Problem
The open question is whether JavaScript with sound types, AOT and a JIT can be competitive with Rust and C, and nothing measures it. bench/ has only lexer and parser benches plus the TS checker comparison; the server benchmark in #794 covers cold start, memory and binary size, not throughput.
Evidence (origin/main @ d4b219507)
Scope
- Kernels: nbody, fannkuch-redux, binary-trees, spectral-norm, mandelbrot, JSON parse/stringify, regex-redux, and an HTTP requests/s test.
- Runtimes:
home run(JIT),home runwith sound-mode emit, AOT TypeScript (when available), native Home, Bun, Node, Deno, Rust and C (same algorithm, idiomatic code). - Report time, peak RSS and variance; pin versions and hardware in the output.
Acceptance criteria
- Suite checked into
bench/with a README and one command to run it. - Results published per release (CI or scheduled job), so every sound/AOT change shows its effect.