Repository navigation
Construct fast dense execution of the original paired collector - #3131
Merged
Mister-Meeseeks merged 1 commit intoOct 17, 2026
Merged
Conversation
Mister-Meeseeks
pushed a commit
that referenced
this pull request
Oct 17, 2026
…at tensor multiplicity two Checks the first paired P104 collector of PR #3131 at actual tensor multiplicity two (n = 18, N = 262144, rank 14, events 2025-2039, native events 2031 and 2033), beyond the predecessor's complete N = 512 tests. Two separately written exact evaluators agree on the entire outgoing value, promoted query and working grid of four deterministic signed fixtures; the five dirty sources are preserved and the private collector returns to literal zero. Both actual wrappers are checked on every address (524,288 forward columns) and inverted. Two fixtures have nonzero rounding error, 5379503/8 and 2615337/4 in working-grid numerator units (grids 3 and 5), inside the exact forward and paired-inverse relative bounds; the cancellation and zero fixtures (grids 25 and 3) have none. The original return recipe passes an ideal round-trip fixture, and layout, phase, lift-order, promotion, restoration, source and receipt controls reject. PR #3131's own nested evaluator did not finish on this fixture in two 600-second attempts; the optional script is kept outside the registered replay and is not claimed as evidence. PR #3134's implementation, lambda > 2, the fast physical trace, clean-stock hosting, all-size refinement and the consumer bill remain open, and the inherited 599/584 screen is unchanged. No new kappa or selected-frontier change. Adds research/p104-tensor-collector-audit/ only. Submitted by @commdata2338 with substantial AI assistance (OpenAI Codex research and implementation; packaged with Claude Code). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG
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
2 participants
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.
I supply a conditional fast execution of the original paired P104 collector, including its dense Gaussian monomial wrappers. For
n=9lambda,N=2^nand linear working precision, I derive timeO(N lambda^(2-8003/6000000))and additional clean spaceO(N lambda^(5991997/3000000))+F. I charge both calls, four wrappers, grid handling, source/query access and cleanup.I use the credited scalar-period identity to turn PR3109's changed-order lift into an exact dense scaled-linear backend. On a sufficient power-of-two input scale, standard character conjugation then supplies whole-field CNOTs and toggles with exact shifts and literal inverse cleanup. A second residue partition preserves the original consecutive nonlinear word. I bind the construction to PR2586 events 2025–2039 through Tekalign's PR3025/3033 paired interface.
I passed eight registered cold commands on Python 3.12.13/macOS arm64. Checks include 512 ideal native columns, 2,048 affine address cases, 2,610 interval plans, original-word payloads through 257 bits, and four complete N=512 collectors matching the predecessor's query, grid and private return. I included large dirty-source cancellation, a zero sum, synthetic tensor wrappers and guard/phase controls; algebra and consumer checks also pass under
-O.The finite executable uses direct semantic backends. It does not benchmark or emit the fast physical C trace; the time/space result is a written construction under the retained C-layer/compiler premises. The input contract uses explicit field-major serialization and charged clean history. I claim no linear-space host, width-one Boolean speedup, new multiplication exponent or full P104 supplier. The frozen five-forward/four-dual profile still has zero-saving bill at least599/584.
Reproduce:
python3.12 scripts/research_package.py --package research/dense-paired-collector.I credit Tekalign's PR3025/3033, Ross Richardson's PR3072, Douglas Colkitt's Fourier-conjugation analysis, ywz65's physical dual, the PR2250/2287/2899/3109 framework and Vasek/flyingviper's frozen-consumer audit. Tekalign's concurrent PR3122 reduces core history and explicitly leaves the wrapper/collector scope open; I retain a conservative collector-space bound and import none of that new code. SOURCES.json retains exact pins and source notices. Canonical coordination: 2750#18986640.
Jose Manuel (
jmbarrancoml), with substantial OpenAI Codex assistance, including analysis subagents. I claim no hosted CI, independent human or proof-assistant verification. Apache-2.0. I leave the selected frontier unchanged.