KAIST researchers introduced Agent‑X, a pure‑software acceleration framework that eliminates prefill and decode bottlenecks on mobile devices, achieving a 1.61× end‑to‑end speedup for on‑device AI agents without any loss in task accuracy.
By combining low‑overhead eBPF data collection with AI‑driven diagnosis and an agent‑based execution layer, the authors present a three‑tier system that shifts mobile optimization from peak performance to sustained energy efficiency, achieving sub‑1% monitoring overhead and up to 20% power savings in real‑world video workloads.
During the 2026 Chinese New Year, Xiaohongshu built a real‑time dynamic interaction system that combined adaptive scheduling, high‑performance particle rendering, and industrial‑grade ASR to deliver synchronized voice greetings and emoji fireworks to over a billion daily active users across heterogeneous mobile devices.
This article explains how the Skyline rendering engine improves WeChat Mini Programs by eliminating first‑frame white screens, accelerating page transitions, and delivering native‑like experiences through data pre‑loading, animation techniques, and detailed migration steps.
This guide outlines comprehensive performance indicators for Android and iOS apps—including startup time, page load speed, responsiveness, resource consumption, stability, network efficiency, interaction quality, background task handling, installation, compatibility, security, and low‑end device adaptation—providing industry‑standard thresholds for each metric.
This article details three major optimization directions—component scheduling, component reuse, and component governance—applied to Taobao live streaming, explaining the technical redesign, task scheduler implementation, RecyclerView view‑holder reuse, performance gains, and future improvement plans.
This article details Ctrip's approach to weak network detection, covering background, data collection, processing, dynamic weighting algorithms, result output, deployment effects, and future plans, and provides practical code examples and threshold settings for improving mobile network performance.
This article details Ctrip's end‑to‑end weak‑network identification model, covering background, metric selection, data collection on iOS and Android, processing pipelines with dynamic weighting, weighted median calculations, success‑rate trends, threshold tuning, and deployment results across multiple platforms.
This article details the design and implementation of a passive weak‑network diagnosis framework for a mobile app, covering background, overall architecture, data collection of HttpRTT and throughput, filtering strategies, weighted median calculations, caching, interface design, threshold definitions, performance metrics, and practical application scenarios.
Optimizing a mobile app’s first frame—by defining scope, measuring latency, using profiling tools, and applying strategies such as pre‑loading, lazy initialization, parallel processing, and skeleton screens—boosts brand perception, conversion rates, and resource efficiency, while requiring continuous monitoring, A/B testing, and anti‑degradation safeguards.
The article explains why first‑frame latency matters for video apps and presents a comprehensive set of optimization methods—including pre‑fetching URLs, intelligent preloading, prerendering, and scenario‑specific tweaks for cold start, short‑video scrolling, page navigation, and long‑video playback—to dramatically reduce start‑up time and improve user experience.
The article analyzes IO‑intensive bottlenecks in mobile apps, explains how MMKV’s mmap‑based storage, rewrite, and expansion mechanisms cause main‑thread stalls, and presents concrete optimizations such as value comparison before write, pre‑expansion, compression, expiration handling, and proper instance management to dramatically reduce latency.
This article presents a detailed summary of the AutoHome client team's performance optimization efforts for both the native mobile app and the H5 web frontend, covering background, specific tool and platform enhancements, metric‑driven improvements, process standardization, and future quality‑boosting initiatives.
Effective mobile page performance requires identifying three jank types—screen tearing, frame drops, and long unresponsiveness—monitoring metrics such as response time, animation latency, idle time, and SM, understanding the CPU‑GPU rendering pipeline, and applying optimizations like hardware acceleration, transform‑based animations, reduced layout thrashing, task slicing, and GPU‑friendly techniques.
A comprehensive battery endurance test using Geekbench 4 on six iPhone models (iPhone 8, SE2, XR, 11, 12, 13) shows that iOS 16.4 dramatically reduces runtime for all devices, with iPhone 12 and 13 losing about an hour compared to earlier iOS versions.
The article explains Baidu App's memory‑control solution that monitors real‑time memory usage, predicts page‑level memory growth, defines safe and dangerous memory watermarks, and triggers proactive or reactive actions to avoid OOM crashes on low‑end iOS devices.
Alipay’s client engineering team details a comprehensive approach to monitoring, measuring, and improving time‑consuming user experiences—especially cold‑start—by employing video frame analysis, ActivityTaskManager, extensive instrumentation, home‑page snapshot techniques, temperature control, patch‑APK injection, AOP‑based diagnostics, and the Spider SDK within a robust APM platform.
This article analyzes why image loading can become a bottleneck in iOS apps like NetEase Cloud Music, identifies factors such as image size, network conditions, and cache behavior, and presents a comprehensive solution that standardizes URL parameters, reuses local images, and extends SDWebImage's manager and cache to achieve up to 50% faster downloads and significant CDN bandwidth savings.
This article analyzes JDHybrid's offline loading architecture for Android and iOS, detailing mechanisms such as direct file loading, request interception, WKURLSchemeHandler, package generation, download tiering, preloading strategies, memory warm‑up, debugging tools, and performance monitoring that together boost first‑screen speed and reduce bandwidth during high‑traffic events.
This article details the background, analysis, and step‑by‑step solutions for reducing package size, improving cold‑start time, and establishing a comprehensive mobile performance testing framework for both Android and iOS versions of the Xueersi Online School app.
Thresh is a Flutter-based cross-platform dynamic solution for frontend developers that integrates the JavaScript Engine Interface (JSI) to streamline communication between JavaScript and Dart, reducing serialization overhead and thread switching, thereby improving startup and rendering performance on mobile devices.
This article explains how a performance optimization team quickly designed a hardware scoring framework for mobile games, detailing the definition of key terms, the background and methodology of hardware scoring, the analysis of major GPU families, and a step‑by‑step guide to creating and applying scoring rules in a tight project timeline.
The design separates the original I/O pool into dedicated network, I/O, and polling thread pools, adds comprehensive monitoring of task duration and frequency, enforces unified polling rules, and automatically tunes pool parameters, resulting in a 76 % reduction in UI lag and easier troubleshooting.
This article presents a method for precisely measuring mobile app response latency by extracting video frames, detecting start and end frames through image markers and OCR, and calculating the time difference, offering a high‑precision, customizable solution for performance evaluation across diverse app scenarios.
The article details Kuaishou's Y-tech AI SDK (YKit) architecture, covering its design for computer vision capabilities, performance optimization strategies for mobile devices, and real-world case studies such as GAN-based effects and intelligent matting, outlining challenges and future directions.
Baidu’s end‑to‑end optimization of commercial landing pages—splitting container initialization, parallelizing billing, preloading critical resources, and employing server‑ and client‑side pre‑rendering—cuts load times below the 1‑second threshold that Baidu’s data shows is essential to prevent a 5% conversion drop, though implementation demands significant development effort.
This article details how a 3D interactive incentive game was created using Oasis Engine, covering scene architecture, ECS design, script development, collision handling, camera follow, skeletal and particle animations, shader effects, lane configuration, mobile orientation, performance optimizations, testing, deployment, and key lessons learned.
This article explores why fast app startup matters, using Gaode Map as a case study to detail techniques such as Universal/App Links, H5 launch page handling, download size reduction, Android .so loading hooks, thread scheduling, task orchestration, and the tooling needed for performance analysis on both iOS and Android.
This article details the design and deployment of JD Mall's mobile app stutter monitoring system, introducing the stutter rate metric, describing data collection, analysis, and reporting modules, presenting iOS and Android implementation code, and summarizing the performance improvements achieved across the platform.
A humorous Zhihu dialogue illustrates how iOS and Android manage RAM and background processes differently, explaining why iOS often feels smoother despite having less memory, by personifying apps like WeChat, Alipay, and Taobao, the answer highlights iOS's stricter resource control versus Android's more generous allocation.
This article introduces the background, core features, network and page performance modules, error monitoring, reporting, and optimization practices of a mobile App Performance Management (APM) system, highlighting its impact on monitoring, diagnostics, and performance improvement for large‑scale applications.
The article explains how Baidu App’s smoothness monitoring distinguishes jitter and stutter, sets 30 ms and 70 ms thresholds for jank detection, and adopts the per‑second and stage‑converted jank rate metrics—derived from jank time—to provide intuitive, bounded measurements while avoiding FPS limitations.
This guide explains why energy consumption is a critical performance metric for iOS apps, introduces core concepts such as idle, active and overhead states, identifies the main power‑hungry components, and walks through Apple’s profiling tools—Energy Impact, Energy Log, Sysdiagnose and MetricsKit—plus concrete optimization techniques for network, location, CPU, GPU and notifications.
To address the slow loading and high bandwidth costs of GIFs in short‑content feeds, the team evaluated APNG, WebP, SharpP, and MP4, found MP4 video conversion offers the best compression, universal support and comparable CPU usage, achieving a 62% size reduction, 90% instant‑open rate and a 5.6% exposure boost.
A detailed case study shows how a mobile game's sudden frame‑rate drop on a Galaxy S9+ was traced to thermal throttling, highlighting the importance of correlating CPU, GPU, memory, and temperature data during performance testing.
The article introduces Shooter, a mobile performance monitoring platform for Android that provides non‑intrusive, lightweight SDKs to monitor app startup, jank, network, native page, and WebView performance, explains the technical solutions, data collection methods, and showcases console visualizations and real‑world empowerment cases.
The project replaces karaoke effects with ASS‑based lyrics on Android, using a backend to generate ASS files and a JNI renderer that rasterizes and composites bitmap fragments; optimizations like transparent‑fragment filtering, opacity early‑out, NEON SIMD, copy elimination, and double‑buffered decoding cut frame time from 52 ms to under 3 ms, raise frame rate from 7 fps to 15 fps, and shrink memory usage from 180 MB to under 10 MB.
Matrix‑ApkChecker, an in‑house APM tool used by WeChat, analyzes Android APKs to identify oversized files, redundant resources, unused assets, and other inefficiencies, providing detailed reports and enabling targeted optimizations such as removing duplicate files, converting PNGs to WebP, and stripping unused libraries, ultimately reducing app size by over 130 KB.
This article explains the concept of response time in mobile applications and provides a detailed, image‑rich workflow—including video recording with ReplayKit, simulated interactions, video parsing, log extraction, timestamp alignment, and stability detection—to accurately measure and reduce response latency.
This article details Suning Finance's systematic approach to mobile login optimization, covering the discovery of root causes, the establishment of end‑to‑end monitoring, step‑by‑step workflow analysis, network link enhancements, request consolidation, and standardized cookie management that together dramatically improved response time and success rates.
Baidu Maps confronts severe performance bottlenecks as its client app scales, analyzing main‑thread overload, third‑party SDK interference, and unstable network latency, and introduces the Nirvana framework to provide internal resource scheduling, monitoring, and isolation, achieving 10‑50% speed gains and noticeable user improvements.
This article explains why H5 pages in mobile apps suffer white‑screen delays and presents a comprehensive set of frontend, client‑side, and offline‑package optimizations—including caching strategies, preloading, WebView pooling, and server‑side rendering—to achieve near‑native startup performance.
This article explains why resource loading optimization is vital for mobile web development, shows how to measure page load speed with Chrome DevTools, breaks down the resource loading lifecycle, and provides practical solutions for queueing, high TTFB, and long download times.
This article explains how 360 Mobile Guard, a massive Android security app built on the RePlugin framework, implements a comprehensive performance monitoring system—including offline code checks, Argus APM tools, online crash and metric collection, diagnostic plugins, and future AI-driven analysis—to identify and resolve performance issues and improve user experience.
VasSonic, a lightweight high‑performance hybrid framework developed by Tencent’s VAS team, evolved from addressing slow H5 page loads in Mobile QQ to implementing terminal optimizations, static and dynamic rendering, parallel loading, and incremental updates, ultimately cutting first‑screen load time by more than 40%.
Meituan Dianping’s iOS development seminar presented expert insights on performance monitoring with Hertz, explored Swift monads through source‑code analysis, tackled dynamic‑library implementation hurdles, and shared cross‑platform code‑reuse strategies, offering practical solutions for functional programming and large‑scale mobile app optimization.
Hertz, Meituan Waimai’s mobile‑app performance monitoring framework, automatically detects development anomalies, generates test reports, and streams production metrics—such as FPS, CPU, memory, UI jank, page‑load time, and network traffic—using platform‑specific instrumentation and configurable modes to pinpoint and resolve performance bottlenecks across the app lifecycle.
This article details Tencent's practical approaches to reducing user‑perceived latency in mobile QQ and Qzone by analyzing server, network, and client delays, employing private protocols, multi‑path connection strategies, real‑time monitoring, and big‑data clustering to identify and fix performance bottlenecks.
This article presents a comprehensive, evolving mobile performance evaluation framework for O2O applications, detailing assessment parameters, multi‑stage automation approaches, uiautomator enhancements, and key technical optimizations to enable timely, extensible, and accurate performance testing of competitive mobile products.
This article explains how Tencent's myAPM platform provides method‑level injection and monitoring for mobile apps, enabling automatic APK size analysis, lag detection, startup performance tracking, and core‑link analysis to help developers identify and eliminate unused code and performance bottlenecks.
The article examines why traditional CDNs provide little speed benefit for mobile users, analyzes last‑mile latency on wireless networks, presents performance calculations, and discusses the operational and cost challenges of deploying edge servers closer to mobile users.
The article presents a low‑cost, high‑precision approach for measuring mobile web first‑screen load time by streaming the device screen to a computer via STF and applying Sikuli image‑recognition scripts to automatically detect when the first screen is fully rendered.
This article details the five major performance problems faced by the Mobile Taobao app, the Android profiling tools used to identify bottlenecks, and the step‑by‑step optimizations—including startup, UI rendering, memory management, and lazy loading—that reduced memory usage by nearly 50% and boosted frame rates by about 20%.
Ctrip's wireless development director shares a comprehensive overview of the app's network architecture, common performance pitfalls, and six concrete optimization techniques that boosted service success rates above 99% and cut latency by up to 200 ms across diverse network conditions.
The article analyzes Ilya Grigorik's findings that traditional CDNs deliver minimal performance gains for mobile users due to high last‑mile latency, outlines latency components, presents comparative data, and discusses the operational challenges of deploying mobile‑optimized edge caching.
This article outlines six practical techniques—background execution, pre‑loading content, caching, UI‑first networking, predictive task start, and input minimization—to make mobile apps feel faster and more responsive, illustrated with real‑world examples from Instagram, Weibo, App Store, WeChat, and more.
