Re: [PATCH v2] Simplify pow with constant

[Wilco Dijkstra <Wilco.Dijkstra@arm.com>]

Jeff Law wrote:
> Right.  exp is painful in glibc, but pow is *dramatically* more painful
> and likely always will be.
>
> Siddhesh did some great work in bringing those costs down in glibc but
> the more code we can reasonably shunt into exp instead of pow, the better.
>
> It's likely pow will always be significantly more expensive than exp.
> It's also likely that predicting when these functions are going to go
> off the fast paths is painful.

With a modern implementation there won't be any slow path - it's completely
unnecessary, and you can get 100x speedup by simply doing things in a
sane way.

Szabolc's version of powf is almost literally doing exp(log(x) * y), so exp is
about twice as fast as pow.

Wilco