c语言实现memcpy

今天到I 公司去面试，面试方式比较特殊，没有笔试，就是2 个面试官，一人一句轮番发问，涉及面很广，涉及到操作系统(MMU 、page out 、process/thread 、semaphore 、interrupt), OOP( 多态、design pattern) 、数据结构( 排序、二叉查找树) 、计算机网络(OSI 5 层) 、C 语言(big/small endian) 、英语口语等等，问了大约一个小时左右。

所有问题都是口头表述，只在纸上写了一个memcpy 程序，用C 语言实现，脑子一发蒙，既然写成了strcpy ，真该死。

回家了查询了一下memcpy 定义，如下：

Void *memcpy(void *dest, const void *src, unsigned int count);

查询msdn, 发现Remark 如下：

memcpy copies count bytes from src to dest ; If the source and destination overlap, the behavior of memcpy is undefined. Use memmove to handle overlapping regions.

以上描述针对dest 和 src 所指的内存地址有重叠的情况，内存地址重叠情况，memcpy 函数处理步骤未定，而memmove 对重叠情况给予处理；

在winXP+visual c++2005 测试 memcpy 函数，程序如下：

 

#include "stdafx.h"

#include <string.h>

int _tmain (int argc , _TCHAR * argv [])

{

       char s [16] = "aabbcc" ;

       char d [16] = {0};

      

       memcpy (s +2, s , 4);

       printf ("%s" , s );

       return 0;

}

结果输出 “aaaabb”, 由此可见windows 平台的c 运行时MSVCRT 的memcpy 函数对重叠部分做了处理，同memmove 的实现。//notes: 如果重叠部分不做处理，应该输出”aaaaaa”

下面我们用c 语言来实现memcpy 函数, 首先我们写出不对内存重叠的处理函数，如下：

void *memcpy_no_handle_overlap (void *dest , void *src , unsigned int count )

{

       if ((NULL ==dest ) || (NULL ==src ))

              return NULL ;

 

       char *d = (char *)dest ;

       char *s = (char *)src ;

 

       //Do normal (Upwards) Copy

       while (count -- > 0)

              *d ++ = *s ++;

       return dest ;

}

测试程序如下：

int _tmain(int argc, _TCHAR* argv[])

{

       char s[16] = "aabbcc";

       char d[16] = {0};

      

       memcpy_no_handle_overlap(s+2, s, 4);

 

       printf("%s", s);

       return 0;

}

输出结果”aaaaaa”

下面讨论处理memory overlapping 情况，如下图：

判断overlapping 条件如下：

If ( (dest <= src) ||                // green region 1

   (dest >=src+count) )           // green region 2

{

       // no memory overlapping

}

Else  // red region 3

{

       // there is overlapping

}

Overlapping 的处理：

我们可以看到memcpy_no_handle_overlap 函数，是从低地址依次赋值到高地址；在处理overlapping 时，如果我们采用同样的方法( 低地址到高地址) ，高地址的值将会被覆盖，所以我们应该从高地址依次到低地址赋值，如下图：

  函数代码如下:

void *memcpy_handle_overlap(void *dest, void *src, unsigned int count)

{

       if ((NULL==dest) || (NULL==src))

              return NULL;

 

       char *d = (char *)dest;

       char *s = (char *)src;

 

       //Check for overlapping buffers:

       if ( (d<=s) || (d>=s+count) )

       {     

              //Do normal (Upwards) Copy

              while (count-- > 0)

                     *d++ = *s++;

       }

       else

       {

              //Do Downwards Copy to avoid propagation

              while (count > 0)

              {

                     *(d+count-1) = *(s+count-1);

                     --count;

              }

 

       }

 

       return dest;

}

测试代码：

int _tmain(int argc, _TCHAR* argv[])

{

       char s[16] = "aabbcc";

       char d[16] = {0};

      

       memcpy_handle_overlap(s+2, s, 4);

 

       printf("%s", s);

       return 0;

}

输出结果为: “aaaabb “

 

最后测试代码如下：

int _tmain(int argc, _TCHAR* argv[])

{

       char s[16] = "aabbcc";

       memcpy_no_handle_overlap(s+2, s, 4);

       printf("memcpy(ignore memory overlapping): %s/n", s);

 

       strcpy(s, "aabbcc");

       memcpy_handle_overlap(s+2, s, 4);

       printf("memcpy(handle memory overlapping): %s/n", s);

 

     strcpy(s, "aabbcc");

       memcpy(s+2, s, 4);

       printf("memcpy( MSVCRT ): %s/n", s);

 

       strcpy(s, "aabbcc");

       memmove(s+2, s, 4);

       printf("memmove( MSVCRT): %s/n", s);

 

       return 0;

}

输出结果为：

memcpy(ignore memory overlapping): aaaaaa

memcpy(handle memory overlapping): aaaabb

memcpy( MSVCRT ): aaaabb

memmove( MSVCRT): aaaabb
————————————————
版权声明：本文为CSDN博主「wowRicky」的原创文章，遵循CC 4.0 BY-SA版权协议，转载请附上原文出处链接及本声明。
原文链接：https://blog.csdn.net/wowricky/article/details/6163767