Re: [RFC PATCH v3 03/10] PM: Introduce an Energy Model management framework

[Quentin Perret]

On Friday 08 Jun 2018 at 11:36:13 (+0200), Juri Lelli wrote:
> On 08/06/18 09:25, Quentin Perret wrote:
> > Hi Dietmar,
> > 
> > On Thursday 07 Jun 2018 at 17:55:32 (+0200), Dietmar Eggemann wrote:
> 
> [...]
> 
> > > IMHO, part of the problem why this might be harder to understand is the fact
> > > that the patches show the use of the 2. init call
> > > 'em_rescale_cpu_capacity()' but not the 1. one 'em_register_freq_domain()'.
> > > I guess that Quentin wanted to keep the set as small as possible.
> > 
> > Yes, this is confusing. I'm now starting to think that patch 10/10 should
> > probably not be part of this patch-set, especially if I don't provide
> > the patches registering the freq domains from the CPUFreq drivers. And
> > it's the only "Arm-specific" patch in this arch-independent patch-set.
> > 
> > So I think I'll drop patch 10/10 for v4 ... That part should be
> > discussed separately, with the rest of the Arm-specific changes.
> 
> Mmm, I would actually vote to at least have one example showing how and
> where the em_register_freq_domain() is going to be used. I had to look
> at the repo you referenced since I think it's quite fundamental piece to
> understand the design, IMHO.

Hmmm I see your point. OK, having an example will help. I'll keep patch
10/10 and add an other one tweaking cpufreq-dt to give an example. But
I'll mark the two as OPTIONAL. I really hope we can reach an agreement
on the core design ideas before discussing too much the details on the
driver side.

There are several valid places where em_register_freq_domain() can be
called. But the exact way of doing so will be platform-dependent, and
driver-dependent, so let's agree on what we want to know from the
drivers first :-)

Thanks,
Quentin