Re: interesting skylake perf tidbit

[Paul Goyette <paul%whooppee.com@localhost>]

On Mon, 18 Jun 2018, maya%netbsd.org@localhost wrote:

> For anyone interested in performance tuning..
>
> https://aloiskraus.wordpress.com/2018/06/16/why-skylakex-cpus-are-sometimes-50-slower-how-intel-has-broken-existing-code/
>
> tl;dr pause in newer skylakes is a lot.
> We call this internally x86_pause and do an entire function call for it
> (could use an intrinsic like __builtin_ia32_pause() too for these
> things...)
>
> It's our choice for SPINLOCK_BACKOFF_HOOK.
>
> (Perhaps a tunable read-mostly SPINLOCK_BACKOFF_* would be worthwhile?)
>
> Also on the same topic: SPINLOCK_RUN_HOOK seems to be unused?

Just curious - does anyone know if (or how) this affects AMD processors?


+------------------+--------------------------+----------------------------+
| Paul Goyette     | PGP Key fingerprint:     | E-mail addresses:          |
| (Retired)        | FA29 0E3B 35AF E8AE 6651 | paul at whooppee dot com   |
| Kernel Developer | 0786 F758 55DE 53BA 7731 | pgoyette at netbsd dot org |
+------------------+--------------------------+----------------------------+