Re: [math-fun] 32-bit reversal speed tests

10 Apr 2014

      * Joerg Arndt <arndt@jjj.de> [Apr 10. 2014 16:28]:
...
[...]
...
129 (thanks for asking) goes by [ 1, 63, 1, 63, 1 ],
again 6 steps.
Nope, 129 takes only 5, via  [ 43 43 43 ], where
43 takes 4 via  [ 21 1 21 ], where
21 takes 3 via [ 9 3 9 ].

Surprisingly, the first deviation from the OEIS sequences
occurs quite a bit earlier than I thought, ...

   *** drum roll ***

uint revbin17(uint x)
{
    uint z;
    z = (x ^ (x >>  2)) & 0b00100010010010001;  x ^= z ^ (z <<  2);
    z = (x ^ (x >>  4)) & 0b00001110000000111;  x ^= z ^ (z <<  4);
    z = (x ^ (x >> 10)) & 0b00000000001111111;  x ^= z ^ (z << 10);
    return x;
}

Best,   jj
...
[...]