Skip to content

Draft: Phase 1 model for native-route parallel realization - #11385

Merged
gunbai-bot[bot] merged 2 commits into
mainfrom
session/crisp-hawk-831-plan
Sep 15, 2026
Merged

gunbai-bot[bot] merged 2 commits into
mainfrom
session/crisp-hawk-831-plan

Conversation

@gunbai-bot

@gunbai-bot gunbai-bot Bot commented Sep 14, 2026

Copy link
Copy Markdown
Contributor

Summary

  • Durable Phase 1 model for native-route prepare sharding (slice 1 construct, slice 2 modeled). Implementation is not in this PR and is not to be landed under operator wind-down.
  • Homes cited, not forked: Placement.LocalChildProcess, width envelope process_memory_aware_spawn_width, width choice select_realization, WorkOperation not extended, instrument scratch/deep-cat-655-span-bisect@5b41958d07d9.
  • Four binding conditions from eager-raven-113: three-arm acceptance; compose on The exclusive row set becomes a keyed collection over a closed key #11374 keyed exclusive rows; worker artifacts under RUNNER_TEMP with digest-joined refuse-closed merge; contiguous-slice assignment.

Not credited

  • 3x, five minutes, the serial projection, semantic equivalence of fierce-lark's merged shard surface.

Test plan

Made with Cursor

Operator wind-down: durable plan with the four GO conditions and the not-credited list; no implementation.

Co-authored-by: Cursor <cursoragent@cursor.com>
@briansrls
briansrls marked this pull request as ready for review September 14, 2026 16:34
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Sep 14, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-09-14T16:38:15.463727Z 5f844aa Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5f844aa505

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Homes accepted as written, not forked:

- `std.realization` — `Placement.LocalChildProcess` (fifth arm). Serial `SourceRootEvalDriver` remains `LocalInProcess`. A prepare worker is a local OS process with its own address space, not a filesystem cache, not a remote host, not an accelerator. Do not nickname Placement as `WorkerKind`. Do not put the arm in `gunbc.compute`. Worker `RealizedStep` = `ExecuteEffect` + `LocalChildProcess` + `Recompute`. Parent merge stays `LocalInProcess` plus `ReadEffect` over worker artifacts (condition 3). Slice 1 does not take a Share cache obligation across siblings.
- `std.realization_width` — envelope is `process_memory_aware_spawn_width` (cores ∩ memory ∩ pids). This is the bound fold, not a second knob. Do not mint `NATIVE_SHARD_COUNT` / `--width`. `NATIVE_SHARD_INDEX` is an observation label only.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Define refusal semantics for unavailable width inputs

When memory or PID observations are unavailable—or the safe memory budget cannot fit one worker—the mandated process_memory_aware_spawn_width is not the claimed cores ∩ memory ∩ pids envelope: dag/std/realization_width.dag returns conservative_fallback_width directly for zero memory/PID inputs and clamps subunit memory capacity to one. An implementation following this plan could therefore return width 4 on a one-core host or admit W=1 when no worker fits, bypassing the later “above envelope refuses” requirement. Specify a feasibility refusal for these states, or name a stricter envelope operation before making this helper binding.

Useful? React with 👍 / 👎.

…behind a fail-closed input contract

Landing side chat ruling (2026-09-15) on #11385: the helper's conservative-fallback and minimum-one arms can answer outside the observed envelope, so the plan states the precondition and the named refusal rather than preserving a known-false binding. Text-only repair; no implementation is owed in this plan-only PR.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DQyBe1FaYb3bFZmpTczetX
@gunbai-bot

gunbai-bot Bot commented Sep 15, 2026

Copy link
Copy Markdown
Contributor Author

Landing side chat verdict (2026-09-15): HOLD for one bounded text repair — the plan bound the width envelope to process_memory_aware_spawn_width as if it were cores ∩ memory ∩ pids with refusal above the envelope, but that helper returns conservative_fallback_width on zero memory/PID observations and clamps to one worker when the budget fits none. Pushed 7e31379c43: the plan now states the fail-closed precondition (nonzero, readable CPU/memory/PID inputs and capacity for ≥1 worker, else a named NoFeasibleRealization), names the helper's fallback and minimum-one arms as inadmissible for this realization, and allows a stricter existing operation to be bound instead. Text only; the side chat authorized landing the successor on a diff read after fresh head/CLEAN/CI/tally verification.

@gunbai-bot
gunbai-bot Bot added this pull request to the merge queue Sep 15, 2026
Merged via the queue into main with commit 7daefb3 Sep 15, 2026
4 checks passed
@gunbai-bot
gunbai-bot Bot deleted the session/crisp-hawk-831-plan branch September 15, 2026 19:08
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.

0 participants