요약·해설과 원문, 전문 번역을 서로 분리했습니다. API 이름, symbol, source path는 원문 표기를 사용합니다.
1. 요약·해설
원문의 핵심 논리와 kernel programming 관점의 보충 설명입니다. 아래의 전문 번역과는 별도로 작성했습니다.
2. 영어 원문 전체
번역 기준이 된 Linux v6.18.37 원문입니다. 줄 번호는 이 버전의 파일 좌표입니다.
원문 전체 펼치기
==========================================
CPU capacity bindings
==========================================
==========================================
1 - Introduction
==========================================
Some systems may be configured to have cpus with different power/performance
characteristics within the same chip. In this case, additional information has
to be made available to the kernel for it to be aware of such differences and
take decisions accordingly.
==========================================
2 - CPU capacity definition
==========================================
CPU capacity is a number that provides the scheduler information about CPUs
heterogeneity. Such heterogeneity can come from micro-architectural differences
(e.g., ARM big.LITTLE systems) or maximum frequency at which CPUs can run
(e.g., SMP systems with multiple frequency domains). Heterogeneity in this
context is about differing performance characteristics; this binding tries to
capture a first-order approximation of the relative performance of CPUs.
CPU capacities are obtained by running a suitable benchmark. This binding makes
no guarantees on the validity or suitability of any particular benchmark, the
final capacity should, however, be:
* A "single-threaded" or CPU affine benchmark
* Divided by the running frequency of the CPU executing the benchmark
* Not subject to dynamic frequency scaling of the CPU
For the time being we however advise usage of the Dhrystone benchmark. What
above thus becomes:
CPU capacities are obtained by running the Dhrystone benchmark on each CPU at
max frequency (with caches enabled). The obtained DMIPS score is then divided
by the frequency (in MHz) at which the benchmark has been run, so that
DMIPS/MHz are obtained. Such values are then normalized w.r.t. the highest
score obtained in the system.
==========================================
3 - capacity-dmips-mhz
==========================================
capacity-dmips-mhz is an optional cpu node [1] property: u32 value
representing CPU capacity expressed in normalized DMIPS/MHz. At boot time, the
maximum frequency available to the cpu is then used to calculate the capacity
value internally used by the kernel.
capacity-dmips-mhz property is all-or-nothing: if it is specified for a cpu
node, it has to be specified for every other cpu nodes, or the system will
fall back to the default capacity value for every CPU. If cpufreq is not
available, final capacities are calculated by directly using capacity-dmips-
mhz values (normalized w.r.t. the highest value found while parsing the DT).
===========================================
4 - Examples
===========================================
Example 1 (ARM 64-bit, 6-cpu system, two clusters):
The capacities-dmips-mhz or DMIPS/MHz values (scaled to 1024)
are 1024 and 578 for cluster0 and cluster1. Further normalization
is done by the operating system based on cluster0@max-freq=1100 and
cluster1@max-freq=850, final capacities are 1024 for cluster0 and
446 for cluster1 (578*850/1100).
cpus {
#address-cells = <2>;
#size-cells = <0>;
cpu-map {
cluster0 {
core0 {
cpu = <&A57_0>;
};
core1 {
cpu = <&A57_1>;
};
};
cluster1 {
core0 {
cpu = <&A53_0>;
};
core1 {
cpu = <&A53_1>;
};
core2 {
cpu = <&A53_2>;
};
core3 {
cpu = <&A53_3>;
};
};
};
idle-states {
entry-method = "psci";
CPU_SLEEP_0: cpu-sleep-0 {
compatible = "arm,idle-state";
arm,psci-suspend-param = <0x0010000>;
local-timer-stop;
entry-latency-us = <100>;
exit-latency-us = <250>;
min-residency-us = <150>;
};
CLUSTER_SLEEP_0: cluster-sleep-0 {
compatible = "arm,idle-state";
arm,psci-suspend-param = <0x1010000>;
local-timer-stop;
entry-latency-us = <800>;
exit-latency-us = <700>;
min-residency-us = <2500>;
};
};
A57_0: cpu@0 {
compatible = "arm,cortex-a57";
reg = <0x0 0x0>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A57_L2>;
clocks = <&scpi_dvfs 0>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <1024>;
};
A57_1: cpu@1 {
compatible = "arm,cortex-a57";
reg = <0x0 0x1>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A57_L2>;
clocks = <&scpi_dvfs 0>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <1024>;
};
A53_0: cpu@100 {
compatible = "arm,cortex-a53";
reg = <0x0 0x100>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A53_1: cpu@101 {
compatible = "arm,cortex-a53";
reg = <0x0 0x101>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A53_2: cpu@102 {
compatible = "arm,cortex-a53";
reg = <0x0 0x102>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A53_3: cpu@103 {
compatible = "arm,cortex-a53";
reg = <0x0 0x103>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A57_L2: l2-cache0 {
compatible = "cache";
};
A53_L2: l2-cache1 {
compatible = "cache";
};
};
Example 2 (ARM 32-bit, 4-cpu system, two clusters,
cpus 0,1@1GHz, cpus 2,3@500MHz):
capacities-dmips-mhz are scaled w.r.t. 2 (cpu@0 and cpu@1), this means that first
cpu@0 and cpu@1 are twice fast than cpu@2 and cpu@3 (at the same frequency)
cpus {
#address-cells = <1>;
#size-cells = <0>;
cpu0: cpu@0 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <0>;
capacity-dmips-mhz = <2>;
};
cpu1: cpu@1 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <1>;
capacity-dmips-mhz = <2>;
};
cpu2: cpu@2 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <0x100>;
capacity-dmips-mhz = <1>;
};
cpu3: cpu@3 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <0x101>;
capacity-dmips-mhz = <1>;
};
};
===========================================
5 - References
===========================================
[1] ARM Linux Kernel documentation - CPUs bindings
Documentation/devicetree/bindings/arm/cpus.yaml
3. 한국어 전문 번역
영어 원문의 문단 순서와 의미를 유지한 전체 번역입니다. 코드, 함수명, symbol과 URL은 원문 표기를 유지합니다.
CPU heterogeneity 정보
1-13한 chip 안에서 power/performance 특성이 다른 CPU를 함께 구성할 수 있습니다. Kernel이 차이를 인식하고 scheduling 결정을 내리려면 devicetree가 추가 정보를 제공해야 합니다.
CPU capacity 정의와 측정
14-41CPU capacity는 scheduler에 CPU heterogeneity를 알려 주는 수치입니다. 차이는 ARM big.LITTLE 같은 microarchitecture 차이 또는 여러 frequency domain을 가진 SMP의 최대 frequency 차이에서 올 수 있습니다. 이 값은 CPU 상대 성능의 first-order approximation입니다.
적절한 benchmark로 capacity를 얻어야 합니다. 특정 benchmark의 유효성이나 적합성을 보장하지는 않지만 single-threaded 또는 CPU-affine이어야 하고, 실행 CPU frequency로 결과를 나누며, 실행 중 dynamic frequency scaling의 영향을 받지 않아야 합니다.
현재 권장 방식은 cache를 enable한 상태에서 각 CPU를 maximum frequency로 두고 Dhrystone을 실행하는 것입니다. DMIPS score를 실행 frequency(MHz)로 나눠 DMIPS/MHz를 구한 뒤 system에서 가장 높은 score를 기준으로 normalize합니다.
Raw benchmark 결과를 frequency-independent relative capacity로 변환합니다.
capacity-dmips-mhz 속성
42-56`capacity-dmips-mhz`는 CPU node의 선택적 u32 속성이며 normalized DMIPS/MHz로 CPU capacity를 나타냅니다. Boot 시 kernel은 해당 CPU의 maximum available frequency를 사용해 내부 capacity 값을 계산합니다.
이 속성은 all-or-nothing입니다. 한 CPU node에 지정했다면 모든 CPU node에 지정해야 하며, 하나라도 빠지면 모든 CPU가 default capacity로 fallback합니다.
cpufreq를 사용할 수 없으면 DT parsing 중 발견한 최고 `capacity-dmips-mhz` 값을 기준으로 이 값 자체를 normalize하여 final capacity를 계산합니다.
Property 완전성 및 cpufreq availability에 따른 계산 경로입니다.
ARM64 6-CPU, 2-cluster 예제
57-194첫 예제는 ARM64 6-CPU system의 A57 cluster0 두 core와 A53 cluster1 네 core를 기술합니다. Scaled DMIPS/MHz 값은 cluster0이 1024, cluster1이 578입니다.
OS가 maximum frequency를 추가 반영합니다. Cluster0은 1100 MHz, cluster1은 850 MHz이므로 cluster0 final capacity는 1024, cluster1은 `578*850/1100 = 446`입니다.
CPU map, PSCI idle state, SCPI DVFS clock, shared L2 cache와 모든 CPU node의 `capacity-dmips-mhz`를 포함한 원문 예제를 그대로 보존합니다.
cpus {
#address-cells = <2>;
#size-cells = <0>;
cpu-map {
cluster0 {
core0 {
cpu = <&A57_0>;
};
core1 {
cpu = <&A57_1>;
};
};
cluster1 {
core0 {
cpu = <&A53_0>;
};
core1 {
cpu = <&A53_1>;
};
core2 {
cpu = <&A53_2>;
};
core3 {
cpu = <&A53_3>;
};
};
};
idle-states {
entry-method = "psci";
CPU_SLEEP_0: cpu-sleep-0 {
compatible = "arm,idle-state";
arm,psci-suspend-param = <0x0010000>;
local-timer-stop;
entry-latency-us = <100>;
exit-latency-us = <250>;
min-residency-us = <150>;
};
CLUSTER_SLEEP_0: cluster-sleep-0 {
compatible = "arm,idle-state";
arm,psci-suspend-param = <0x1010000>;
local-timer-stop;
entry-latency-us = <800>;
exit-latency-us = <700>;
min-residency-us = <2500>;
};
};
A57_0: cpu@0 {
compatible = "arm,cortex-a57";
reg = <0x0 0x0>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A57_L2>;
clocks = <&scpi_dvfs 0>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <1024>;
};
A57_1: cpu@1 {
compatible = "arm,cortex-a57";
reg = <0x0 0x1>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A57_L2>;
clocks = <&scpi_dvfs 0>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <1024>;
};
A53_0: cpu@100 {
compatible = "arm,cortex-a53";
reg = <0x0 0x100>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A53_1: cpu@101 {
compatible = "arm,cortex-a53";
reg = <0x0 0x101>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A53_2: cpu@102 {
compatible = "arm,cortex-a53";
reg = <0x0 0x102>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A53_3: cpu@103 {
compatible = "arm,cortex-a53";
reg = <0x0 0x103>;
device_type = "cpu";
enable-method = "psci";
next-level-cache = <&A53_L2>;
clocks = <&scpi_dvfs 1>;
cpu-idle-states = <&CPU_SLEEP_0 &CLUSTER_SLEEP_0>;
capacity-dmips-mhz = <578>;
};
A57_L2: l2-cache0 {
compatible = "cache";
};
A53_L2: l2-cache1 {
compatible = "cache";
};
};
DMIPS/MHz와 maximum frequency가 scheduler의 final capacity로 결합됩니다.
ARM32 4-CPU, 2-cluster 예제
195-232두 번째 예제는 ARM32 4-CPU system입니다. CPU0과 CPU1은 1 GHz, CPU2와 CPU3은 500 MHz에서 동작하며 `capacity-dmips-mhz`는 앞 두 CPU가 2, 뒤 두 CPU가 1입니다.
같은 frequency에서 CPU0/1이 CPU2/3보다 두 배 빠르다는 뜻입니다. 실제 final capacity는 이 microarchitectural ratio와 각 maximum frequency를 함께 반영합니다.
cpus {
#address-cells = <1>;
#size-cells = <0>;
cpu0: cpu@0 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <0>;
capacity-dmips-mhz = <2>;
};
cpu1: cpu@1 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <1>;
capacity-dmips-mhz = <2>;
};
cpu2: cpu@2 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <0x100>;
capacity-dmips-mhz = <1>;
};
cpu3: cpu@3 {
device_type = "cpu";
compatible = "arm,cortex-a15";
reg = <0x101>;
capacity-dmips-mhz = <1>;
};
};
CPU binding 참조
233-238CPU node 자체의 공통 속성은 ARM Linux Kernel CPU binding `Documentation/devicetree/bindings/arm/cpus.yaml`을 참조합니다.
요약과 해설
cpu-capacity.txt:1-238Benchmark 측정, all-or-nothing 속성과 frequency 기반 normalization을 설명합니다.