11d,0e07d00b2h  ;  {oop('org/xrq/test/design/singleton/LazySingleton')}
 0x0000000002931303: mov   r10,qword ptr [r12+r11*8+0b0h]
 0x000000000293130b: mov   qword ptr [rax],r10
 0x000000000293130e: mov   dword ptr [rax+8h],0e07d00b2h
                        ;  {oop('org/xrq/test/design/singleton/LazySingleton')}
 0x0000000002931315: mov   dword ptr [rax+0ch],r12d
 0x0000000002931319: mov   rbp,rax      ;*new ; - org.xrq.test.design.singleton.LazySingleton::getInstance@6 (line 14)
 0x000000000293131c: mov   rdx,rbp
 0x000000000293131f: call  2907c60h     ; OopMap{rbp=Oop off=132}
                        ;*invokespecial <init>
                        ; - org.xrq.test.design.singleton.LazySingleton::getInstance@10 (line 14)
                        ;  {optimized virtual_call}
 0x0000000002931324: mov   r10,rbp
 0x0000000002931327: shr   r10,3h
 0x000000000293132b: mov   r11,7ada9e428h  ;  {oop(a 'java/lang/Class' = 'org/xrq/test/design/singleton/LazySingleton')}
 0x0000000002931335: mov   dword ptr [r11+58h],r10d
 0x0000000002931339: mov   r10,7ada9e428h  ;  {oop(a 'java/lang/Class' = 'org/xrq/test/design/singleton/LazySingleton')}
 0x0000000002931343: shr   r10,9h
 0x0000000002931347: mov   r11d,20b2000h
 0x000000000293134d: mov   byte ptr [r11+r10],r12l
 0x0000000002931351: lock add dword ptr [rsp],0h ;*putstatic instance
                        ; - org.xrq.test.design.singleton.LazySingleton::getInstance@13 (line 14)
 0x0000000002931356: jmp   29312bfh
 0x000000000293135b: mov   rdx,703e80590h  ;  {oop('org/xrq/test/design/singleton/LazySingleton')}
 0x0000000002931365: nop
 0x0000000002931367: call  292fbe0h     ; OopMap{off=204}
                        ;*new ; - org.xrq.test.design.singleton.LazySingleton::getInstance@6 (line 14)
                        ;  {runtime_call}
 0x000000000293136c: jmp   2931319h
 0x000000000293136e: mov   rdx,rax
 0x0000000002931371: jmp   2931376h
 0x0000000002931373: mov   rdx,rax      ;*new ; - org.xrq.test.design.singleton.LazySingleton::getInstance@6 (line 14)
 0x0000000002931376: add   rsp,10h
 0x000000000293137a: pop   rbp
 0x000000000293137b: jmp   2932b20h     ;  {runtime_call}
[Stub Code]
 0x0000000002931380: mov   rbx,0h      ;  {no_reloc}
 0x000000000293138a: jmp   293138ah     ;  {runtime_call}
[Exception Handler]
 0x000000000293138f: jmp   292fca0h     ;  {runtime_call}
[Deopt Handler Code]
 0x0000000002931394: call  2931399h
 0x0000000002931399: sub   qword ptr [rsp],5h
 0x000000000293139e: jmp   2909000h     ;  {runtime_call}
 0x00000000029313a3: hlt
 0x00000000029313a4: hlt
 0x00000000029313a5: hlt
 0x00000000029313a6: hlt
 0x00000000029313a7: hlt

0x0000000002931351: lock add dword ptr [rsp],0h ;*putstatic instance
                        ; - org.xrq.test.design.singleton.LazySingleton::getInstance@13 (line 14)

【1】位测试和修改指令（BTS、BTR、BTC）
【2】交换指令（XADD、CMPXCHG、CMPXCHG8B）
【3】自动假设有LOCK前缀的XCHG指令
【4】下列单操作数的算数和逻辑指令：INC、DEC、NOT、NEG
【5】下列双操作数的算数和逻辑指令：ADD、ADC、SUB、SBB、AND、OR、XOR
一个加锁的指令会保证对目标操作数所在的内存区域加锁，但是系统可能会将锁定区域解释得稍大一些。
软件应该使用相同的地址和操作数长度来访问信号量（用作处理器之间发送信号的共享内存）。例如，如果一个处理器使用一个字来访问信号量，其它处理器就不应该使用一个字节来访问这个信号量。
总线锁的完整性不收内存区域对齐的影响。加锁语义会一直持续，以满足更新整个操作数所需的总线周期个数。但是，建议加锁访问应该对齐在它们的自然边界上，以提升系统性能：
【1】任何8位访问的边界（加锁或不加锁）
【2】锁定的字访问的16位边界
【3】锁定的双字访问的32位边界
【4】锁定的四字访问的64位边界
对所有其它的内存操作和所有可见的外部事件来说，加锁的操作都是原子的。所有取指令和页表操作能够越过加锁的指令。加锁的指令可用于同步一个处理器写数据而另一个处理器读数据的操作。