[PATCH] x86_64: Add log2 with FMA

[Paul Zimmermann]

       Hi H.J.,

> On Skylake, it improves log2 bench performance by:
> 
>         Before       After     Improvement
> max     208.779      63.827       69%
> min     9.977        6.55         34%
> mean    10.366       6.8191       34%

for what concerns accuracy, after a limited search, I get a small improvement
for the largest known error. It was 0.554114 ulp before (obtained for
x=0x1.0b53197bd66c8p+0), and I find 0.547517 ulp now (obtained for
x=0x1.140c8ac38f327p+0).

Paul