Tencent Meeting C++ Cross-Platform Compilation Acceleration Practice

As a naturally cross-platform product, Tencent Meeting adopted a homogeneous architecture approach from the first line of code—one unified architecture serving all scenarios. Over the past year, Tencent Meeting iterated and optimized 20,000 features, stably supporting hundreds of millions of users. The client alone has over 1 million lines of upper-layer business logic code. After optimization, compilation time on Windows has been reduced to as fast as 10 seconds, becoming a benchmark for C++ cross-platform projects in the industry.

The article details the compilation acceleration strategies: First, it analyzes two feasible approaches for module compilation exemption—module independent operation and module pre-compilation. The pre-compilation solution was chosen for higher feasibility. The key challenges addressed include: where to execute the build (using Coding CI), how to incrementally publish products (using git diff with pathspec to identify modified modules), where to upload pre-compiled products (Tencent Cloud Generic repository), and how to use pre-compiled products (automatic matching and source compilation switching).

The article further explains the implementation details: using Git commit ID as module version numbers for automatic version management, creating a manifest file (base_manifest.json) to record module product information, and implementing product/source compilation switching for different platforms (CMake for C++, Gradle for Android). For IDE integration, it covers automatic CMake regeneration on Windows, manual regeneration prompts for Xcode, and LLDB source mapping for breakpoint debugging.

Performance results show significant improvements: Android reduced from 17 minutes to 3 minutes (82% efficiency gain), Windows from 20 minutes to 10 seconds (99% gain), iOS from 12 minutes to 3 minutes (75% gain), and macOS from 16 minutes to 4 minutes (75% gain).