kernel/cpu power

This is intended as a way to document how to apply cpu power limits to AMD Ryzen CPUs.

power-profiles-daemon

powerprofilesctl Show current status
powerprofilesctl set power-saver Change mode to power-saver
powerprofilesctl list-actions List all possible actions and their status
powerprofilesctl configure-action amdgpu_dpm --enable Enable amdgpu_dpm action
powerprofilesctl configure-battery-aware --disable Disable dynamic changes on battery status change.

ls /sys/devices/system/cpu/cpufreq/policy0/*amd*
echo 0 > /sys/devices/system/cpu/cpufreq/policy*/boost Disable boost
for i in /sys/devices/system/cpu/cpufreq/policy{0..15}/scaling_max_freq ; do echo 3500000 > $i ; done Set max freq to 3.5GHz
for i in /sys/devices/system/cpu/cpufreq/policy{0..15}/scaling_min_freq ; do echo 400000 > $i ; done Set min freq to 400MHz
for i in /sys/devices/system/cpu/cpu{1,3,5,7,8,9,10,11,12,13,14,15}/online; do echo 0 > $i ; done Power off cores
cat /sys/devices/system/cpu/online Check online cores
while : ; do : ; done Put load on a single thread.
``

Core 0 can not be shutted down as some architectures may have some special dependency on a certain CPU.

echo 0 | sudo tee /sys/devices/system/cpu/cpu4/online

Possible states

cat /sys/devices/system/cpu/hotplug/states

Change a target

echo 140 > /sys/devices/system/cpu/cpu4/hotplug/target

List governors

sudo cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors

Set governor

echo performance | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

cat /proc/interrupts

cat /proc/stat
---
cpu  863551 3429 857793 8085941 31541 33767 19313 0 33353 0

Meaning:

user: normal processes executing in user mode
nice: niced processes executing in user mode
system: processes executing in kernel mode
idle: twiddling thumbs
iowait: In a word, iowait stands for waiting for I/O to complete. But there are several problems:
Cpu will not wait for I/O to complete, iowait is the time that a task is waiting for I/O to complete. When cpu goes into idle state for outstanding task io, another task will be scheduled on this CPU.
In a multi-core CPU, the task waiting for I/O to complete is not running on any CPU, so the iowait of each CPU is difficult to calculate.
The value of iowait field in /proc/stat will decrease in certain conditions. So, the iowait is not reliable by reading from /proc/stat.
irq: servicing interrupts
softirq: servicing softirqs
steal: involuntary wait
guest: running a normal guest
guest_nice: running a niced guest
https://www.kernel.org/doc/Documentation/filesystems/proc.txt