先把原则写在最前

1. 非必要不调节，多数情况下预设的模式就够了。

底层的调节需要综合考虑性能、能耗、温度等等，最难的就是达到均衡，预设模式经过专业的工程师调节，力求体验。非专业用户可以调到激进，但可能出现副作用，比如用舒适温度换性能。

2. 如果核芯解放后出现异常发热或者卡顿的情况，建议回归预设效能（比如平衡模式）。

如果核芯解放调节后效果很好，欢迎分享论坛。

3基本原理概述

不同游戏对性能的调用有需求有差别，所以对不同游戏可以做不同调节。

核芯解放翻译如下，供大家参考:

Temperature: "2","3","4","5","6

温控, 数字越大越高温

dev/cpu_dma_latency "0","1"

是否让cpu进入睡眠

/sys/devices/system/cpu/cpufreq/policy0/scaling_governor :"0","1"

/sys/devices/system/cpu/cpufreq/policy4/scaling_governor

/sys/devices/system/cpu/cpufreq/policy7/scaling_governor

cpu运行速度的policy

0: 预设

1:最高速

MIN_ONLINE_CPU_CLUSTER_BIG :"2", "3"

大核至少运行的核心数, 3表示统统要运行, 2表示两个另一个可以OFF

MIN_ONLINE_CPU_CLUSTER_LITTLE : "2", "3", "4"

大核至少运行的核心数,4表示统统要运行, 2表示两个另2个可以OFF

SCHED_PREFER_SPREAD :"2", "3", "4"

让程序各个线程分配运行方式

0 Disable

1 在小核内积极散落运行

2 在大,小核内各自积极散落运行

3 在大小核内综合积极散落运行

4 在大小核内综合积极散落运行, 并可互选核心内空闲cpu

SCHED_UCLAMP_BOOST :"0", "10", "20", "30", "40"

让系统评估TASK BUSY程度,最低不能低于设定值 <--数值越大，让系统认为更BUSY, 会调高CPU 速度或是往大核超大核安排运行

MPCTLV3_LPM_BIAS_HYST:"0", "33", "66", "99"

计算CPU是否进入LOW POWER MODE的时间

数值越大越慢进入

sys/module/msm_performance/parameters/cpu_min_freq 0-3",

sys/module/msm_performance/parameters/cpu_min_freq 4-6",

sys/module/msm_performance/parameters/cpu_min_freq 7",

指定各个核心最小频率

sys/module/msm_performance/parameters/cpu_max_freq 0-3",

sys/module/msm_performance/parameters/cpu_max_freq 4-6",

sys/module/msm_performance/parameters/cpu_max_freq 7",

指定各个核心最大频率

/dev/cpuset/top-app/cpu : "1", "3", "15", "240", "127", "255"

/dev/cpuset/system-background/cpus: "1", "3", "15", "240", "127", "255"

/dev/cpuset/foreground/cpus : "1", "3", "15", "240", "127", "255"

/dev/cpuset/background/cpus: "1", "3", "15", "240", "127", "255"

最上层app, 前景app, 背景app, 背景service的cpu核心指定使用

使用cpu核心的bit mask,

例如

255=0xFF 表示 8 核心全用,

240 = 0xF0, 表示大四核使用

15 =0x0F, 表示只使用小核

/sys/class/kgsl/kgsl-3d0/min_pwrlevel :"9"-"0"

锁定GPU最低速 9:315Mhz 0: 840Mhz

/sys/class/kgsl/kgsl-3d0/max_pwrlevel

锁定GPU最高速 9:315Mhz 0: 840Mhz

/sys/class/kgsl/kgsl-3d0/idle_timer: "80","500","1000","2500","5000","10000"

gpu睡眠前空闲等待时间

单位 ms (毫秒)

/proc/sys/kernel/sched_downmigrate

/proc/sys/kernel/sched_upmigrate

/proc/sys/kernel/sched_group_downmigrate

/proc/sys/kernel/sched_group_upmigrate

核心运行工作排程

若工作运行忙碌超过upmigrate 数字(percentage), 将工作往大核或超大核排程

若工作运行忙碌低于dowbmigrate 数字(percentage), 将工作往大核或小核排程

/sys/devices/system/cpu/cpuX/sched_load_boost : "0","10","20","30","40","50","60","70"

预设某cpu的忙碌程度

scheduler根据cpu忙碌程度来调控频率以及大小和排程

将此值设大一点就会比较高速也可能比较容易往大核跑

/sys/devices/system/cpu/cpu0/cpufreq/schedutil/hispeed_load : "90", "80", "70", "60", "50", "40"

/sys/devices/system/cpu/cpu0/cpufreq/schedutil/hispeed_freq : "1209", "1497", "1804"

/sys/devices/system/cpu/cpu4/cpufreq/schedutil/hispeed_load

/sys/devices/system/cpu/cpu4/cpufreq/schedutil/hispeed_freq

/sys/devices/system/cpu/cpu7/cpufreq/schedutil/hispeed_load

/sys/devices/system/cpu/cpu7/cpufreq/schedutil/hispeed_freq

设定当cpu忙碌程度到达一定值时立刻跳到设定的频率

/sys/class/devfreq/soc:qcom,cpu-cpu-llcc-bw/bw_hwmon/io_percent

/sys/class/devfreq/soc:qcom,cpu-llcc-ddr-bw/bw_hwmon/io_percent

设定cpu/L3 cache/ddr/之间bus 速度百分比,数值越低跳得越高速

/sys/class/devfreq/18590100.qcom,cpu%d-cpu-l3-lat/min_freq : "3000","4032","6144","7104","8832","10560","12480","14208","15168","15936"

指定cpu/L3 memory bus speed频率

/sys/class/devfreq/18590100.qcom,cpu0-cpu-l3-lat/mem_latency/ratio_ceil: "400","1000","2000","4000","8000","16000"

设定memory bus的频率天花板, 数值越高越容易往上升频

STORAGE_CLK_SCALING_DISABLE STORAGE_CLK_GATING_DISABLE

UFS clock是否要关闭或调频

IO1,IO2

各cpu L3 cache是否要休眠