[PATCH v2] AArch64: Optimise SVE scalar callbacks

[Wilco Dijkstra]

Hi Joe,

> Instead of using SVE instructions to marshall special results into the
> correct lane, just write the entire vector (and the predicate) to
> memory, then use cheaper scalar operations.
>
> Geomean speedup of 16% in special intervals on Neoverse with GCC 14.

LGTM - pushed.

Reviewed-by: Wilco Dijkstra  <Wilco.Dijkstra@arm.com>


> I also think we need to be judicious about overriding stack hardening. We
> do it on some startup routines due correctness (since the code can not use
> stack hardening at the moment); but for optimization I think we should use
> what compiler is using by default.

Distro compiler defaults are pretty awful and skewed towards wasting a lot
of performance for no gain (I wonder when people will turn on the insane
clear all registers to zero before returning!). Enabling PAUTH would be much
better than paying the overhead of -fstack-protector-strong (which appears
to be the default in some distros).

This is best done as a different patch, so I'm happy with leaving this out for now.
However it is something we need to figure out going forward - there are many
exceptions for stack-protector since it causes many correctness issues itself!

Cheers,
Wilco