将不使用的对象置为NULL可以加快回收吗

     一直听说将不用的对象置为NULL,可以加迅被GC回收,但又听见另一种观点说置NULL不会被GC.倒底会不会加速呢,本人不太懂GC的机制,但也可以知道置NULL有没有用,写个代码测试一下就OK了.贴代码:

public class LoopObject
{
	public static void main(String[] args)
	{
		LoopObject lo;
		for (int i = 0; i < 100000; i++)
		{
			lo = new LoopObject();
		}

	}
}

javac LoopObject.java
为了公平,执行了五次运行命令,每执行一次,复制记录一下:
java -verbosegc LoopObject
结果如下:
[GC 896K->161K(5056K), 0.0013781 secs]
[GC 896K->161K(5056K), 0.0013644 secs]
[GC 896K->161K(5056K), 0.0013768 secs]
[GC 896K->161K(5056K), 0.0013536 secs]
[GC 896K->161K(5056K), 0.0013715 secs]


被置为空后的代码:

public class LoopObject
{
	public static void main(String[] args)
	{
		LoopObject lo;
		for (int i = 0; i < 100000; i++)
		{
			lo = new LoopObject();
			lo = null;
		}
	}
}

javac LoopObject.java
为了公平,执行了五次运行命令,每执行一次,复制记录一下:
java -verbosegc LoopObject
结果如下:
[GC 896K->161K(5056K), 0.0013662 secs]
[GC 896K->161K(5056K), 0.0013609 secs]
[GC 896K->161K(5056K), 0.0013539 secs]
[GC 896K->161K(5056K), 0.0013664 secs]
[GC 896K->161K(5056K), 0.0014042 secs]

结果对比,用前段代码的GC回收时间减去后一段代码的回收时间:
0.0000119、0.0000035、0.0000229、-0.0000128、-0.0000327
从结果上看，感觉差不多。。。。。。可能是例子太简单的，那么再写个复杂的代码

import java.util.Hashtable;

public class LoopObject
{
	public static void main(String[] args)
	{
		Hashtable<Object, Object> ht;
		for (int i = 0; i < 100000; i++)
		{
			ht = new Hashtable<Object, Object>();
		}
	}
}

javac LoopObject.java
执行了三次运行命令,每执行一次,复制记录一下:
java -verbosegc LoopObject
结果如下:
[GC 896K->161K(5056K), 0.0013631 secs]
[GC 1057K->161K(5056K), 0.0004871 secs]
[GC 1057K->161K(5056K), 0.0002996 secs]
[GC 1057K->161K(5056K), 0.0001650 secs]
[GC 1057K->161K(5056K), 0.0001509 secs]
[GC 1057K->161K(5056K), 0.0001557 secs]
[GC 1057K->161K(5056K), 0.0001597 secs]
[GC 1057K->161K(5056K), 0.0001625 secs]
[GC 1057K->161K(5056K), 0.0002215 secs]
[GC 1057K->161K(5056K), 0.0001568 secs]

[GC 896K->161K(5056K), 0.0013520 secs]
[GC 1057K->161K(5056K), 0.0005030 secs]
[GC 1057K->161K(5056K), 0.0003084 secs]
[GC 1057K->161K(5056K), 0.0001522 secs]
[GC 1057K->161K(5056K), 0.0001523 secs]
[GC 1057K->161K(5056K), 0.0001546 secs]
[GC 1057K->161K(5056K), 0.0001626 secs]
[GC 1057K->161K(5056K), 0.0001473 secs]
[GC 1057K->161K(5056K), 0.0002252 secs]
[GC 1057K->161K(5056K), 0.0001502 secs]

[GC 896K->161K(5056K), 0.0013446 secs]
[GC 1057K->161K(5056K), 0.0004839 secs]
[GC 1057K->161K(5056K), 0.0003243 secs]
[GC 1057K->161K(5056K), 0.0001527 secs]
[GC 1057K->161K(5056K), 0.0001585 secs]
[GC 1057K->161K(5056K), 0.0001537 secs]
[GC 1057K->161K(5056K), 0.0001669 secs]
[GC 1057K->161K(5056K), 0.0002167 secs]
[GC 1057K->161K(5056K), 0.0001568 secs]
[GC 1057K->161K(5056K), 0.0001760 secs]

import java.util.Hashtable;

public class LoopObject
{
	public static void main(String[] args)
	{
		Hashtable<Object, Object> ht;
		for (int i = 0; i < 100000; i++)
		{
			ht = new Hashtable<Object, Object>();
			ht = null;
		}
	}
}

重新编译后并执行
javac LoopObject.java
执行了三次运行命令,每执行一次,复制记录一下:
java -verbosegc LoopObject

[GC 896K->161K(5056K), 0.0013605 secs]
[GC 1057K->161K(5056K), 0.0004848 secs]
[GC 1057K->161K(5056K), 0.0003503 secs]
[GC 1057K->161K(5056K), 0.0001571 secs]
[GC 1057K->161K(5056K), 0.0001541 secs]
[GC 1057K->161K(5056K), 0.0001685 secs]
[GC 1057K->161K(5056K), 0.0001570 secs]
[GC 1057K->161K(5056K), 0.0001546 secs]
[GC 1057K->161K(5056K), 0.0001561 secs]
[GC 1057K->161K(5056K), 0.0001341 secs]

[GC 896K->161K(5056K), 0.0014766 secs]
[GC 1057K->161K(5056K), 0.0004738 secs]
[GC 1057K->161K(5056K), 0.0003499 secs]
[GC 1057K->161K(5056K), 0.0001562 secs]
[GC 1057K->161K(5056K), 0.0001534 secs]
[GC 1057K->161K(5056K), 0.0001450 secs]
[GC 1057K->161K(5056K), 0.0001547 secs]
[GC 1057K->161K(5056K), 0.0001607 secs]
[GC 1057K->161K(5056K), 0.0001585 secs]
[GC 1057K->161K(5056K), 0.0001300 secs]

[GC 896K->161K(5056K), 0.0013647 secs]
[GC 1057K->161K(5056K), 0.0004726 secs]
[GC 1057K->161K(5056K), 0.0003491 secs]
[GC 1057K->161K(5056K), 0.0001531 secs]
[GC 1057K->161K(5056K), 0.0001437 secs]
[GC 1057K->161K(5056K), 0.0001537 secs]
[GC 1057K->161K(5056K), 0.0001530 secs]
[GC 1057K->161K(5056K), 0.0001282 secs]
[GC 1057K->161K(5056K), 0.0001714 secs]
[GC 1057K->161K(5056K), 0.0001593 secs]

从这次的结果中不难看出，多加了一行置空的代码，还真有点效果，只是不知道这些效果对系统性能有没有什么提高呢

评论

1 楼 yuejingjiahong 2013-01-17  

在搞并发下是有的