Dot performance dashboard

Warm, in-process timing of @knowvah/dot-engine vs the native dot oracle over the dot test corpus. Peer to PARITY.md — PARITY tracks correctness, PERF tracks speed. A report, not a gate. The fidelity target is ≤3× native.

Regenerate: npm run build:js && node test/corpus/bench.mjs && node test/corpus/perf-dashboard.mjs.

Method

• Port: the shipped bundle (dist/index.js) loaded once in a pool of resident, JIT-primed worker threads; pure renderSvg() is timed (best-of-N), so the measured region excludes all process/transpile/module-load startup — the warm steady state a long-lived consumer sees. Light graphs run at full pool; heavy graphs (native > 2s) are timed serially by default — measured cross-talk inflates a concurrent big render's single sample materially (≈66% on 2620). Set BENCH_HEAVY_POOL>1 for a faster, noisier scan.
• Native: dot -Tsvg best-of-3 (min).
• Budget: target ≤3× native. Per-render cap 1351530ms (SIGKILL → over-cap, i.e. a true synchronous hang).
• Caveat: light graphs are timed under up-to-6-way load; for a precise single number re-run BENCH_POOL=1 BENCH_IDS=<id> node test/corpus/bench.mjs.

Summary

• Rated inputs: 779 · within ≤3× native: 762 (97.8%)
• ok (≤3×): 761 · slow (>3×): 18 · over-cap (hang): 0 · errored: 0 · oracle-error: 9
• ratio (port/native): p50 0.01× · p90 0.23× · max 6.48×

Ratio distribution

band	count
≤1× (port ≥ native, warm)	737
1–2×	11
2–3×	14
3–4×	11
4–6×	5
6–10×	1
>10×	0
over-cap (≥1351530ms, possible hang)	0

Over budget — slower than 3× native (worst first)

id	native ms	port ms (warm)	ratio
2108	12849	83220	6.48×
2471	3269	17929	5.48×
2095	261	1155	4.42×
graphs-b100	8646	37691	4.36×
graphs-b104	8635	35254	4.08×
graphs-b103	1210	4883	4.04×
2095_1	22309	88075	3.95×
linux.x86-root_circo	245	943	3.85×
nshare-root_twopi	247	900	3.65×
2646	97463	352989	3.62×
2743	201	721	3.59×
graphs-root	214	757	3.54×
nshare-root_circo	247	809	3.28×
linux.x86-root_twopi	248	805	3.25×
2371	73639	235360	3.2×
graphs-b29	585	1852	3.17×
2620	371	1122	3.02×

Over-cap / possible hang

None — no input exceeded the per-render cap (no hangs).