Repository navigation
Tooling: 44× faster production verify_factor via packed-row equality with identical receipts - #2194
Merged
Mister-Meeseeks merged 1 commit intoOct 15, 2026
Conversation
Wire Ross Richardson's injective packed-row central-factor check into the shipped scripts/experiments/complex_centers.py path used by audit_complex_centers / verify-certificates. Factor receipts stay fieldwise identical; certificates/dyadic-central-factor.json changes only that source pin. Sibling pin-restore tests accept the new complex_centers hash. Package research/central-factor-production/ with baseline, verify, benchmark and RESULT.json (Linux/CPython 3.12.3 medians: factor 44.06×, audit 1.72×). Prepared by Wentao Li with substantial AI assistance.
This was referenced Oct 15, 2026
Merged
kumagi
added a commit
to kumagi/integer-mult-bounds
that referenced
this pull request
Oct 20, 2026
…tree The packed central-factor audit (batch batch952) refuses to start on the current tree: admit_sources() rejects at the first stale pin (reproduced independently; also reported in CrocSwap#1166). Five pinned files legitimately changed in main after the recorded base 47481c8 ("Record publication and merge confirmation for all eighteen PRs"), each via an identical-receipt tooling integration: - certificates/dyadic-central-factor.json (c22e42d, PR CrocSwap#2194 integration regenerated the certificate with current source hashes; the central_factor bytes are unchanged) - scripts/exclusion_circuit.py (d7d8bfc, d8029ee) - scripts/experiments/complex_centers.py (f0c58a2) - scripts/experiments/disjoint_tensor.py (3e917c5) - scripts/shared_point_circuit.py (ff889cc) verify.py compares its regenerated complete receipt byte-for-byte against the current ROOT/certificates/dyadic-central-factor.json, so the full audit could never pass with the stale set regardless of the workflow trigger. After the repin the complete historical audit passes on the unmodified current tree: PASS_CENTRAL_FACTOR_PACKED, receipt_sha256 2aabc7e3..., with adapter pins (packed.py, verify.py) untouched, and test_packed.py passes in both -B and -O -B modes.
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.
Change and scope
At base
0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2,scripts/experiments/complex_centers.py:verify_factorstill used the historical nested entrywise loop over all scatter/gather pairs. Ross Richardson’s opt-in adapterresearch/central-factor-packed/already proved an injective packed-row algorithm with identical receipts. This PR wires that equivalent check into the production source used byscripts/audit_complex_centers.py/make verify-certificates.Only the source pin for
scripts/experiments/complex_centers.pychanges insidecertificates/dyadic-central-factor.json. Every coefficient is still checked; factor generators, side/phase obligations and the retained historical2^-31language are unchanged.Tooling improvement; no new kappa. Selected frontier untouched.
Result
On Linux / CPython 3.12.3 (Linux-6.12.94+-x86_64-with-glibc2.39), five alternating fresh-process pairs gave median wall time:
Timing is host-descriptive and is not a correctness gate. Every
verify_factorreceipt matches the baseline checker fieldwise.Package:
research/central-factor-production/.Validation
Expected:
Default verify is read-only. Coefficient-mismatch and temp-copy corruption controls reject;
requiregates remain under-O. Fullmake verify, selected cold replay, Lean, and multiplication-theorem obligations were not rerun for this isolated historical-checker change.Overlap and credit
Closest priors (different delivery): Ross Richardson’s
research/central-factor-packed/(opt-in packed adapter; this is the production wiring), CrocSwap/integer-mult-bounds#756, CrocSwap/integer-mult-bounds#2117 (samedyadic-central-factor.json, different pin keydisjoint_tensor.py). No competing open PR on productionverify_factorat refresh.Douglas R. Colkitt authored the dyadic central factor and original checker. Ross Richardson authored the packed algorithm. Prepared by Wentao Li (VictorLiwentao, @Is__Victor) with substantial AI assistance, under Apache-2.0.