Effective bit-17 restart to carry 22
A hard reset at bit 17 remains operationally effective only when no later bit-18 reset overwrites it; under that condition it should reconstruct carry 22 exactly.
Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.
What was tested?
A hard reset at bit 17 remains operationally effective only when no later bit-18 reset overwrites it; under that condition it should reconstruct carry 22 exactly.
Why the test is meaningful
In ripple addition, the latest reset dominates all earlier carry history. Calling every bit-17 reset useful would double-count records later reset at bit 18. The correct estimand is therefore the effective subset: reset17 ∧ ¬reset18.
E17 = R17 ∧ ¬R18c18 = value(R17) on E17c22 = G_{18:21} ∨ (P_{18:21} ∧ c18)How it was tested
Join the exact reset atlas to the connected suffix masks, retain only records whose latest reset is bit 17, reconstruct carry 22 lane by lane and repeat the frozen procedure on an independent 300-record cohort.
What happened
The original cohort contained 233 effective bit-17 resets among 1,361 records (17.12%); the fresh cohort contained 64 among 300 (21.33%). Carry-22 reconstruction had zero semantic errors. Fresh reset values split 28 zeros and 36 ones, while mean nonlinear support at carry 22 was 2.484 monomials.
Exactness and statistical controls
Exact lane equality, not correlation, was the primary gate. Original and fresh cohorts were reported separately, and bit-18 dominance was applied before any rate calculation. No independent probabilities were multiplied.
What the result means
V41 validates a narrower, causally correct restart capability. The bit-17 reset survives to the requested boundary only in the effective subset; its downstream Boolean complexity remains nonzero.
Limitations
- The effective fraction is distribution-dependent.
- ANF support is not native execution cost.
- No prospective selector or ASIC saving was measured.
Evidence trail
Rebuild reset17/reset18 masks, select the latest reset, ripple the exact suffix to carry 22 and require zero lane errors in original and untouched fresh cohorts.
Canonical variants
SUBENGINE-V41ASUBENGINE-V41BSUBENGINE-V41CSource: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.