As the name implies FDE.AI - hints at the presence of AI. AI idea is that he will understand the script using the device, and based on that adjust the various settings of the system in real time. Also, the AI logic is closely tied to the temperature of the device.
All AI actions are written to the log. The main plus of the AI -
no power consumption.
What is implemented: GPU turbo: When you run the AI determine whether GPU Turbo can work on your device, and write it in the log. AI will monitor the load level on the video card. If there is a high load for a long time, then the AI will raise the frequency of the accelerator to the maximum by about 30 seconds before the next inspection. As soon as the load decreases, the value returned to stock. All this does not increase the frame rate, but will make it more stable, because the frequency of the accelerator traffic controller does not have to jump over the frequencies at high load, which ultimately have a positive impact on the FPS indicator of stability. In logic there is no reference to the name of the application - the GPU load is measured directly, it does not matter that it loads (game benchmark, etc.), thus making it a universal logic. GPU Turbo will not conflict with the logic of Huawei with the same name. When the device temperature rises threshold GPU Turbo, to avoid overheating.
The logic is only active when ON screen. When switched off, the frequency of the screen back in stock value. It works on most video cards Adreno and Mali.
CPU turbo: When you run the AI will determine whether the CPU Turbo can work on your device, and write it in the log. AI will watch the level of CPU load. At high load for a long time to raise responsiveness AI regulator CPU and cause it to keep the frequency of the above 'high speed freq.' (About half of the maximum frequency of the CPU) for ~ 30 seconds before the next inspection. In logic there is no reference to the name of the application - CPU load is measured directly, it does not matter that it loads (game benchmark, etc.), thus making it a universal logic. The higher the device temperature - the higher the threshold CPU Turbo (to avoid overheating).
The logic is only active when ON screen. When the screen is off settings return to their initial values. It works with interactive, ondemand, schedutil and based on their CPU regulators.
Frame-buffer boost: When you run the AI determine whether Framebuffer boost can work on your device, and write it in the log. At the moment, this logic will be used only if there is support for GPU Turbo. Framebuffer boost - the acceleration frame buffer. When accelerating rise rate frame buffer. When using the active GPU (load more than 30%) framebuffer boost is activated automatically, and as a result of the refresh rate is the maximum stable.
The logic is only active when ON screen. When switched off, the frequency of the screen back in stock value. Supported on most devices with SoC Qualcomm.
Force DOZE: The option is available for devices running version 6.0 and higher. When the screen device is forced into the deep sleep mode (Doze), then, after 3 seconds, the forced operation mode is canceled and Doze continues in the normal mode, leaving the device to sleep.
Machine learning: At the moment, machine learning (MO) is used only if there is a GPU or CPU Turbo support. Defense will examine how you use your device and customize some system parameters in real time. For example, if you play a lot - the AI will adjust the system for better performance. If you simply write and watch videos - AI will adjust the system for better energy efficiency. If you often do both - AI will decide for himself what to do, for this and made machine learning. The application can be configured FDE constant MO selection preference towards performance or energy efficiency.
Dynamic customization VM: Every ~ 30 seconds checked RAM usage. Depending on the data obtained VM adjust some settings. As a result, should improve the management of cache in memory.
The logic is active only when the ON display on devices with 4+ GB of RAM.
throttling switch: This option can be enabled in the application FDE (default is OFF). It activates or deactivates for throttling the CPU and graphics accelerator. When throttling is turned off - the unit may overheat, but the performance will be at the maximum even during heavy use.
Temperature controller: AI checks the temperature of the device and notifies the log in if it overheats. While this is all that makes this logic.
"Chat" with AI: (Text on a single line - 96 characters)
Starting with version 7 application has the opportunity to correspond with the AI. This can be viewed either as another method of setting the AI, or just for fun. OCR is searched by keywords. The option is still in development. Settings in the "chat" reflect the settings in the application menu.
What is working:
- Display of the message (#help)
- Display short of Statistics (#stats)
- AI setting for the add. performance (#performance)
- AI setting for the add. Energy Saving (#powersave)
- AI Setting the 'balance' (#adaptive) mode
- Setting the AI to keep the turbo TUST heavy games (#hardgaming)
- Display the list of games, appreciated in AI as "severe" (#gamelist)
- Clears the chat / log (#clean #chat)
- Close background applications (#clean #ram)
- Cleans the cache files of applications (#clean #cache)
- Optimizes cache ART (#art)
- Activates a dangerous flag FS, which increases the rate of writing into the memory with the risk of loss of data and memory corruption (#nobarrier)
- Turn off the launch script runoff Qualcomm settings (#dontqcom)
- Skip the use of additional tweaks build.prop for the GPU during the boot process (#skipgputweak)
- Responds to hi / until some other words on the subject (#hi, #bye)
- Jokes (#joke)
- Responds to hvalbu (well done, like you, love, etc.)
- He knows who Daddy
- Other ..