Skip to content

Asynchronous whole-owner replicas: kappa 0.001357701757628940 (+1.9236% over #496) - #505

Closed
Dugongue wants to merge 2 commits into
CrocSwap:mainfrom
Dugongue:tn15
Closed

Dugongue wants to merge 2 commits into
CrocSwap:mainfrom
Dugongue:tn15

Conversation

@Dugongue

Copy link
Copy Markdown
Contributor

Result and scope

This adds a new asynchronous replica schedule giving conditional kappa = 0.001357701757628940 (67885087881447/50000000000000000), 1.923642% above #496. The complex supplier is b = 0.001359547617819233570818. The unchanged #493 bit supplier is c = 0.001565442026981656908596 and is nonbinding.

The inherited all-size compiler, mutable-basis/common-label, #484 Lemma A realization, restored-row, precision/tape, and analytic-transfer contracts remain explicit. This is a checked finite derivative under those conditions, not an unconditional multiplication theorem. #494's larger computation uses a different disputed split-residual interface and is not claimed surpassed.

New mechanism

Remove replica lockstep while keeping every complete local invocation atomic. Shuffle the eight nine-stage controllers by (stage + 3*replica/7, stage, replica), preserving every replica's exact local, bridge, idle-completion and terminal order. Whole helpers may borrow pending coordinates from a different replica according to that owner's actual last touch and deadline.

This opens4,539 same-pair/different-replica uses that the stage-only test rejects. Each helper still has exactly one whole physical owner; all 800 families per invocation are distinct. Each borrowed spectator is restored before the next complete invocation. Disjoint residual blocks plus an explicit remainder reproduce each original pending child. No gains are added numerically and no cross-family residual split is used.

The local gauge80-kernel42 word is unchanged. All40,320 helper roles fit within7,680 live families (8,258 physical). The changed profile is {1:148240,2:66816,3:7032,4:19728,5:26109,6:144,7:2209,8:1224,16:1,48:408}, with rank mass 558,144 and deficit 2,496. The maximum child 48 retains depth2 because73^2>2*48^2.

The prefix is freshly recounted:72 local sweeps and 72 individual bridge sweeps give 205bits;512 are charged, with an extra2*512 in the two-level row bill. The row coefficient is16,467<20,161. Both exact moment engines include the inherited bad-cover fallback. Six bootstrap steps feed the strict47-inequality assembly; the next kappa grid point and zero-eta/zero-beta/unit-beta controls are rejected.

Reproduction

Added folder: research/asynchronous-whole-owners.

python -B research/asynchronous-whole-owners/verify.py --work /absolute/path/to/new-output

Python3.11+ standard library only; offline; no optimizer or compiler required. This reconstructs the parent local dependency, emits the asynchronous template, independently audits all physical owner charts and exact phase/event projections, checks19 deliberately broken controllers, and recomputes the full changed finite invoice and exact assembly. The frozen witness replaces all search-time MILP assumptions. The unchanged #493 bit operator is cited as a pinned dependency; its published profile is repriced here, but its operator is not redundantly reverified.

The serialized version explicitly restricts replica-local events to their named replica. An old decoder that ignores these new fields cannot interpret this template. Two final whole-stock events run only once.

Sources and what this combines

Retained research

PROOF.md and SCHEDULING-LEMMA.md give the sufficient composition argument. ADVERSARIAL-REVIEW.md checks bridge-sensitive intervals and explicit event semantics. RESEARCH.md records the positive scheduling route and the matroid-component, joint gauge/kernel, sparse-shear and retiming searches, distinguishing checked local identities from failed global improvements. Exact source hashes and frozen witnesses are included for future workers. No optimality or exhaustive impossibility claim is made.

Public snapshot frozen through #497 before packaging. No claim of freshness beyond that snapshot.

DreamingOfClouds added a commit to DreamingOfClouds/integer-mult-bounds that referenced this pull request Oct 11, 2026
…e, nits)

- M1: the stream model is sourced to CrocSwap#442 section 2 (distinct slots ->
  distinct physical classes, sequential co-ownership across cover classes);
  CrocSwap#458 section 3 is cited only for g = d sigma^-1 and the fixed-route
  bijection, and its one-helper rule is stated as CrocSwap#458's case. Lemma A is
  proved in CrocSwap#442's stream model, which is inherited.
- M2: CrocSwap#505 coverage stated for the template emitted by replaying CrocSwap#505
  (4ee769c) here, which does not hash-match CrocSwap#505's EXPECTED (aa741871 vs
  276b69e3); (E), (F), per-replica order, prefix bound and invoice not
  re-checked.
- Nits: hne for data wires and centre copies; loop_order check is
  declarative; checker layout assumptions stated; commute_of_not_mem axiom
  probe (24 named results); anchor coverage rests on this package's
  checkers; CrocSwap#512 rests on its own report.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mister-Meeseeks pushed a commit that referenced this pull request Oct 12, 2026
…in Lean 4

research/e8-lemma-a-lean/: the realization part of Lemma A (assumed from
#458 s3 in #484) as a general theorem over any semiring, modules, charts and
stage count: one owner + whole invocation + no other actor + Lemma G's exact
endpoint + the checked factorization C_[w] x I => the helper's family ends as
chart^-1 (C x id) chart of itself, spectators untouched; necessity of the
one-owner rule; disjoint-block residuals commute and gather into the idle
child; #505's asynchronous borrowing as a proved window variant. Reuses the
Lemma G package (Lake path dependency, same Mathlib pin). New instance check
code/check_owners.py (literal owner injectivity on the emitted templates).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mister-Meeseeks pushed a commit that referenced this pull request Oct 12, 2026
- Model: units are physical streams (family, cover class) as in #442/#458
  and the template's route_semantics. New LemmaA/Streams.lean:
  streamOwner_injective(_of_disjoint), sweep_unique, lemmaA_streams (one
  owner per stream from disjoint footprints on shared families); plus
  lemmaA_piece and lemmaA_async (Lemma A feeding #505's window variant).
- code/check_owners.py rewritten for stream units (data wires alone,
  disjoint footprints on shared families, centre copies in distinct private
  slots, atomic calls; 3 mutants): all four #484 words pass, including the
  32 gauge80 / 16 kernel79 slot-sharing calls and the anchor.
- code/check_windows.py (new, independent): every hosted block inside its
  pending child's open window, blocks disjoint, family-group windows
  disjoint (2 mutants); all four words and #505's emitted async template.
- verify.py: runs Lemma G's check_extended_labels.py ((F) is read in the
  extended labels), fails on a missing LEMMA-A.json, checks I-helper counts
  against EXPECTED, prints the owner and window results; --async T.json.gz.
- Docs: unit model stated, coverage per word, #505 replay results and
  what is not re-run, commutative-semiring wording, credits #433/#442/#500.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@Mister-Meeseeks

Copy link
Copy Markdown
Contributor

Thank you @Dugongue. This contribution is now retained on main in the scoped 100-PR review and contributor credits.

Contribution: Asynchronous scheduling of complete whole-owner invocations.

Review and limits: Each replica's local order and owner deadlines remain fixed; spectators are restored between calls. No residual is split and no general scheduling optimum follows. Missing metadata pin repaired on a copy; adapter JSON line endings explain the template pin exactly, and dependent emitters/audits reproduce every original expected value.

The complete original source, notices and relevant execution evidence are retained through the pinned source archive. The review distinguishes fresh local runs, source-bound historical Linux CI and prior exact-head review; it does not claim 100 fresh full mathematical audits. Closing this historical PR as reconciled. This preserves parallel and cumulative credit without replacing the selected frontier or promoting priced research targets to completed constructions.

Reviewed head: 4ee769c8e28837482272b5f4742497da8e2fc9d4. Publication: 64039f1bd3b6a523b96083fd858ad040a75e3030.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants