posix_memalign performance regression in 2.38?

[DJ Delorie]

Adhemerval Zanella Netto <adhemerval.zanella@linaro.org> writes:
> It seems to be caused by 24cdd6c71debfd10a9f7cb217fe2a2c4c486ed6f, where
> posix_memalign now calls chunk_ok_for_memalign (which takes most of time).
>
> DJ, any idea on how we can improve this?

This was not unexpected, as we're adding more logic to the memalign
family of functions in order to use less memory.  We can do some
optimizations in this area but it will remain a balancing act between
speed and memory usage.

The old code was terrible wrt memory usage.