Tooling: 3.12× faster inplace-side prefix replay with identical receipts

certificates/inplace-side.json

@@ -5609,7 +5609,7 @@
     "docs/research/inplace-side.md": "c228f131d95f3ea348455c13378c19ed6e3c75af7e14cf68ea0ada445ab3ac72",
     "scripts/audit_inplace_side.py": "16d295f9f9250d35f6f1de2843d5e13c7adbf88d595ebcd0b65b00d1a938acd9",
     "scripts/experiments/core_limits.py": "a31a27cc24cf0349a507d1141a2b965ea25f1c60f355d7e13c581305a12d3a43",
-    "scripts/experiments/inplace_side.py": "4e7cb3df2e263f6c7f5907c55e3ee04c08926e6dbc984bbda1722a27b0a48ff6",
+    "scripts/experiments/inplace_side.py": "4397b84f3ac52554bf14303171410a2799af0bf8dae11ec64594bb5c011e93a2",
     "scripts/experiments/inplace_side_witnesses.json": "90fe710f5daa35cc9e4892b29d712d66110b7b2b4c404da95a960ea7041618b9"
   },
   "status": "SPECIFIED IN-PLACE FAMILY EXCLUDED; NO STRONGER KAPPA",

research/inplace-side-performance/README.md

@@ -0,0 +1,121 @@
+# Faster historical inplace-side prefix replay
+
+**Tooling improvement; no new kappa.** The historical
+`scripts/experiments/inplace_side.py:verify_prefix` check rejects a specified
+n-role Gauss–Jordan side family using three pinned witnesses (`h = 42, 46, 50`).
+Each elimination column previously rebuilt `sorted(unused)` to locate the least
+unused row with the pivot bit set. The production checker now scans an unused
+flag array from index 0 and tests bits with a column mask. The lexicographic
+rule, XOR updates, switch witnesses, deficit arithmetic and every receipt field
+are unchanged.
+
+The selected frontier, `upstream/` and the mathematical rejection in
+[docs/research/inplace-side.md](../../docs/research/inplace-side.md) are not
+reopened. Only the source pin for `scripts/experiments/inplace_side.py` changes
+inside `certificates/inplace-side.json`.
+
+## Reproduce
+
+Python **3.12**, standard library only, from the repository root:
+
+```sh
+python3.12 -I -B research/inplace-side-performance/verify.py
+python3.12 -I -O -B research/inplace-side-performance/verify.py
+python3.12 -I -B -m unittest tests.test_inplace_side tests.test_inplace_side_performance -v
+python3.12 -B scripts/research_package.py --package research/inplace-side-performance --check-only
+python3.12 -B scripts/research_package.py --package research/inplace-side-performance
+```
+
+Expected verifier line:
+
+```text
+PASS_INPLACE_SIDE_PREFIX_FAST witnesses=3 h=[42, 46, 50] paths=[151, 415, 767] ... no_new_kappa
+```
+
+Expected registered replay:
+
+```text
+PASS_REGISTERED_RESEARCH_PACKAGE research/inplace-side-performance outcome=completed
+```
+
+Normal verification is read-only. To regenerate the certificate after editing the
+production source:
+
+```sh
+python3.12 -I -B research/inplace-side-performance/verify.py --write-certificate
+```
+
+Before/after wall times (descriptive; not a pass gate):
+
+```sh
+python3.12 -I -B research/inplace-side-performance/benchmark.py --rounds 5
+```
+
+Expected: `PASS_IDENTICAL_INPLACE_PREFIX_RECEIPTS`. Samples are retained in
+[RESULT.json](RESULT.json). On this Linux host / CPython 3.12.3, five alternating
+fresh-process pairs gave median **17.061 s → 5.466 s (3.12×)** for the three
+pinned witnesses. Timing is descriptive, not a correctness gate.
+
+## Why the receipts stay identical
+
+For each column `col`, the original code selected
+
+```text
+min{ i in unused : rows[i] has bit col }
+```
+
+by materializing `sorted(unused)` and taking the first hit. Scanning
+`i = 0, 1, …` over a boolean unused table and stopping at the first hit yields
+the same index, because the integers are totally ordered. Column tests
+`rows[i] & (1 << col)` equal `rows[i] >> col & 1` on these nonnegative row
+masks. The subsequent row-XOR updates therefore run on the same pivot sequence
+and produce the same residual matrix, the same certified nonorthogonal path
+list and the same deficit upper bound.
+
+[baseline.py](baseline.py) is a byte-identical copy of
+`scripts/experiments/inplace_side.py` at base
+`0b1cccf86e2f5132ed286a26ef0a681a495740ea` (blob
+`db3de6f1d962ccc3a4eae31f83c72bba995ba44e`). The verifier compares every field
+of every witness receipt against that baseline, checks shared rejection
+behavior, rejects a greatest-unused mutant on all three witnesses, and requires
+the working certificate to differ from the base certificate only in the single
+`inplace_side.py` source pin.
+
+## Task card, overlap and limits
+
+Chooser seed `1cd1f71746c2d200dd617bc3dd17d8a6` drew `bit-mechanisms`; live
+helper-contract and register-model surfaces were saturated. Expertise override
+to `replay-runtime` (verifier-engineering). Scope **refined** to this fixed
+historical screen inside `make verify-certificates`.
+
+Closest priors (different objects; no code reused here):
+
+- [CrocSwap #756](https://github.com/CrocSwap/integer-mult-bounds/pull/756) —
+  packed-column balanced-Fano replay (2.91×) on a different checker.
+- [CrocSwap #1584](https://github.com/CrocSwap/integer-mult-bounds/pull/1584) —
+  finite-bit block-core frame interning.
+- [CrocSwap #1076](https://github.com/CrocSwap/integer-mult-bounds/pull/1076) /
+  [#1075](https://github.com/CrocSwap/integer-mult-bounds/pull/1075) —
+  quadratic-affine / quadratic-code tooling.
+- Local `find_overlap.py` for `inplace side` / `inplace_side` / `verify_prefix`
+  reported no competing open PR on this object at refresh time.
+
+Limits: timing is host-descriptive; OS caches were not flushed. Full
+`make verify`, selected-supplier cold replay, Lean and multiplication-theorem
+obligations were not rerun for this isolated historical-checker change.
+
+## Open next questions
+
+- Can the same least-unused scan replace `sorted(unused)` inside `compile_side`
+  without changing gate lists on `h = 6, 10` and, if identical, on `h = 50`?
+- Should CI grow a soft regression timer for this screen, or keep RESULT.json
+  descriptive only?
+
+## Credit
+
+Douglas R. Colkitt authored the inplace-side rejection, witnesses and original
+checker. Prepared by Wentao Li
+([VictorLiwentao](https://github.com/VictorLiwentao),
+[@Is__Victor](https://x.com/Is__Victor), wentaol4@illinois.edu)
+with substantial AI assistance, under Apache-2.0. No independent human
+peer-review or formal proof-assistant claim.

research/inplace-side-performance/RESULT.json

@@ -0,0 +1,62 @@
+{
+  "ai_assistance": "Substantial AI assistance for profiling, implementation, packaging, benchmarking and writing. Not independent human peer review.",
+  "base_commit": "0b1cccf86e2f5132ed286a26ef0a681a495740ea",
+  "baseline_seconds": [
+    16.747456939999665,
+    17.098737376999907,
+    17.06069083300008,
+    16.890133087999857,
+    17.154922305999662
+  ],
+  "engines": {
+    "baseline": "research/inplace-side-performance/baseline.py",
+    "optimized": "scripts/experiments/inplace_side.py"
+  },
+  "environment": {
+    "implementation": "CPython",
+    "isolated_host": false,
+    "os_caches_flushed": false,
+    "platform": "Linux-6.12.94+-x86_64-with-glibc2.39",
+    "python": "3.12.3 (main, Aug 31 2026, 10:18:26) [GCC 13.3.0]"
+  },
+  "fixture": "scripts/experiments/inplace_side_witnesses.json",
+  "medians": {
+    "baseline": 17.06069083300008,
+    "optimized": 5.465907218000211,
+    "speedup": 3.121291700088902
+  },
+  "no_new_kappa": true,
+  "optimized_seconds": [
+    5.465907218000211,
+    5.394471960999908,
+    5.672483955000189,
+    5.443477510999855,
+    5.647890871000072
+  ],
+  "receipt_fields_compared": [
+    "all_internal_rational_frames_allowed",
+    "bad_endpoints",
+    "certified_nonorthogonal_paths",
+    "deficit_upper",
+    "disjoint_switches",
+    "excess_per_invocation_lower",
+    "h",
+    "n",
+    "no_positive_deficit",
+    "prefix_columns",
+    "scope"
+  ],
+  "rounds": 5,
+  "scope": "Fixed historical inplace-side prefix check on the three pinned witnesses (h=42,46,50). Not a selected-supplier cold replay or multiplication runtime.",
+  "source_sha256": {
+    "baseline.py": "4e7cb3df2e263f6c7f5907c55e3ee04c08926e6dbc984bbda1722a27b0a48ff6",
+    "scripts/experiments/inplace_side.py": "4397b84f3ac52554bf14303171410a2799af0bf8dae11ec64594bb5c011e93a2",
+    "scripts/experiments/inplace_side_witnesses.json": "90fe710f5daa35cc9e4892b29d712d66110b7b2b4c404da95a960ea7041618b9"
+  },
+  "status": "PASS_IDENTICAL_INPLACE_PREFIX_RECEIPTS",
+  "witness_h": [
+    42,
+    46,
+    50
+  ]
+}

research/inplace-side-performance/SOURCES.json

@@ -0,0 +1,48 @@
+{
+  "schema_version": 1,
+  "base_commit": "0b1cccf86e2f5132ed286a26ef0a681a495740ea",
+  "files": [
+    {
+      "path": "docs/research/agent-task-catalog.json",
+      "usage": "provenance",
+      "git_blob": "06c96f5579bb67ed1b86d334e78fbed7da6945f1",
+      "sha256": "6686ad58f1fbdbc87134a78901b60fd53a20d3984355e91defbbebc9e9fa8075"
+    },
+    {
+      "path": "certificates/selected-result.json",
+      "usage": "provenance",
+      "git_blob": "ae6a99f28aa468649eb66a361e764d769363bf11",
+      "sha256": "10b9e340fd8e354a37a0c4813c90e5de1dd42044b5f6aca4a3fc5a42b685fa32"
+    },
+    {
+      "path": "certificates/inplace-side.json",
+      "usage": "provenance",
+      "git_blob": "8d3240c59c26c1839c2cf7a9b063fd30cbbe5a84",
+      "sha256": "1a2020c346e7d039e924a8c782dc6b3f0e94039eb1bec2b2644fc3de9d41ccdb"
+    },
+    {
+      "path": "docs/research/inplace-side.md",
+      "usage": "provenance",
+      "git_blob": "a0bb57cacd27d7c2921f6060f59b42aeb1355311",
+      "sha256": "c228f131d95f3ea348455c13378c19ed6e3c75af7e14cf68ea0ada445ab3ac72"
+    },
+    {
+      "path": "scripts/audit_inplace_side.py",
+      "usage": "provenance",
+      "git_blob": "2e9108178ee2e07d17f52d7cf79e079e91ba5feb",
+      "sha256": "16d295f9f9250d35f6f1de2843d5e13c7adbf88d595ebcd0b65b00d1a938acd9"
+    },
+    {
+      "path": "scripts/experiments/inplace_side.py",
+      "usage": "provenance",
+      "git_blob": "db3de6f1d962ccc3a4eae31f83c72bba995ba44e",
+      "sha256": "4e7cb3df2e263f6c7f5907c55e3ee04c08926e6dbc984bbda1722a27b0a48ff6"
+    },
+    {
+      "path": "scripts/experiments/inplace_side_witnesses.json",
+      "usage": "input",
+      "git_blob": "9d9970245a71705da7b84122a8be5b44755b5005",
+      "sha256": "90fe710f5daa35cc9e4892b29d712d66110b7b2b4c404da95a960ea7041618b9"
+    }
+  ]
+}

research/inplace-side-performance/TASK.json

@@ -0,0 +1,70 @@
+{
+  "schema_version": 1,
+  "task_id": "replay-runtime",
+  "priority_revision": 3,
+  "base_commit": "0b1cccf86e2f5132ed286a26ef0a681a495740ea",
+  "chooser_seed": "1cd1f71746c2d200dd617bc3dd17d8a6",
+  "original_card": {
+    "id": "replay-runtime",
+    "question": "Can one measured replay bottleneck be sped up without dropping coverage?",
+    "starting_points": [
+      "scripts/run_isolated_tests.py",
+      "scripts/verify_replay_coverage.py",
+      "certificates/selected-result.json"
+    ],
+    "methods": [
+      "Profile one isolated replay and optimize the measured bottleneck.",
+      "Design source-bound reuse with explicit dependency invalidation and negative controls."
+    ],
+    "acceptance_gate": "Report before/after wall time on the same fixture, retained checks and invalidation behavior.",
+    "first_step": {
+      "command": [
+        "python3",
+        "-B",
+        "scripts/research_baseline.py"
+      ],
+      "expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY",
+      "inspect": "Read the selected reproduction command and choose a single replay stage for profiling.",
+      "scope": "Positive control / source-bound orientation only; this is not a new construction or full cold replay."
+    },
+    "first_experiment": "Measure one unchanged fixture, preserve check coverage, and test one optimization with invalidation controls.",
+    "bottleneck_link": "Faster trustworthy replays increase review throughput and let mechanism candidates be checked sooner.",
+    "allowed_variations": [
+      "Profile one stage or design one source-bound cache boundary."
+    ],
+    "scope_change_triggers": [
+      "Skipping coverage or comparing different fixtures to claim speedup."
+    ],
+    "overlap_terms": [
+      "replay runtime",
+      "cache invalidation",
+      "isolated tests"
+    ]
+  },
+  "scope_status": "refined",
+  "actual_scope": "Can the historical inplace-side prefix checker (scripts/experiments/inplace_side.py:verify_prefix on the three pinned h=42/46/50 witnesses) be sped up without changing any receipt field, rejection condition or the regenerated certificates/inplace-side.json except its single source pin?",
+  "scope_change_reason": "The catalogue card asks for one measured replay bottleneck with before/after wall time and retained checks. The OS-random bit-mechanisms draw (seed 1cd1f71746c2d200dd617bc3dd17d8a6) was saturated on helper-contract/register-model work; expertise override selected verifier-engineering. Profiling make verify-certificates pointed at this fixed historical screen rather than isolated-test harness caching.",
+  "bottleneck_link": "Side result on review throughput: the inplace-side certificate is part of verify-certificates. A faster identical replay does not change kappa or the binding complex supplier.",
+  "budget_minutes": 60,
+  "stop_rule": "Ship an equivalence-preserving verify_prefix change with frozen baseline, identical witness receipts, rejection/mutant controls, certificate pin discipline, registered replay and a measured before/after median. Do not reopen the mathematical rejection or run the selected cold frontier replay.",
+  "outcome": "completed",
+  "evidence": "PASS_INPLACE_SIDE_PREFIX_FAST via research/inplace-side-performance/verify.py (also under -O); PASS_IDENTICAL_INPLACE_PREFIX_RECEIPTS via benchmark.py; certificates/inplace-side.json differs from base only in the inplace_side.py source pin.",
+  "author": "Wentao Li",
+  "github": "VictorLiwentao",
+  "email": "wentaol4@illinois.edu",
+  "ai_assistance": "Substantial AI assistance for orientation, overlap search, profiling, implementation, packaging, benchmarking and writing. Not independent human peer review.",
+  "open_questions": [
+    {
+      "question": "Can the same least-unused scan replace sorted(unused) pivots inside compile_side without changing gate lists?",
+      "why_it_matters": "compile_side is not on the certificate hot path here, but shares the same lexicographic rule; unifying it would remove a second O(n log n) sort per column on small controls.",
+      "first_step": "Diff gate lists for h=6 and h=10 between set/sort and flag-scan compile_side, then time h=50 if identical.",
+      "blocked_by": "Out of this session's stop rule; not needed for the verify_prefix receipt claim."
+    },
+    {
+      "question": "Should verify-certificates adopt a microbenchmark regression gate for this screen?",
+      "why_it_matters": "A future edit could restore sorted(unused) without failing equivalence tests.",
+      "first_step": "Decide whether RESULT.json medians belong in CI or remain descriptive.",
+      "blocked_by": "Maintainer CI-budget preference."
+    }
+  ]
+}

research/inplace-side-performance/VERIFY.json

@@ -0,0 +1,54 @@
+{
+  "schema_version": 1,
+  "status": "ready",
+  "python": "3.12",
+  "requirements": [],
+  "inputs": [
+    "scripts/experiments/inplace_side.py",
+    "scripts/experiments/inplace_side_witnesses.json",
+    "scripts/audit_inplace_side.py",
+    "certificates/inplace-side.json",
+    "docs/research/inplace-side.md",
+    "tests/test_inplace_side.py"
+  ],
+  "commands": [
+    {
+      "argv": [
+        "{python}",
+        "-I",
+        "-B",
+        "{package}/verify.py"
+      ],
+      "timeout_seconds": 180,
+      "checks": "Exact baseline copy of the base-commit source; identical verify_prefix receipts on all three pinned witnesses and small h=6/10 controls; shared rejection of reused switches, wrong prefixes and out-of-range switches; greatest-unused mutant rejected on every witness; certificates/inplace-side.json matches certificate() and differs from the base certificate only in the inplace_side.py source pin."
+    },
+    {
+      "argv": [
+        "{python}",
+        "-I",
+        "-O",
+        "-B",
+        "{package}/verify.py"
+      ],
+      "timeout_seconds": 180,
+      "checks": "Same equivalence and certificate discipline under -O (require() gates, not assert)."
+    },
+    {
+      "argv": [
+        "{python}",
+        "-I",
+        "-B",
+        "-m",
+        "unittest",
+        "discover",
+        "-s",
+        "tests",
+        "-p",
+        "test_inplace_side*.py",
+        "-v"
+      ],
+      "timeout_seconds": 180,
+      "checks": "Existing inplace-side unit tests plus focused performance-package controls."
+    }
+  ]
+}