小白请教几个关于Java虚拟机内存分配策略的问题

最近在看周志明所著的《深入理解Java虚拟机》,有几个问题不太明白,希望对虚拟机有研究的哥们儿帮我解答一下。
先说一下我进行试验的环境:

操作系统:Mac OS X 10.11.6 EI Capitan

Java环境: java version "1.8.0_92"

Java(TM) SE Runtime Environment (build 1.8.0_92-b14)

Java HotSpot(TM) 64-Bit Server VM (build 25.92-b14, mixed mode)

1. 我发现在年轻代中,即使我只是一个空类,没有进行任何的内存分配工作,Eden区域内存还是会被占用一部分,例子如下:

public class BlankedClassSize {

  public static void main(String[] args) {

  // TODO Auto-generated method stub
  }
}

虚拟机参数设置:

-verbose:gc -XX:+UseSerialGC -XX:+PrintGCDetails -Xmx20M -Xms20M -Xmn10M

虚拟机GC情况如下:

Heap

def new generation  total 9216K, used1164K
[0x00000007bf000000, 0x00000007bfa00000, 0x00000007bfa00000)

eden space 8192K,  14% used [0x00000007bf000000, 0x00000007bf1231d0, 0x00000007bf800000)

from space 1024K,   0% used [0x00000007bf800000, 0x00000007bf800000, 0x00000007bf900000)

to   space 1024K,   0% used [0x00000007bf900000, 0x00000007bf900000, 0x00000007bfa00000)

tenured generation   total 6144K, used 0K [0x00000007bfa00000, 0x00000007c0000000, 0x00000007c0000000)

the space 6144K,   0% used [0x00000007bfa00000, 0x00000007bfa00000, 0x00000007bfa00200, 0x00000007c0000000)

Metaspace       used 2666K, capacity 4486K, committed 4864K, reserved 1056768K

class space    used 289K, capacity 386K, committed 512K, reserved 1048576K

问题1:这里是什么占用了内存?

2.当年轻代进行一次GC后,因为年轻代GC是采用复制算法收集内存的,那么正常情况下进行一次Minor GC后,Eden区域占存为什么不是零?是什么占用着内存?例子如下:

public class MinorGC {

    private static final int_1MB = 1024*1024;
    
    public static void main(String[] args) {

        // TODO Auto-generated method stub

        testEdenGC();

    }

    public static void testEdenGC(){

        byte[] allocation1,allocation2,allocation3,allocation4,allocation5,allocation6;

        allocation1=newbyte[_1MB];

        allocation2=newbyte[_1MB];

        allocation3=newbyte[_1MB];

        allocation4=newbyte[_1MB];

        allocation5=newbyte[_1MB];

        allocation6=newbyte[_1MB];

    }

}

虚拟机参数设置:

-verbose:gc -XX:TenuringThreshold=1 -Xms20M -Xmx20M -Xmn10M -XX:SurvivorRation=3 -XX:+PrintGCDetails -XX:+UseSerialGC

虚拟机GC情况:

[GC (Allocation Failure) DefNew: 6029K->1354K(8192K), 0.0032985 secs] 6029K->5450K(18432K), 0.0033216 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]

Heap

def new generation   total 8192K, used 2440K [0x00000007bec00000, 0x00000007bf600000, 0x00000007bf600000)

eden space 6144K,  17% used [0x00000007bec00000, 0x00000007bed0f748, 0x00000007bf200000)

from space 2048K,  66% used [0x00000007bf400000, 0x00000007bf5529b0, 0x00000007bf600000)

to   space 2048K,   0% used [0x00000007bf200000, 0x00000007bf200000, 0x00000007bf400000)

tenured generation   total 10240K, used 4096K [0x00000007bf600000, 0x00000007c0000000, 0x00000007c0000000)

the space 10240K,  40% used [0x00000007bf600000, 0x00000007bfa00040, 0x00000007bfa00200, 0x00000007c0000000)

Metaspace       used 2666K, capacity 4486K, committed 4864K, reserved 1056768K

class space    used 289K, capacity 386K, committed 512K, reserved 1048576K

问题2:根据6026K->1354K(8192K)可知,进行GC后,allocation1转入survivor1,allocation2~4转入老年区,此时eden中还有1354K内存被占用,是什么占用的?根据6029K->5450K(18432K),知道GC后,Java堆内存减少了579K,减少的这部分是谁的内存(明显新建的几个对象都可达,不至于被回收)?

3. 当给新对象分配内存的时候,因为年轻代空间不足,会发生一次minor GC,将年轻代中依然存活的对象复制到老年区,给新对象腾出空间,根据我的实验,此时若老年区内存不够将年轻代对象复制过去但是够新对象的存储,(此时会发生担保失败)则会直接将新对象放进老年代,例子如下:

public class MinorGC {

  private static  final int_1MB = 1024*1024;

  public static void main(String[] args) {

    // TODO Auto-generated method stub

    testAllocation();
  }

  public static void testAllocation(){

    byte[] allocation1,allocation2,allocation3,allocation4;

    allocation1 = newbyte[2*_1MB];

    allocation2 = newbyte[2*_1MB];

    allocation3 = newbyte[2*_1MB];

    allocation4 = newbyte[4*_1MB];

  }
}

 虚拟机参数设置如下:

-verbose:gc -Xms16M -Xmx16M -Xmn10M -XX:+PrintGCDetails  -XX:SurvivorRatio=8  -XX:+HeapDumpOnOutOfMemoryError -XX:MaxTenuringThreshold=15 -XX:+UseSerialGC

GC情况如下:

[GC (Allocation Failure) [DefNew (promotion failed) : 7144K->7474K(9216K), 0.0028055 secs][Tenured: 4096K->4096K(6144K), 0.0015627 secs] 7144K->6473K(15360K), [Metaspace: 2660K->2660K(1056768K)], 0.0044048 secs] [Times: user=0.00 sys=0.00, real=0.00 secs]

Heap

def new generation   total 9216K, used 6555K [0x00000007bf000000, 0x00000007bfa00000, 0x00000007bfa00000)

eden space 8192K,  80% used [0x00000007bf000000, 0x00000007bf666ed0, 0x00000007bf800000)

from space 1024K,   0% used [0x00000007bf900000, 0x00000007bf900000, 0x00000007bfa00000)

to   space 1024K,   0% used [0x00000007bf800000, 0x00000007bf800000, 0x00000007bf900000)

tenured generation   total 6144K, used 4096K[0x00000007bfa00000, 0x00000007c0000000, 0x00000007c0000000)

the space 6144K,  66% used [0x00000007bfa00000, 0x00000007bfe00020, 0x00000007bfe00200, 0x00000007c0000000)

Metaspace       used 2666K, capacity 4486K, committed 4864K, reserved 1056768K

class space    used 289K, capacity 386K, committed 512K, reserved 1048576K

根据上面输出结果,可以看出,GC过程中,年轻代7144K->;7144K(9216K),并没有将年轻代中的对象转移走,而老年代中4096K->4096K(6144K),tenured generation   total 6144K, used 4096K,说明allocation4直接被转移到老年代中了。

问题3:难道这种情况下虚拟机会自动设置大对象阈值??将新对象直接放入老年去?

以上三个问题让我这个虚拟机小白百思不得其解,还希望大家帮我解释一下,在此谢过。

posted @ 2016-10-05 19:33  001010  阅读(532)  评论(0编辑  收藏  举报