“Deploying C++ modules to 100s of millions of lines of code"

on YouTube ↗01h 00m

" summary

Google migrates hundreds of millions of lines of C++ to Clang modules and header PCM files, using module maps and a changed protoc generator to modularize protocol buffers and reduce reparsing. Average compile time improves ~10% with up to ~50% gains in the 99th percentile, but naive rollout doubles distributed-build load because more module inputs force extra staging and recompiles. Fixes include ~10k header repairs, modularizing STL dependencies, include-scanning for modules, and splitting libraries; the approach works but exposes new build-system and tooling costs.

AI-written summary · about

" chapters (11)

1. 0:00Introduction to Google C++
2. 2:54Google build system basics
3. 6:06Motivation for C++ modules
4. 8:10Analyzing slow build causes
5. 12:25Modules implementation plan
6. 13:56Modules versus traditional builds
7. 17:02Rolling out modules for protox
8. 22:12Technical challenges and fixes
9. 31:41Initial rollout results
10. 38:01Optimizing the build process
11. 40:57Lessons learned and conclusion

" related talks

• 01h 01mC++ Modules and Large-Scale Development (Part 1)John Lakos · NDC TechTown 2019
• 01h 07mBuilding C++ ModulesBoris Kolpackov · CppCon 2017
• 58mSo, You Want to Use C++ Modules … Cross-Platform?Daniela Engert · C++ on Sea 2023
• 01h 00mSo, you want to use C++ Modules ... cross-platform?Daniela Engert · NDC TechTown 2023
• 01h 30mC++ Modules and Large-Scale DevelopmentJohn Lakos · ACCU 2018
• 34mPractical C++20 Modules and the future of tooling around C++ ModulesCameron DaCamara · Pure Virtual C++ 2020
• 59mC++ ModulesNathan Sidwell · CppCon 2018
• 01h 30mC++ Modules and Large-Scale DevelopmentJohn Lakos · ACCU Autumn 2019 Belfast