Repository navigation
perf: key macOS StopBurst RSS baseline by hardware class - #374
Merged
Merged
Conversation
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Hosted macOS StopBurst has two evidenced RSS modes: M1 retains 93.58 MB, while the qualified M2 Pro cohort median is 103.19 MB. Select the M2 Pro override only for
hosted-macos+transcript:StopBurst, using the CPU brand string already logged by the measurement job. Unknown CPUs and missing overrides use the existing strict 93.58 MB baseline. The group receipt includes CPU model, detected class, baseline source and applied numeric baseline.Refs #365. AC3 stays open for ten hosted macOS main passes after merge.
The +10% regression formula, strict 150 MB per-invocation ceiling, wall maximum, telemetry checks, retry/advisory policy and every other group/runner baseline are unchanged. No product source, runtime dependency, workflow or fixture changes.
Evidence and derivation:
All three raw fingerprints are Apple M2 Pro (Virtual), 5 CPU, 14336 MB, image macos26 20261102.0473.1. Every outcome is retained, including all three RSS misses. Each p95 is independently reconstructed as rank 38 of 40 logical-run peaks across all follow-ups, matching the reported value. Median 103.1875 MB rounds once to 103.19 at the existing two-decimal precision; no pooling, trimming, discarded samples or outcome filter. The supplied fresh retained summary contained five M1 attempts and no additional M2 run.
The exact 20-file direct hook/worker/spool/harness/fixture blob matrix matches comparison main f606b0b for all three checkouts. This is the same direct set used in task 215, including bootstrap, native/RSS/Record/metrics support and timing helpers. It is not a claim of identical dependency closures: runtime dependency
src/memory/capture.tsat b0c64d4 predates the exact-0600 no-op chmod guard from the console-cache change; the owner explicitly accepted that difference for this cohort. Its only delta is unconditional chmod versus stat/mode-guarded chmod; the other two capture.ts blobs match main. The older M2 jobs 129497662298 and 129908133298 remain excluded for direct-source mismatch, regardless of their outcome.Provenance caveat retained: the first two jobs each contain one incomplete/unknown RSS envelope under the existing abandonment advisory (logical run 9/follow-up 5 and run 11/follow-up 2). Neither observation is filtered or described as a complete worker peak. Their rank-38 p95 values come from known cold-call peaks. This PR adds no provenance exception.
Raw log SHA256, in table order: 6ba8284468726f100428ec84521276d1c708a21f124c4aeaebd842786b23ccc0; 69b62dce3bfcd8dfba6e5f324db3d9eda203a1d82fe04bec3ed1cccd014bd446; f8d85bd5becfcf4581fb948f14e8741bb7e039f2148ede5be76d4799dcda67ca. Local audit artifacts: scratchpad/temp/task226/{qualification.md,source-blobs.json,cohort-audit.json,fresh-retained-log-audit.json,capture-dependency.diff}.
Validation on final head e3de743, pinned Bun 1.4.2 in a remote Linux runner: typecheck passes; boundary passes (172); README facts pass; focused budget/timing tests pass 49/49 (533 assertions); full suite passes 3,679 with 26 skipped and 0 failed (50,485 assertions, 3,705 tests across 337 files, 1,346.57 seconds); pack dry-run passes (182 files). Verification exits 0. Full raw log SHA256: 66cc92f2a2d5b630e6cece29aae40aa9e96330f26ca21680b44878a0623305d5, at scratchpad/temp/task226/full-e3de743.log. Local focused budget-attempt tests also pass 39/39 (272 assertions), and diff check passes. Independent read-only source-cohort and implementation reviews found no blocking issue. Docs check reports the expected stale README/contracts stamps after covered source changes; this change does not advance release stamps. Hosted checks on both OSes and hosted hardware-class receipts are unverified/pending; no future main-run AC3 pass is claimed.