[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAD=FV=VDCbWU4ukYwJUsTKMfEz9+55rmdLL1a39JWcPzjUmZCQ@mail.gmail.com>
Date: Mon, 2 Nov 2020 16:41:57 -0800
From: Doug Anderson <dianders@...omium.org>
To: Quentin Perret <qperret@...gle.com>
Cc: Lukasz Luba <lukasz.luba@....com>,
LKML <linux-kernel@...r.kernel.org>,
Linux PM <linux-pm@...r.kernel.org>,
Linux Doc Mailing List <linux-doc@...r.kernel.org>,
"open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS"
<devicetree@...r.kernel.org>,
Linux ARM <linux-arm-kernel@...ts.infradead.org>,
Daniel Lezcano <daniel.lezcano@...aro.org>,
Rob Herring <robh+dt@...nel.org>,
Amit Kucheria <amitk@...nel.org>,
Jonathan Corbet <corbet@....net>,
Dietmar Eggemann <Dietmar.Eggemann@....com>,
morten.rasmussen@....com, Matthias Kaehlcke <mka@...omium.org>,
Rajendra Nayak <rnayak@...eaurora.org>,
"Rafael J. Wysocki" <rafael@...nel.org>,
Sudeep Holla <sudeep.holla@....com>,
Viresh Kumar <viresh.kumar@...aro.org>,
Stephen Boyd <sboyd@...nel.org>, Nishanth Menon <nm@...com>
Subject: Re: [PATCH v3 0/4] Clarify abstract scale usage for power values in
Energy Model, EAS and IPA
Hi,
On Mon, Nov 2, 2020 at 5:54 AM Quentin Perret <qperret@...gle.com> wrote:
>
> On Monday 02 Nov 2020 at 08:54:38 (+0000), Lukasz Luba wrote:
> > Gentle ping to Quentin and Daniel for sharing opinion on this patch set.
> > If you are OK, then I could use this as a base for next work.
>
> One or two small nits, but overall this LGTM. Thanks Lukasz.
>
> > As you probably know I am working also on 'sustainable power' estimation
> > which could be used when there is no DT value but it comes from FW.
> > That would meet requirement from Doug, when the DT cannot be used,
> > but we have sustainable levels from FW [1].
>
> Cool, and also, I'd be happy to hear from Doug if passing the sustained
> power via sysfs is good enough for his use-case in the meantime?
It does sound like sysfs could be made to work for us, but it's
definitely a workaround. If the normal way to set these values was
through sysfs then it would be fine, but I think most people expect
that these values are just setup properly by the kernel. That means
anyone using our board with a different userspace (someone running
upstream on it) would need to figure out what mechanism they were
going to use to program them. There's very little advantage here
compared to a downstream patch that just violates official upstream
policy by putting something bogoWatts based in the device tree.
My current plan of record (which I don't love) is basically:
1. Before devices are in consumer's hands, accept bogoWatts numbers in
our downstream kernel.
2. Once devices are in consumers hands, run the script I sent out to
generate some numbers and post them upstream.
If, at some point, there's a better solution then I'll switch to it,
but until then that seems workable even if it makes me grumpy.
-Doug
Powered by blists - more mailing lists