Skip to content

Ordinary BIT saving 0.003483969976 (+0.24491667%) via owned G20 relations - #5512

Merged
Mister-Meeseeks merged 2 commits into
CrocSwap:mainfrom
ai-nanachi:codex/bit-g20-nine-owned-relations
Oct 25, 2026
Merged

Mister-Meeseeks merged 2 commits into
CrocSwap:mainfrom
ai-nanachi:codex/bit-g20-nine-owned-relations

Conversation

@ai-nanachi

Copy link
Copy Markdown
Contributor

This changes the complete native BIT program and admits the conditional ordinary saving 435496247/125000000000 = 0.003483969976, up 0.24491667% from the completed PR #4814 endpoint 1737729/500000000. The exact relative gain is 1063997/434432250. This is an ordinary BIT component improvement; the selected whole multiplication result is unchanged.

The package combines the previously private G20 shadow/rank4 producer with Ali Romman's nine owned capture-relation method from PR #5421, independently rederived on the actual G20 word. G20's actual generators, native source, proof, immutable inputs and retained predecessor notices are published in this PR; it is not treated as a prior public or externally accepted result.

  • The local word accepts 649 arbitrary original functions and explicitly initializes/returns 11 owned arrays to zero. Nine complete rank8 helper targets are retired; all four partial rank4 helper targets remain. Exact rational FRAME induction and paired HEAD/TAIL controls cover 660 independent RAW/DIRTY entry columns.
  • The literal 18-call caller acts on 11,238 independent originals plus 11 arbitrary DIRTY owned entry columns. It pays 56,997 recursive jobs, removes 162 rank1/rank7 pairs and debits 162 rank8 targets. Every current-source descriptor binds the actual complete prefix, owner, epoch, version and geometry. Owned storage receives no productive credit.
  • The same physical endpoint serves both thick-call directions, and every final job is a physical involution on the same domain. The complete program therefore supplies inverse/transpose/inverse-transpose with fresh paid owned storage, without reversing destructive erasures. Native routes, all-column and whole-ROW/spectator assumptions remain explicit.
  • The new 517-bit envelope, complete positive scans/parking/routes/normalizers, eleven owned banks per compiled or suspended instance, fresh compiler/prime/cover/stock bounds and three finite completed backend levels are charged. The admitted ordinary endpoint is distinct from the raw crossing (0.0034839853304,0.0034839853305).

Reproduce with Python 3.12, standard library only:

python3 -B scripts/research_package.py --package research/bit-g20-nine-owned-relations

Validation passed: registered metadata and replay; separate literal/QQ/positive-series auditors and semantic failure controls; fresh primary and separate-audit cold replays under normal Python and -O; exact committed-byte/source-history preflight; committed export integrity and Git-free cold replay. All runtime inputs are package-local. Canonical proof payloads are compared rather than gzip headers or timing. The final documentation receipt retains the exact source-validated ancestor and the unchanged mathematical source coverage.

The construction retains the inherited native/all-column, owned-erasure, mixed-target finite compiler, completed ordinary backend, precision/fixed-tape and analytic framework premises. The enormous uniform program is algorithm-defined. No unconditional or full formal multiplication theorem, external-human audit, selected-frontier change, or optimality claim is made. Open next questions and precise coverage are in TASK.json/PROOF.md.

Credit: Ali Romman (PR #5421); DreamingOfClouds/eumemic and the PR #589/PR #555 word lineage; Cinna Davis's controller/ROW/cover work; the PR #4760/PR #4814 productive-helper and ordinary lineage; rohanarun's PR #185/#199/#204 finite bootstrap; and the unchanged exact interval/recursive sources and all retained notices. Prepared for ai-nanachi with substantial OpenAI Codex assistance. Separate internal AI reviews are distinguished from external acceptance. New work is Apache-2.0.

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