[PATCH] x86_64: Optimize ffsll function code size.

[Florian Weimer]

* Adhemerval Zanella Netto:

> On 27/07/23 13:24, Florian Weimer via Libc-alpha wrote:
>> * Sunil Pandey:
>> 
>>> Ffsll is one of the benchmark tests in the phoronix test suite, not
>>> sure how much it matters to the application. Lots of people involved
>>> in phoronix benchmark testing/tracking and this kind of random perf
>>> behavior wastes their time.
>> 
>> That's a good point.  I've seen similar reports before (sadly I don't
>> recall if they were specifically about ffsll).
>> 
>> Regarding the mechanics of fixing it, if the instruction ordering and
>> sizing is so sensitive, should this be an assembler implementation
>> instead?  And will the fix even work for distributions that build with
>> --enable-cet, considering that there's going to be an additional 4-byte
>> NOP at the start of the function?

> Sigh... do we really need to care about this synthetic benchmark that is
> exercising a fallback path since compiler will most likely issue the
> inline builtin? And even this is really important, tune function alignment
> and size to fit on a cacheline should be done by the compiler, specially
> in the case where we can implement by using a builtin.

I think avoiding wasting people's time with spurious benchmark
differences is useful.  Compared to things like the PID cache, this one
seems pretty harmless.

Maybe we should increase function alignment to cover the CET case, too?

Thanks,
Florian