[PATCH 8/9] i386: Move hypot implementation to C

[Paul Zimmermann]

> > In fact we need to handle underflow:
> > 
> >   long double lx = x;
> >   long double ly = y;
> >   double r = sqrtl (lx * lx + ly * ly);
> 
> You probably also need a math_narrow_eval here to ensure the value checked 
> for overflow and returned has definitely been properly narrowed to double 
> first.

right, If x = y = 0x1.fffffffffffffp1023, then lx * lx + ly * ly will not
overflow as long double, nor sqrtl(.), but the conversion to double will.
This should probably be added as a test case (if not already).

Paul