Re: [PATCH v1 6/7] arm64: dts: sdm845: Increase alert trip point to 95 degrees

From: Matthias Kaehlcke <mka@chromium.org>
To: Amit Kucheria <amit.kucheria@linaro.org>
Cc: LKML <linux-kernel@vger.kernel.org>,
	linux-arm-msm <linux-arm-msm@vger.kernel.org>,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	Viresh Kumar <viresh.kumar@linaro.org>,
	Eduardo Valentin <edubezval@gmail.com>,
	Andy Gross <andy.gross@linaro.org>,
	Taniya Das <tdas@codeaurora.org>,
	Stephen Boyd <swboyd@chromium.org>,
	Douglas Anderson <dianders@chromium.org>,
	David Brown <david.brown@linaro.org>,
	Rob Herring <robh+dt@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	"open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS" 
	<devicetree@vger.kernel.org>
Subject: Re: [PATCH v1 6/7] arm64: dts: sdm845: Increase alert trip point to 95 degrees
Date: Fri, 11 Jan 2019 10:30:23 -0800	[thread overview]
Message-ID: <20190111183023.GE261387@google.com> (raw)
In-Reply-To: <CAHLCerPN3dxd=XBEkU2cMacFDkZjNhXEw7BDYNTEyfSUszSUzg@mail.gmail.com>

On Fri, Jan 11, 2019 at 03:54:23PM +0530, Amit Kucheria wrote:
> On Thu, Jan 10, 2019 at 6:45 AM Matthias Kaehlcke <mka@chromium.org> wrote:
> >
> > Hi Amit,
> >
> > On Thu, Jan 10, 2019 at 05:30:55AM +0530, Amit Kucheria wrote:
> > > 75 degrees is too aggressive for throttling the CPU. After speaking to
> > > Qualcomm engineers, increase it to 95 degrees.
> > >
> > > Signed-off-by: Amit Kucheria <amit.kucheria@linaro.org>
> > > ---
> > >  arch/arm64/boot/dts/qcom/sdm845.dtsi | 16 ++++++++--------
> > >  1 file changed, 8 insertions(+), 8 deletions(-)
> > >
> > > diff --git a/arch/arm64/boot/dts/qcom/sdm845.dtsi b/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > > index c27cbd3bcb0a..29e823b0caf4 100644
> > > --- a/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > > +++ b/arch/arm64/boot/dts/qcom/sdm845.dtsi
> > > @@ -1692,7 +1692,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert0: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> > > @@ -1713,7 +1713,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert1: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> > > @@ -1734,7 +1734,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert2: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> > > @@ -1755,7 +1755,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert3: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> > > @@ -1776,7 +1776,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert4: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> > > @@ -1797,7 +1797,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert5: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> > > @@ -1818,7 +1818,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert6: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> > > @@ -1839,7 +1839,7 @@
> > >
> > >                       trips {
> > >                               cpu_alert7: trip0 {
> > > -                                     temperature = <75000>;
> > > +                                     temperature = <95000>;
> > >                                       hysteresis = <2000>;
> > >                                       type = "passive";
> > >                               };
> >
> > The change itself looks good to me, however I wonder if it would be
> > worth to eliminate redundancy and merge the current 8 thermal zones
> > into 2, one for the Silver and one for the Gold cluster (as done by
> > http://crrev.com/c/1381752). There is a single cooling device for
> > each cluster, so it's not clear to me if there is any gain from having
> > a separate thermal zone for each CPU. If it is important to monitor
> > the temperatures of the individual cores this can still be done by
> > configuring the thermal zone of the cluster with multiple thermal
> > sensors.
> 
> Reducing the number of thermal zones to 2 (by grouping 4 sensors per
> zone) is not possible due a limitation of the thermal framework[1]. It
> is something that we want to address. Previous attempts to fix this
> were rejected for various reasons. Eduardo was going to share a way to
> have more flexible mapping between sensors and zones after discussions
> at LPC.

I wasn't aware of this limitation, thanks for the clarification! With
this I understand that for now we indeed need the 8 thermal zones with
all the redundant information :(

> <nag> Eduardo, do you have anything we can review? </nag> :-)
> 
> Having said that, we'll need some aggregation functions when we add
> multiple sensors to a zone (e.g. max, mean) to reflect the zone. This
> will lose information about hotspots and prevent things like idle
> injection on a particular CPU that is causing most of the heat in the
> aggregated zone. So IMHO, it might be useful to have information about
> the hotspots (i.e TZ per sensor) and aggregated values (ambient
> temperature) that can be fed to the thermal policy.

Ok, it seems for now we need the 8 thermal zones in any case, when
support for multiple sensors becomes available we can evaluate whether
it's worth to change that or not.

Cheers

Matthias