Keynote
summary
Goal is Ruby 3x3: make Ruby 3 three times faster than Ruby 2.0 using real-world benchmarks and constraints. Implementation uses an MJIT that emits C (via RTL/LTL), compiles with GCC/Clang, and loads shared objects with worker threads and precompiled headers to cut compile time. Results: optcarrot (NES emulator) runs ~2.8× faster with ~1.16× memory, some microbenchmarks hit ~6×, and Rails still doesn’t run yet. Rule set includes configurable low memory and no third-party dependency for long-term maintainability.
AI-written summary · about
chapters (16)
- 0:00Welcome and Ruby update
- 2:19Ruby’s 25-year survival
- 6:07Breaking compatibility dangers
- 9:29Rewriting rarely works
- 11:39Usability over raw speed
- 13:56Ruby 3x3 performance goal
- 17:423x3 rules: memory and deps
- 21:523x3 rules: maintainability
- 23:39Microbenchmarks vs real-world Ruby
- 25:52Standard Ruby 3x3 benchmarks
- 28:48JIT compilers for Ruby
- 30:39New Ruby VM with LTL and JIT
- 35:43How VM boosts performance
- 37:54Early results and challenges
- 39:46Ruby’s future and community
- 41:12Q&A: Performance and flexibility
