Re: [tip:sched/core] [sched] 2ae891b826: hackbench.throughput 6.2% regression

[Qais Yousef <qyousef@layalina.io>]

On 02/25/25 23:35, Chen Yu wrote:
> On 2025-02-25 at 13:42:20 +0000, Qais Yousef wrote:
> > On 02/25/25 21:15, Chen Yu wrote:
> > > On 2025-02-25 at 13:27:05 +0100, Peter Zijlstra wrote:
> > > > On Tue, Feb 25, 2025 at 05:31:34PM +0800, Chen Yu wrote:
> > > > 
> > > > > 
> > > > > But consider that the 6% regression is not that high, and the user might customize
> > > > > base_slice via debugfs on-demand, we can keep an eye on this and revist it in the
> > > > > future(we have encountered some SPECjbb regression due to over-preemption).
> > > > 
> > > > You can specify a per-task slice using sched_attr::sched_runtime. Also
> > > > see commit 857b158dc5e8 ("sched/eevdf: Use sched_attr::sched_runtime to
> > > > set request/slice suggestion")
> > > > 
> > > >
> > > 
> > > Thanks, we'll have a try during the next test cycle.
> > 
> > Could you also try with HRTICK enabled?
> > 
> > 	echo HRTICK | sudo tee /sys/kernel/debug/sched/features
> 
> Sure. Is HRTICK supposed to encourage preemption? I thought
> hackbench might be sensitive to preemption(downgrading).

Yes my bad, I don't know why I read this as preemption was happening later not
earlier. Please ignore me then.

Thanks

--
Qais Yousef