myMemcpy

__forceinline   //因为通常Size已知，内联后编译器可以优化掉大部分无用代码
void* myMemcpy(char* Dst, const char* Src, size_t Size)
{
    void* start = Dst;
    for ( ; Size >= sizeof(__m256i); Size -= sizeof(__m256i) )
    {
        __m256i ymm = _mm256_loadu_si256(((const __m256i* &)Src)++);
        _mm256_storeu_si256(((__m256i* &)Dst)++, ymm);
    }

#define CPY_1B *((uint8_t * &)Dst)++ = *((const uint8_t * &)Src)++
#define CPY_2B *((uint16_t* &)Dst)++ = *((const uint16_t* &)Src)++
#define CPY_4B *((uint32_t* &)Dst)++ = *((const uint32_t* &)Src)++
#if defined _M_X64 || defined _M_IA64 || defined __amd64
#   define CPY_8B *((uint64_t* &)Dst)++ = *((const uint64_t* &)Src)++
#else
#   define CPY_8B _mm_storel_epi64((__m128i *)Dst, _mm_loadu_si128((const __m128i *)Src)), ++(const uint64_t* &)Src, ++(uint64_t* &)Dst
#endif
#define CPY16B _mm_storeu_si128((__m128i *)Dst, _mm_loadu_si128((const __m128i *)Src)), ++(const __m128i* &)Src, ++(__m128i* &)Dst

    switch (Size)
    {
    case 0x00:                                          break;
    case 0x01:  CPY_1B;                                 break;
    case 0x02:          CPY_2B;                         break;
    case 0x03:  CPY_1B; CPY_2B;                         break;
    case 0x04:                  CPY_4B;                 break;
    case 0x05:  CPY_1B;         CPY_4B;                 break;
    case 0x06:          CPY_2B; CPY_4B;                 break;
    case 0x07:  CPY_1B; CPY_2B; CPY_4B;                 break;
    case 0x08:                          CPY_8B;         break;
    case 0x09:  CPY_1B;                 CPY_8B;         break;
    case 0x0A:          CPY_2B;         CPY_8B;         break;
    case 0x0B:  CPY_1B; CPY_2B;         CPY_8B;         break;
    case 0x0C:                  CPY_4B; CPY_8B;         break;
    case 0x0D:  CPY_1B;         CPY_4B; CPY_8B;         break;
    case 0x0E:          CPY_2B; CPY_4B; CPY_8B;         break;
    case 0x0F:  CPY_1B; CPY_2B; CPY_4B; CPY_8B;         break;
    case 0x10:                                  CPY16B; break;
    case 0x11:  CPY_1B;                         CPY16B; break;
    case 0x12:          CPY_2B;                 CPY16B; break;
    case 0x13:  CPY_1B; CPY_2B;                 CPY16B; break;
    case 0x14:                  CPY_4B;         CPY16B; break;
    case 0x15:  CPY_1B;         CPY_4B;         CPY16B; break;
    case 0x16:          CPY_2B; CPY_4B;         CPY16B; break;
    case 0x17:  CPY_1B; CPY_2B; CPY_4B;         CPY16B; break;
    case 0x18:                          CPY_8B; CPY16B; break;
    case 0x19:  CPY_1B;                 CPY_8B; CPY16B; break;
    case 0x1A:          CPY_2B;         CPY_8B; CPY16B; break;
    case 0x1B:  CPY_1B; CPY_2B;         CPY_8B; CPY16B; break;
    case 0x1C:                  CPY_4B; CPY_8B; CPY16B; break;
    case 0x1D:  CPY_1B;         CPY_4B; CPY_8B; CPY16B; break;
    case 0x1E:          CPY_2B; CPY_4B; CPY_8B; CPY16B; break;
    case 0x1F:  CPY_1B; CPY_2B; CPY_4B; CPY_8B; CPY16B; break;
    }
#undef CPY_1B
#undef CPY_2B
#undef CPY_4B
#undef CPY_8B
#undef CPY16B
    return start;
}