字节对齐

【C语言】字节对齐(内存对齐)

数据对齐 

1)平台原因(移植原因):不是所有的硬件平台都能访问任意地址上的任意数据,某些硬件平台只能在某些地址处取某些特定类型的数据,否则抛出硬件异常
2)硬件原因:经过内存对齐之后,CPU的内存访问速度大大提升。


1.  对齐原则:

【原则1】数据成员对齐规则:结构(struct)(或联合(union))的数据成员,第一个数据成员放在offset为0的地方,以后每个数据成员的对齐按照#pragma pack指定的数值和这个数据成员自身长度中,比较小的那个进行。

【原则2】结构(或联合)的整体对齐规则:在数据成员完成各自对齐之后,结构(或联合)本身也要进行对齐,对齐将按照#pragma pack指定的数值和结构(或联合)最大数据成员长度中,比较小的那个进行。

【原则3】结构体作为成员:如果一个结构里有某些结构体成员,则结构体成员要从其内部最大元素大小的整数倍地址开始存储。

备注:数组成员按长度按数组类型长度计算,如char t[9],在第1步中数据自身长度按1算,累加结构体时长度为9;第2步中,找最大数据长度时,如果结构体T有复杂类型成员A,该A成员的长度为该复杂类型成员A的最大成员长度。

小结:当#pragma pack的n值等于或超过所有数据成员长度的时候,这个n值的大小将不产生任何效果。


 【注意】(对齐位数跟处理器位数和编译器都有关)VS, VC等编译器默认是#pragma pack(8),所以测试我们的规则会正常;注意gcc默认是#pragma pack(4),并且gcc只支持1,2,4对齐。套用三原则里计算的对齐值是不能大于#pragma pack指定的n值。

2. 自然对齐存放变量的地址要是该变量数据类型大小的整数倍。如:存放int型数据的地址一定要是4的倍数,存放short型数据的地址一定要是2的倍数。

 

3. 改变缺省的对界条件(指定对界):

  • 使用伪指令#pragma pack (n),C编译器将按照n个字节对齐。
  • 使用伪指令#pragma pack(),取消自定义字节对齐方式。

举例一

例1:

1

2

3

4

5

6

7

8

9

10

#pragma pack(1)

struct AA

{

    int a;   //长度4 < 1 按1对齐;偏移量为0;存放位置区间[0,3]

    char b;  //长度1 = 1 按1对齐;偏移量为4;存放位置区间[4]

    short c; //长度2 > 1 按1对齐;偏移量为5;存放位置区间[5,6]

    char d;  //长度1 = 1 按1对齐;偏移量为6;存放位置区间[7]

    //整体存放在[0~7]位置区间中,共八个字节。

};

#pragma pack()

 结果:8个字节

 

例2:

1

2

3

4

5

6

7

8

9

#pragma pack(2)

struct AA

{

    int a;   //长度4 > 2 按2对齐;偏移量为0;存放位置区间[0,3]

    char b;  //长度1 < 2 按1对齐;偏移量为4;存放位置区间[4]

    short c; //长度2 = 2 按2对齐;偏移量要提升到2的倍数6;存放位置区间[6,7]

    char d;  //长度1 < 2 按1对齐;偏移量为7;存放位置区间[8];共九个字节

};

#pragma pack()

 结果:10个字节

 

例3:

1

2

3

4

5

6

7

8

9

#pragma pack(4)

struct AA

{

    int a;   //长度4 = 4 按4对齐;偏移量为0;存放位置区间[0,3]

    char b;  //长度1 < 4 按1对齐;偏移量为4;存放位置区间[4]

    short c; //长度2 < 4 按2对齐;偏移量要提升到2的倍数6;存放位置区间[6,7]

    char d;  //长度1 < 4 按1对齐;偏移量为7;存放位置区间[8];总大小为9

};

#pragma pack()

 结果:12个字节

 

例4:

1

2

3

4

5

6

7

8

9

#pragma pack(8)

struct AA

{

    int a;   //长度4 < 8 按4对齐;偏移量为0;存放位置区间[0,3]

    char b;  //长度1 < 8 按1对齐;偏移量为4;存放位置区间[4]

    short c; //长度2 < 8 按2对齐;偏移量要提升到2的倍数6;存放位置区间[6,7]

    char d;  //长度1 < 8 按1对齐;偏移量为7;存放位置区间[8],总大小为9

};

#pragma pack()

 结果:12个字节  

 

例5:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

struct EE  //8个字节对齐

{

    int a;      //长度4 < 8 按4对齐;偏移量为0;存放位置区间[0,3]

    char b;     //长度1 < 8 按1对齐;偏移量为4;存放位置区间[4]

    short c;    //长度2 < 8 按2对齐;偏移量由5提升到6;存放位置区间[6,7]

 

    struct FF   //结构体内部最大元素为int,由于偏移量为8刚好是4的整数倍,所以从8开始存放接下来的struct FF

    {

        int a1;     //长度4 < 8 按4对齐;偏移量为8;存放位置区间[8,11]

        char b1;    //长度1 < 8 按1对齐;偏移量为12;存放位置区间[12]

        short c1;   //长度2 < 8 按2对齐;偏移量为13,提升到2的倍数14;存放位置区间[14,15]

        char d1;    //长度1 < 8 按1对齐;偏移量为16;存放位置区间[16]

    };              //整体对齐系数 = min((max(int,short,char), 8) = 4,将内存大小由17补齐到4的整数倍20

     

    char d;         //长度1 < 8 按1对齐;偏移量为21;存放位置区间[21]

                    //整体对齐系数 = min((max(int,short,char), 8) = 4,将内存大小由21补齐到4的整数倍24

};

 

1

2

3

4

5

6

7

8

9

10

11

12

13

struct B

{

    char e[2];      //长度1 < 8 按2对齐;偏移量为0;存放位置区间[0,1]

    short h;        //长度2 < 8 按2对齐;偏移量为2;存放位置区间[2,3]

                    //结构体内部最大元素为double,偏移量为4,提升到8,所以从8开始存放接下来的struct A

    struct A

    {

        int a;      //长度4 < 8 按4对齐;偏移量为8;存放位置区间[8,11]

        double b;   //长度8 = 8 按8对齐;偏移量为12,提升到16;存放位置区间16,23]

        float c;    //长度4 < 8,按4对齐;偏移量为24,存放位置区间[24,27]

    };

    //整体对齐系数 = min((max(int,double,float), 8) = 8,将内存大小由28补齐到8的整数倍32

};

举例二

代码1:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

#include <stdio.h>

typedef struct

{

    int aa1;   //4个字节对齐 1111 

    char bb1;  //1个字节对齐 1 

    short cc1; //2个字节对齐 011 

    char dd1;  //1个字节对齐 1 

}testlength1;

int length1 = sizeof(testlength1); //4个字节对齐,占用字节1111 1011 1000,length = 12

 

typedef struct

{

    char bb2;  //1个字节对齐 1 

    int aa2;   //4个字节对齐 01111 

    short cc2; //2个字节对齐 11 

    char dd2;  //1个字节对齐 1 

}

testlength2;

int length2 = sizeof(testlength2); //4个字节对齐,占用字节1011  1111 1000,length = 12

 

typedef struct

{

    char bb3;   //1个字节对齐 1 

    char dd3;   //1个字节对齐 1 

    int aa3;    //4个字节对齐 001111 

    short cc23; //2个字节对齐 11 

}testlength3;

int length3 = sizeof(testlength3); //4个字节对齐,占用字节1100 1111 1100,length = 12

 

typedef struct

{

    char bb4;  //1个字节对齐 1 

    char dd4;  //1个字节对齐 1 

    short cc4; //2个字节对齐 11 

    int aa4;   //4个字节对齐 1111 

} testlength4;

int length4 = sizeof(testlength4); //4个字节对齐,占用字节1111 1111,length = 8

 

int main(void)

{

    printf("length1 = %d.\n",length1);

    printf("length2 = %d.\n", length2);

    printf("length3 = %d.\n", length3);

    printf("length4 = %d.\n", length4);

    return 0;

} 

VS2017输出结果:

 

 代码2:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

#include <stdio.h>

#pragma pack(2)

 

typedef struct

{

    int aa1;   //2个字节对齐 1111 

    char bb1;  //1个字节对齐 1 

    short cc1; //2个字节对齐 011 

    char dd1;  //1个字节对齐 1 

} testlength1;

int length1 = sizeof(testlength1); //2个字节对齐,占用字节11 11 10 11 10,length = 10

 

typedef struct

{

    char bb2;    //1个字节对齐 1

    int aa2;     //2个字节对齐 01111 

    short cc2;   //2个字节对齐 11 

    char dd2;    //1个字节对齐 1 

} testlength2;

int length2 = sizeof(testlength2); //2个字节对齐,占用字节10 11 11 11 10,length = 10

 

typedef struct

{

    char bb3;   //1个字节对齐 1 

    char dd3;   //1个字节对齐 1 

    int aa3;    //2个字节对齐 11 11 

    short cc23; //2个字节对齐 11 

}testlength3;

int length3 = sizeof(testlength3); //2个字节对齐,占用字节11 11 11 11,length = 8

 

typedef struct

{

    char bb4;  //1个字节对齐 1 

    char dd4;  //1个字节对齐 1 

    short cc4; //2个字节对齐 11 

    int aa4;   //2个字节对齐 11 11

}testlength4;

int length4 = sizeof(testlength4); //2个字节对齐,占用字节11 11 11 11,length = 8

 

int main(void)

{

    printf("length1 = %d.\n", length1);

    printf("length2 = %d.\n", length2);

    printf("length3 = %d.\n", length3);

    printf("length4 = %d.\n", length4);

    return 0;

}

VS2017输出结果:

  

代码3:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

#include<iostream>

using namespace std;

 

typedef struct bb

{

    int id;             //[0]....[3]

    double weight;      //[8].....[15]      原则1

    float height;       //[16]..[19],总长要为8的整数倍,补齐[20]...[23]     原则3

}BB;

 

typedef struct aa

{

    char name[2];    //[0],[1]

    int  id;         //[4]...[7]          原则1

    double score;    //[8]....[15]    

    short grade;     //[16],[17]        

    BB b;            //[24]......[47]          原则2

}AA;

 

int main()

{

    AA a;

    cout << sizeof(a) << " " << sizeof(BB) << endl;

    return 0;

}

VS2017输出结果: 48 24

 

代码4:

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

#include<iostream>

using namespace std;

 

#pragma pack(2)

typedef struct bb

{

    int id;            

    double weight;     

    float height;      

}BB;

 

typedef struct aa

{

    char name[2];   

    int  id;       

    double score;   

    short grade;         

    BB b;          

}AA;

 

int main()

{

    AA a;

    cout << sizeof(a) << " " << sizeof(BB) << endl;

    return 0;

}

VS2017输出结果:32 16  

 

参考资料

1. 5分钟搞定内存字节对齐
2.  快速理解字节对齐问题
3.  关于面试题中结构体内存对齐计算总结

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值