Research/jp8 diagonal gauge seed35

[mythrilium]

Result and scope

Research construction with an exact scoped optimality certificate; no new kappa or complete supplier-cost improvement.

On the historical PR626 JP8 cleanup, a restored power-of-two diagonal change of 203 live coordinates saves 6,154 real-component scalar primitives per orientation, including all 406 boundary-scale macros (916 primitives). Forward/backward totals become 251,078 / 251,126 from 257,232 / 257,280. The forward positive prefix envelope grows from 64 to 70 bits; backward remains 101, within the inherited 192-bit allowance.

The exact integer primal/dual witness certifies objective 10,326 against 13,403 at zero gauge, globally over constant positive power-of-two diagonal gauges of the fixed chronological cleanup, original per-add Horner lowering, and separately paid boundaries. It is not global circuit optimality. The larger macro count may worsen the generic adapter allowance; no complete supplier invoice or practical speedup is claimed.

Sources and overlap

• Original research base: 66849e1475c930fd7732b1a0067ffee7e1e95f60; OS assignment seed 35, complex supplier.
• Publication refresh: bce54fb3f895cbe5cd4c0eb477c3afeed100190f, selecting Conditional kappa 0.001672660197076918: rare-pattern T16 exchanges #1094 at κ = 0.001672660197076918. This package concerns frozen Conditional κ = 0.001635527380375392: coordinated owner repair on PR623 #626, not the current frontier.
• The original selected record is retained byte-for-byte inside the package; all six other consumed source files remain unchanged on refreshed main. Exact source hashes and PR heads are in SOURCES.json.
• Earlier local claim, candidate selection and completed coverage are in CLAIM.json. No pinned sprint issue existed during the research session.
• Credit Jeremy Kothe / Conditional κ = 0.00163542231500688: K-triangle skeleton atom (492 helpers per call) on the PR617 cold-ended witness #622/Conditional κ = 0.001635490664920455: improved owner map (ck 8 / jp8 64) for the PR622 K-triangle witness #623, Chafik Boukhalfa / Conditional κ = 0.001635527380375392: coordinated owner repair on PR623 #626, icekylinx, DreamingOfClouds, Dominik Scholz, Dugongue, Rohan Arun, eumemic and retained predecessors.
• Distinct concurrent work includes Research construction: rescale JP8 coefficient work across magnitudes (no new kappa) #739/Research: save 62 CK primitives per orientation by relocating one scale (no new kappa) #750/Research: JP8 coefficient reuse across all scalar phases (no new kappa) #755/Research: CK dyadic ladders save 13,770 / 67,146 primitives (no new kappa) #769 and later Research: factor center quarters and certify the adapter-cost limit (no kappa) #786/Research: batch JP8 compensation, saving 80,150 / 81,464 primitives (no new kappa) #793/Scoped negative result: exact 73-primitive factory scale optimum (no new kappa) #949/Research note: scale deferral saves 66 beyond one JP8 fan-in block (no new kappa) #984. In particular Research: batch JP8 compensation, saving 80,150 / 81,464 primitives (no new kappa) #793 reports larger scalar savings through coefficient routes; this contribution adds a different simultaneous live-register gauge and exact unbounded-domain class optimum. No pending code is imported or savings claimed additive.

Only research/complex-jp8-diagonal-gauge/ is added. Selected certificates and upstream/ remain unchanged.

Reproduce

python3 -I -B research/complex-jp8-diagonal-gauge/check.py

Expected:
PASS_JP8_GAUGE_COMPLETE saved=6154/6154 optimal_cost=10326 tests=10 inherited_controls=25 kappa=unchanged

The final publication replay passed: solver-free exact primal/dual check, whole scalar programs on 716 native plus eight arbitrary private-centre input columns, all 846 distinct coefficient primitive boundaries, separate integer-dyadic cleanup replay, ten tests with twelve adverse mutations, and unchanged native emitter/auditor with 25 inherited controls. The uploaded Git tree matches the locally verified tree exactly.

Earlier verification also covered optimized-parent execution, relocated offline replay, equivalent gzip recompression, source/receipt corruption, deterministic rediscovery and the research-base selected-record check. See VALIDATION.md for exact coverage and limits. Full repository/cold frontier replay, global invoicing, hosted CI, Lean and independent human review are not claimed.

Inherited all-size/compiler, physical/common-label realization, hosting, precision/tape/recovery, bit and analytic obligations remain. Open next questions cover complete adapter pricing and composition with coefficient reuse.

Prepared for Luis Leon (@mythrilium) with substantial OpenAI Codex assistance in research, implementation, verification and writing. Apache-2.0; predecessor notices retained.

[Mister-Meeseeks]

100-PR batch review at fd33c640407dda0867604722b612e6c68b7cd6e2.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: Restored JP8 constant diagonal gauge with exact integer primal/dual optimum, both arbitrary-dirt orientations, complete scalar costs and inherited native/frame controls. Saves 6,154 real primitives per orientation; extra macro/adaptor costs still require a global invoice.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.