出处:
http://blog.csdn.net/lazyking/article/details/5319580
什么是placement new?
所谓placement new就是在用户指定的内存位置上构建新的对象,这个构建过程不需要额外分配内存,只需要调用对象的构造函数即可。
举例来说:
class foo{};
foo* pfoo = new foo;
pfoo指向的对象的地址你是不能决定的,因为new已经为你做了这些工作。第一步分配内存,第二步调用类的构造函数。
而placement new是怎么做的呢,说白了就是把原本new做的两步工作分开来。第一步你自己分配内存,第二步你调用类的构造函数在自己分配的内存上构建新的对象。
placement new的好处:
1)在已分配好的内存上进行对象的构建,构建速度快。
2)已分配好的内存可以反复利用,有效的避免内存碎片问题。
关于Placement new的标准用法,网上有一篇文章 讲得很清楚,我这里再转述一下。
<!--
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> class Foo
{
char cc;
float f;
public :
void print() { std::cout << " ADDR: " << this << std::endl; }
void set_f( float _f ) { std::cout << " set f val : " << _f << std::endl; f = _f; }
void get_f() { std::cout << " get f val : " << f << std::endl; }
};
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> class Foo
{
char cc;
float f;
public :
void print() { std::cout << " ADDR: " << this << std::endl; }
void set_f( float _f ) { std::cout << " set f val : " << _f << std::endl; f = _f; }
void get_f() { std::cout << " get f val : " << f << std::endl; }
};
1)分配内存
<!--
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> char * buff = new char [ sizeof (Foo) * N ];
memset( buff, 0 , sizeof (Foo) * N );
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> char * buff = new char [ sizeof (Foo) * N ];
memset( buff, 0 , sizeof (Foo) * N );
2)构建对象
<!--
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> Foo * pfoo = new (buff)Foo;
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> Foo * pfoo = new (buff)Foo;
3)使用对象
<!--
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> pfoo -> print();
pfoo -> set_f( 1.0f );
pfoo -> get_f();
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> pfoo -> print();
pfoo -> set_f( 1.0f );
pfoo -> get_f();
4)析构对象,显式的调用类的析构函数。
<!--
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> pfoo ->~ Foo();
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> pfoo ->~ Foo();
5)销毁内存
<!--
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> delete [] buff;
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> delete [] buff;
上面5个步骤是标准的placement new的使用方法。
对于buff这块内存可以反复使用,只要重复2)、3)、4)步骤即可。
placement new还可以解决的一个问题是建立带参数的构造函数对象数组。
代码示例如下:
<!--
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> 1 class CPong
2 {
3 public :
4 CPong( int m ) : v(m) { std::cout << " CPong ctor. " << std::endl; }
5
6 private :
7 int v;
8 };
9 char * pong = new char [ sizeof (CPong) * 10 ];
10 CPong * pp = (CPong * )pong;
11
12 for ( int i = 0 ; i < 10 ; ++ i )
13 {
14 new (pp + i)CPong(i);
15 }
16
17 for ( int j = 0 ; j < 10 ; ++ j )
18 {
19 pp[j]. ~ CPong();
20 }
21
22 delete [] pong;
Code highlighting produced by Actipro CodeHighlighter (freeware)
http://www.CodeHighlighter.com/
--> 1 class CPong
2 {
3 public :
4 CPong( int m ) : v(m) { std::cout << " CPong ctor. " << std::endl; }
5
6 private :
7 int v;
8 };
9 char * pong = new char [ sizeof (CPong) * 10 ];
10 CPong * pp = (CPong * )pong;
11
12 for ( int i = 0 ; i < 10 ; ++ i )
13 {
14 new (pp + i)CPong(i);
15 }
16
17 for ( int j = 0 ; j < 10 ; ++ j )
18 {
19 pp[j]. ~ CPong();
20 }
21
22 delete [] pong;