devicetree.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: lukasz.luba@arm.com
To: kgene@kernel.org, krzk@kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-samsung-soc@vger.kernel.org, linux-kernel@vger.kernel.org,
	devicetree@vger.kernel.org, linux-pm@vger.kernel.org
Cc: myungjoo.ham@samsung.com, kyungmin.park@samsung.com,
	cw00.choi@samsung.com, robh+dt@kernel.org, mark.rutland@arm.com,
	b.zolnierkie@samsung.com, lukasz.luba@arm.com,
	dietmar.eggemann@arm.com
Subject: [PATCH 2/3] ARM: dts: exynos: Add Exynos5422 CPU dynamic-power-coefficient information
Date: Mon, 27 Jan 2020 21:54:52 +0000	[thread overview]
Message-ID: <20200127215453.15144-3-lukasz.luba@arm.com> (raw)
In-Reply-To: <20200127215453.15144-1-lukasz.luba@arm.com>

From: Lukasz Luba <lukasz.luba@arm.com>

Add dynamic power coefficient into CPU nodes which let CPUFreq subsystem
register the Energy Model (EM) for the CPUs.

The 'dynamic-power-coefficient' is used for calculating the dynamic power
according to the equation in documentation [1].  The Energy Model (EM)
framework relies on calculated power and cost for each OPP. The OPP power
values come from CPUFreq driver, which registered required callback
function. The simple implementation of a CPUFREQ driver, like cpufreq-dt,
uses 'dev_pm_opp_of_register_em()' which relay on
'dynamic-power-coefficient' to calculate the power of requested OPP for the
EM [2].

The calculated values might be checked in
/sys/kernel/debug/energy_model/pd*/

$ grep . /sys/kernel/debug/energy_model/pd1/cs*/*
/sys/kernel/debug/energy_model/pd1/cs:1000000/cost:558
/sys/kernel/debug/energy_model/pd1/cs:1000000/frequency:1000000
/sys/kernel/debug/energy_model/pd1/cs:1000000/power:310
/sys/kernel/debug/energy_model/pd1/cs:1100000/cost:558
/sys/kernel/debug/energy_model/pd1/cs:1100000/frequency:1100000
/sys/kernel/debug/energy_model/pd1/cs:1100000/power:341
/sys/kernel/debug/energy_model/pd1/cs:1200000/cost:558
/sys/kernel/debug/energy_model/pd1/cs:1200000/frequency:1200000
/sys/kernel/debug/energy_model/pd1/cs:1200000/power:372
/sys/kernel/debug/energy_model/pd1/cs:1300000/cost:674
/sys/kernel/debug/energy_model/pd1/cs:1300000/frequency:1300000
/sys/kernel/debug/energy_model/pd1/cs:1300000/power:487
/sys/kernel/debug/energy_model/pd1/cs:1400000/cost:675 ...

$ grep . /sys/kernel/debug/energy_model/pd0/cs*/*
/sys/kernel/debug/energy_model/pd0/cs:1000000/cost:200
/sys/kernel/debug/energy_model/pd0/cs:1000000/frequency:1000000
/sys/kernel/debug/energy_model/pd0/cs:1000000/power:154
/sys/kernel/debug/energy_model/pd0/cs:1100000/cost:260
/sys/kernel/debug/energy_model/pd0/cs:1100000/frequency:1100000
/sys/kernel/debug/energy_model/pd0/cs:1100000/power:220
/sys/kernel/debug/energy_model/pd0/cs:1200000/cost:260
/sys/kernel/debug/energy_model/pd0/cs:1200000/frequency:1200000
/sys/kernel/debug/energy_model/pd0/cs:1200000/power:240
/sys/kernel/debug/energy_model/pd0/cs:1300000/cost:260
/sys/kernel/debug/energy_model/pd0/cs:1300000/frequency:1300000
/sys/kernel/debug/energy_model/pd0/cs:1300000/power:260
/sys/kernel/debug/energy_model/pd0/cs:200000/cost:130 ...

To provide a proper value of the 'dynamic-power-coefficient' the real power
can be measured using a dedicated hardware, i.e. INA2xx. The Odroid-XU3
hwmon sensors have been used to capture the power value during a sysbench
test running on single core and at each possible OPP. The measured values
were divided by 2, since the dynamic power is typically half of the
consumed power (the second half is static power). Next, the approximation
was made and the power model derived, showing the 'C' value of routhly X.
Check the example equations in drivers/opp/of.c [2].
Thus, i.e. the power = 1.0Watt at 1GHz => 0.5W dynamic power =>
dynamic-power-coefficient = 400

Using this simple technique we can provide and needed coefficient.  The
approximation does not have to be super precised. The proportion is
important and the difference between power consumed by different CPUs
running at the same frequency, which is then used in Energy Aware Scheduler
algorithms. An example power values on Odroid-XU3:

(LITTLE CPU)
/sys/kernel/debug/energy_model/pd0/cs:1000000/frequency:1000000
/sys/kernel/debug/energy_model/pd0/cs:1000000/power:154
(big CPU)
/sys/kernel/debug/energy_model/pd1/cs:1000000/frequency:1000000
/sys/kernel/debug/energy_model/pd1/cs:1000000/power:310

In Odroid-XU3 case the derived coefficient value for 'big' CPU has:
dynamic-power-coefficient = <310>;
while the 'LITTLE':
dynamic-power-coefficient = <128>;

[1] Documentation/devicetree/bindings/arm/cpus.yaml
[2] https://elixir.bootlin.com/linux/v5.4/source/drivers/opp/of.c#L1044

Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
---
 arch/arm/boot/dts/exynos5422-cpus.dtsi | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/arch/arm/boot/dts/exynos5422-cpus.dtsi b/arch/arm/boot/dts/exynos5422-cpus.dtsi
index 1b8605cf2407..c9a0dc99d2fb 100644
--- a/arch/arm/boot/dts/exynos5422-cpus.dtsi
+++ b/arch/arm/boot/dts/exynos5422-cpus.dtsi
@@ -31,6 +31,7 @@
 			operating-points-v2 = <&cluster_a7_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <539>;
+			dynamic-power-coefficient = <128>;
 		};
 
 		cpu1: cpu@101 {
@@ -43,6 +44,7 @@
 			operating-points-v2 = <&cluster_a7_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <539>;
+			dynamic-power-coefficient = <128>;
 		};
 
 		cpu2: cpu@102 {
@@ -55,6 +57,7 @@
 			operating-points-v2 = <&cluster_a7_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <539>;
+			dynamic-power-coefficient = <128>;
 		};
 
 		cpu3: cpu@103 {
@@ -67,6 +70,7 @@
 			operating-points-v2 = <&cluster_a7_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <539>;
+			dynamic-power-coefficient = <128>;
 		};
 
 		cpu4: cpu@0 {
@@ -79,6 +83,7 @@
 			operating-points-v2 = <&cluster_a15_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <1024>;
+			dynamic-power-coefficient = <310>;
 		};
 
 		cpu5: cpu@1 {
@@ -91,6 +96,7 @@
 			operating-points-v2 = <&cluster_a15_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <1024>;
+			dynamic-power-coefficient = <310>;
 		};
 
 		cpu6: cpu@2 {
@@ -103,6 +109,7 @@
 			operating-points-v2 = <&cluster_a15_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <1024>;
+			dynamic-power-coefficient = <310>;
 		};
 
 		cpu7: cpu@3 {
@@ -115,6 +122,7 @@
 			operating-points-v2 = <&cluster_a15_opp_table>;
 			#cooling-cells = <2>; /* min followed by max */
 			capacity-dmips-mhz = <1024>;
+			dynamic-power-coefficient = <310>;
 		};
 	};
 };
-- 
2.17.1


  parent reply	other threads:[~2020-01-27 21:55 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-27 21:54 [PATCH 0/3] Enable Odroid-XU3/4 to use Energy Model and Energy Aware Scheduler lukasz.luba
2020-01-27 21:54 ` [PATCH 1/3] ARM: exynos_defconfig: Enable SCHED_MC lukasz.luba
2020-01-31 12:47   ` Krzysztof Kozlowski
2020-01-31 15:59     ` Lukasz Luba
2020-01-31 20:33       ` Krzysztof Kozlowski
2020-01-27 21:54 ` lukasz.luba [this message]
2020-01-31 13:05   ` [PATCH 2/3] ARM: dts: exynos: Add Exynos5422 CPU dynamic-power-coefficient information Krzysztof Kozlowski
2020-01-31 16:42     ` Lukasz Luba
2020-01-27 21:54 ` [PATCH 3/3] ARM: exynos_defconfig: Enable Energy Model framework lukasz.luba
2020-01-31 13:16   ` Krzysztof Kozlowski
2020-01-31 17:30     ` Lukasz Luba
2020-01-31 20:41       ` Krzysztof Kozlowski
2020-02-05 12:49         ` Lukasz Luba
2020-02-06 12:59           ` Krzysztof Kozlowski
2020-02-06 14:15             ` Lukasz Luba
2020-01-31 13:30   ` Bartlomiej Zolnierkiewicz
2020-01-31 13:31     ` Krzysztof Kozlowski
2020-01-31 13:47       ` Bartlomiej Zolnierkiewicz
2020-01-31 13:48         ` Bartlomiej Zolnierkiewicz
2020-01-31 17:38     ` Lukasz Luba

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200127215453.15144-3-lukasz.luba@arm.com \
    --to=lukasz.luba@arm.com \
    --cc=b.zolnierkie@samsung.com \
    --cc=cw00.choi@samsung.com \
    --cc=devicetree@vger.kernel.org \
    --cc=dietmar.eggemann@arm.com \
    --cc=kgene@kernel.org \
    --cc=krzk@kernel.org \
    --cc=kyungmin.park@samsung.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pm@vger.kernel.org \
    --cc=linux-samsung-soc@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=myungjoo.ham@samsung.com \
    --cc=robh+dt@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).