On Tue, 9 Jan 2024 at 12:56, Dietmar Eggemann <dietmar.eggem...@arm.com> wrote:
> On 08/01/2024 14:48, Vincent Guittot wrote:
> > Now that cpufreq provides a pressure value to the scheduler, rename
> I.e. that thermal (e.g. IPA governor) switches from average
> (rq->avg_(thermal/hw).load_avg) (1) to instantenous (cpu_pressure) (2).
> I rememeber a related dicussion during LPC 2018 :-)
> > arch_update_thermal_pressure into HW pressure to reflect that it returns
> > a pressure applied by HW (i.e. with a high frequency change) and not
> > always related to thermal mitigation but also generated by max current
> > limitation as an example. Such high frequency signal needs filtering to be
> > smoothed and provide an value that reflects the average available capacity
> > into the scheduler time scale.
> So 'drivers/cpufreq/qcom-cpufreq-hw.c' is the only user of (1) right
> now. Are we expecting more users here? If this stays the only user,
> maybe it can do the averages by itself and we can completely switch to (2)?

Yes I expect more users for this high frequency change which is
typically 1ms and below. And I prefer to keep PELT averaging
calculation inside scheduler instead of letting driver doing their own
averaging which will be most of the time not aligned with what
scheduler wants

> [...]

Reply via email to