Re: ::pow(T, n) vs std::pow(T, n) for non-constant n

[Richard Guenther <rguenth at tat dot physik dot uni-tuebingen dot de>]

Gabriel Dos Reis wrote:
> Richard Guenther <rguenth@tat.physik.uni-tuebingen.de> writes:
>
> | exponent.  This is why I suggest for __builtin_pow not to dispatch to
> | the library for non-const arguments, but to an internal routine
> | modeled after the v3 binary algorithm.
>
> There definitely should be a __builtin_powi() and the like.  But I
> don't like that std::pow(, 4) and std::pow(, n) (with n == 4) should
> give different results.

With -ffast-math this would be acceptable.  But I don't like std::pow(x, 
 n) and ::pow(x, n) to differ either.  One could improve the binary v3 
algorithm with a table dispatch to optimal code for exponents < 256 like 
the builtin function does.

Richard.