Comprehensive Guide to iOS App Startup Optimization for Sohu Video

In early 2021 the Sohu Video iOS team launched a ten‑month startup optimization effort that cut the app's launch time from the 2‑second range to about 1 second, a 46% reduction, by applying a series of systematic performance improvements.

Measuring launch time was performed with Xcode Instruments (flame graphs) and a custom tool called HHTimeProfile , which records per‑function execution times to 0.1 ms precision and generates flame graphs for both the pre‑main and didFinishLaunching phases.

Pre‑main phase optimizations included:

Converting 19 third‑party dynamic libraries to static libraries (using use_frameworks! :linkage => :static in the Podfile), reducing pre‑main latency by ~200 ms.

Removing ~450 redundant classes and categories, and eliminating unused SDKs.

Refactoring +load methods and C/C++ static initializers into explicit initialization code, saving 50‑100 ms.

Pre‑warming system dynamic libraries via a push‑notification extension, yielding a 20‑500 ms gain depending on device.

Binary reordering to colocate frequently called symbols, theoretically saving ~100 ms.

Prioritizing startup items and implementing a self‑registration mechanism using custom Mach‑O sections, cutting another ~200 ms.

Applying a time‑slicing strategy to split large tasks into smaller chunks, reducing perceived latency by ~200 ms.

Asynchronously fetching system settings (e.g., push‑notification flags), which lowered synchronous work by ~9 ms.

Adopting a double‑root view architecture to bypass unnecessary UI initialization when the app is launched via push, saving ~100 ms.

Data from real devices (iPhone 6 Plus, iPhone 7 Plus) showed average launch time reductions of 500 ms on older hardware and modest gains on newer models. Table examples illustrate the impact of push‑triggered launches versus normal launches.

Device

Normal Launch (ms)

Push Launch (ms)

iPhone 6 Plus

2307

1792

iPhone 7 Plus

631

574.5

The article concludes that launch speed is critical for user retention—each second of delay can cost ~7% of users—and that continuous profiling, incremental improvements, and monitoring (dashboards, daily reports) are essential to maintain a healthy startup performance as iOS evolves.