Common-sense acceleration of your MLOC build
summary
Describes concrete steps that cut multi-hour builds to minutes: mounting /tmp as a ram disk for link-heavy IO, benchmarking SSD vs 15k‑RPM SAS and preferring RAM-backed storage, using the gold linker with threaded links and LTO, bootstrapping and profile-bootstrap gcc to optimize tokenizers, tuning NFS (MTU, read/write sizes, read‑only mounts) and placing ccache on local fast storage, and integrating ccache with distcc. Measurements show tens‑of‑percent link reductions and an extreme case from ~2 hours down to ~10 minutes.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 7:40Vocabulary (contd)
- 14:20Accelerating the work
- 15:14make the source checkout FAST
- 19:21Use a RAM disk
- 30:05Optimize file servers
- 34:17Use gold linker
- 37:19Optimize the compiler for "YOU"
- 42:47Optimize the compiler (cont'd)
- 57:37Using ccache with distcc
related talks
46mBuilding and Packaging Modern C++
32mSweeten your Yocto Build Times with Icecream
16mMind the Gap (Between Your Code and Your Toolchain)
05mTips for Speeding Up Cargo Build in CI
53mReal-World Yocto: Getting the Most out of Your Build System
27mImproving the Usability of Remote Caching
47mUsing Incredibuild to Accelerate Static Code Analysis and Builds
30mThe Bitter and the Sweet of Running a Planet-Scale Build &...