Repository navigation
feat(wordram): execute bounded Chan–He recursion natively #1512
Description
Complete traversal component of #1446 under unchanged #1394. Parent implementation worker codex-chanhe-native-t3-28da2261 owns this scope; #1482 frontend and #1493 encodings remain disjoint.
Implement a fixed-register finite native program that consumes compact injective three-set descriptors and runtime fuel, executes the existing two-prime modulus and complete heavy/packed-node programs, and traverses exactly the existing ChanHe.nodes in preorder using an explicit bounded memory stack. Charge all stack/descriptor/index/address work and loops, skip exactly the mathematical zero-fuel/empty branches, output exact packed array pointers/count/common length and preserve live input/output memory. Derive local arithmetic/memory fits from an explicit polynomial envelope and bound charged work by the ternary node bound times actual per-node costs. No assumed node oracle or runtime-sized code. Include satisfiable boundary/negative clients, six gates/static checks, source/model/encoding/resource mapping and independent/outside review requests.
Dependencies: #1506 native modulus and #1508 complete native heavy/node packing, both above corrected #1497. Frontend State.init composition, the (n+2)^κ size bridge, global positive O(log(n+2)) width and final genuinely subquadratic preprocessing theorem remain parent #1446 work after this component. Neither this issue nor a stack kernel alone completes #1446 or #1394.
Activity
edemaine commentedon Oct 28, 2026
active —
codex-chanhe-native-t3-28da2261claims full traversal component in provided isolated checkoutt3-28da2261, actual brancht3/native-chan-he-recursionfrom #1506 a5a5391. I will integrate exact #1508 locally and implement native bounded stack traversal with charged complete node calls; #1482 frontend remains exclusivelycodex-native-frontend-9249e0b5. No merge/approval or full #1446 completion is claimed.edemaine commentedon Oct 28, 2026
active —
codex-chanhe-native-t3-28da2261: fixed native DFS code and worklist invariants are implemented locally. Exact seven-word native transfers, the complete zero-fuel controller (173 instructions), actual modulus/node calls with all argument delivery, and 24-instruction output delivery elaborate. Executed regressions check zero/empty roots, signed singletons and a depth-two branch with two different exact packed arrays. General live-step/loop correctness and resource-envelope composition remain in progress; no PR or completion claim yet. No global Axioms audit is running.edemaine commentedon Oct 28, 2026
active —
codex-chanhe-native-t3-28da2261: traversal source/proofs are preserved while I finish the two review clarifications on frontend #1527 in the same isolated checkout (all six gates pass; serial static lint rerun in progress). Next is resuming the full worklist loop invariant. The native child/visit/output contracts are proved locally, but this component is not yet submitted or complete; #1446/#1394 stay open.edemaine commentedon Oct 28, 2026
codex-chanhe-native-t3-28da2261 active on t3/native-chan-he-recursion: complete fixed-128-register root traversal now passes a strict targeted build. The proof discharges actual pop/search/histogram/packing/push executions, exact finite preorder/readback, bounded stack, protected frontend prefix, polynomial arena envelope and charged transition bound. A fully admitted live client and full validation remain before submission. I am preserving this local work while addressing requested existing-PR cleanups on #1497/#1465 before another publication. #1446 still requires forest/State.init, derived logarithmic width and global subquadratic-time composition; #1394 remains open.
edemaine commentedon Oct 28, 2026
codex-chanhe-native-t3-28da2261 active: bounded-root proof/cost checkpoint is preserved on t3/native-chan-he-recursion; strict root and admitted live-client checks passed. The #1446 child forest branch now has an actual complete fixed-code parameter/frontend/prime/reservation/root-loop program, an exact frontend-entry-to-existing-allNodes bridge, and a successful complete empty-input interpreter regression (including all 20 roots and the computed prime count/reservations). These are implementation progress, not a completed initialized-input correctness/width/time theorem. Targeted artifact validation is running; queued modulus doc and then root all-six validation follow the released band slot.
edemaine commentedon Oct 28, 2026
awaiting review — codex-chanhe-native-t3-28da2261: complete bounded-root native traversal is submitted in #1535, stacked on #1527 with unchanged #1508 as a sibling prerequisite. All six/static checks pass; exact validation/source/model/ABI and remaining-scope record is in the PR. Outside coordinator review requested; independent nonauthoring math/API review is requested through the coordinator. Full initialized-input forest, global width and subquadratic preprocessing composition remain active under #1446; #1394 remains open.
SamuelSchlesinger commentedon Oct 29, 2026
Completed: actual bounded native recursion, stack execution, preorder output and charged costs. Landed through #1570; verified on main with accepted source, contributor credit and API follow-ups retained. The broader #1394 objective, concrete fast ET backend and complete ordinary-3SUM/1.9992 endpoint remain open.
feat(3sum): integrate native preprocessing and conditional host (csli…