Building Self Driving Cars with Bazel
summary
BMW uses Bazel to build driver‑assistance and self‑driving software, consolidating many repos into a monorepo‑style workspace with git submodules, handling 23 million lines of code, ~2,000 developers and >20,000 CI builds per day. Migration starts with a 5–6‑engineer prototype team, then a month‑long rollout. Bazel delivers 10x speedup for unit tests and ~12x for target builds via remote cache and remote execution, integrates ROS, MATLAB codegen and heterogeneous cross toolchains, and produces cars built with Bazel from 2021.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:04Speakers' backgrounds
- 4:57BMW chose Bazel
- 7:46Bazel for safety
- 12:23Migration strategy
- 16:14Performance gains
- 20:16C++ toolchain challenges
- 24:17Python toolchain lessons
- 27:45ROS middleware integration
- 34:10Bazel's future in BMW
- 35:35Q&A
related talks
20mDriving Intelligence: The Software Shift Shaping Tomorrow’s Cars
29mPerfect Sandboxing in Bazel
32mThe Classics Never Go Out of Style: An Empirical Study of Downgrades from Bazel
33mCI/CD for Autonomous Drive and other complex systems - What we have learned about...
17mperspective on automotive software and Eclipse SDV
18mSoftware is the New Fuel, AI the New Horsepower - Pioneering New Paths
33mDaimler's Disruptive Product Development Evolution
46mHow to architect for continuous delivery?