Benchmarks

Comparison docs for compile time, binary size, and runtime. Regenerate before publishing numbers.

Guide	Comparison
JIT	Cold / warm / daemon compile latency on macOS ARM64
Polyglot compilers	`in compile` median wall time vs native compiler per language (`in/native` ratio)
Self-host vs native	Owned Rust front (`in build` on `main.rs`) vs `cargo build --release` (full crate)

./scripts/bench-self-host.sh && python3 scripts/render-self-host-bench-md.py
./scripts/bench_polyglot_compilers.v
in bench   # JIT samples (see jit.md)

How to read ratios: On polyglot rows, in/native < 1 means inauguration finished faster than that native check for the same sample. On self-host, the front-only in build wall time is compared to an incremental Cargo release rebuild (not a clean cargo clean cold build).