As you can see from the name FDE.AI - hints at the presence of a real AI. The meaning of the AI is that it will understand the usage scenarios and on this basis will adjust the various settings of the device in real time. The AI is tied to the temperature of the device - so there will be no acceleration if the device overheats. The function is in development.
All AI actions are written to the log.
fde.txt. Well, the main plus AI -
no power consumption.
What is implemented: RAM monitoring: For all devices with 3GB RAM or less, the number of free RAM with a uniquely calculated response algorithm for each device is monitored. If the AI understands that there is too little free RAM, and LMK does not cope with the work, the AI will clear the RAM of the unused cache.
FStrimmer: To keep the file system in shape, the AI will execute fstrim commands for all partitions once or twice a day, depending on how you use the device.
ZRAM protection: In the old kernels there is a bug when using zRAM - the device works well for about a day, then it starts wildly because zRAM is filled to the limit, and the bug is that zRAM is not automatically released. AI will solve this problem by clearing zRAM. As a result, devices with a similar bug should not lag after prolonged use of zRAM.
GPU turbo: For starters, the AI will determine if the FDE GPU turbo can work on your device and write about it in the log
fde.txt. The AI will monitor the load on the video accelerator. If there is a high load for a long time, then the AI will raise the accelerator frequency to the maximum. As soon as the load falls, the real-time stock values are returned. All this will not make your games faster, but will make them.
more stable, because the accelerator frequency controller will not have to jump in frequencies under high load, which ultimately should lead to a stable FPS index. In logic, there is no binding to the name of the application - the GPU load is measured directly, no matter what loads it (game, rendering, benchmark, etc.), thereby making the logic universal. FDE GPU turbo will not conflict with Huawei's similar logic.
CPU turbo: For starters, the AI will determine if the FDE CPU turbo can work on your device and write about it to the log
fde.txt. The AI will monitor the CPU load level. With a high load for a long time, the AI will increase the responsiveness of the CPU and force the CPU frequency to keep above 'high speed freq.' (this is about half the maximum frequency of the CPU) for 30 seconds. In logic, there is no binding to the name of the application - the CPU load is measured directly, no matter what loads it (game, rendering, benchmark, etc.), thereby making the logic universal.
Machine learning: At the moment machine learning (MO) will be used only if there is support for GPUturbo. MO will study how you use the device and adjust some system parameters in real time. For example, if you play a lot - AI will tune the system for better performance. If you just chat and watch the video - the AI will tune the system for better power saving. If you often do both, the AI will decide what to do, machine learning is done for this.
There will be more ..