Performance
Benchmark execution times tracked across commits — tree-walker vs bytecode VM
Avg Speedup
—
tree-walker / bytecode VM
Overview — Bytecode VM Times
Methodology
- Each benchmark is run 3 times per execution mode; the median time is recorded.
- Two modes are compared: tree-walker (original interpreter) and bytecode VM (compiled, stack-based).
- Timing uses Python's
time.perf_counter() for sub-millisecond accuracy.
- Benchmarks run automatically every 6 hours on a dedicated build server.
Benchmark source files:
bench/
on GitHub.
Lattice