有趣的代码

今天刷论坛时发现下面这段代码

i = 0x5f3759df - ( i >> 1 );               // what the fuck?

哈哈，刚看到这行时我也不禁说了一句：“what the fuck?”

完整的代码如下：

/*
** float q_rsqrt( float number )
*/
float Q_rsqrt( float number )
{
    long i;
    float x2, y;
    const float threehalfs = 1.5F;

    x2 = number * 0.5F;
    y  = number;
    i  = * ( long * ) &y;                       // evil floating point bit level hacking
    i  = 0x5f3759df - ( i >> 1 );               // what the fuck?
    y  = * ( float * ) &i;
    y  = y * ( threehalfs - ( x2 * y * y ) );   // 1st iteration
//  y  = y * ( threehalfs - ( x2 * y * y ) );   // 2nd iteration, this can be removed

#ifndef Q3_VM
#ifdef __linux__
    assert( !isnan(y) ); // bk010122 - FPE?
#endif
#endif
    return y;
}

这是《雷神之锤III》中的代码(quake3-1.32b/code/game/q_math.c)。是平方根倒数速算法。把浮点数当做整数来处理，还有这种操作？（what the fuck?）不禁感叹前人的智慧。

联想到以前读Redis源码时，里面也有有趣的（diao diao的）注释。

"don't play with this unless you benchmark! ... just believe me, it works"
（不要优化这段代码，除非你做了性能测试。…… 相信我）

完整的代码在这里。

#define HSIZE (1 << (HLOG))

/*
 * don't play with this unless you benchmark!
 * the data format is not dependent on the hash function.
 * the hash function might seem strange, just believe me,
 * it works ;)
 */
#ifndef FRST
# define FRST(p) (((p[0]) << 8) | p[1])
# define NEXT(v,p) (((v) << 8) | p[2])
# if ULTRA_FAST
#  define IDX(h) ((( h             >> (3*8 - HLOG)) - h  ) & (HSIZE - 1))
# elif VERY_FAST
#  define IDX(h) ((( h             >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
# else
#  define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
# endif
#endif
/*
 * IDX works because it is very similar to a multiplicative hash, e.g.
 * ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1))
 * the latter is also quite fast on newer CPUs, and compresses similarly.
 *
 * the next one is also quite good, albeit slow ;)
 * (int)(cos(h & 0xffffff) * 1e6)
 */

短短的几行代码，堪称智慧的结晶。这也是计算机有趣的地方。