Date:   Thu, 30 Aug 2018 13:50:31 +0100
From:   Patrick Bellasi <patrick.bellasi@....com>
To:     Quentin Perret <quentin.perret@....com>
Cc:     peterz@...radead.org, rjw@...ysocki.net,
        linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
        gregkh@...uxfoundation.org, mingo@...hat.com,
        dietmar.eggemann@....com, morten.rasmussen@....com,
        chris.redpath@....com, valentin.schneider@....com,
        vincent.guittot@...aro.org, thara.gopinath@...aro.org,
        viresh.kumar@...aro.org, tkjos@...gle.com, joel@...lfernandes.org,
        smuckle@...gle.com, adharmap@...eaurora.org,
        skannan@...eaurora.org, pkondeti@...eaurora.org,
        juri.lelli@...hat.com, edubezval@...il.com,
        srinivas.pandruvada@...ux.intel.com, currojerez@...eup.net,
        javi.merino@...nel.org
Subject: Re: [PATCH v6 05/14] sched/topology: Reference the Energy Model of
 CPUs when available

On 30-Aug 11:47, Quentin Perret wrote:
> On Thursday 30 Aug 2018 at 11:00:20 (+0100), Patrick Bellasi wrote:
> > Dunno... but, in any case, probably we don't care about using EAS until
> > the boot complete, isn't it?
> 
> So, as of now, EAS will typically start soon after CPUFreq. I don't see a
> point in starting before, and I'm not sure if starting it later is
> really what we want either ... It probably makes sense to start the
> DVFS-related features (CPUFreq, EAS, ...) together.
> 
> > Just to better understand, what is the most common activation path we expect ?
> > 
> >   1. system boot
> >   2. CPUs online
> >   3. CPUFreq initialization
> >   4. EM registered
> >   X. ???
> 
> X is sort of arch dependent (like the EM registration actually). For
> arm/arm64, the arch topology driver (see drivers/base/arch_topology.c)
> will rebuild the sched_domains once the capacities of CPU are
> discovered. In previous versions, I had a patch that did that explicitly:
> 
> https://lore.kernel.org/lkml/20180724122521.22109-14-quentin.perret@arm.com/
> 
> However, I dropped this patch for v6 because I rebased the series on top
> of Morten's automatic flag detection patches, which happen to already do
> that for me. More precisely, Morten's patches will rebuild the sched
> domains if the SD_ASYM_CPUCAPACITY flag needs to be set somewhere in the
> hierarchy. EAS depends on this flag anyway, so I guess it's fine to rely
> on this rebuild to start EAS at the same time.
> 
> I have been wondering for a while if such a 'loose' way of starting EAS
> was alright or not. My conclusion was that it should be OK for now,
> because that should suit all the users I can see in the short/medium
> term. An alternative could be to introduce something like a notifier in
> the EM framework in order to let other subsystems know that the EM is
> complete or something along those lines. And then we could register a
> callback on this notifier to rebuild the sched_domains. But that's added
> complexity, which isn't really needed (yet). If that needs to be done to
> accomodate the needs of new users/archs, we can still implement that
> later :-)
> 
> >   N. partition_sched_domains
> >         build_perf_domains
> > 
> > IOW, who do we expect to call build_perf_domains for the first time ?
> 
> On arm/arm64, when the arch_topology driver calls rebuild_sched_domains.

Great, I missed that point, which you actually call out in the cover
letter. Thanks also for the more comprehensive explanation above!

> Thanks !
> Quentin

Cheers Patrick

-- 
#include <best/regards.h>

Patrick Bellasi

Powered by blists - more mailing lists