bench(packages): package matrix re-run after the fixing phase

[proggeramlug]

Re-run of the package-performance matrix (benchmarks/packages/, scripts/package_bench.py, from #11466) to close the fixing phase. Data and docs only: no compiler, runtime or harness change, and no version bump.

Commit measured: 9e29f59d43 (the merge of #11645) on both hosts. Perry was built from that commit separately on each host.

• Instructions plus a Linux RSS sample: perrymaster, 2026-09-30 19:32–20:09 UTC.
• Wall, cold start and RSS: the M1 bench mini, 2026-09-30 20:04–21:15 UTC.
• Node 26.5.1 and Bun 1.3.14 on both hosts, the same pins as the baseline. No workload was load-flagged on either host.
• Baseline: the Phase-1 run at 36420d2e56.

Summary (details, the full before/after matrix and the attribution are in benchmarks/packages/RERUN-2026-09-30.md):

• Instructions: the Perry/Node geomean went from 19.7× to 12.4× over the 38 workloads that were correct in both runs. The median change in Perry instructions is −32 %.
• Wall-clock: the geomean went from 18.7× to 12.0× over 30 workloads. The median change is −36 %.
• Correctness: 8 workloads that failed at the baseline now pass, and mongodb/insert_find (mongodb 7.6.0: insertOne after deleteMany rejects with "reading 'state'" and the process hangs #11460, known_hang) now runs correctly. 2 Perry failures remain, and both have package-free repros in the doc:
  • node-forge/rsa_sign: this is undefined inside a do…while body in non-class functions.
  • fastify/inject: assignSocket is a no-op on a CJS util.inherits subclass of ServerResponse.
  • Both still reproduce on origin/main 9e05951071. Issues have not been filed yet.
• Still above 2× Node on instructions: 44 of the 47 correct workloads. The top offenders and what they map to:
  • mongodb (GC and property reads)
  • cron/luxon (getter property reads: perf: property access is 13x-112x node warm — two static reads cost 162 instructions against node's 12, and a const-key read costs 7.6x the same read spelled statically #10761, perf: class getters/setters are 430× (setter) to ~1,300× (getter) slower than Node (no accessor inline cache; every store runs the full OrdinarySet walk) #10498)
  • lru-cache and redis (private members: perf: private class members #x/#m() are ~1,200× slower than Node, 130× slower than public (marker-string format + by-name brand lookup per access) #10501)
  • node-forge (method calls on function-constructor prototypes: perf: obj.method() through the runtime dispatcher is ~3,000× slower than Node (two O(own-keys) string-compare scans per call before any cache) #10502, perf: function-constructor objects (new F(), x instanceof F) are 78–566× slower than Node (F.prototype and builtin identity re-derived from side tables per operation) #10507)

Files:

• results/*.json, REPORT.md and report_notes.md are regenerated the way bench(packages): Phase 1 package-performance harness + first report (Perry vs Node vs Bun) #11466 stored them. The previous versions are in git history at 24a5262bd6.
• profile/rerun-2026-09-30/: call-graph attribution for the top six workloads plus mongodb/insert_find.
• RERUN-2026-09-30.md: the before/after comparison and the re-ranking.

Not run or not done:

• The database workloads have no wall-clock or RSS numbers from the mini, because it has no database servers (same as the baseline).
• There was no full Phase-3 --callgraph sweep; only the top offenders were profiled. PROFILE.md still describes the 36420d2e56 attribution.
• mongodb/insert_find is still marked known_hang in the manifest. It was measured separately with --include-known-hangs at the same commit.
• No lint gates were run beyond check_file_size.sh. The diff touches only benchmarks/packages/** and changelog.d/.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

Note

Currently processing new changes in this PR. This may take a few minutes, please wait...

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: ec33dadf-60b7-4eb6-a6d1-f7d6e343b458

📥 Commits

Reviewing files that changed from the base of the PR and between 9e05951 and 1b2201b.

📒 Files selected for processing (14)

• benchmarks/packages/REPORT.md
• benchmarks/packages/RERUN-2026-09-30.md
• benchmarks/packages/profile/rerun-2026-09-30/callgraph-mongodb-insert_find.json
• benchmarks/packages/profile/rerun-2026-09-30/callgraph-mongodb-insert_find.md
• benchmarks/packages/profile/rerun-2026-09-30/callgraph.json
• benchmarks/packages/profile/rerun-2026-09-30/callgraph.md
• benchmarks/packages/report_notes.md
• benchmarks/packages/results/compile-linux-symbols.json
• benchmarks/packages/results/compile-linux.json
• benchmarks/packages/results/compile-mac.json
• benchmarks/packages/results/instr.json
• benchmarks/packages/results/profile.json
• benchmarks/packages/results/results.json
• benchmarks/packages/results/wall.json

 ____________________________________________
< This is what it sounds like when bugs cry. >
 --------------------------------------------
  \
   \   (\__/)
       (•ㅅ•)
       / 　 づ

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Autopilot is currently an internal CodeRabbit preview.

---

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.}