Re: [PATCH 0/5] sched: Lazy preemption muck

[Mike Galbraith <efault@gmx.de>]

On Mon, 2024-10-07 at 09:46 +0200, Peter Zijlstra wrote:
> Hi!
>
> During LPC Thomas reminded me that the lazy preemption stuff was not there yet.
>
> So here goes, robot says it builds, and I checked both a regular and PREEMPT_RT
> build boots and can change the mode.
>
> Please have a poke.

My box seems content.

I'm gonna miss VOLATILE (voluntary;) when it's eventually whacked, but
not a lot.. or at all general case, security and whatnot over time have
beaten up high end switchers far worse.

tbench 8 30s - single run, box=i7-4790

master static
  voluntary     3620.45 MB/sec   1.000   mean3 3613.45  stddev3 14.08
  voluntary     2706.54 MB/sec    .747   +cpu_mitigations
  voluntary     4028.72 MB/sec   1.112   nostalgia (4.4.231)
  preempt       3449.35 MB/sec    .952
  none          3631.99 MB/sec   1.003

master dynamic
  none          3548.19 MB/sec    .980
  voluntary     3495.77 MB/sec    .965
  full          3476.50 MB/sec    .960
  lazy          3473.95 MB/sec    .959
  laziest       3492.09 MB/sec    .964

master dynamic rt
  full          2352.58 MB/sec    .649
  lazy          2986.28 MB/sec    .824
  laziest       2977.63 MB/sec    .822

distro kernel dynamic
  none          1744.51 MB/sec    .481
  none          2189.74 MB/sec    .604  mitigations=off