Question

I have the following class:

public class SeqGenerator {

    int last = 0;
    volatile int lastVolatile = 0;

    public int getNext() {
        return last++;
    }

    public synchronized int getNextSync() {
        return last++;
    }

    public int getNextVolatile() {
        return lastVolatile++;
    }

    public void caller() {
        int i1 = getNext();
        int i2 = getNextSync();
        int i3 = getNextVolatile();
    }

}

When I look at the disassembled code I don’t see the difference between the representation of three methods getNext(), getNextSync() and getNextVolatile() .

public int getNext();
  Code:
   0:   aload_0
   1:   dup
   2:   getfield    #2; //Field last:I
   5:   dup_x1
   6:   iconst_1
   7:   iadd
   8:   putfield    #2; //Field last:I
   11:  ireturn

public synchronized int getNextSync();
  Code:
   0:   aload_0
   1:   dup
   2:   getfield    #2; //Field last:I
   5:   dup_x1
   6:   iconst_1
   7:   iadd
   8:   putfield    #2; //Field last:I
   11:  ireturn

public int getNextVolatile();
  Code:
   0:   aload_0
   1:   dup
   2:   getfield    #3; //Field lastVolatile:I
   5:   dup_x1
   6:   iconst_1
   7:   iadd
   8:   putfield    #3; //Field lastVolatile:I
   11:  ireturn

public void caller();
  Code:
   0:   aload_0
   1:   invokevirtual   #4; //Method getNext:()I
   4:   istore_1
   5:   aload_0
   6:   invokevirtual   #5; //Method getNextSync:()I
   9:   istore_2
   10:  aload_0
   11:  invokevirtual   #6; //Method getNextVolatile:()I
   14:  istore_3
   15:  return

How the JMV can distinguish between these methods?

The generated code is the same of these methods and also of their callers. How the JVM performs the synchronization?

Answer 1

The synchronized keyword applied to a method just sets the ACC_SYNCHRONIZED flag on that method definition, as defined in the JVM specification § 4.6 Methods. It won’t be visible in the actual bytecode of the method.

The JLS § 8.4.3.6 synchronized Methods discusses the similarity of defining a synchronized method and declaring a synchronized block that spans the whole method body (and using the same object to synchronize on): the effect is exactly the same, but they are represented differently in the .class file.

A similar effect happens with volatile fields: It simply sets the ACC_VOLATILE flag on the field (JVM § 4.5 Fields). The code that accesses the field uses the same bytecode, but acts slightly different.

Also please note that using only a volatile field here is not threadsafe, because x++ on a volatile field x is not atomic!