Re: [PATCH] x86_64: Implement AVX2 version of strlcpy/wcslcpy function

[Noah Goldstein <goldstein.w.n@gmail.com>]

On Mon, Jul 3, 2023 at 11:30 AM Paul Eggert <eggert@cs.ucla.edu> wrote:
>
> On 2023-06-30 15:21, Sunil Pandey wrote:
> > Attached is strcpy/wcslcpy microbenchmark data based on Noah
> > strlcpy/wcslcpy microbenchmark patch.
>
> Although it's helpful to know that the proposed patch improves
> microbenchmark scores, that's not enough to justify it. Let's see
> benchmarks of real programs. If they don't show significant wins, let's
> not bother.
>
> Programs that use strlcpy, by and large, don't use it in
> performance-sensitive areas, and their developers and users are far more
> worried about security than about performance. Making the implementation
> harder to audit will likely be a net negative for these applications.
> This doesn't sound a like a win at all.
>
> Plus, who uses wcslcpy? Why bother to tune it if nobody uses it?

Think we should look into dropping optimized strcpy/wcscpy family
in general? For the most part don't see them in perf sensitive areas
anyways (generally people that care about perf maintain the length
and use mem* functions).