Re: [RFC PATCH 1/7] PM: Introduce em_pd_get_higher_freq()

From: Quentin Perret <quentin.perret@arm.com>
To: Patrick Bellasi <patrick.bellasi@arm.com>
Cc: Douglas Raillard <douglas.raillard@arm.com>,
	linux-kernel@vger.kernel.org, linux-pm@vger.kernel.org,
	mingo@redhat.com, peterz@infradead.org, dietmar.eggemann@arm.com
Subject: Re: [RFC PATCH 1/7] PM: Introduce em_pd_get_higher_freq()
Date: Fri, 21 Jun 2019 11:22:03 +0100	[thread overview]
Message-ID: <20190621102201.kuid47xl5pi72c4y@queper01-lin> (raw)
In-Reply-To: <20190621101704.pw7oluum4eqgdgzp@queper01-lin>

On Friday 21 Jun 2019 at 11:17:05 (+0100), Quentin Perret wrote:
> On Thursday 20 Jun 2019 at 14:04:39 (+0100), Patrick Bellasi wrote:
> > On 19-Jun 17:08, Douglas Raillard wrote:
> > > Hi Patrick,
> > > 
> > > On 5/16/19 2:22 PM, Patrick Bellasi wrote:
> > > > On 16-May 14:01, Quentin Perret wrote:
> > > > > On Thursday 16 May 2019 at 13:42:00 (+0100), Patrick Bellasi wrote:
> > > > > > > +static inline unsigned long em_pd_get_higher_freq(struct em_perf_domain *pd,
> > > > > > > +	unsigned long min_freq, unsigned long cost_margin)
> > > > > > > +{
> > > > > > > +	unsigned long max_cost = 0;
> > > > > > > +	struct em_cap_state *cs;
> > > > > > > +	int i;
> > > > > > > +
> > > > > > > +	if (!pd)
> > > > > > > +		return min_freq;
> > > > > > > +
> > > > > > > +	/* Compute the maximum allowed cost */
> > > > > > > +	for (i = 0; i < pd->nr_cap_states; i++) {
> > > > > > > +		cs = &pd->table[i];
> > > > > > > +		if (cs->frequency >= min_freq) {
> > > > > > > +			max_cost = cs->cost + (cs->cost * cost_margin) / 1024;
> > > > > >                                                                           ^^^^
> > > > > > ... end here we should probably better use SCHED_CAPACITY_SCALE
> > > > > > instead of hard-coding in values, isn't it?
> > > > > 
> > > > > I'm not sure to agree. This isn't part of the scheduler per se, and the
> > > > > cost thing isn't in units of capacity, but in units of power, so I don't
> > > > > think SCHED_CAPACITY_SCALE is correct here.
> > > > 
> > > > Right, I get the units do not match and it would not be elegant to use
> > > > it here...
> > > > 
> > > > > But I agree these hard coded values (that one, and the 512 in one of the
> > > > > following patches) could use some motivation :-)
> > > > 
> > > > ... ultimately SCHED_CAPACITY_SCALE is just SCHED_FIXEDPOINT_SCALE,
> > > > which is adimensional. Perhaps we should use that or yet another alias
> > > > for the same.
> > > 
> > > Would it be a good idea to use SCHED_FIXEDPOINT_SCALE in energy.c ?
> > > Since it's not part of the scheduler, maybe there is a scale covering a wider scope,
> > > or we can introduce a similar ENERGY_FIXEDPOINT_SCALE in energy_model.h.
> > 
> > Well, in energy_model.c we have references to "capacity" and
> > "utilization" which are all SCHED_FIXEDPOINT_SCALE range values.
> > That symbol is defined in <linux/sched.h> and we already pull
> > in other <linux/sched/*> headers.
> > 
> > So, to me it seems it's not unreasonable to say that we use scheduler
> > related concepts and it makes more sense than introducing yet another
> > scaling factor.
> > 
> > But that's just my two cents ;)
> 
> Perhaps use this ?
> 
> https://elixir.bootlin.com/linux/latest/source/include/linux/energy_model.h#L43
> 

Nah, bad idea actually ... Sorry for the noise