Re: [PATCH v2] freezer,sched: Use saved_state to reduce some spurious wakeups

[Peter Zijlstra]

On Fri, Sep 08, 2023 at 03:30:43PM -0700, Elliot Berman wrote:
> 
> 
> On 9/8/2023 3:08 PM, Peter Zijlstra wrote:
> > On Fri, Sep 08, 2023 at 01:08:07PM -0700, Elliot Berman wrote:
> > 
> >>> Perhaps we should start off by doing the below, instead of making it
> >>> more complicated instead. I suppose you're right about the overhead, but
> >>> run a hackbench just to make sure or something.
> >>>
> >>
> >> I ran perf bench sched message -g 40 -l 40 with the v3 patch [1]. After 60
> >> iterations each, I don't see a significant difference on my arm64 platform:
> >> both samples ~normal and ~eq variance w/t-test p-value: 0.79.
> >>
> >> We also ran typical high level benchmarks for our SoCs (antutu,
> >> geekbench, et. al) and didn't see any regressions there.
> > 
> > So if you would've made this 2 patches, the first removing the ifdef,
> > then the changelog for that patch would be a good place to mention it
> > doesn't measurably regress things.
> 
> No problem, easily done.
> 
> > As a bonus, it then makes your other changes smaller too ;-)
> 
> Did you mean that each commit is smaller but overall delta is the same
> or something else? 

That.

> I still wanted to update comments on saved_state in
> kernel/sched/core.c as it gives good explanation of what is going on. I
> have split the commit but want to make sure I make the changes you were
> thinking :-)

well, it's nearly 1am, I'm not thinking very much :-) Changing those
comments seems fine when you add the freezer thing.