Skip to content

feat(wordram): extract Chan–He heavy sets and pack one node natively #1481

Closed

Description

@edemaine

Independent implementation component of full native preprocessing #1446, under #1394. Delegator codex-chanhe-native-t3-28da2261 retains full milestone integration; a distinct isolated worker will claim this issue with its actual branch before editing.

For three injective signed-encoded integer arrays and a supplied positive modulus, implement the complete native one-node work: compute needed residues/histograms with existing native counting APIs, extract each exact ChanHe.heavy subset into compact arrays, and construct the exact existing Node.oneArray over an explicit common length at least 8m² (or equivalent exact length using N ≥ 2M). Include clearing/padding, signed negation/tags, all input/output/index work, a concrete nonoverlapping memory layout/frame, fixed code/registers, all local fit/intermediate bounds, and a charged O(n₀+n₁+n₂+M+N) bound. Reuse existing mathematical heavy/bucket/arrXY/arrZ/oneArray, not a parallel reduction. No host-level output or assumed algorithmic procedure.

Scope is new native heavy/light/node-packing modules and tests, with directly needed histogram/cardinality bridges. Do not edit ExactTriangle, frontend/sorting/parameters or prime-search modules. Prime choice and recursive traversal belong to the parent. Parent will derive global width/layout/time; this complete one-node component does not complete #1446/#1394. Preserve source credit, check actual pinned source quotations, six gates/static checks, abstract no-impact assessment and independent/outside review requests.

Activity

  1. self-assigned this
    on Oct 28, 2026
  2. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    available — scoped by codex-chanhe-native-t3-28da2261 for isolated implementation; no worker has started. Parent #1446 retains integration and will assign a distinct worker before edits. Length clarification: construct exactly Node.oneArray for 4N cells with N ≥ 2M; the full-preprocessing specialization is N=2m², hence common output length 8m².

  3. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    active — codex-native-node-packing-9f0c8d49 (implementation worker under codex-chanhe-native-t3-28da2261) claims #1481 on t3/native-chan-he-node-packing, worktree C:/Users/edemaine/.t3/worktrees/complexitylib/t3-9f0c8d49, starting at 1a45c8b. Scope is the COMPLETE native residues/histograms, three compact exact heavy outputs and exact Node.oneArray over 4N cells (N ≥ 2M), with concrete disjoint layout/frame, fixed code/registers, fits and charged linear cost. Only new scoped modules/bridges/tests will be edited. Prime/frontend/recursion/global composition remain parent-owned; #1446/#1394 stay open.

  4. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    active — codex-native-node-packing-9f0c8d49: histogram/cardinality and exact-heavy-list bridges plus the fixed direct placement scan are drafted. The source pin was verified locally; the existing Node.oneArray tags/negative-Z conventions are retained. Initial isolated cache seeding is still copying Mathlib; C: fell from ~11 GiB to ~2.6 GiB free, so I am compressing only this checkout's copied artifacts and controlling further cache growth before elaboration. No validation success or component completion is claimed.

  5. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    active — codex-native-node-packing-9f0c8d49 resumes the same complete implementation after the provider event stream closed unexpectedly. Existing branch/worktree/files are preserved. The 32n+8 native placement scan elaborates; exact semantic readback and the 12N+5 four-column padding loop are being checked, followed by full three-count/three-placement composition, layout/fit proofs and tests. Disk pressure has cleared. Per delegator coordination, only targeted checks run while ancestor repair audits occupy the serial window; no PR or aggregate/global audit will start before the corrected prime-search base is supplied. #1481/#1446/#1394 are not complete.

  6. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    active — codex-native-node-packing-9f0c8d49: rebased only my provided branch onto corrected prime-search head 0e05605; all implementation files preserved, cache bootstrap rerun and Complexity.lean regenerated. Strict targeted gates pass for histogram bridges, exact heavy/arrXY/arrZ readback, 32n+8 placement, 12N+5 padding, and the actual composed count/place/descriptor-write routine (66n+5M+45). Completing the three-round node layout, runtime setup, exact Node.oneArray readback and total resource theorem. The aggregate window is released but no full audit has started; publication remains held for the delegator's #1469 hosted-CI release.

  7. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    active — codex-native-node-packing-9f0c8d49, branch t3/native-chan-he-node-packing, rebased onto corrected 0e05605fb. The complete fixed native program and exact heavy/oneArray/frame contract pass targeted strict builds. Charged cost is 66(n₀+n₁+n₂)+15M+12N+228; concrete full-program regressions cover empty, unique/heavy, negative and boundary inputs. Finishing theorem-level admissibility clients and validation; coordinating the global Axioms window with #1446 before all six gates. Full frontend/recursion/global width-time composition remains with the parent.

  8. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    awaiting review — codex-native-node-packing-9f0c8d49: complete component published in #1508 at ba4cab2a6096c4d692e2d7a75b59d7591ea5162e, stacked on corrected #1497. All six gates pass together, including global Axioms; validation, source mapping, boundary clients and abstract assessment are recorded in the PR. Exact heavy/oneArray construction, full local fits/frame and charged linear cost are proved. Outside review requested; coordinator independent mathematical/API review requested. #1446/#1394 frontend, traversal and global composition remain open. Global audit window is released.

  9. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    Changes requested — outside review of #1508 at ba4cab2 identifies one blocking test/claim mismatch: the invalid-promises block does not exercise the component. Replace it with program-linked counterexamples, or remove it and correct the PR description as the reviewer permits. Review and inline details: #1508 (comment). Native owner notified; exact-head repair and focused delta review are required before integration.

  10. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    codex-chanhe-native-t3-28da2261 active: read exact ba4cab2 outside CHANGES REQUESTED and all inline findings. I am repairing only the three disconnected tests, the unsupported ordering prose and the PR validation claim, on local t3/native-node-review-fixes-28da2261 based at ba4cab2, pushing normally to the existing t3/native-chan-he-node-packing PR branch after all six/static checks. Native code and public contracts remain unchanged. Original child implementation is preserved; no overlapping remote edits. General fit-lemma placement will be recorded with concrete #1512 consumers; optional square-helper API churn is deferred.

  11. edemaine commented on Oct 28, 2026

    @edemaine
    CollaboratorAuthor

    codex-chanhe-native-t3-28da2261 awaiting review: blocking test/description repair pushed normally to #1508 at f6f22da. All six gates and static checks pass; native code/public library contracts unchanged. Focused outside delta review requested in the PR; unchanged mathematical review reused. Global audit window released. Full #1446/#1394 composition remains open.

  12. SamuelSchlesinger commented on Oct 29, 2026

    @SamuelSchlesinger
    Collaborator

    Completed: full native heavy extraction and exact Chan–He node packing. 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.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions