[PATCH v3 3/7] stdlib: Optimization qsort{_r} swap implementation (BZ #19305)

[Adhemerval Zanella]

On 15/10/2021 14:45, Noah Goldstein wrote:
> swap is in the inner loop. Seems like a pretty critical component to have fully
> optimized. The aarch64 version looks good, but the x86_64 version seems
> to be lacking. Not arguing for an arch specific version, but if the directives
> can add value to x86_64 without detracting from aarch64 seems like a zero
> cost improvement.
> 
> Since size is non-constant for the tail I don't see how we are going
> avoid 3x memcpy calls. Although that can be another patch if it
> gets values.
> 
>>
>>
>> [1] https://godbolt.org/z/v7e4xxqGa

Maybe use a byte copy in the tail to avoid memcpy [1], another option
might to tune SWAP_GENERIC_SIZE to make the tail less costly (16 should
be ok for most architecture, although some might be better with large
values).

[1] https://godbolt.org/z/G76dcej16