Skip to content

feat(WordRAM): observe uniform native method answers - #2267

Open
SamuelSchlesinger wants to merge 1 commit into
codex/ram-memory-queue-adapter-c2-20261101from
work/c4-observation-1730-20261102
Open

SamuelSchlesinger wants to merge 1 commit into
codex/ram-memory-queue-adapter-c2-20261101from
work/c4-observation-1730-20261102

Conversation

@SamuelSchlesinger

@SamuelSchlesinger SamuelSchlesinger commented Nov 2, 2026 •

Copy link
Copy Markdown
Collaborator

On hold — coordinator consolidation. Under the #2205 decision, this PR will not be extended or queued. The single framed Backend implementation and consolidated queue migration will absorb the applicable work with source credit. Existing branches, tests and review findings are retained until that absorption is validated; this is not a completed migration.

The framed realization adapter now accepts a uniform result carrier and an operation-indexed answer, so argument-bearing operations can return Unit without echoing their arguments. One shared native proof retains the exact source-program equation, full-memory semantics, every-memory-state decoder obligation, uniform code selection, framing and payment. The original whole-label adapter signatures remain specializations. This extends the existing Realizes/Module/SubroutineBase path; it introduces no additional compiler, judgment or memory model.

The actual two-operation queue now returns only the existing two-word Option encoding. Enqueue returns none observed as Unit; dequeue returns its actual optional word. Native staging/control remain charged: local work is 17 for enqueue and dispatcher work + 2 for dequeue, with G ≤ 7*M+71, amortized body charge 144*factor and potential 48*factor*rearCount. The selected factor changes, so this is no empirical speedup claim. Argument delivery remains outside the body clock, halts are separate, and local linear dirty clearing remains distinct from faithful source allocation. Broader #2205/#2213 and their M1 gate remain open.

Validation: all six default scripts/check.sh gates pass at 407b09095ad10203a22399ccd2924f790fb5881e (2 threads), with exact tested-source guards. Focused strict checks (3,602 jobs) and hygiene pass, including same-code shifted/maxword enqueue, immediate/transfer/empty dequeue, initialized FIFO, capacity-zero and dirty-clear cases. New collision tests distinguish faithful common-result encoding from noninjective answer observation. Selected abstracts/catalogue/rendering have no impact. Stacked on #2252; approved parent heads are preserved.

Fixes #2256. Original Samuel Schlesinger and C2 supplier credit is retained. AI-assisted implementation/self-review: OpenAI Codex codex-c4-observation-1730-20261102; fresh outside-authoring-fleet API/correctness review is required.

Generalize the existing framed adapter to a uniform result carrier with
operation-indexed observations and exact source equations. Keep the old
whole-label adapters as specializations of one shared native proof.

Migrate the actual queue to faithful optional-word answers without an
enqueue payload echo. Preserve full memory, framing, initialization and
uniformity; charge local work explicitly. Recompute the work envelope
7*M+71 and amortized body bound144*factor, with preparation/halts and
literal allocation retaining their distinct contracts. No speedup claim
is made across separately selected realization factors.

All six scripts/check.sh gates pass on source tree af52de6
(session19537,2threads). Focused3602/native/observation tests, hygiene,
old public signatures and all unrelated source entries are verified.
Selected abstracts/catalogue/rendering: no impact.

Refs #2256, #2205, #2213. Existing source and supplier credit retained.
AI-assisted authoring/self-review: OpenAI Codex
(codex-c4-observation-1730-20261102). Outside-authoring-fleet review
required; no author approval.

Copy link
Copy Markdown
Collaborator Author

Authoring worker codex-c4-observation-1730-20261102 requests one fresh outside-authoring-fleet review of exact 407b09095 for C4. Please cover the new uniform-result/answer boundary and actual-state decoder, the whole-label specialization, and the genuine queue consumer's output/staging/work changes. Reuse unchanged parent execution/init/frame/source coverage. The tested source tree is af52de6a; all six local gates and collision/maxword/empty/FIFO regressions pass. The body-only clock and changed-factor limit are intentional. Our Codex authoring fleet cannot approve this public API/consumer change.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant