Repository navigation
Native player main thread 3-7 ms faster on Midway: held batch decision + render thread on by default (PRD-613) - #577
Merged
Merged
Conversation
…PRD-743 surface) PRD-743 added IAssetLoader.cache, prefetch and startBackgroundFill, and the playtest bridge's state sample reads cache.enabled. The native player's loader had none of them, so every native playtest observation threw TN_INSPECT_GAME_BRIDGE (reading 'enabled'), and a game passing assets.cache or assets.backgroundFill threw TN_NATIVE_ASSET_OPTIONS_UNSUPPORTED. The native loader now reports the cache as disabled with core's native reason, and prefetch and the fill resolve at once. Red-green: native_engine_player_imports red on the options refusal, green after.
… structure does RenderDatabase::batch judged a projecting projection decision again every frame: a full scene walk, content classes and the group maps, only to read the rigs' verdicts. On Midway that was about 11% of main-thread samples at cockpit and rear gunner. The decision now holds while hierarchyVersion and structureStamp (moved from the bindings into projection/plan.h, shared with hostProjectionReport) hold, and while every rig's lane key (lane reason, similarity, mirror sign) holds; the 60-frame cadence still re-judges it. TN_BATCH_DECISION_CACHE=0 (diagnostic) restores the old cadence for paired arms. ctest: skinned_projecting_cadence red then green; skinned_crowd_cpu caught the rig-scale and material cases; native-engine label 516/517 (target_graph is the UI-overlay configure).
Phase 3 passed on Midway at 99ad82d: main-thread CPU per frame falls by 3.4-7.5 ms per pose in three alternating pairs (render thread plus the held batch decision against today's develop path), parity PASS in every run, no steady-state fallback, visual judge NEUTRAL at all three poses. TN_RENDER_THREAD=0 turns it off; a GPU timing run keeps it off, since it has no synchronized timing snapshot (TN_RENDER_THREAD=1 with GPU timing still fails closed). Red: a Midway flight with the variable unset on the binary before this change reports no renderThreadCpuMs.
Contributor
Author
Pixels unchanged: visual judge NEUTRAL (PRD-613 Phase 3)Panels, left to right: BEFORE A1 and BEFORE A2 ( Takeoff Cockpit Rear gunner Verdict (fresh read-only Opus subagent, criteria set before judging): NEUTRAL at all three poses. No missing objects and no change to pose, material, shadow or flicker. Pixel counts over 1% grey, across 7 off-runs and 10 on-runs:
Follow-up outside this PRD: some state at those two spots is sampled off the fixed-step clock. |
dflt-green (render thread on with no flag, after the switch) reports renderThreadCpuMs in 4,574 frame lines, fallbacks 1, parity PASS; dflt-red on 99ad82d reports none. The PRD moves to docs/PRDs/done/. PRD-610 AC-16 stays open for the owner: cull and project stay on the main thread.
jonit-dev
added
prd:100% — ready
jonit-dev
changed the title
jonit-dev
marked this pull request as ready for review
October 27, 2026 11:54
…e and player changes
Contributor
|
The merge queue stopped on
Fix, push, re-enqueue. |
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Oct 27, 2026
jonit-dev
enabled auto-merge
October 27, 2026 12:56
Contributor
|
The merge queue stopped on
Fix, push, re-enqueue. |
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Oct 27, 2026
Contributor
|
The merge queue stopped on
Fix, push, re-enqueue. |
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
Oct 27, 2026
jonit-dev
added a commit
that referenced
this pull request
Oct 27, 2026
…its generated records Resolves the four conflicts develop's #577/#579 left against this branch, and recomputes the two generated records on the merged tree (neither parent's values are right for it). - .gitleaksignore: keep develop's allowlist entry for #574's kit-loop KPI field id (the branch had nothing there). - core-assets.mjs: union. This lane's idempotent `dispose()` (final release must not empty the reusable cache) plus develop's PRD-743 report surface (`cache`, `prefetch`, `startBackgroundFill`) that the playtest bridge's state sample reads. Both sides' new player-imports checks pass against it. - scaffold.spec.ts: recompute PRD_201_PARENT_SCAFFOLD_HASHES on the merged tree. The kits this lane touched (collectathon, adventure, kart, starter) are a third state, not either parent's; the untouched kits take develop's values. - native-coverage-2026-08-28.md: regenerated by `native:coverage` on the merged tree (48 contract targets executed, 3 blocked as configured). Source digest 22ac6e9e…; src/engine/ 83.90% against its 82.22% floor. Floor tables were identical on both sides. Verified locally on the merged tree: typecheck, lint, `budgets`, `check:docs`; unit phase 13,555 passed (17 failures are machine-only: 14 need Node 20 and pass under v20.19.6, 1 needed the Wasm engine and 2 the browser, both green once built); scaffold hash spec green; native C++ board green; Wasm engine 4,192,083 B against the 4 MiB budget (2,221 B under, this lane's 2,178 under held). Not re-run locally: the 21 runtime-native package tests that need the `build/tn-linux/*-test` targets, which `native:build` does not build — CI's test-native job owns them.
This was referenced Oct 27, 2026
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.
What is this PR about?
The native desktop player's main (game) thread does less work per frame. Two changes do this:
RenderDatabase::batchre-ranprojection::decideon every frame, with a full scene walk and regroup, only to read the skinned rigs' verdicts. The decision now holds while the scene structure (hierarchyVersion+structureStamp) and every rig's lane key (lane reason, similarity, mirror sign) hold. The 60-frame cadence still re-judges it.structureStampmoved intoprojection/plan.h, sohostProjectionReportand the renderer share one rule.TN_RENDER_THREAD=0turns it off, and a GPU timing run keeps it off.Also in this PR:
cache/prefetch/startBackgroundFill, so native playtest observations threwTN_INSPECT_GAME_BRIDGE … reading 'enabled'.developsince Hook reference search to the new data repository and its release shape #579.Why does it matter?
Midway (desktop, matched CPU, 3 alternating pairs against today's
developpath) runs 3.4 to 7.5 ms less main-thread CPU per frame at takeoff, cockpit and rear gunner. Instructions per frame fall from 40-48 M to 29-34 M (about −27%). Parity PASS in every run. The visual judge returns NEUTRAL at all three poses (sheets in the PR comment).Done (PRD-613,
prd:100%)native_engine_skinned_projecting_cadence: red then green.ctest -L native-engine516/517: the only red istarget_graph, from this checkout's UI-overlay configure.Still open, outside this PR