一天一个CRT函数 memcpy

人快不行了，最近连续性失眠，憔悴啊。这样透支生命，还没活到30岁就会没命的！我还要一直写C++程序呢~~

4.memcpy/wmemcpy

现在M$已经不推荐使用此函数，取而代之的是memcpy_s, wmemcpy_s.这样就解决了缓冲区溢出问题。安全性也就提高了，M$有那么多攻击它的人，能不防备么？作为使用库的程序员而言，这也带来了无尽的快乐！来看看实现吧

   1: /****

   2: T* memcpy(T *pDest, T *pSrc, unsigned int uLen)

   3: 

   4: Purpose:

   5:       把pSrc赋值到pDest缓冲区中，如果pDest和pSrc内存重叠，则行为未定义

   6: 

   7: Entry:

   8:        T *pDest - pDest 

   9:        T *pSrc - pSrc

  10:        unsigned int uLen - max number of bytes to search

  11: 

  12: Exit:

  13:        pDest叠加后的地址

  14: 

  15: Uses:

  16: 

  17: Exceptions:

  18: 

  19: ***/

  20: template<typename T>

  21: inline T* tMemCpy(T *pDest, T *pSrc, unsigned int uLen)

  22: {

  23:     while( uLen )

  24:     {

  25:         *pDest++ = *pSrc++;

  26:         --uLen;

  27:     }

  28:  

  29:     return pDest;

  30: }

  31:  

  32: template<typename T>

  33: inline int tMemcpyS(T *pDest, unsigned int uSizeInBytes, T *pSrc, unsigned int uCnt)

  34: {

  35:     if( uCnt == 0 )

  36:     { 

  37:         /* nothing to do */

  38:         return 0;

  39:     }

  40:  

  41:     /* validation section */

  42:     _ASSERT(pDest != NULL);

  43:     if( pSrc == NULL || uSizeInBytes < uCnt )

  44:     {

  45:         _ASSERT(pSrc != NULL);

  46:         _ASSERT(uSizeInBytes >= uCnt);

  47:  

  48:         /* useless, but prefast is confused */

  49:         return -1;

  50:     }

  51:  

  52:     tMemCpy(pDest, pSrc, uCnt);

  53:     return 1;

  54: }

看到了吧，其实memcpy_s也是利用memcpy实现的，只是对缓冲区进行了有效性检查。当然memcpy是没有考虑内存重叠的，而memmove倒是一个不错的选择

测试

   1: volatile tChar str1[] = _T("aabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbccaabbcc");

   2: volatile tChar str2[1024] = {0};

   3:  

   4: volatile const DWORD dwCount = 10000000;

   5:  

   6:  

   7: DWORD dwLast = 0;

   8: {

   9:     CCYPerformance timer(dwLast);

  10:     for(DWORD i = 0; i < dwCount; ++i)

  11:     {

  12:         CY_CRT::tMemCpy(str2, str1, sizeof(str1) / sizeof(str1[0]));

  13:     }

  14: }

  15: cout << dwLast << endl;

  16:  

  17: dwLast = 0;

  18: {

  19:     CCYPerformance timer(dwLast);

  20:     for(DWORD i = 0; i < dwCount; ++i)

  21:     {

  22:         ::memcpy((void *)str2, (const void *)str1, sizeof(str1) / sizeof(str1[0]));

  23:     }

  24: }

  25: cout << dwLast << endl;

对这个差距来说，汇编确实强大~呵呵，没有可比性！