RE: [PATCH v2] arm64: dts: ls1088a: add one more thermal zone node

[Andy Tang <andy.tang@nxp.com>]

Hi Shawn,

> -----Original Message-----
> From: Shawn Guo <shawnguo@kernel.org>
> Sent: 2019年3月20日 16:19
> To: Andy Tang <andy.tang@nxp.com>
> Cc: Daniel Lezcano <daniel.lezcano@linaro.org>; mark.rutland@arm.com;
> devicetree@vger.kernel.org; linux-pm@vger.kernel.org;
> linux-kernel@vger.kernel.org; Leo Li <leoyang.li@nxp.com>;
> edubezval@gmail.com; robh+dt@kernel.org; rui.zhang@intel.com;
> linux-arm-kernel@lists.infradead.org
> Subject: Re: [PATCH v2] arm64: dts: ls1088a: add one more thermal zone node
> 
> On Fri, Mar 08, 2019 at 09:57:09AM +0000, Andy Tang wrote:
> >
> >
> > > -----Original Message-----
> > > From: Daniel Lezcano <daniel.lezcano@linaro.org>
> > > Sent: 2019年3月8日 17:28
> > > To: Andy Tang <andy.tang@nxp.com>; Shawn Guo <shawnguo@kernel.org>
> > > Cc: Leo Li <leoyang.li@nxp.com>; robh+dt@kernel.org;
> > > mark.rutland@arm.com; linux-arm-kernel@lists.infradead.org;
> > > devicetree@vger.kernel.org; linux-kernel@vger.kernel.org;
> > > linux-pm@vger.kernel.org; rui.zhang@intel.com; edubezval@gmail.com
> > > Subject: Re: [PATCH v2] arm64: dts: ls1088a: add one more thermal
> > > zone node
> > >
> > > On 08/03/2019 03:07, Andy Tang wrote:
> > > >
> > > >
> > > >> -----Original Message-----
> > > >> From: Daniel Lezcano <daniel.lezcano@linaro.org>
> > > >> Sent: 2019年3月7日 17:15
> > > >> To: Andy Tang <andy.tang@nxp.com>; Shawn Guo
> > > >> <shawnguo@kernel.org>
> > > >> Cc: Leo Li <leoyang.li@nxp.com>; robh+dt@kernel.org;
> > > >> mark.rutland@arm.com; linux-arm-kernel@lists.infradead.org;
> > > >> devicetree@vger.kernel.org; linux-kernel@vger.kernel.org;
> > > >> linux-pm@vger.kernel.org; rui.zhang@intel.com;
> > > >> edubezval@gmail.com
> > > >> Subject: Re: [PATCH v2] arm64: dts: ls1088a: add one more thermal
> > > >> zone node
> > > >>
> > > >>>>> PS: In order to keep consistency to the first thermal-zone
> > > >>>>> node, there will be "WARNING: line over 80 characters" warnings.
> > > >>>>>
> > > >>>>>  arch/arm64/boot/dts/freescale/fsl-ls1088a.dtsi |   43
> > > >>>> +++++++++++++++++++++--
> > > >>>>>  1 files changed, 39 insertions(+), 4 deletions(-)
> > > >>>>>
> > > >>>>> diff --git a/arch/arm64/boot/dts/freescale/fsl-ls1088a.dtsi
> > > >>>>> b/arch/arm64/boot/dts/freescale/fsl-ls1088a.dtsi
> > > >>>>> index 661137f..9f52bc9 100644
> > > >>>>> --- a/arch/arm64/boot/dts/freescale/fsl-ls1088a.dtsi
> > > >>>>> +++ b/arch/arm64/boot/dts/freescale/fsl-ls1088a.dtsi
> > > >>>>> @@ -129,19 +129,19 @@
> > > >>>>>  	};
> > > >>>>>
> > > >>>>>  	thermal-zones {
> > > >>>>> -		cpu_thermal: cpu-thermal {
> > > >>>>> +		ccu {
> > > >>>>
> > > >>>> Is this change really necessary?  What does 'ccu' stand for?
> > > >>> I think so. ccu stands for core cluster unit. cpu is too general.
> > > >>> On some platforms, there are more than one core clusters.
> > > >>> At least we should change it to "core cluster" if short form is
> > > >>> not
> > > appropriate.
> > > >>
> > > >> If the sensor is a the cluster level, 'cluster' is enough. IMHO,
> > > >> no need to give a description of what contains the cluster,
> > > >> otherwise you will end up with a 'core-gpu-cluster-l2' name.
> > > > If cluster is specific to core, we can use cluster instead. But I don't think
> so.
> > > > Cluster may refer to "core cluster", "GPU cluster" etc.
> > > > So, I think "core-cluster" is ok.
> > > > If core was divided to several clusters, we can name it as
> > > > "core-cluster1",
> > > "core-cluster2" etc.
> > > > If GPU was divided to several clusters we can name it as
> > > > "gpu-cluster1",
> > > "gpu-cluster2" etc.
> > >
> > >
> > > Hi Andy,
> > >
> > > I think there is a confusion around the 'cpu' term and 'cluster'.
> > >
> > > ARM would like to see the 'cluster' word to disappear, so whenever
> > > possible we should avoid it.
> > >
> > > From the hardware side, 'CPU' is usually used to describe the
> > > physical chip containing the cores+cache.
> > >
> > > From the software side, 'CPU' is usually used to describe the
> > > logical process unit, aka a core or a hyper-thread.
> > >
> > > As we are in the DT, so describing the hardware, the CPU refers to
> > > the group
> > > cores+caches.
> > >
> > > From my POV, using 'cpu' for the group of cores and 'gpu' for the
> > > graphic sounds ok, and so far that is what is used for the other platforms.
> > >
> > > If you change the name, that may give the feeling there is something
> > > special with those thermal zones.
> >
> > Thanks Daniel for your detailed explanations.
> >
> > But as you said 'CPU' is usually used to describe the physical chip.
> 
> Here is how I would understand Daniel's comments:
> 
>   CPU = cores + caches
>   physical chip = SoC = CPU + GPU + peripherals ...

Agree.

> 
> > So if we name it as CPU, it sounds like this temperature sensor is monitoring
> the whole chip.
> > That's not true in our case.
> >
> > Take ls2088a for example:
> > In ls2088a SoC, there are 7 temperature sensors. Please note that they are all
> located in SoC.
> > The placement of the temperature sensors are showed below:
> >
> > Sensor ID		placement
> > 1			DDR controller 1
> > 2			DDR controller 2
> > 3			DDR controller 3
> > 4			core cluster 1
> > 5 			core cluster 2
> > 6 			core cluster 3
> > 7 			core cluster 4
> >
> > Apparently using CPU or CPU-cluster is not appropriate. Core-cluster is better.
> 
> So using CPU is appropriate for me, less confusing, more consistent with other
> platforms.
What about core cluster? We can't name it cpu0, cpu1 etc I think.

BR,
Andy
> 
> Shawn