Android Weekly Issue 2025-02

Word count: 1.2kReading time: 7 min

 2025/01/12

Android Weekly is a weekly newsletter focused on the Android technical ecosystem, updated every Monday. This newsletter delves into high-quality technical content in areas such as Android system architecture, performance optimization, cross-platform development, and AI, providing continuous knowledge updates and technical insights for developers.

You can subscribe and read through platforms like WeChat Official Account, Zhihu, Juejin, personal blogs, and Zhubai.

Technical Articles

Tools, not Rules: become a better Android developer with Compiler Explorer: This article introduces a tool called Compiler Explorer, which helps Android developers understand optimization mechanisms during compilation. It allows developers to explore how Java and Kotlin code is converted into assembly instructions and the various optimizations performed by the ART compiler. Through specific code examples, the article demonstrates how Compiler Explorer helps in understanding compiler inner workings to write more efficient Android code.
The Most Detailed Guide to AOSP Compilation and Debugging to Date: A comprehensive guide on AOSP compilation and debugging, covering the full download and compilation of Android 13 source code in an Ubuntu environment. It includes two ways to import source code into Android Studio, introduces the source tree structure, build products, code searching, and individual module compilation, as well as system debugging methods like breakpoint debugging for Java code.
Chat | Solving Bugs Like Sherlock Holmes: When solving bugs, one should be like Sherlock Holmes: collect sufficient data and tools, and look deep into principles rather than surface symptoms. It emphasizes having a broad vision to see the essence of problems, continuously learning from solved bugs, and objectively prioritizing high-risk issues while considering compromises for incurable problems.
Is High CPU Usage Always Efficient?: This article discusses why high CPU usage doesn’t necessarily mean high efficiency. Through experiments, the author analyzes performance differences between pessimistic locks (synchronized) and spinlocks (CAS, spin_lock) and provides optimizations for spinlocks. The conclusion highlights the “Memory Wall” as a core cause of performance issues.
AI Agent Technical Whitepaper (Google, 2024): This post introduces the basic architecture and principles of generative AI “Agents.” An Agent is an application that uses tools and orchestration layers to extend LLM capabilities, allowing for complex tasks and external system interaction. It explores the three core components: the model, tools (Extensions, Functions, Data Stores), and the orchestration layer, alongside methods to improve performance like in-context learning and fine-tuning.
LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects: A 48-page survey paper by vivo AI Lab and CUHK MMLab covering over 200 references. It systematically summarizes the development history, technical frameworks, application scenarios, and future challenges of LLM-driven phone automation agents. It highlights how LLMs overcome limitations like poor versatility and weak intent understanding in traditional automation.
Android Performance: Memory Optimization (Theory Part): This article covers foundational theory for Android memory optimization, including memory allocation, single-process limits, GC algorithms, reference types, and optimization strategies. It stresses the importance of avoiding OOMs to improve system efficiency and user experience.
Android Performance: Memory Optimization (Practice Part): A practical summary of memory optimization in Android, detailing common memory detection tools, OOM issues like Bitmap usage and memory leaks, and monitoring methods. It also touches on business-level optimizations such as device grading and service downgrading.
Preparing Your Apps for 16KB Page Size Devices: Helping developers prepare for the shift to 16KB page sizes in future Android devices.
Getting Started with CameraX in Jetpack Compose: An overview of creating camera previews in Compose using Android’s newer CameraX and Jetpack Compose libraries.
Practice and Mechanism Study of HarmonyOS App Signing: Introducing the signing mechanism for single-framework HarmonyOS apps, deconstructing the practice and underlying principles of signature verification.
Translation: How I Program with LLMs: The author shares personal experience with using LLMs for autocompletion, technical searches, and conversational programming, discussing new trends like readable tests and smaller package structures.
Discussing What Power Consumption Is: An article on mobile device power consumption from a software perspective. It covers CPU load, sensors, and screen usage, advocating for analysis from the process/thread level down to specific functions and tasks, linking key structures like Looper messages and Binder calls to power drain.
SPE Profiling and Its Usage: Introducing Arm Statistical Profiling Extension (SPE), which captures more precise CPU pipeline information (instruction addresses, data addresses, latency) compared to traditional PMUs.
Android Performance: Binding RenderThread to Big CPU Cores: A guide on how to bind threads to specific CPU cores in Android. It covers sched_setaffinity, gathering CPU frequency info to categorize cores, identifying RenderThread, and providing JNI implementation for Java invocations.
Performance Weekly 2025-01-04 Issue 3: The first performance-focused weekly of 2025.
Exploration of ARSC Compression After TargetSdk 30 Upgrade: Describing how Kuaishou re-compressed resources.arsc on TargetSdk 30 to achieve significant package size savings.
My Two Years in a Performance Team: An introduction to a frontend performance team’s journey, covering metrics definition, guarding mechanisms, and specific optimization techniques across the entire lifecycle.
Thoroughly Understanding Android Broadcast Knowledge (Part 1): A systematic exploration of principles, classifications, receivers, and senders within the Android broadcast mechanism.
Thoroughly Understanding Android Broadcast Knowledge (Part 2): Detailing the sending and receiving process, including queue management and interactions between senders, the distribution center, and receivers.

Non-Technical Articles

Orange’s 2024 Year-end Summary: In Pursuit of Happiness: Reflecting on a year of English learning, running, traveling, relationships, and work, discussing how to peacefully face life’s events to find happiness.
Sugar Revolution — Why Am I Sleepy After Eating Sweet Potatoes?: Exploring blood sugar management, analyzing how different foods impact curves, and sharing successful health improvements through continuous monitoring.
Reflections on the English Learning Process: Summarizing specific methods for building learning materials, reading, listening, and vocabulary acquisition after six months of serious study.
vivo X200 Pro User Experience: A comprehensive review covering purchase selection, daily usage, system operation, imaging, and battery life.
Dos and Don’ts for 2025: Pursuing life simplification by “subtracting” phone time and external judgments while “adding” writing, reading, and exercise. It advocates the “20+20” principle for home organization.
Reflections on Time Management: Experience sharing on time management and content output.
What We Learned from Large Language Models in 2024: Reviewing major LLM advancements in 2024, including performance breakthroughs, efficiency gains, the rise of multimodal models, and prompt-driven app generation, while noting challenges like rising complexity and knowledge distribution.
Independent Developer Weeklog 85: 2024 Year-end Summary.

Tools

E-book - Large Language Models: A technical history and current status of LLMs in industry and academia, discussing the challenges of “democratizing” core technical details.
Windsurf: A Detailed Look at Future-Ready AI Programming Tools: Windsurf offers advanced context awareness, multi-model AI, and real-time collaboration with its innovative “Flows” and “Cascade” features, promising a better AI-human collaboration paradigm.