Позначка: android

Google is developing a new feature called AutoFDO, designed to enhance the speed and battery life of Android smartphones.

Previously, discussions centered around the Galaxy S26 simulator, which converts any iPhone or Android device into a Samsung interface. Meanwhile, Android 17 has introduced a tool named DeliQueue.

AutoFDO operates by utilizing real-world instruction execution patterns to guide the compiler. These patterns are the most commonly executed instruction paths during actual code usage and are captured by recording the processor’s branch history.

According to Google, AutoFDO will function in the Android kernel by default, reverting to traditional methods if a process deviates from predefined patterns. The updates will be rolled out in the latest kernel versions of Android16-6.12, Android15-6.6, and Android17-6.18. This optimization aims to speed up the user interface, improve app switching, extend battery life, and enhance device responsiveness.

The team behind the Android LLVM tool announced the kernel update with AutoFDO — an automatic feedback-driven optimization. Smartphones in standby mode make thousands of decisions that demand significant processing resources.

AutoFDO directs the compiler along “the most common paths” of execution, reducing workload and freeing up more computing power for other tasks. Consequently, energy consumption is lowered, which extends battery life.

Google explains that during standard software compilation, the compiler makes numerous small decisions, such as whether to inline a function or which variant of a conditional operator to use, based on statistical hints from the code. While these methods are beneficial, they do not always accurately predict code execution in real-world phone usage.

“Although data can be collected from network-connected devices, we synthesize it in a lab environment for the kernel, using representative workloads like launching the 100 most popular apps. A sampling profiler gathers this data, identifying which code segments are ‘hot’ and which are ‘cold.’ When we recompile the kernel with these profiles, the compiler can make much smarter optimization choices tailored to realistic Android workloads,” Google highlights.

Initial tests by the company noted a 2.1% improvement in loading times, a 4.3% boost in launching idle apps, and significant enhancements in other areas that might not be immediately noticeable to the average user. These patterns were developed based on the 100 most popular smartphone apps to simulate real-world usage. Subsequently, they were fine-tuned for the most frequently used code sections.

Google has released the first beta of Android 17: what’s new?

“Although data can be collected from network-connected devices, we synthesize it in a lab environment for the kernel, using representative workloads like launching the 100 most popular apps. A sampling profiler gathers this data, identifying which code segments are ‘hot’ and which are ‘cold.’ When we recompile the kernel with these profiles, the compiler can make much smarter optimization choices tailored to realistic Android workloads,” Google highlights.

What are your thoughts on this article? Votes: Quite fascinating Interesting Frustrating Quite intriguing… Annoying, even.