选项 -farray-widen-compare

数组宽比较优化：支持使用宽数据类型对原数组指针（指向的数组元素为窄类型）解引用，达到一次比较多个元素的效果，从而提高程序性能。

针对如下代码场景：

#define my_min(x, y) ((x) < (y) ? (x) : (y))
uint32_t func (uint32_t n0, uint32_t n1, const uint32_t limit, const uint8_t * a, const uint8_t * b)
{
  uint32_t n = my_min(n0, n1);
  while (++n != limit)
    if (a[n] != b[n])
      break;
  return n;
}

添加选项后，可以优化为如下代码：

#define my_min(x, y) ((x) < (y) ? (x) : (y))
uint32_t func (uint32_t n0, uint32_t n1, const uint32_t limit, const uint8_t * a, const uint8_t * b)
{
  uint32_t n = my_min(n0, n1);
  for (++n; n + sizeof(uint64_t) <= limit; n += sizeof(uint64_t))
  {
    uint64_t k1 = *((uint64_t*)(a+n));
    uint64_t k2 = *((uint64_t*)(b+n));
    if(k1 != k2)
    {
       int lz = __builtin_ctzll(k1 ^ k2);
       n += lz/8;
       return n;
     }
    }
    for (;n != limit; ++n)
    {
      if (a[n] != b[n])
        break;
     }
     return n;
}