最快StringToLowerCase()

Question

我想优化以下代码：

#ifdef USE_SSE
#define STRING_PREFETCH_TBL(ptr)            \
        _mm_prefetch(ptr, _MM_HINT_T0);     \
        _mm_prefetch(ptr+64, _MM_HINT_T0);  \
        _mm_prefetch(ptr+128, _MM_HINT_T0); \
        _mm_prefetch(ptr+192, _MM_HINT_T0)
#else

#define STRING_PREFETCH_TBL(ptr)
#endif


__declspec(align(128)) const char TblToLower[] = 
{
    0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,
    35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,97,98,
    99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,91,
    92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,
    119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,
    144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,
    170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,
    197,198,199,200,201,202,203,204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219,220,221,222,223,224,
    225,226,227,228,229,230,231,232,233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254,255,

};

void StrToLowerCase( const char* src, char* dst )
{
    STRING_PREFETCH_TBL(TblToLower);

    while(*src)
    {
        *dst = TblToLower[*src];
        dst++;
        src++;
    }
    *dst = '\0';
}

Answer 1

这是Visual 2012在没有SSE支持的情况下为我生成的/O2：

_StrToLowerCase:
  00000000: 8B 54 24 04        mov         edx,dword ptr [esp+4]
  00000004: 8B 44 24 08        mov         eax,dword ptr [esp+8]
  00000008: 8A 0A              mov         cl,byte ptr [edx]
  0000000A: 84 C9              test        cl,cl
  0000000C: 74 19              je          00000027
  0000000E: 2B D0              sub         edx,eax
  00000010: 0F BE C9           movsx       ecx,cl
  00000013: 8D 40 01           lea         eax,[eax+1]
  00000016: 0F B6 89 00 00 00  movzx       ecx,byte ptr _TblToLower[ecx]
            00
  0000001D: 88 48 FF           mov         byte ptr [eax-1],cl
  00000020: 8A 0C 02           mov         cl,byte ptr [edx+eax]
  00000023: 84 C9              test        cl,cl
  00000025: 75 E9              jne         00000010
  00000027: C6 00 00           mov         byte ptr [eax],0
  0000002A: C3                 ret

...and这是与SSE一起使用的：

_StrToLowerCase:
  00000000: 8B 54 24 04        mov         edx,dword ptr [esp+4]
  00000004: B8 00 00 00 00     mov         eax,offset _TblToLower
  00000009: 8A 0A              mov         cl,byte ptr [edx]
  0000000B: 0F 18 08           prefetcht0  [eax]
  0000000E: B8 40 00 00 00     mov         eax,offset _TblToLower+40h
  00000013: 0F 18 08           prefetcht0  [eax]
  00000016: B8 80 00 00 00     mov         eax,offset _TblToLower+80h
  0000001B: 0F 18 08           prefetcht0  [eax]
  0000001E: B8 C0 00 00 00     mov         eax,offset _TblToLower+0C0h
  00000023: 0F 18 08           prefetcht0  [eax]
  00000026: 8B 44 24 08        mov         eax,dword ptr [esp+8]
  0000002A: 84 C9              test        cl,cl
  0000002C: 74 19              je          00000047
  0000002E: 2B D0              sub         edx,eax
  00000030: 0F BE C9           movsx       ecx,cl
  00000033: 8D 40 01           lea         eax,[eax+1]
  00000036: 0F B6 89 00 00 00  movzx       ecx,byte ptr _TblToLower[ecx]
            00
  0000003D: 88 48 FF           mov         byte ptr [eax-1],cl
  00000040: 8A 0C 02           mov         cl,byte ptr [edx+eax]
  00000043: 84 C9              test        cl,cl
  00000045: 75 E9              jne         00000030
  00000047: C6 00 00           mov         byte ptr [eax],0
  0000004A: C3                 ret

对我来说已经很理想了。所以我试着：

1. 内联函数以避免函数调用的开销
2. 研究给函数的输入字符串；如果您发现有许多重复出现的字符串(和长字符串)，您可能需要考虑缓存结果。

作为最下面的说明，我可能会将函数的主循环编写为

while ( *src ) *dst++ = TblToLower[*src++];

..and将TblToLower表声明为函数内部的static，以提高(read: reduce)的可见性范围。也许这甚至提高了缓存的局部性？