{ "version": 1, "captured": "2026-05-20", "captured_via": "bench/run.sh -n 5", "note": "Baseline for bench/check.py regression detection. The language-invariant thresholds (rc/bump <= 1.3x throughput, p99/median <= 5x latency) are NOT the regression-check tolerances; the per-metric tolerances below are tuned to absorb run-to-run noise on a quiet developer machine. To update after an intentional change, re-run bench/run.sh and replace the values, recording the reason in the commit body that ships the baseline bump. The latency gate history: max_us and p99_9_us were removed on the 2026-05-20 recapture (Gitea #15 / #16) because tail-of-distribution latency metrics are dominated by OS-level jitter (THP defrag, scheduler preemption, IRQ load), not allocator behaviour, and produced 3+ consecutive false-positive REGRESSION rows on byte-identical no-op milestones; see docs/specs/0047-bench-harness-recalibration.md. At the 2026-05-28 kernel-extension-mechanics audit close, latency.explicit_at_rc.p99_us and latency.explicit_at_rc.p99_over_median were ALSO removed after a 6-invocation bencher characterisation found the explicit-arm p99 cv at 18.6% (vs implicit-arm cv 2.9%) with within-invocation 4-run spread up to 1.87x \u2014 structurally identical to the 2026-05-20 metric-removal precedent. The explicit-arm now gates only on median_us (which has cv 0.37% across 6 invocations, the actual allocator signal). The implicit-arm p99 / p99_over_median is kept because it is stable (cv 2.9%) and provides the contrast that lets future audits distinguish allocator regressions from machine jitter.", "throughput": { "bench_list_sum": { "bump_s": { "baseline": 0.053075, "tolerance_pct": 10 }, "rc_s": { "baseline": 0.144645, "tolerance_pct": 10 }, "rc_over_bump": { "baseline": 2.73, "tolerance_pct": 8 }, "bump_rss_kb": { "baseline": 97884.0, "tolerance_pct": 5 }, "rc_rss_kb": { "baseline": 193820.0, "tolerance_pct": 5 } }, "bench_tree_walk": { "bump_s": { "baseline": 0.039941, "tolerance_pct": 10 }, "rc_s": { "baseline": 0.099691, "tolerance_pct": 10 }, "rc_over_bump": { "baseline": 2.5, "tolerance_pct": 8 }, "bump_rss_kb": { "baseline": 55196.0, "tolerance_pct": 5 }, "rc_rss_kb": { "baseline": 108956.0, "tolerance_pct": 5 } }, "bench_closure_chain": { "bump_s": { "baseline": 0.008045, "tolerance_pct": 25 }, "rc_s": { "baseline": 0.032188, "tolerance_pct": 20 }, "rc_over_bump": { "baseline": 4.0, "tolerance_pct": 15 }, "bump_rss_kb": { "baseline": 16024.0, "tolerance_pct": 15 }, "rc_rss_kb": { "baseline": 39652.0, "tolerance_pct": 10 } }, "bench_hof_pipeline": { "bump_s": { "baseline": 0.051527, "tolerance_pct": 10 }, "rc_s": { "baseline": 0.143245, "tolerance_pct": 10 }, "rc_over_bump": { "baseline": 2.78, "tolerance_pct": 8 }, "bump_rss_kb": { "baseline": 97628.0, "tolerance_pct": 5 }, "rc_rss_kb": { "baseline": 193572.0, "tolerance_pct": 5 } }, "bench_compute_collatz": { "bump_s": { "baseline": 0.056148, "tolerance_pct": 12 }, "rc_s": { "baseline": 0.056289, "tolerance_pct": 12 }, "rc_over_bump": { "baseline": 1.0, "tolerance_pct": 10 }, "bump_rss_kb": { "baseline": 13840.0, "tolerance_pct": 15 }, "rc_rss_kb": { "baseline": 14000.0, "tolerance_pct": 15 } }, "bench_list_sum_explicit": { "bump_s": { "baseline": 0.052832, "tolerance_pct": 10 }, "rc_s": { "baseline": 0.158251, "tolerance_pct": 10 }, "rc_over_bump": { "baseline": 3.0, "tolerance_pct": 8 }, "bump_rss_kb": { "baseline": 97436.0, "tolerance_pct": 5 }, "rc_rss_kb": { "baseline": 141796.0, "tolerance_pct": 8 } } }, "latency": { "explicit_at_rc": { "median_us": { "baseline": 218.8, "tolerance_pct": 15 } }, "implicit_at_rc": { "median_us": { "baseline": 303.5, "tolerance_pct": 15 }, "p99_us": { "baseline": 469.0, "tolerance_pct": 20 }, "p99_over_median": { "baseline": 1.54, "tolerance_pct": 20 } } } }