Optimizing Production Performance with MRI JIT
summary
Claims that MRI JIT finally speeds real Rails apps on Ruby 3.0 and gives measurable wins when tuned: use Ruby 3.0.0 (3.0.1 is broken, 3.0.2 differs), increase jit_max_cache=10000, set --jit-call-threshold=10000, disable or account for TracePoint/zeitwerk and ractor interactions, wait for jit_compaction and watch jit verbose logs, and accept temporary slowdowns during C compilation; reported benchmarks show JIT surpassing interpreter performance after these concrete fixes and warmup steps.
AI-written summary · about
chapters (9)
- 0:00Start
- 2:03Characteristics of the Mgt
- 9:10Key Ideas
- 14:44Change the Jit Max Cache
- 22:55Optimization Switches
- 23:32Instance Variable Cache
- 24:18Send Cache
- 24:56Jit Compaction
- 28:38Conclusion
