[PATCH 08/12] math: Optimize f{max, min}imum{f, l, f128} order sign comparison

[Adhemerval Zanella Netto]

On 21/01/26 10:13, Wilco Dijkstra wrote:
> Hi Adhemerval,
> 
> I would expect these to use the builtin like this (and use similar code as the
> magnitude variants otherwise):
> 
>  FLOAT
>  M_DECL_FUNC (__fmaximum) (FLOAT x, FLOAT y)
>  {
>   if (__glibc_likely (!isunordered (x, y)))
>     {
> #if M_USE_BUILTIN (FMAX)
>        return M_SUF (__builtin_fmax) (x, y);
> #else
>       if (__glibc_unlikely (x == y))
> 	return signbit (x) ? x : y;
>       return isgreater (x, y) ? x : y;
> #endif
>     }
> 
> Same for the other non-magnitude cases.

Indeed this is a good suggestion, I will add them.