Repository navigation
Conversation
…e-optimized frames
Two observations establish the failure mode: CrocSwap#202's canonical check requires the sha256 of a container it regenerates (exact_complex_flow_lift.py:409 and :419, required by source-assisted-v4/verify.py:162) and fails on unchanged inputs, while CrocSwap#194 passes in the same environment; and PR212's pinned archive rebuilds byte for byte as a tar payload but not as a container. The repository already has the fixes in use -- CrocSwap#194 decompresses before comparing, CrocSwap#217 pins payload hashes and CrocSwap#207 checks semantic fields -- so this adds a scanner that locates the pattern, the audit with file and line evidence, and one CONTRIBUTING.md paragraph stating the rule once. Generated with Codebuff Co-Authored-By: Codebuff <noreply@codebuff.com>
Two observations establish the failure mode: CrocSwap#202's canonical check requires the sha256 of a container it regenerates (exact_complex_flow_lift.py:409 and :419, required by source-assisted-v4/verify.py:162) and fails on unchanged inputs, while CrocSwap#194 passes in the same environment; and PR212's pinned archive rebuilds byte for byte as a tar payload but not as a container. The repository already has the fixes in use -- CrocSwap#194 decompresses before comparing, CrocSwap#217 pins payload hashes and CrocSwap#207 checks semantic fields -- so this adds a scanner that locates the pattern, the audit with file and line evidence, and one CONTRIBUTING.md paragraph stating the rule once. Generated with Codebuff Co-Authored-By: Codebuff <noreply@codebuff.com>
Two observations establish the failure mode: CrocSwap#202's canonical check requires the sha256 of a container it regenerates (exact_complex_flow_lift.py:409 and :419, required by source-assisted-v4/verify.py:162) and fails on unchanged inputs, while CrocSwap#194 passes in the same environment; and PR212's pinned archive rebuilds byte for byte as a tar payload but not as a container. The repository already has the fixes in use -- CrocSwap#194 decompresses before comparing, CrocSwap#217 pins payload hashes and CrocSwap#207 checks semantic fields -- so this adds a scanner that locates the pattern, the audit with file and line evidence, and one CONTRIBUTING.md paragraph stating the rule once. Generated with Codebuff Co-Authored-By: Codebuff <noreply@codebuff.com>
Two observations establish the failure mode: #202's canonical check requires the sha256 of a container it regenerates (exact_complex_flow_lift.py:409 and :419, required by source-assisted-v4/verify.py:162) and fails on unchanged inputs, while #194 passes in the same environment; and PR212's pinned archive rebuilds byte for byte as a tar payload but not as a container. The repository already has the fixes in use -- #194 decompresses before comparing, #217 pins payload hashes and #207 checks semantic fields -- so this adds a scanner that locates the pattern, the audit with file and line evidence, and one CONTRIBUTING.md paragraph stating the rule once. Generated with Codebuff Co-Authored-By: Codebuff <noreply@codebuff.com> (cherry picked from commit 24e51c7)
|
Thank you @hcg890. This contribution is now retained on main in the scoped 100-PR review and contributor credits. Contribution: Paid XOR butterflies with cascade-optimized frames. Review and limits: Prior exact-head changed bit construction, strict complex adapter, packing and assembly replay. Original dependency-version gate and adapted environment are explicitly distinguished. The complete original source, notices and relevant execution evidence are retained through the pinned source archive. The review distinguishes fresh local runs, source-bound historical Linux CI and prior exact-head review; it does not claim 100 fresh full mathematical audits. Closing this historical PR as reconciled. This preserves parallel and cumulative credit without replacing the selected frontier or promoting priced research targets to completed constructions. Reviewed head: |
Result
Conditional κ = 0.000684696826673891
Exact value: 684696826673891 / 10^18
Improvement over PR211: 0.1241%
This combines a new 440-butterfly XOR rewrite with PR211’s cascade-optimized frames, using the existing PR200 bit construction, PR197 completed-residual packing and retained PR193 complex supplier.
What contributes
New scalar rewrite: 440 XOR butterflies. Each replaces
(a+d)+(b+c)with(a+c)+(b+d)at two single-use pair nodes feeding the same four-sum. This changes 880 scalar controls and 880 graph nodes while preserving the carried value. The unused pair has no later consumer, and reverse execution restores every temporary.Inherited frame optimization: PR211’s 6,191 entries. Three entries, at operations 15154, 16230 and 16881, overlap the rewritten pair operations. They are explicitly replaced by the admitted butterfly pair frames, and the complete combined chronology and geometry are checked.
Inherited terminal elimination, packing and complex supplier. The existing 34-sink transformation and completed-residual packing are retained and repriced on the combined word. Their savings are not claimed as new butterfly improvements.
The measured κ improvement is for this complete composition versus PR211. Separate mechanism gains are not added together.
Combined cost and validation
The final paid profile has rank mass 4,055,136, deficit 5,808 and largest child 22. Nine complete core invocations retain 45,842 banks and 154,026 incidences, charging 7,093,890 literal XORs and 5,701,209,426 selector calls. Assembly uses three finite bootstrap levels.
The complete replay and profile/price binding pass:
Scope
The proposed consequence is conditional:
T(n) = O(n (log n)^(1−κ))
Weighted compilation, separated completed-core residuals, restored-row and ordinary-selector contracts, routing, prime supply, precision/recovery and fixed-tape analytic transfer remain inherited hypotheses. These finite certificates do not establish an unconditional theorem or practical multiplication speed. SOURCES.json, README.md and PROOF.md provide exact dependencies, reproduction instructions and the argument.
Lineage and credits
Rohan Arun (@rohanarun): PR211 cascade frames and PR185/187 predecessors. Dugongue: PR207 frame refinement and PR186 banks. Chafik Boukhalfa (@chafreaky): PR200/202. Evan McKinney (@evmckinney9): PR197 packing. Avi Eisenberg (@ikeboy): PR193 adaptation. @icekylinx and @eumemic: flow and paired-cube antecedents. James Chang (@jamesyc), Jean-Guillaume Dumas and Bruno Grenet: accumulation antecedents. Maxime Fleury (@maxime-fleury): PR204’s independent combination model. Original notices and earlier credits are retained.
Assisted with ChatGPT.