Build site: T-504, T-505, T-515, T-516, T-517 blocked by T-527 (their open criteria are T-527's work since the owner re-opened them under it)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

context/plans/build-site.md

@@ -567,8 +567,8 @@ the mid-pan dumps, never by a screenshot. No device run while the owner is at th
 |------|-------|---------|-------------|-----------|--------|--------|--------|
 | T-502 | The spike and the compositor-side scroll (rounds 1-137): patches 0040/0041/0042, the daemon's job pipeline, band reads on the band surface, the displayport, fixed/sticky rows, the composition fix for wide pages | cavekit-gpu-rendering.md | R2, R3 (part) | — | yes | L | DONE (commits through `bc4a8eaf`/`b30b9989`: test pan 27-34 fps, animation 26-31, HN 36-37, desktop Wikipedia 13-14 vs software 9-12.4) |
 | T-503 | Fixed and sticky at minimum cost: opaque stuck layers kept (per opaque rect, the visible region's gaps), others read as columns, fixed header + footer as two bands; the draw-culling fix kept; verify and dumps on the long-static page, a header+footer page and desktop Wikipedia | cavekit-gpu-rendering.md | R4, R2 | T-502 | yes | M | DONE 2026-10-18 (rounds 138-150): opaque fixed/stuck layers kept per covered rect, the rest read as column chunks packed on the band surface, header + footer as two bands; the draw culling fixed; a sticky layer the engine laid out stuck counted (probed sticky ranges); the root composition covers what the card shows (Hacker News' blank bottom strip); verify made side-effect free with a structural test: 0 structurally wrong rows on the test page, `jihad-sticky-mix.html`, desktop Wikipedia and Hacker News. Pans: test page 27-34 fps, sticky-mix 15-17.5 (software 0.9-3.1), Wikipedia 16.4-18.5 (was 13-14; software 9-12.4), Hacker News 34-37.5, animation 24-32 |
-| T-504 | Zoom keeps up: a scripted pinch and double-tap on Wikipedia and the test page with the frame timeline; first frame at the new zoom ≤ 250 ms; no grey at rest; tile uploads off the GPU's queue | cavekit-gpu-rendering.md | R6 | T-502 | yes | L | PARTIAL 2026-10-18 (rounds 162-186: the zoom hold; the grey strip of a zoomed-in settle fixed; engine 0044 (tile textures converted at 1x1 and pooled: the Adreno's first write after a draw) and 0045 (no DEAA for axis-aligned scales: shader compiles in a zoom's first frame); the layers painted at the zoom's resolution above 1.1; double taps scripted (`t`), `JIHAD_GPU_ZOOM_DUMP`. Met: the scripted pinches and double taps with the timeline, the card covered at rest (dumps), uploads that do not wait, sharp zoomed text. Open: the first frame within 250 ms while the page's script holds the main thread (0.8-1.7 s on Wikipedia after its load) and one double tap in ten at ~400 ms: frames made off the main thread, with T-505); 2026-10-19, T-515: the zoom under script met (Wikipedia after load 10/10 within 225 ms); the double taps' slow ones are GPU draws: T-516; 2026-10-19 (rounds 334-341, `docs/rounds/t502-2026-10-17/dtap-r334-341.txt`): scripted double taps 2 of 18 over 250 ms (286, 367), the second core unplugged after an idle pause; the scripted tap skips the touch path's governor ramp, so a finger's tap is for T-514; a zoom's express frame now gives up uncomposed main-lane frames at the old zoom. Task token budget spent (the plugin's 45k): left PARTIAL for the owner) |
-| T-505 | Real-page pans to the chart: desktop Wikipedia ≥ 24 fps; the 56 ms+ gaps traced (the main loop uncapped) and the avoidable ones removed; the per-page table against `NOGPU=1` | cavekit-gpu-rendering.md | R3 | T-503 | yes | L | PARTIAL 2026-10-18 (rounds 187-201: the per-page table against `NOGPU=1` (the GPU path ahead on every page: the test page 30-35 against 2.7-3.2 fps, sticky-mix 17-19 against 2.4-3.1, desktop Wikipedia 19-26 against 2.7-2.9, Hacker News 31-37 against 6.4-10.4; the software path pans inside its painted region and publishes few frames); the 56 ms+ gaps traced to the page's restyle, the displayport moves' paints and misc events (stall sampler); overscan, 0.15 of the height above and below the view (engine 0046): grey-free pans, verify clean on four pages, at 10% of the rate (Wikipedia 16.5-24, the test page 25-32). Open: Wikipedia ≥ 24 fps, which needs the frames made off the main thread (T-515)); 2026-10-19, T-515: the main thread's stalls covered (19-21 fps, median gap 45 ms); the rest is the compositor thread's ~45 ms a Wikipedia frame: T-517; 2026-10-19 evening (rounds 349-357): the 10 s pan 23.2-24.1 fps; the page's tile uploads, partial uploads, deferred uploads, a topped-up pool and the express lane on its own slot measured and not kept; the band frame is GPU-bound (r357). BLOCKED by the circuit breaker: the owner's finger round (T-514) and the bench's sensord load (the owner's call) decide) |
+| T-504 | Zoom keeps up: a scripted pinch and double-tap on Wikipedia and the test page with the frame timeline; first frame at the new zoom ≤ 250 ms; no grey at rest; tile uploads off the GPU's queue | cavekit-gpu-rendering.md | R6 | T-502 | yes | L | BLOCKED by T-527 (2026-10-25: the double tap out's first frame (R6) is T-527's work; the owner re-opened it under T-527) — PARTIAL 2026-10-18 (rounds 162-186: the zoom hold; the grey strip of a zoomed-in settle fixed; engine 0044 (tile textures converted at 1x1 and pooled: the Adreno's first write after a draw) and 0045 (no DEAA for axis-aligned scales: shader compiles in a zoom's first frame); the layers painted at the zoom's resolution above 1.1; double taps scripted (`t`), `JIHAD_GPU_ZOOM_DUMP`. Met: the scripted pinches and double taps with the timeline, the card covered at rest (dumps), uploads that do not wait, sharp zoomed text. Open: the first frame within 250 ms while the page's script holds the main thread (0.8-1.7 s on Wikipedia after its load) and one double tap in ten at ~400 ms: frames made off the main thread, with T-505); 2026-10-19, T-515: the zoom under script met (Wikipedia after load 10/10 within 225 ms); the double taps' slow ones are GPU draws: T-516; 2026-10-19 (rounds 334-341, `docs/rounds/t502-2026-10-17/dtap-r334-341.txt`): scripted double taps 2 of 18 over 250 ms (286, 367), the second core unplugged after an idle pause; the scripted tap skips the touch path's governor ramp, so a finger's tap is for T-514; a zoom's express frame now gives up uncomposed main-lane frames at the old zoom. Task token budget spent (the plugin's 45k): left PARTIAL for the owner) |
+| T-505 | Real-page pans to the chart: desktop Wikipedia ≥ 24 fps; the 56 ms+ gaps traced (the main loop uncapped) and the avoidable ones removed; the per-page table against `NOGPU=1` | cavekit-gpu-rendering.md | R3 | T-503 | yes | L | BLOCKED by T-527 (2026-10-25: desktop Wikipedia ≥ 24 fps with the sensors and the gaps' sources (R3) are T-527's work; the owner re-opened it under T-527) — PARTIAL 2026-10-18 (rounds 187-201: the per-page table against `NOGPU=1` (the GPU path ahead on every page: the test page 30-35 against 2.7-3.2 fps, sticky-mix 17-19 against 2.4-3.1, desktop Wikipedia 19-26 against 2.7-2.9, Hacker News 31-37 against 6.4-10.4; the software path pans inside its painted region and publishes few frames); the 56 ms+ gaps traced to the page's restyle, the displayport moves' paints and misc events (stall sampler); overscan, 0.15 of the height above and below the view (engine 0046): grey-free pans, verify clean on four pages, at 10% of the rate (Wikipedia 16.5-24, the test page 25-32). Open: Wikipedia ≥ 24 fps, which needs the frames made off the main thread (T-515)); 2026-10-19, T-515: the main thread's stalls covered (19-21 fps, median gap 45 ms); the rest is the compositor thread's ~45 ms a Wikipedia frame: T-517; 2026-10-19 evening (rounds 349-357): the 10 s pan 23.2-24.1 fps; the page's tile uploads, partial uploads, deferred uploads, a topped-up pool and the express lane on its own slot measured and not kept; the band frame is GPU-bound (r357). BLOCKED by the circuit breaker: the owner's finger round (T-514) and the bench's sensord load (the owner's call) decide) |
 | T-506 | Animation, new content and self-scroll measured against the chart (DOM edit and typing ≤ 106 ms) | cavekit-gpu-rendering.md | R5 | T-502 | yes | S | DONE 2026-10-18 (rounds 202-203: the animation page 26-31 fps reading only its changed rows; typing 22-62 ms and DOM edits 26-70 ms from the input to the first published frame with changed rows (an input-latency probe on the inject channel); the self-scrolling page followed at 10.7-13.9 fps) |
 | T-507 | Every other card event on the GPU path: rotation, resize, select popups, caret, link and find highlights, thumbnails, navigation without stale frames | cavekit-gpu-rendering.md | R9 | T-502 | yes | M | DONE 2026-10-18 (rounds 204-212 on `jihad-card-events.html`: rotation to landscape and back, the keyboard's resize, the select list and pick, the caret, find-in-page, link navigation with back, forward and reload, the thumbnail; the thumbnail's view row and the popups' origin corrected for overscan; the test hook gains `<`, `>`, `R`, `O<orientation>`; the inject channel `thumb <path>`) |
 | T-508 | Video and Flash on the GPU path: `<video>` (VP8, Theora, H.264) to the end, styled video (rounded, fade, under a modal), Flash draw/animate/input | cavekit-gpu-rendering.md | R7 | T-502 | yes | L | DONE 2026-10-19 (rounds 213-228: VP8, Theora, VP9 and H.264 to the end on the GPU path, at the clips' rates; styled video (rounded, fade, modal); Flash draws, animates and takes keys; `push-engine-update.sh` now carries FFmpeg's libavcodec/libavutil, which a device updated by it lacked (H.264, AAC and MP3 were refused on both paths); new pages `jihad-video-styled.html`, `jihad-anim-opacity.html`, `jihad-canplay.html`) |
@@ -578,9 +578,9 @@ the mid-pan dumps, never by a screenshot. No device run while the owner is at th
 | T-512 | Fit to keep: diagnostics pruned and bounded, the knobs listed, the engine work folded into the patch queue, the docs and README | cavekit-gpu-rendering.md | R11 | T-503…T-511, T-517 | no | M | DONE 2026-10-19 (`e8b0e425`: the `[jihad-gpu]` diagnostics behind one gate each in the engine (0057, `mozilla/gfx/JihadGpuLog.h`) and the daemon (`render/goanna/JihadGpuLog.h`), three startup lines in a shipped daemon, the default checks of all three interfaces pass; every switch in `tools/device/t502/README.md`; the 57 patches reproduce the pin and each has its ENGINE-SOURCE row; RENDERING-OPTIONS, the README and PICKUP carry the evening's numbers. The review (R11's third) is T-513) |
 | T-513 | Review of Tier 87 (Codex by direct invocation, the inspector as backup) and fixes | cavekit-gpu-rendering.md | R11 | T-512 | no | M | DONE 2026-10-19 (Codex out of credits: two inspectors, 12 findings, all fixed, `context/impl/impl-review-findings-gpu-2.md`; twenty card cycles flat (R10); a fast flick and two cards in turn pass the verify mode once its reference is composed behind the job (R2); the three default checks on the bench) |
 | T-514 | [OWNER] Finger round on the GPU path in all three interfaces: pan, flick, pinch, double-tap | cavekit-gpu-rendering.md | R12 | T-513 | yes | S | DONE 2026-10-24 by the emulated finger (T-524; R12 at [~]: "finds it smooth" stays the owner's word) |
-| T-515 | Frames made off the main thread: the card's scroll and zoom read on an IPC thread, the compositor composing and the frame published while the main thread runs the page's script or a paint (the open criteria of T-504, the first frame at a new zoom within 250 ms under script, and T-505, desktop Wikipedia ≥ 24 fps) | cavekit-gpu-rendering.md | R3, R6 | T-505 | yes | L | PARTIAL 2026-10-19 (rounds 240-267, the express lane: each card's YAP commands read on a thread of the proxy's own, the daemon's frame lock (`render/goanna/JihadFrameLock.h`; the main thread lets go of it in the pump and the layer paint), an express thread that composes the layer tree as it stands and publishes, engine 0048 (the paint and the compose post apart, jobs from any thread). Met: the zoom under script (Wikipedia after load 10/10 within 160-225 ms, held off median 681 and three never framed; a page holding the main thread 600 ms in 700 within 45-91 ms, held off median 450); pans on that page 26-32 fps (2.4-3.8); Wikipedia's 2.2 s restyle under frames every ~45 ms. Open, and no longer the main thread's: desktop Wikipedia 19-21 fps, the compositor thread's ~45 ms a frame (T-517); double taps 40-449 ms, GPU draws of the other resolution's tiles (T-516). `docs/rounds/t502-2026-10-17/express-r240-267.txt`) |
-| T-516 | The zoom's first frame on every gesture: a zoom out after a zoom in's repaint draws the new resolution's tiles at ~0.55, 20-35 ms a draw on the Adreno (a 179 ms composite, round 263); found and removed, card double taps both ways within 250 ms in runs of ten | cavekit-gpu-rendering.md | R6 | T-515 | yes | M | PARTIAL 2026-10-19 (rounds 270-280: the slow frames were not draws: the first readback after a pause ran 100-390 ms (30-40 usual) with the CPU swizzle 20-122 ms (7 usual): the clocks after two idle seconds (the CPU floor released, the second core unplugged, the GPU idled down) and the engine's other threads. Kept: the YAP reader raises the clock hold on a scroll, zoom or smart-zoom request (`JIHAD_CLOCK_HOLD_MS`), the compositor and express threads at nice -6 (`JIHAD_GPU_FRAME_NICE`), the clocks on the timeline. Double taps (12 a run): before 4-8 over 250 ms (max 479), after 0-2 (261-323; median 158). Tried and dropped: an 8 s hold (no change: the GPU still idles), a warm-up frame on the smart-zoom request (its own slow frame queued ahead of the zoom's: worse). Open: the GPU's ramp after a pause) |
-| T-517 | Desktop Wikipedia's pan frame under 41 ms: the compositor thread's ~45 ms a frame (3-7 ms of CPU a composite, the column passes, the readback's wait for the GPU) broken down and cut, whichever lane asks; ≥ 24 fps in the scripted pan; the same floor under a large animated area: the 480 px WebGL cube ≥ 24 fps on the card (T-511's open part; measured option: the adapter invalidating only the changed strip, +12 %, round 304) | cavekit-gpu-rendering.md | R3, R8 | T-515 | yes | L | PARTIAL 2026-10-19 (rounds 314-332, `docs/rounds/t502-2026-10-17/floor-r314-320.txt`. Kept: no cache clean of the frame rows on write-through segments; frame windows (a frame moves within its segment instead of 3.7 MB of rows being moved: `render/browserserver/Src/JihadFrameWindow.h`, the adapter's `BrowserOffscreen`, engine 0053), with both lanes keeping every slot's stale rows and the express lane building on its own slot when that measures faster. The long test page 30.5-36.4 fps (median gap 26 ms, was 29). Open: Wikipedia 19-21 fps: its two see-through sticky panels are composed again every frame in 3-4 chunks of the band surface, ~5 ms of CPU a pass; a rule keeping see-through sticky columns once proven still did not apply there (round 333: the contents panel changes as the page scrolls) and was reverted; next: one tree traversal for all of a frame's regions; rounds 342-348: engine 0055 (a band frame's regions in one composite) and 0056 (tiles outside the clip skipped before their lock): chunked frames compose 10.6 ms where they took 22.5; Wikipedia's 10 s pan 23.4-23.9 fps (was 20.0-21.7), typical frame 35-38 ms; the gaps left are the page's own work in the compositor's queue and on the other core; 2026-10-19 evening (rounds 349-357, `floor-r314-320.txt`): the page's layer transactions on the compositor thread timed (tile uploads), every cut after 0055/0056 measured and recorded in `dead-ends.md`; BLOCKED with T-505) |
+| T-515 | Frames made off the main thread: the card's scroll and zoom read on an IPC thread, the compositor composing and the frame published while the main thread runs the page's script or a paint (the open criteria of T-504, the first frame at a new zoom within 250 ms under script, and T-505, desktop Wikipedia ≥ 24 fps) | cavekit-gpu-rendering.md | R3, R6 | T-505 | yes | L | BLOCKED by T-527 (2026-10-25: what it left (the compositor floor, the double taps) is T-527's work; the owner re-opened it under T-527) — PARTIAL 2026-10-19 (rounds 240-267, the express lane: each card's YAP commands read on a thread of the proxy's own, the daemon's frame lock (`render/goanna/JihadFrameLock.h`; the main thread lets go of it in the pump and the layer paint), an express thread that composes the layer tree as it stands and publishes, engine 0048 (the paint and the compose post apart, jobs from any thread). Met: the zoom under script (Wikipedia after load 10/10 within 160-225 ms, held off median 681 and three never framed; a page holding the main thread 600 ms in 700 within 45-91 ms, held off median 450); pans on that page 26-32 fps (2.4-3.8); Wikipedia's 2.2 s restyle under frames every ~45 ms. Open, and no longer the main thread's: desktop Wikipedia 19-21 fps, the compositor thread's ~45 ms a frame (T-517); double taps 40-449 ms, GPU draws of the other resolution's tiles (T-516). `docs/rounds/t502-2026-10-17/express-r240-267.txt`) |
+| T-516 | The zoom's first frame on every gesture: a zoom out after a zoom in's repaint draws the new resolution's tiles at ~0.55, 20-35 ms a draw on the Adreno (a 179 ms composite, round 263); found and removed, card double taps both ways within 250 ms in runs of ten | cavekit-gpu-rendering.md | R6 | T-515 | yes | M | BLOCKED by T-527 (2026-10-25: the GPU's ramp after a pause is T-527's work; the owner re-opened it under T-527) — PARTIAL 2026-10-19 (rounds 270-280: the slow frames were not draws: the first readback after a pause ran 100-390 ms (30-40 usual) with the CPU swizzle 20-122 ms (7 usual): the clocks after two idle seconds (the CPU floor released, the second core unplugged, the GPU idled down) and the engine's other threads. Kept: the YAP reader raises the clock hold on a scroll, zoom or smart-zoom request (`JIHAD_CLOCK_HOLD_MS`), the compositor and express threads at nice -6 (`JIHAD_GPU_FRAME_NICE`), the clocks on the timeline. Double taps (12 a run): before 4-8 over 250 ms (max 479), after 0-2 (261-323; median 158). Tried and dropped: an 8 s hold (no change: the GPU still idles), a warm-up frame on the smart-zoom request (its own slow frame queued ahead of the zoom's: worse). Open: the GPU's ramp after a pause) |
+| T-517 | Desktop Wikipedia's pan frame under 41 ms: the compositor thread's ~45 ms a frame (3-7 ms of CPU a composite, the column passes, the readback's wait for the GPU) broken down and cut, whichever lane asks; ≥ 24 fps in the scripted pan; the same floor under a large animated area: the 480 px WebGL cube ≥ 24 fps on the card (T-511's open part; measured option: the adapter invalidating only the changed strip, +12 %, round 304) | cavekit-gpu-rendering.md | R3, R8 | T-515 | yes | L | BLOCKED by T-527 (2026-10-25: the compositor job's floor (R3) is T-527's work; the owner re-opened it under T-527) — PARTIAL 2026-10-19 (rounds 314-332, `docs/rounds/t502-2026-10-17/floor-r314-320.txt`. Kept: no cache clean of the frame rows on write-through segments; frame windows (a frame moves within its segment instead of 3.7 MB of rows being moved: `render/browserserver/Src/JihadFrameWindow.h`, the adapter's `BrowserOffscreen`, engine 0053), with both lanes keeping every slot's stale rows and the express lane building on its own slot when that measures faster. The long test page 30.5-36.4 fps (median gap 26 ms, was 29). Open: Wikipedia 19-21 fps: its two see-through sticky panels are composed again every frame in 3-4 chunks of the band surface, ~5 ms of CPU a pass; a rule keeping see-through sticky columns once proven still did not apply there (round 333: the contents panel changes as the page scrolls) and was reverted; next: one tree traversal for all of a frame's regions; rounds 342-348: engine 0055 (a band frame's regions in one composite) and 0056 (tiles outside the clip skipped before their lock): chunked frames compose 10.6 ms where they took 22.5; Wikipedia's 10 s pan 23.4-23.9 fps (was 20.0-21.7), typical frame 35-38 ms; the gaps left are the page's own work in the compositor's queue and on the other core; 2026-10-19 evening (rounds 349-357, `floor-r314-320.txt`): the page's layer transactions on the compositor thread timed (tile uploads), every cut after 0055/0056 measured and recorded in `dead-ends.md`; BLOCKED with T-505) |
 
 ## Tier 86 — Content blocking: ads and trackers (2026-10-15; added to the site, not a regeneration)