bench(public): regenerate the Node/Bun baseline at v0.5.1335 and restore the README markers

[proggeramlug]

Unblocks the public-baseline gate. lint has been red on main and on every open PR, for two independent reasons — I only knew about one of them until the run actually completed.

1. The artifact was stale — and the gate was right

source_fingerprint covers Cargo.toml, and #7302 changed panic = "unwind" → panic = "abort" in both the release and dist profiles. That is a real, benchmark-relevant change, so the demand for a regeneration was correct rather than a fingerprinting artifact. (My first hypothesis was that the version bump invalidated it on every merge; that is already normalised by _CARGO_VERSION_RE, so it was wrong.)

Regenerated on the pinned quiet M1 Max — the host the artifact's own host.cpu names — with pinned node v22.23.1 / bun 1.3.14 / zig 0.15.2, all five wait_for_quiet gates passed, at 2ba59501b (v0.5.1335). All five components present, validate_public passes.

2. public_baseline.py check could never pass

#6736 rewrote README.md as a marketing landing page and deleted the <!-- public-node-bun:start --> / <!-- public-node-bun:end --> markers, replacing the generated block with a hand-maintained table. _replace_block therefore raised README generated markers are missing on every invocation, independent of the artifact. No amount of regenerating would have fixed it.

That is worth naming as a pattern rather than a typo: a gate whose subject was deleted still runs, still reports, and can never go green — so it stops being a gate and becomes a permanent reason to bypass. Same family as #7582.

The markers are restored under the sentence that promises them — "We publish everything, including the workloads where V8's JIT still beats us — no cherry-picked table can survive an open harness" — inside a collapsed <details>, so the landing page keeps the shape #6736 deliberately chose while the published numbers are once again derived from the artifact instead of hand-maintained. The generated table is what makes that sentence true: it carries prime_sieve 30 ms vs 6, and matrix_multiply 87 ms vs 34, as losses.

python3 benchmarks/public_baseline.py check now exits 0.

What the run measured

Two results worth surfacing, both with output verified against the Bun reference 20/20:

• image_convolution 302.5 ms — 3.1× faster than bun (947.2), faster than Rust (429.8), within 8% of Zig (279.8).
• json_pipeline at 100 records: 77.0 ms — ahead of bun (90.8) and node (128.1), within 5% of Rust and Zig, at 8.5 MB RSS against bun's 30.3.

And one that is not good, filed as #7592: the same JSON workload at 500,000 records is 60,358 ms against bun's 618 ms — 97.6×. It is a scaling cliff rather than a constant factor (Perry grows 784× across the fixture increase where bun grows 6.8×, Rust 8.9×, Zig 12.0×). Correct output, 20/20 — purely performance.

That row is now published by this PR, which is the point of the mechanism.

Note for follow-up

README.md's hand-written table cites honest_bench/REPORT.md for its JSON row and shows the 100-record variant; the same report's 500k variant is the 97.6× row above. Its absolute numbers are also stale — the JSON row reads 39 ms where today measures 77 ms, though bun (51 → 90.8) and Rust (34 → 73.4) moved by the same factor, so that is measurement conditions rather than a Perry regression, and the relative ordering holds. Reconciling that table is a product decision about the landing page and is left to the owner; this PR does not touch it.

Summary by CodeRabbit

• Documentation

  • Updated the documented release version to 0.5.1337.
  • Added a collapsible full benchmark suite section covering methodology, correctness, and Perry, Node.js, and Bun results.
  • Added a changelog entry documenting refreshed public baselines and JSON workload performance findings.

• Benchmark Updates

  • Refreshed benchmark measurements, memory usage, metadata, rankings, and aggregate results across application patterns, JSON, and polyglot workloads.
  • Confirmed all recorded runs matched the reference output.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: c51e63d6-0a2f-400f-b120-07f7c6276f04

📥 Commits

Reviewing files that changed from the base of the PR and between 2bf04b2 and 292b4c7.

⛔ Files ignored due to path filters (1)

• Cargo.lock is excluded by !**/*.lock

📒 Files selected for processing (13)

• CLAUDE.md
• Cargo.toml
• README.md
• benchmarks/app-patterns/results/matrix-20260807-130603.md
• benchmarks/honest_bench/REPORT.md
• benchmarks/honest_bench/results/metadata.json
• benchmarks/honest_bench/results/results.json
• benchmarks/honest_bench/results/summary.txt
• benchmarks/json_polyglot/RESULTS.md
• benchmarks/polyglot/RESULTS_AUTO.md
• benchmarks/results/public-node-bun-v1.json
• benchmarks/suite/results/RESULTS.md
• changelog.d/7593-public-baseline-v0-5-1335.md

---

📝 Walkthrough

Walkthrough

The pull request refreshes benchmark evidence for the 2026-08-07 run, adds a published performance summary, records the baseline regeneration, and updates the workspace version from 0.5.1336 to 0.5.1337.

Changes

Benchmark refresh

Layer / File(s)	Summary
Honest benchmark measurements benchmarks/honest_bench/...	Updated benchmark metadata, executable paths, reports, summaries, and per-run measurements for image convolution and JSON pipelines. All recorded runs retain matching output validation.
Polyglot benchmark reports benchmarks/json_polyglot/RESULTS.md, benchmarks/polyglot/RESULTS_AUTO.md	Refreshed dates, runtime versions, medians, variance, extrema, memory values, and unavailable Hermes entries.
Runtime comparison reports benchmarks/app-patterns/results/*, benchmarks/suite/results/RESULTS.md	Added app-pattern comparison results and refreshed Node/Bun suite evidence, classifications, metadata, and aggregate counts.
Published benchmark summary README.md, changelog.d/7593-public-baseline-v0-5-1335.md, CLAUDE.md, Cargo.toml	Added the published benchmark summary, documented baseline regeneration and the JSON regression, and updated the documented workspace version to 0.5.1337.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• PerryTS/perry#7285: Updates the same public Node/Bun benchmark artifacts and generated reports.
• PerryTS/perry#7476: Refreshes the same JSON and polyglot benchmark result files.
• PerryTS/perry#7479: Tracks a performance backlog based on the refreshed benchmark results.

Suggested labels: parity

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch baseline-v0.5.1336

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}