Ask Me Anything with Chris Lattner
summary
A wide-ranging Q&A covers practical takeaways: tradeoffs between publication and building robust prototypes, why LLVM/Clang/Swift/MLIR grew (modularity, permissive license, active community), concrete testing tools like Csmith and EMI and reducers such as C-Reduce, and specific EDA projects like Yosys, Verilator, OpenROAD. Recommendations include prioritizing stable C APIs for ecosystem growth, using internships to bridge academia and industry, applying ML to heuristics (e.g. inliner decisions) while keeping determinism, and tracing language changes via Swift Evolution and progressive lowering with high-level IRs.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:10Lessons from academia and industry
- 3:30Developing a research prototype
- 6:20Compiler testing
- 9:02Bugs reported by developers
- 10:25Swift
- 13:05Backwards compatibility
- 16:46llvm evolution
- 18:08Open source community
- 20:28Open source tools
- 22:49Why is llvm so popular
- 24:57Machine learning in compilers
- 27:35Software development process
- 30:45Future problems
related talks
38mFireside chat
19mClangBuiltLinux: What's Next?
32mGetting started with LLVM using Swift
48mCross-Domain Compilation: Exploiting Synergies Across the CS Community
53mLLVM- 19m"ClangBuiltLinux: What's Next?"
31m“New standards to the rescue: the view through an IDE’s glasses”
01h 31m"Thoughts on language evolution"