Back to projects
Java · Evidence-driven Java performance lab
Java Performance Lab
Correctness-first Java 21 case study with JMH, Java Flight Recorder, allocation and GC profiling, virtual threads, and reproducible evidence.
The challenge
Performance optimizations quickly become misleading without proven behavioral equivalence, controlled measurement conditions, and raw evidence.
The result
A multi-module lab that verifies two implementations against the same contract, measures them in forked JVMs, and diagnoses runtime effects with JFR.
What this project demonstrates
Reviewable decisions instead of slogans.
- Contract and differential tests before every performance claim
- JMH with forks, warm-up, parameters, and the GC profiler
- Custom JFR events and memory-conscious streaming analysis
- Versioned raw evidence and a deliberately bounded optimization decision
Technologies