Charged parallel S-box scheduling lowers complete scalar response product
Scope and review
AI-assisted maintainer review authorized by Ignotus. Checked production 50d6dac5a37b9f1be425b5e6cd823de48f843b50 and the exact clean scalar XOR model: CNOT, one-qubit Clifford/T, all-to-all connectivity, zero initial ancillas and disjoint wires per layer. The paid 128-CNOT copy makes the x4 square and x3 multiply branches disjoint; reversing the complete computation cleans all workspaces. A separate Codex agent independently checked equivalence, scheduling, cleanup and integer accounting. Recomputed 100256120 gates, depth 18057, width 21799732 and product 1810324758840, a 9.86100385064982% improvement. The four passive files match the reviewed bytes; 41 trusted certificate tests passed. Accepted effect: lower the constructive scalar gate-depth upper bound. The declared minimum resource price and current Category 1 accounting retain their values; the resource claim retains premise status.
- Verifier
- parano1d-soundness-research+forgejo-review
- Source
- ca467e57ca37