Promote takum16 to strict SW-bitexact (independent 400-bit witness, 62->63/83)#1229
Merged
Merged
Conversation
Closes #1228 takum16 moves structural -> strict SW-bitexact. It reuses the proven make_takum_decoder(16) (value = exp(ell/2), ell = (-1)^S*(c+m)); the bit-precise claim is backed by a NEW independent second witness (conformance/vectors/witness_takum16_bitexact.py): ell is re-derived EXACTLY from the bit layout (dyadic rational), value = exp(ell/2) computed at 400-bit mpmath, rounded to nearest-even f64. Over ALL 65536 codes: the generator decoder == high-precision witness with 0 mismatches; 0 codes ambiguous near a rounding midpoint; min gap to a midpoint = 4.2995803e-5 ULP, STABLE at 400/800/1600-bit (a real gap, not numerical noise). Same criterion that promoted takum8 (#1223). takum32/64 HONESTLY stay structural (2^32 / 2^64 not exhaustively witnessable; need an external libtakum oracle for a full-domain gap proof). Encoding-conformance only; no quality/silicon claim. catalog stays 83; decode-HW/compute-HW unaffected. Conformance gate verdict UNCHANGED vs master (zero new drift). INDEX: bitexact 62->63, structural 15->14, selfconsistent 6.
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
This was referenced Jun 30, 2026
…omote-takum16-bitexact # Conflicts: # docs/NOW.md
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-07-01 11:50:38 UTC
Summary
Seal Status
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #1228
Что
takum16 переходит из structural в строгий SW-bitexact. Использует уже доказанный
make_takum_decoder(16)(value = exp(ell/2)); bit-precise подкреплён НОВЫМ независимым вторым свидетелемconformance/vectors/witness_takum16_bitexact.py.Доказательство (независимый второй свидетель, без внешнего libtakum)
ellвосстанавливается ТОЧНО из бит-раскладки (диадическая дробь),exp(ell/2)при 400-битной mpmath (независимый высокоточный оракул),По ВСЕМ 65536 кодам:
→ каждый код однозначно correctly-rounded f64; abs_error=0 на каждом записанном векторе. Тот же критерий, что повысил takum8 (#1223).
Честность / границы
Файлы (+203/−22, дрейф 0 0)
gen_all_formats.py (ветка takum16) · INDEX_all_formats.json · takum16 pack · structural_reason в takum32/64 · witness_takum16_bitexact.py (новый) · docs/NOW.md (свежая запись).