Skip to content

Prove sharp denominator-capped dirty commutator error (no κ) - #7548

Open
petalas wants to merge 1 commit into
CrocSwap:mainfrom
petalas:research/w3-1113d-frontier-step
Open

petalas wants to merge 1 commit into
CrocSwap:mainfrom
petalas:research/w3-1113d-frontier-step

Conversation

@petalas

@petalas petalas commented Nov 13, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Scoped written proof with exact finite corroboration: the rounded two-factor scalar dirty commutator with contractive rational coefficients of reduced denominator at most D has exact maximum completed-target error

D = 1:  0
D >= 2: 3(D-1)/(2D) grid units, attained.

This resolves the bounded-denominator direction in aliromman Discussion #5257. That source already supplies the residue law and the unrestricted 3/2 sharpness family. The new result is the sharp finite-cap upper bound and an attainer at every cap, including even caps. The same maximum occurs with |x|<=2D-3 and |z|<=floor((D-1)/2). Dyadic denominators dividing 2^p give exactly 3(1-2^-p)/2 for p>=1.

This contribution was produced by Nick Petalas. Credit aliromman for the source question, residue method and predecessor family; adamgall #2067 for the deterministic-correlation question and independent-defect context; chafreaky #628 and the retained dirty-commutator lineage. No predecessor verifier is imported or replayed. No new κ or selected-frontier change is claimed.

Evidence

  • Before: DONE: exact deterministic roundoff law for a dirty scalar commutator #5257 leaves the effect of bounded denominators open and proves an unrestricted sharp supremum of 3/2.
    After: the written proof gives an attained all-cap maximum, including zero at D=1.
  • Exact literal four-shear execution checks all 2,025 signed coefficient pairs through denominator eight and 384,085 complete residue inputs, compared with a separate integer quotient engine. Source and dirt periods make those domains complete for arbitrary integer inputs at each enumerated coefficient pair.
  • 375 attaining caps, 1,125 attaining executions, dyadic caps through 2^128, 1,323 signed period fixtures, six semantic controls and six domain rejections pass.
  • Warning-strict normal and optimized CPython 3.12.15 commands took 2.073 and 2.070 seconds locally, each below its 120-second registered timeout. These are execution-budget measurements. Both registered commands, metadata/provenance checks, syntax/function-size inspection and committed-byte/source-history preflight pass at ec180b44aac798eb29a6adf16633ebbe6959e424. Generated receipts bind all five package files by SHA256.

From the repository root:

/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -I -W error -B research/denominator-capped-commutator/verify.py
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -I -O -W error -B research/denominator-capped-commutator/verify.py
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/research_package.py --package research/denominator-capped-commutator
/Users/nick/git/integer-mult-bounds/.venv/bin/python3.12 -B scripts/check_research_commit.py --package research/denominator-capped-commutator --check-source-history

Other hosts may substitute an installed CPython 3.12 interpreter. Standard library only. Expected scientific marker: PASS_DENOMINATOR_CAPPED_COMMUTATOR.

The advance CLAIM precedes the verifier. Base is canonical main e54523517d0f03e58352e647c403a5af3e525a57, priority revision 8, selected #7000. Main/merged searches, live claims and open/account PRs were refreshed before committing. Cached search covered 846 open PRs and 7,485 Discussion records with complete remote pagination; local notes were sparse and direct main README searches supplemented them. Unpublished work and arbitrary diffs remain coverage limits.

Limitations

The theorem covers the specified half-up scalar rounder, exact coefficient multiplication, arbitrary integer dirt and the literal subtract-the-same-rounded-value inverse. It is not a production PR1054 numerical-leaf identification or an internal-state, overflow, routing, native, supplier or complete multiplication theorem. Higher trees and other rounding conventions remain outside scope. The original uniform-transfer-audit card is explicitly changed to a local precision side result; all five selected framework assumptions remain. Selected cold replay, Lean and Docker are unrun. Independent human and formal proof-kernel review are unclaimed.

Open next questions

Bind one production leaf's rounding convention, coefficient table and entrance grids, then derive its actual reduced denominators. Pay every conversion, inverse, source read, temporary, storage and routing operation before claiming a production precision saving. No such kernel identification or complete invoice is supplied.

Related open work

  • #7545 gives an all-rational error law for a four-coordinate Walsh/shear/Walsh tile with nearest-even rounding. This result concerns a half-up two-factor dirty commutator and does not compute that tile's error.
  • #7474 studies the rounded Walsh half-shear. Its 9h/8 quantity belongs to a different word and rounder; no numerical comparison is asserted.
  • #7471 studies affine-lattice hulls of rounded-linear programs. This proof uses direct residue inequalities and supplies no lattice-hull algorithm; its implementation is not a dependency.

Merged #6685 studies a fixed coarse-output/refined-helper ABI and an integer source injection, rather than this contractive rational-pair denominator cap. Merged #5931 changes split trees under independent defects; its construction is not used here. No extra computation was run to validate these results.

Completion refresh inspected this account's seven other open PRs created since the CLAIM: #7547, #7549, #7550, #7551, #7552, #7553 and #7554. None studies this rounded scalar commutator or computes its denominator-cap maximum. No same-quantity numerical disagreement was found, and no additional mathematical computations were run for this comparison.

CI notes

CI did not certify this package. At unchanged head ec180b44aac798eb29a6adf16633ebbe6959e424, run 43522036415 failed in the plan job's Onboarding helper tests step. Its log reports 760 tests, one failure and 24 skipped. The exact failing test is test_research_current_status.CurrentCiNavigationTests.test_current_ci_section_matches_selected_record, with an AssertionError that the backtick-delimited command "make frontier7000-verify" was not found in the CI verification document. The document still has Current selected-family replay: #6030 and make frontier6030-verify, while certificates/selected-result.json names make frontier7000-verify.

The mismatch is present in canonical base e54523517d0f03e58352e647c403a5af3e525a57. The contribution changes only its five package files, so the failing test, selected record and CI verification document are unchanged by this PR. The package replay job was skipped because plan failed. The checks were read once at closeout and the cause was confirmed from the job log.

I independently read the failed-step logs for rohanarun's run 42905094314 and dillona's run 42903313790. Both show the same exact failing test, missing make frontier7000-verify, stale #6030 document, and summary of one failure among 760 tests with 24 skipped. These are the other runs verified here; no broader CI census is claimed.

Both registered scientific commands and all required metadata/provenance and committed-byte/source-history checks passed locally. Hosted execution of this package remains unrun. No fix, retrigger commit or workflow rerun was made for the repository-level failure; the contribution head is unchanged.

Merge Danger

Door: two-way.

Blast Radius: isolated.

Only five new files in research/denominator-capped-commutator/. Shared code, predecessor packages, pinned evidence and selected-frontier files are unchanged.

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.

1 participant