mirror of
https://github.com/MobileGL-Dev/MobileGL
synced 2026-11-19 16:01:57 +09:00
[Docs] (.claude/skills, .github/workflows, CMakeLists.txt, docs/archive, docs/Disaggregated, docs/MobileVA, docs/Performance, MG_Backend, MG_Backend/DirectGLES, MG_Backend/DirectVulkan, MG_Backend/Record, MG_Gbm, MG_Impl, MG_Impl/EGLImpl, MG_Impl/GLImpl, MG_Impl/GLXImpl, MG_Impl/Pipe, MG_IntegrationTest, MG_IntegrationTest/Harness, MG_IntegrationTest/Scenarios, MG_Pipe, MG_Pipe/generated, MG_Remote, MG_Remote/Client, MG_Remote/Server, MG_Remote/Transport, MG_Remote/Wire, MG_State/GLState, MG_Test/Pipe, MG_Test/Util, MG_Test/Wire, MobileGL, scripts, scripts/data, tools/device_bench, userguide): archive the planning notes and agent contracts under docs/archive
- DEBTS.md and OPEN-QUESTIONS.md stay maintained, now in docs/Disaggregated.
This commit is contained in:
@@ -0,0 +1,88 @@
|
||||
# BRIEF-P5B — verb migration under inproc, Minecraft-first
|
||||
|
||||
**Goal, one sentence:** fully separate-thread rendering of REAL workloads as fast as possible —
|
||||
every Minecraft trace, and the four A/B trace cases, rendered by the apply thread under
|
||||
`MOBILEGL_TRANSPORT=inproc` — by migrating the class-C slots the census measured, in the order
|
||||
it measured them, with the inproc lane's abort count and the traces' first-blocker list as the
|
||||
only gates (ID-66, ID-68).
|
||||
|
||||
Read `PACKAGE-PREAMBLE.md` first, then `MobileGL/MG_Remote/CONTRACT-P5B.md` (the wire contract,
|
||||
every row of it), then this file.
|
||||
|
||||
## 1. What c0b landed (head `p5b/c0b`, see `p5b-results/c0b-v1.md`)
|
||||
|
||||
Every package compiles and links on day one against real headers:
|
||||
|
||||
- the wire rows: `draw_vbo` grew `kDrawIsIndirect` + the `MGPDrawIndirect` second tail; five
|
||||
rows appended at opcodes 72..76 (`bind_shader_image`, `patch_parameter`, `bind_stream_output`,
|
||||
`set_storage_block_binding`, `copy_framebuffer_to_texture`); `MGPCopyRegion` grew its two GL
|
||||
names; the clear discriminants have one spelling in `MGPipeTypes.h`;
|
||||
- the codec: encode/decode arms for every row, tail cross-checks, the name blob;
|
||||
- `WireVerbSink` has one method per row (13 new + the widened `OnDrawVbo`), and
|
||||
`ServerVerbSink` overrides every one with a `Fatal{UnmigratedVerb, "<GL slot>"}` stub;
|
||||
- the client emit table's class-C list is PARTITIONED per package
|
||||
(`MGR_UNMIGRATED_{D1,I1,T2,F1,TAIL}_SLOTS`, `kEmittedSlots{D1,I1,T2,F1}`, ownership
|
||||
`static_assert`s 19 / 7 / 7 / 11 / 20) so four packages edit four disjoint regions;
|
||||
- the stamp rows (`FieldOwnership.def`), the generated tables, the pinned counts, and a
|
||||
round-trip unit case per row in `PipeWireCodecTest.cpp`.
|
||||
|
||||
## 2. The four packages
|
||||
|
||||
Each package owns its slots, its list and count in `EmitTables.cpp`, its stub bodies in
|
||||
`PipeApplier.cpp`, and its cases in `PipeWireCodecTest.cpp` + the integration lane. CONTRACT-P5B
|
||||
§2 gives every slot's record fields, sink dispatch and refusal names; §8 the two pre-granted
|
||||
edits to c0b's files.
|
||||
|
||||
| package | measured slots (census entries) | companions (same rows, owned too) | rows | model |
|
||||
|---|---|---|---|---|
|
||||
| **d1** — indexed / instanced / multi-draw / indirect | `DrawElements` 56 (+ every Minecraft trace), `MultiDrawElementsBaseVertex` 18, `DrawArraysInstancedBaseInstance` 4, `DrawElementsBaseVertex` 4, `MultiDrawArrays` 2, `MultiDrawElements` 2, `DrawArraysInstanced` 2, `MultiDrawElementsIndirect` 2, `MultiDrawArraysIndirectCount` 2, `DrawElementsInstancedBaseVertex` (trace) | `DrawElementsInstanced`, `DrawElementsInstancedBaseInstance`, `DrawElementsInstancedBaseVertexBaseInstance`, `DrawRangeElements`, `DrawRangeElementsBaseVertex`, `MultiDrawArraysIndirect`, `MultiDrawElementsIndirectCount`, `DrawArraysIndirect`, `DrawElementsIndirect` | `draw_vbo` (59) | Claude, premium |
|
||||
| **i1** — image / compute / barrier / copy-image / storage block | `BindImageTexture` 138, `DispatchCompute` 49, `CopyImageSubData` 30, `MemoryBarrier` 18, `ShaderStorageBlockBinding` 4 | `DispatchComputeIndirect`, `MemoryBarrierByRegion` | `bind_shader_image` (72), `launch_grid` (60), `memory_barrier` (61), `resource_copy_region` (53), `set_storage_block_binding` (75) | Claude |
|
||||
| **t2** — XFB spans, XFB object bind, patch parameter | `BeginTransformFeedback` 95, `PatchParameteri` 43, `BindTransformFeedback` 2 | `EndTransformFeedback`, `PauseTransformFeedback`, `ResumeTransformFeedback`, (`DeleteTransformFeedback`: no row, see contract) | `begin/end/pause/resume_stream_output` (62..65), `bind_stream_output` (74), `patch_parameter` (73) | Claude |
|
||||
| **f1** — clears, framebuffer-sourced copies, mips | `ClearBufferiv` 12, `ClearBufferuiv` 8, `ClearBufferfv` 6, `ClearBufferfi` 1, `ClearNamedFramebufferfv` 2, `CopyTexImage2D` 3, `GenerateMipmap` 2 | `ClearNamedFramebufferfi/iv/uiv`, `CopyTexSubImage2D` | `clear` (57), `copy_framebuffer_to_texture` (76), `generate_mipmap` (54) | codex astra — additive and mechanical |
|
||||
|
||||
Per slot the work is the same three edits: the emitter (class C → B in `EmitTables.cpp`, the
|
||||
pre-verb hook order kept), the `ServerVerbSink` body (the backend call the contract names), and
|
||||
a lane case that proves the record crossed. Nothing in the backends, unless the contract names
|
||||
the site (i1's mirror skip, f1's optional mip-storage re-derivation) - and then only behind
|
||||
`#if MOBILEGL_BUILD_DISAGGREGATED`.
|
||||
|
||||
## 3. The per-package gate
|
||||
|
||||
- **the inproc lane's abort count for your slots → 0**, and each affected scenario reaches its
|
||||
NEXT first blocker, reported by name (`Fatal{UnmigratedVerb, "…"}`, `Fatal{UnmigratedEmulation,
|
||||
"…"}`, or a wrong-answer failure - whichever comes first). Run
|
||||
`~/w7/p5b-c0b-census.sh <your tree> <outdir>` and diff `results.json` against c0b's baseline
|
||||
(`~/w7/p5b-c0b-census-logs/results.json`: 432 passed / 185 skipped / 505 aborted / 27 failed,
|
||||
all 505 aborts `UnmigratedVerb`, the per-slot table identical to `census-classC.md`'s).
|
||||
- **d1 additionally: every Minecraft trace reaches its next first blocker under inproc**,
|
||||
reported by name per fixture (the census's trace table), via the same per-entry private-log
|
||||
method as `p6_census_traces.{sh,py}`; a trace that RENDERS is reported with its SSIM.
|
||||
- the integrator's quick gate: unit lanes green (c0b's baseline + yours), `integration-split`
|
||||
inproc 22/22, push `integration-gpu` 1128/1128 name-for-name, G1 0/0/0/0 `.text +0`.
|
||||
- no reviews between rounds; one codex review at P5b's close.
|
||||
|
||||
## 4. File ownership
|
||||
|
||||
| file | owner in P5b | others |
|
||||
|---|---|---|
|
||||
| `MG_Pipe/MGPipeTypes.h`, `PipeCalls.def`, `PipeFields.def`, `FieldOwnership.def`, `Coverage.def`, `FillPoints.def`, `scripts/gen_pipe*.py`, `MG_Remote/CONTRACT-P5B.md` | c0b / integrator | t2: `kCapBackendOwnsXfbCapture` (contract §6.5); i1: the `GetProgramForDispatch` row (§6.9). Anything else: ask. |
|
||||
| `MG_Remote/Wire/PipeWireCodec.{h,cpp}` | w1's file; c0b landed the arms | a package edits the layout arm of ITS row only, with the integrator |
|
||||
| `MG_Remote/Server/PipeApplier.{h,cpp}` | v1's file | each package replaces ITS stub bodies (the `// ---- d1/i1/t2/f1 ----` blocks) and may add ITS tallies |
|
||||
| `MG_Remote/Client/EmitTables.{h,cpp}` | c1's file | each package edits ITS `MGR_UNMIGRATED_*_SLOTS`, ITS `kEmittedSlots*`, ITS emitters |
|
||||
| `MG_Remote/Client/SlotCaps.h` | c1's file | t2: the `EndTransformFeedback` probe → `kCapBackendOwnsXfbCapture` |
|
||||
| `MG_Test/Wire/PipeWireCodecTest.cpp`, `MG_Test/Wire/RemoteClientTest.cpp` | shared | append your cases below c0b's; the partition pins (`UnmigratedSlotCount()` etc.) move with your counts |
|
||||
| `MG_IntegrationTest/**` | t1's file | each package adds ITS lane cases / scenario arms |
|
||||
| `MG_Backend/DirectGLES/**`, `MG_Backend/DirectVulkan/**` | untouched by c0b | only the sites the contract names, only behind `#if MOBILEGL_BUILD_DISAGGREGATED` |
|
||||
| `~/w7/notes/p5b/p5b-results/<slug>-v1.md` | you | |
|
||||
|
||||
## 5. The loop that follows
|
||||
|
||||
1. The four packages land, each on its own gate. The integrator merges in landing order,
|
||||
regenerates (`gen_pipe.py`, `gen_pipe_field_ownership.py` - never hand-merge `generated/`),
|
||||
runs the quick gate.
|
||||
2. **Re-census** on the merged head: the lane (`p5b-c0b-census.sh`) and the traces. The new
|
||||
first-blocker table is the next round's package list, same shape, same gates.
|
||||
3. Migrate the new first blockers (wave 3's tail is P9/P10-shaped: readbacks, queries, syncs)
|
||||
and the residual-input debts `rsp` points at where a verb's pull is what blocks it.
|
||||
4. Repeat until the four A/B trace cases render under inproc on the Redmi. That is P5b's exit;
|
||||
P6's spawn transport follows (ID-68).
|
||||
@@ -0,0 +1,101 @@
|
||||
# P5b 包通用前言 — 每个包开工前必读
|
||||
|
||||
P5b = verb migration under `inproc`, Minecraft-first(ID-66/ID-68)。目标一句话:让真实负载
|
||||
(首先是每一条 Minecraft trace)在 `MOBILEGL_TRANSPORT=inproc` 下由 apply 线程渲染,
|
||||
越快越好。四个包并行迁移 25 个实测 class-C slot,门只有一个:inproc lane 的 abort 数与
|
||||
trace 的 first-blocker 表。
|
||||
|
||||
## 你在哪、读什么
|
||||
|
||||
- 集成树 `~/w7/pipe`,分支 `feat/disaggregated`。**不要直接在里面干活。**
|
||||
你的工作树:`cd ~/w7/pipe && git worktree add ~/w7/p5b-<你的 slug> -b p5b/<你的 slug> refs/heads/feat/disaggregated`
|
||||
- **契约优先于 brief,按这个顺序读:**
|
||||
1. `MobileGL/MG_Remote/CONTRACT-P5.md`(P5 的四张表与 R-1…R-17,仍然全部有效)。
|
||||
2. `MobileGL/MG_Remote/CONTRACT-P5B.md`(P5b 的增补:25 个 slot 每一行的 wire record、
|
||||
sink 签名、拒绝形状、stamp;§6 是 c0b 的裁定与推翻条件;§8 是文件所有权)。
|
||||
3. `~/w7/notes/p5b/BRIEF-P5B.md`(四个包各自的 slot、文件、门、以及之后的循环)。
|
||||
4. 普查:`~/w7/notes/p6/census-classC.md`(你的 slot 的命中数、"下一个 blocker"的估计)。
|
||||
5. c0b 的报告:`~/w7/notes/p5b/p5b-results/c0b-v1.md`(落了什么、每条裁定、数字、债)。
|
||||
6. P5 的材料按需:`~/w7/notes/p5/PACKAGE-PREAMBLE.md`、`INTEGRATOR-DECISIONS.md`
|
||||
(ID-57 的"按名拒绝"形状、ID-66 的流程规则、ID-68 的普查裁定)、
|
||||
`p5-results/c1-v3.md`(五个 class-B emitter 是怎么写的)、`v1-v3.md`(`ServerVerbSink`)。
|
||||
|
||||
## 流程规则(用户,2026-09-16,ID-66)
|
||||
|
||||
- **快,不做过度验证。** 不做轮次之间的对抗审查;整个 P5b 收尾时由 Codex 做一次。
|
||||
- 一个包在**自己的门 + 集成者的快门**上落地:unit 四车道、`integration-split`(inproc)、
|
||||
push monolith 逐名一致(G2/G14)、G1。没有 verifier 包。
|
||||
- 你的门(BRIEF §3):**lane 里你的 slot 的 abort 数 → 0,受影响的每个场景到达它的下一个
|
||||
first blocker,按名报告**;d1 另加:每一条 Minecraft trace 在 inproc 下到达下一个 first
|
||||
blocker,按名报告。
|
||||
- 设备 A/B 在后台记录,永不阻塞。
|
||||
|
||||
## 第一天必踩的两个坑(P5 原样继承)
|
||||
|
||||
1. **worktree 的 submodule 是符号链接**。抄 `~/w7/p5b-c0b-setup.sh` 与
|
||||
`~/w7/p5b-c0b-submodlinks.sh`(把 `p5b-c0b` 换成你的 slug)。
|
||||
**跑任何 git 命令之前 `submodlinks.sh off`,跑构建之前 `on`。**
|
||||
2. **任何报构建结果的脚本,先确认 `CTestTestfile.cmake` 存在再相信它。**
|
||||
3. (新)**后台任务要 `setsid nohup … < /dev/null &`**:从 Windows 侧的 `wsl.exe` 起的
|
||||
shell 退出时会带走普通的 `nohup … &`。
|
||||
|
||||
## 构建配方(不变)
|
||||
|
||||
```
|
||||
export CCACHE_BASEDIR=/home/swung/w7
|
||||
COMMON="-G Ninja -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ \
|
||||
-DCMAKE_C_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
|
||||
-DCMAKE_BUILD_TYPE=Release -DMOBILEGL_LOG_ACTIVE_LEVEL=MOBILEGL_LOG_LEVEL_INFO \
|
||||
-DMOBILEGL_BUILD_TEST=ON -DMOBILEGL_BUILD_INTEGRATION_TEST=ON -DMOBILEGL_BUILD_BENCHMARK=OFF \
|
||||
-DMOBILEGL_ITEST_EGL_VENDOR=/usr/share/glvnd/egl_vendor.d/50_mesa.json \
|
||||
-DMOBILEGL_ITEST_VK_ICD=/usr/share/vulkan/icd.d/lvp_icd.json \
|
||||
-DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DCMAKE_POLICY_VERSION_MINIMUM=3.5"
|
||||
export GLIBC_TUNABLES=glibc.malloc.tcache_count=0
|
||||
```
|
||||
- pull = 无额外 flag;push = `-DMOBILEGL_PIPE_PUSH=ON`;
|
||||
verify = `-DMOBILEGL_PIPE_VERIFY=ON -DMOBILEGL_ITEST_REQUIRE_GPU=ON`;
|
||||
split = `-DMOBILEGL_PIPE_PUSH=ON -DMOBILEGL_BUILD_DISAGGREGATED=ON -DMOBILEGL_BUILD_DISAGGREGATED_INPROC=ON`。
|
||||
- `-j 24`。`~/w7/p5b-c0b-gate.sh` 是可抄的骨架(四目录 + 生成器 + 纯度 + G1 + 四条 unit 车道),
|
||||
`~/w7/p5b-c0b-census.sh <tree> <outdir>` 是把 P6 普查跑在任意树上的版本
|
||||
(`results.json` 里每条 entry 的 status 与第一行 `Fatal{…}`)。
|
||||
|
||||
## 每个包都要自测的五件事
|
||||
|
||||
1. **G1**:`python3 scripts/symbol_report.py --before ~/w7/p5-before-libMobileGL.so --after build-linux/libMobileGL.so --threshold 0`
|
||||
必须 **0 增 / 0 删 / 0 resize / 0 重命名、`.text` +0**(基线 `.READY` = `4595163c`)。
|
||||
**P5b 最容易破的门**:你加的一切都必须 push/split-only。后端里的任何新函数、新字符串、
|
||||
新成员都要在 `#if MOBILEGL_BUILD_DISAGGREGATED`(或至少 `MOBILEGL_PIPE_PUSH`)之内。
|
||||
2. `ctest -L unit` 四目录(c0b 头上的基线见 c0b-v1.md §2)。
|
||||
3. `integration-split` 在 build-split、`MOBILEGL_TRANSPORT=inproc` 下 **22/22**;
|
||||
push `integration-gpu` 1128/1128 且与 `~/w7/p5-before-ctest-names.txt` 逐名相同(G2/G14)。
|
||||
4. **你的普查**:`~/w7/p5b-c0b-census.sh ~/w7/p5b-<slug> ~/w7/p5b-<slug>-census-logs`,
|
||||
然后按 slot 数 `Fatal{UnmigratedVerb, "…"}`;c0b 头(`f5676103`,含 v1 round 3)上的基线是
|
||||
**432 passed / 185 skipped / 505 aborted / 27 failed,505 个 abort 全部是 `UnmigratedVerb`**,
|
||||
逐 slot 表与 `census-classC.md` 相同(P6 的 426/185/511/27 里那 6 个
|
||||
`StageSnapshotTooNarrow` 已被 v1 r3 修掉,见 c0b-v1.md §2.5)。基线的 `results.json` 在
|
||||
`~/w7/p5b-c0b-census-logs/`。你的 slot 的计数必须归零,多出来的 first blocker 必须按名列出。
|
||||
d1 另跑 `~/w7/notes/tools/p6_census_traces.sh` 的同等物(trace 的 first-blocker 表)。
|
||||
c0b 头上的其余基线:unit 1817 × 3、split 2084;`integration-split` inproc 22/22(2 skipped);
|
||||
push `integration-gpu` 1128/1128;G1 `.text` 10806611 +0、27814 符号 0/0/0/0。
|
||||
5. `python3 scripts/gen_pipe.py --check` / `gen_pipe_field_ownership.py --check`
|
||||
(若你碰了任何 `.def` — 除 CONTRACT-P5B.md §8 明确授权的两处外你不该碰)。
|
||||
|
||||
## 铁律
|
||||
|
||||
- 提交信息 `[Type] (Scope): description`,**单行,绝不加 Co-Authored-By 或任何署名行**。
|
||||
- **只改你拥有的文件**(CONTRACT-P5B.md §8)。需要改别人的,向集成者提。
|
||||
- 目录行的 opcode **永不移动**;新调用追加,退役的调用保留槽位。
|
||||
- **不要拉 GitHub LFS**。fixture 从 `~/w7/pipe/tools/trace_replay/fixtures` 拷(全部已 hydrate)。
|
||||
- **不要碰任何 adb 设备。**
|
||||
- 临时文件 `~/w7/p5b-<slug>-*`,收工时只留报告引用的那几个。
|
||||
**别在别的任务正在写日志时用通配 `rm`。**
|
||||
- **绝不把源文件通过"引号 heredoc 套引号 heredoc"写出去。**
|
||||
- 性能只记录不设门;inproc 是 lockstep,**慢是预期**(R-1 的栅栏税)。
|
||||
|
||||
## 收工
|
||||
|
||||
1. 在你的分支上提交(多个单行提交没问题)。
|
||||
2. 写 `~/w7/notes/p5b/p5b-results/<slug>-v1.md`:落了什么、每一条契约没做的裁定
|
||||
(带证据 + 推翻它需要什么)、五件自测的数字、你的 slot 归零后的**下一个 first blocker 表**
|
||||
(按场景、按 trace,逐名)、以及留给后续轮次的账。
|
||||
3. 返回 ≤200 词:提交区间、验收数字、其他包最需要知道的三件事。
|
||||
@@ -0,0 +1,150 @@
|
||||
# P5b — `inproc` verb migration(`37fc4fdb..82683d4a`)
|
||||
|
||||
> 本页是该阶段的**阶段汇总**:阶段表行、实测与落地形状的完整文本(2026-09-24 从上层索引移入,原文照录)都在这里;上层只留摘要——[`ROADMAP.md`](../../../Disaggregated/ROADMAP.md)、[`MEASUREMENTS.md`](../../../Disaggregated/MEASUREMENTS.md)、[`ARCHITECTURE.md`](../../../Disaggregated/ARCHITECTURE.md)。文中 `file:line` 与"今天""当前"按各段写成时的头理解;文中的 `MEASUREMENTS.md §N` / `ARCHITECTURE.md §17.x` 编号保留,对照表见 [`../README.md`](../README.md)。
|
||||
|
||||
## 摘要
|
||||
|
||||
- 落地:以 class-C 动态普查决定顺序(Minecraft 优先);d1 19 / i1 7 / t2 6 / f1 11 槽迁移 + sync 五槽 + 具名 blit + GLES mip descriptor;r1 / r2 收掉 P5 收官审查;发射表 A=2 / B=54 / C=15。
|
||||
- 门:主机全门 `348d22a4` complete;`integration-split` 107/107;合并 inproc 普查零回退;79 trace 72 / 6 / 1。
|
||||
- **Redmi 出口**(`82683d4a`,APK `p5bcodex2`):正确性 8/8——四条 A/B trace 双后端首次在设备上以独立 apply 线程渲染;**barrier tax 首测**(split − push 逐线程 CPU p50)+5.9% – +18.2%。
|
||||
- 唯一收官审查 0 blocker / 2 major / 1 minor,定向修复。本页末尾归档了 P5 / P5b 出口时记下的债务表(2026-09-22 口径)。
|
||||
|
||||
## 阶段表行(原 `ROADMAP.md`)
|
||||
|
||||
- **阶段**:**P5b** `inproc` verb migration
|
||||
- **状态**:✅ `37fc4fdb..82683d4a`
|
||||
- **落地什么 / 范围**:class-C 普查决定顺序;d1 19 / i1 7 / t2 6 / f1 11 槽迁移;sync 五槽;具名 blit;GLES mip descriptor;r1 / r2;收官审查三项修复。发射表 A=2 / B=54 / C=15
|
||||
- **验收门 / 证据**:主机全门 `348d22a4` complete;`integration-split` 107/107;合并普查零回退、79 trace 72/6/1;**Redmi 正确性 8/8 + 四臂 A/B、barrier tax 首测**。§7
|
||||
|
||||
## 实测:P5b(主机门 `348d22a4`,设备源头 `82683d4a`)(原 `MEASUREMENTS.md` §7)
|
||||
|
||||
### 7.1 迁移与包证据
|
||||
|
||||
`348d22a4` 含 d1 / i1 / t2 / f1、r1 / r2、五槽 sync、具名 blit 与 GLES mip storage;发射表 A=2 / B=54 / C=15(目录 76 条,槽数与 opcode 数不是同一统计量)。
|
||||
|
||||
| 包 | 已落地行为与证据边界 |
|
||||
|---|---|
|
||||
| d1 | 19 个索引 / 实例 / multi-draw / indirect 槽经 `draw_vbo` 过线;独立包普查 77 个 Minecraft 后端用例 28 passed、49 first blockers,全部越过原 draw 首阻塞(默认 stage 32 MiB) |
|
||||
| i1 | 七个 image / compute / barrier / copy-image / storage-block 槽;原首阻塞 239 归零;CopyImage 的 client-shadow 镜像被跳过,`GetTexImage` 前端回退仍可能读旧 shadow |
|
||||
| t2 | 六个 XFB / 曲面细分槽;原首阻塞实测 140(`BeginTransformFeedback` 95 + `PatchParameteri` 43 + `BindTransformFeedback` 2);包内基线 490 / 191 / 390 / 78,原 432 passes 全保留 |
|
||||
| f1 | 11 个 clear / copy / mip 槽;原 34 个首阻塞归零;bound named-clear 接线,unbound 仍具名拒绝 |
|
||||
| r1 / r2 | P5 收官 #1–13;r1 定向 10 passed / 0 skipped;client-only push suppression 两个像素控制变红 |
|
||||
| 具名 blit | `BlitNamedFramebuffer` 经 scoped read/draw binding 发布,退出恢复公开绑定;两后端四个 split 像素 entry 通过 + 15 unit |
|
||||
| GLES mip storage | server 只验证 applier descriptor 的 Levels/extent;三个 mip 像素控制 3/3;iris-BSL GLES 单例 SSIM 0.997496(默认 stage 32 MiB);registry 只 Find 不 mint(barrier 债);RGB 三通道 CPU fallback 保持 Fatal |
|
||||
| sync | 五条现有 opcode 接线,wire 只传句柄,native fence 归 apply 线程;原 LOCAL guard 使 trace 的 `FenceSync` 没真正执行,所以 d1 旧普查里没有可扣除的 FenceSync 首阻塞数 |
|
||||
|
||||
包报告:`~/w7/notes/p5b/p5b-results/{d1-codex-v1,i1-v1,t2-codex-v1,f1-v1,r1-codex-v1,r2-v1,blit-codex-v1,mip-codex-v1,sync-codex-v1}.md`。
|
||||
|
||||
### 7.2 主机收尾门与合并普查(`348d22a4`,`~/w7/p5b-final-host-348d22a4/`;同名重跑 @ `3c80cd62` 见本节末)
|
||||
|
||||
| 车道 | Selected | Passed | Skipped | Failed |
|
||||
|---|---:|---:|---:|---:|
|
||||
| unit-linux / unit-push / unit-verify | 1817 × 3 | 1500 / 1782 / 1795 | 317 / 35 / 22 | 0 |
|
||||
| unit-split | 2132 | 2122 | 10 | 0 |
|
||||
| gpu-linux-monolith / gpu-push-monolith / gpu-split-monolith | 1148 / 1148 / 1178 | 895 / 953 / 953 | 253 / 195 / 225 | 0 |
|
||||
| split(`integration-split`,inproc) | 107 | 105 | 2 | 0 |
|
||||
| verify / audit | 950 / 10 | 804 / 9 | 146 / 1 | 0 |
|
||||
|
||||
G1 27,814 符号 0/0/0/0、`.text` 10,806,611 → 10,806,611;G5 两族 + pin 自测;生成器 / 纯度检查;G2 / G14 rc=0;E1 / E3(a)(精选 4 case)rc=0;E2 OpenRA 2/2、draw-drop SSIM 0.000036 / 758 records、pull-library 拒绝与库 SHA 恢复一致;smoke 16 core + 12 private。retrace-push **79/79**;retrace-verify 在 WSL 重启前完成 4/79,其余 75 条以 SHA-256 钉住的同一 verify 库串行续跑,合并 **79/79**;对账无缺步、无非零退出(`reconciliation-20260916.md`),`complete=true`、exit 0。一次 shell 异常(追加 runner 行触发 `-test-dir: command not found`)保留在原始输出,不冒充测试失败。
|
||||
|
||||
**P5b 基线。** 合并 inproc 普查(`~/w7/p5b-final-census-348d22a4/`):integration lane 显式 32 MiB,**1267 selected = 811 passed / 203 skipped / 62 aborted / 191 failed**;旧 1149 名全部保留、新增 118;旧 432 passes 全保留、零回退;旧 505 abort → 265 passed / 18 skipped / 58 aborted / 164 failed;旧 27 failed → 1 passed / 26 failed。完整 trace 显式 256 MiB(容纳单次 128 MiB 上传,不是默认容量修复):**79 = 72 passed / 6 aborted / 1 failed**(主跑止于 73/79,三轮续跑补齐;`create-indirect` DirectVulkan 在 llvmpipe 上内存膨胀 >60 GiB RSS 被守护杀死,两次复现,记 failed)。7 条未过项首阻塞:`rd12` DirectGLES `Fatal{InitialBytesNotCarried,"resource_respecify"}`、DirectVulkan `Fatal{BarrierTimeout,"Present"}`;`iris-photon`、`iris-derivative`、`create-indirect` 三个 DirectGLES `Fatal{UnmigratedEmulation,"texture-remint-pull"}`;`iris-bsl-esc-menu-854` DirectGLES `InitialBytesNotCarried`;`create-indirect` DirectVulkan 内存守护。通过项含 `improved-transparency-minecraft-26.3` DirectGLES SSIM 1.0 / DirectVulkan 0.999914、`iris-iterationrp` DirectVulkan 0.995833、`iris-bsl-esc-menu-854` DirectVulkan 0.998402。逐 trace 见 `joint-codex-v1.md` 的 census 块与 `identity.json` / `counts.json` / `trace-transitions.json`。
|
||||
|
||||
**同名重跑 @ `3c80cd62`(P7 wave 3 包 E1,2026-09-22,WSL `~/w7/p7-census-e1`,lavapipe)。** 方法逐字沿用上段普查的 helper(`~/w7/p5b-final-census-348d22a4/artifacts/census-helper.py`):取 monolith 条目的命令与 ENVIRONMENT,`MOBILEGL_IPC_STAGE_MB=32` 与 `MOBILEGL_ITEST_REQUIRE_GPU=1` 作环境底、条目自己的 ENVIRONMENT 覆盖之,再强制 `MOBILEGL_TRANSPORT`、私有 `MOBILEGL_LOG_FILE_PATH`、tcache 关,状态映射相同;名单 = 上段 1267 个名字(头上 0 个缺失)。跑器 `~/w7/e1bin/{wa27,walane}.py`,证据 `~/w7/e1-census/{wa27,lane-inproc2}-3c80cd62/`;lane 普查单并发墙钟 426 s。
|
||||
|
||||
| 读数 | P5b `348d22a4` | 头 `3c80cd62` | delta |
|
||||
|---|---|---|---|
|
||||
| 27 个 wrong-answer 同名,inproc | 1 passed / 26 failed | **24 passed / 3 aborted / 0 failed** | 23 转绿;3 条由错答变为具名停止 |
|
||||
| 同上,spawn(P5b 未测) | — | 24 passed / 3 aborted / 0 failed | 与 inproc 逐名相同 |
|
||||
| 合并 lane 普查 1267 同名,inproc | 811 / 203 skipped / 62 aborted / 191 failed | **1051 / 172 / 37 / 7** | failed→passed 187、aborted→passed 62、skipped→passed 31;passed→非绿 40、failed→aborted 4 |
|
||||
| trace(CI split 子集),inproc | 79 = 72 / 6 aborted / 1 failed(含 `rd12` × 2,256 MiB) | **77 = 77 / 0 / 0**(默认 32 MiB;spawn 同为 77 / 0 / 0) | `rd12` 已 `ci: false` 出分母;其余 7 格全活([`notes/p7/iris-census-3c80cd62.md`](../p7/iris-census-1135c664.md) §6) |
|
||||
|
||||
27 条按 §6.2 的家族:
|
||||
|
||||
| 家族 | 条 | 头(inproc = spawn) | 机制 / 读法 |
|
||||
|---|---:|---|---|
|
||||
| `LayeredAttachmentShapeScenario` | 14 | **14 绿** | 未二分到提交;DirectVulkan 的 7 条在 `integration-magma-full-{split,spawn}` 的同名注册条目上也绿 |
|
||||
| packed depth/stencil `GetTexImage` | 3 | **3 绿** | DirectGLES `Split/Spawn.ForcedDs` 与 DirectVulkan `Split/Spawn.Full` 的注册条目同绿 |
|
||||
| framebuffer `HandleRecycleScenario` | 5 | **2 绿**(DG `Handles`、DV `AbaControlHandles`)/ **3 具名停止** | 停止的 3 条都是 `integration-monolith-control` 里 `MOBILEGL_PIPE_PUSH=0` 的对照臂:DG `Legacy` → `Fatal{PipeLegacyMemosDisabled, "MOBILEGL_TRANSPORT is not monolith and kMGPipeSubsystemSamplers (bit 11) is clear …"}`(`MobileGL/MG_Backend/DirectGLES/Managers.cpp:4289` 经 `StopOnArmlessPipeSubsystem`,`MobileGL/MG_Backend/DirectGLES/Managers.cpp:2915-2918`);DV `Legacy` / `AbaControl` → server `Fatal{UnmigratedVerb, "Magma:clear-framebuffer-record"}`(`MobileGL/MG_Backend/DirectVulkan/Renderer/VulkanRenderer.cpp:8008`:push 全关时 transport 下没有 framebuffer record),spawn 臂 client 另记 `Fatal{ReadbackDeclined, "ReadPixels"}`。旋钮矛盾的响亮停止,不是错答;push 打开的两条同名兄弟全绿 |
|
||||
| `PrimitivesGeneratedNoXfbScenario` | 3 | **3 绿** | DV `Split/Spawn.Full` 的 `CountsADrawMadeWithNoCaptureSpan` 同绿(`TheRerouteIsActuallyArmed…` 在 full 车道因不钉旋钮而 skip,按设计) |
|
||||
| `TextureParamsWithoutASamplerView` | 1 | **绿** | DG `Split` / `Spawn` 注册条目同绿——ROADMAP 债务表的「待核」关闭 |
|
||||
| `P4aFinalFixScenario` FBO/RBO delete | 1 | **绿** | 与 P5b 同 |
|
||||
|
||||
lane 的 44 条非绿(37 aborted + 7 failed)**全部是 `integration-monolith-control` 条目、且自己的 ENVIRONMENT 钉了非默认 `MOBILEGL_PIPE_PUSH`**(0 / 0x7f / 0x1ff / 0x5ff / 0x9ff / 0x1fff / 0x8000000000001fff)——普查方法把 monolith 对照臂也强制上了 transport,这 44 条读的是「这组旋钮在 split 下没有臂」:35 条具名停止(18 × `Magma:clear-framebuffer-record`、17 × `PipeLegacyMemosDisabled`),2 条 `Fatal{UnmigratedPipeInput, "GetBufferBindingPoint@DispatchCompute"}`(掩码 0x1fff),7 条断言失败在单体专用仪器上(4 × `CsoContentAddressingScenario` 在 client 日志里找带 `cso[csom= csob=]` 的统计窗口,而 split 下 CSO 在 server 铸造、那一行在 server 日志;3 × DV `HandleRecycle.AbaControlHandles` 的 buffer / VAO 用例是负控,期望 STALE,wire 路径读到 FRESH——即像素正确、负控不起作用)。这 44 条之外的 1223 条全部 passed 或 skipped。其中 40 条在 P5b 是 passed,头上转为上述读法;没有一条是默认旋钮下的产品回归。
|
||||
|
||||
### 7.3 唯一收官审查与定向修复
|
||||
|
||||
审查 `p5b-close-codex-review.md`(`348d22a4`,diff base `37fc4fdb`,只读):**0 blocker / 2 major / 1 minor**。
|
||||
|
||||
| 项 | 问题 | 处置 |
|
||||
|---|---|---|
|
||||
| Major 1 | user-index span 在段内但 Size 可能短于 Count × IndexSize | `a021e3cc`:encoder / decoder / sink 共用 shape/extent gate(单 range、宽度 1/2/4、`Uint64(Count) × IndexSize ≤ Size`),三端短 span 均拒绝、段末 exact-fit 通过;定向 9/9 |
|
||||
| Major 2 | `ClientWaitSync` 保留 64-bit timeout 但 applied/reply 等待一律 30 s | `82683d4a`:预算 = ceil(timeout ns / 1e6) + 30,000 ms,有限 chunk;普通容量等待 / `FenceWaitServer` 仍 30 s,shutdown 可唤醒。**该 30 s 常量后来在 CI 修复轮改成 120 s(`68b55ea3`),原因见本表下注** |
|
||||
| Minor 3 | ReadPixels exact-size reply 的未知 status 穿过 production helper | `82683d4a`:tight 与 bounce 路径拒绝 status=3,`Fatal{ReplyStatusInvalid}` |
|
||||
|
||||
定向 `RemoteClientTest` 7/7(0 ns、1 ns、60 s、UINT64_MAX 预算,`FenceWaitServer`、shutdown、未知状态)。合并后 quickgate 里 split 单元车道仅有的 2 个失败(`PipeWireCodecTest.UserIndexSpan*`)是测试日志捕获缺陷(库按进程截断日志,fork 子进程增量读读空),`7cb29d46` 修复后 3/3。本阶段不再做第二轮全门或审查;原始全门源头仍为 `348d22a4`,三项修复只以定向证据补齐。
|
||||
|
||||
> 上表 Major 2 的 30 s 是 `82683d4a` 当时的值,保留为历史记录。**当前值是 120 s**(`68b55ea3`,`ClientSession.cpp` 的 `kBarrierTimeoutMs`),因为 lavapipe 在 texture-handle 注册(`vkCreateImageView` → `llvmpipe_register_texture`)上可合法阻塞 apply 线程 29.3–39.7 s,30 s 的看门狗会把合法慢路径间歇判红;详因见 `ARCHITECTURE.md` §17.1。`RemoteClientControls.FenceWaitBudgetHonorsGlTimeoutAndFiniteTransportChunks` 的期望值(30000/30001/90000)已同步改为 120000/120001/180000。
|
||||
|
||||
### 7.4 Redmi 出口与四臂 A/B(`82683d4a`,APK `p5bcodex2`)
|
||||
|
||||
Redmi `2f7cbe2e`,证据根 `MobileGL/.trace-work/p5b-redmi/p5bcodex2/2f7cbe2e/`;全部组合显式 `MOBILEGL_IPC_STAGE_MB=256`。**钉频口径变更**:2026-09-11 的厂商 GPU 上限(1050 MHz)已消失,本次 deterministic pin 为 **1100 MHz**;与 1050 MHz 时代的活动不可比钟频,只有同场四臂配对可比。
|
||||
|
||||
**正确性 8/8**:`improved-transparency-minecraft-26.3`、`minecraft-1.21.4-in-world`(GLES SSIM 0.999995)、`minecraft-1.21.4-fabric-sodium-in-world`、`minecraft-1.21.4-fabric-iris-bsl-in-world` × 双后端,inproc split 臂,逐组钉频取证 + PNG / SSIM 阈值 + inproc + apply 线程 + 正 wire 记录三证,skip 即失败。**四条 A/B trace 首次在 Redmi 上以独立 apply 线程渲染——P5b 出口判据达成。**
|
||||
|
||||
**四臂 A/B(barrier tax 首测)**:32 组中 24 组带 200 帧尾验证全绿;三次重复取 wall-time 均值最低者、尾 200 帧(`ab-tables.md`)。逐线程 CPU p50:
|
||||
|
||||
| trace | 后端 | push−pull | **barrier tax(split−push)** | barrier tax 帧 p50 |
|
||||
|---|---|---:|---:|---:|
|
||||
| improved-transparency-26.3 | DirectGLES | +15.6% | **+10.3%** | +0.1% |
|
||||
| improved-transparency-26.3 | DirectVulkan | +8.8% | **+13.1%** | +13.3% |
|
||||
| minecraft-1.21.4-in-world | DirectGLES | +21.4% | **+18.2%** | +0.2% |
|
||||
| minecraft-1.21.4-in-world | DirectVulkan | +11.8% | **+17.9%** | +19.7% |
|
||||
| minecraft-1.21.4-fabric-sodium | DirectGLES | +10.9% | **+5.9%** | −0.1% |
|
||||
| minecraft-1.21.4-fabric-sodium | DirectVulkan | +7.6% | **+8.3%** | +4.3% |
|
||||
|
||||
splitctl−push 全部在 ±1.6% 内(增量来自 inproc 传输与 barrier,不是 APK / 构建差);p99 同向放大(sodium DirectVulkan 帧 p99 +107.7% 为离群尾帧)。inproc 臂 peak RSS 389 MiB – 2.5 GiB,全角色映射 560.5 MiB(256 MiB × 角色视图是虚拟映射容量,不是 RSS 增量)。
|
||||
|
||||
**fixture 受限的 8 组**:`iris-bsl-in-world` 只有 123 个 benchmark 帧,低于 200 帧尾规则,四臂 × 双后端验证全部失败(pull 同失败,是 fixture 上限不是回归);123 帧完整序列按同法折算成补充表 `ab-tables-bsl-123frame-supplement.md`:barrier tax CPU p50 DirectGLES **+7.1%**、DirectVulkan **+9.1%**,不混入上表。性能仍只记录、不作阻塞门。
|
||||
|
||||
## 落地形状(原 `ARCHITECTURE.md` §17.5)
|
||||
|
||||
### 17.5 P5b:class-C verb 迁移
|
||||
|
||||
- **顺序由动态普查决定**(ID-68):79 个 trace × 集成车道在 inproc 下的首阻塞统计(`~/w7/notes/p6/census-classC.md`),所有 Minecraft trace 首阻塞 = `DrawElements` → Minecraft 优先,四包并行(d1 draws、i1 image/compute、t2 XFB/tess、f1 clear/copy/mip)+ 尾包(sync、具名 blit、GLES mip)。
|
||||
- **一次迁移 = 一个发射器 + 一个 sink 体**:c0b 先给 25 个已测量槽铺 wire 记录(`Wire/PipeWireCodec` 的行)、`WireVerbSink` 分派与 `ServerVerbSink` 具名 Fatal stub;包只在 `EmitTables.cpp` 把槽从 C 挪到 B、在 `PipeApplier.cpp` 填 sink 体(apply 线程、只从记录与 server 状态取输入)。这些槽到达的是 sink,不是 `MGPipeApply*`。
|
||||
- **规则 D**:记录把 GL 调用逐字带在句柄旁边(draw 的 mode/count/type/offset/instance/base 字段、clear 的值类与 drawbuffer 下标、named-framebuffer 的句柄形式),后端保留它的 barrier-pulled 读;未测量的形式**按名拒绝**(`+RENDERBUFFER`、`+UNBOUND` 等)而不是猜。
|
||||
- **draw 家族**:19 个索引 / 实例 / multi-draw / indirect 槽全部下沉到 `draw_vbo`;用户索引 span 的 shape/extent 门在 encoder / decoder / sink 三端共用(单 range、宽度 1/2/4、`Uint64(Count) × IndexSize ≤ Size`)。
|
||||
- **sync**:`FenceSync` / `ClientWaitSync` / `GetSyncStatus` / `WaitSync` / `DeleteSync` 五条现有 opcode 接线,client 铸造 Fence 句柄,wire 只过 `{slot,gen}`,native fence 归 apply 线程;顺带修掉 `MGL_BACKEND_SLOT_PTR_LOCAL` 对 split 恒返回 nullptr。
|
||||
- **具名 blit**:`BlitNamedFramebuffer` 经作用域化 client-shadow read/draw 绑定发布,降为现有 bound backend 调用,退出恢复公开绑定。**GLES mip storage**:前端先定义并发布层级,server 只验证 applier descriptor 的 Levels/extent;registry 身份解析只 Find 不 mint。
|
||||
- **P5b 留下的 inproc 依赖**:具名 blit 的 scoped client binding + barrier;mip descriptor 的 barrier-held registry 查询;FBO death 的 inproc mailbox。三者跨地址空间都不成立——它们只是 §17.6 清单里的三行。
|
||||
|
||||
## 里程碑记录(原 `ROADMAP.md`)
|
||||
|
||||
- **P5b 出口(2026-09-16,已达成)**:主机全门 `348d22a4` complete;79 trace 72/6/1;设备源头 `82683d4a`(APK `p5bcodex2`)Redmi 正确性 8/8——四条 A/B trace 双后端首次在设备上 inproc 渲染;barrier tax 首测 split−push 逐线程 CPU p50 +5.9% – +18.2%。
|
||||
|
||||
## 债务表全文(原 `ROADMAP.md`「P5 / P5b 出口记录的债务」,2026-09-22 口径)
|
||||
|
||||
|
||||
| 债务 | 去向 / 当前口径 |
|
||||
|---|---|
|
||||
| P5 的 27 个普通 inproc wrong-answer(历史计数) | **已结(2026-09-22,`3c80cd62` 同名重跑,`MEASUREMENTS.md` §7.2;inproc 与 spawn 逐名相同)**:原「22 → P4b/P7 texture readback」→ **19 绿**(14 layered、3 packed depth/stencil、2 framebuffer recycle),余 **3 条 framebuffer recycle 由错答变为具名停止**——三条都是 `integration-monolith-control` 里 `MOBILEGL_PIPE_PUSH=0` 的对照臂(DG `Legacy` → `Fatal{PipeLegacyMemosDisabled}`;DV `Legacy` / `AbaControl` → `Fatal{UnmigratedVerb, "Magma:clear-framebuffer-record"}`),旋钮在 split 下没有臂,不是 readback 欠账;原「3 → P7 query」→ **3 绿**;1 inspection(`TextureParamsWithoutASamplerView`)→ **绿,「待核」关**;1 FBO/RBO delete → 绿。同一重跑的 1267 名合并 lane 普查 811/203/62/191 → 1051/172/37/7,44 条非绿全是钉了非默认 `MOBILEGL_PIPE_PUSH` 的对照臂 |
|
||||
| `rsp` residual inputs | **P5f 已关闭残余读取**:字段归属 41 record / 6 derived / 0 barrier / 16 fatal;strict 与双块 marker 表均为空,双后端逐帧 RSP 门 2 PASS / 0 skip,Redmi 36 个统计窗口 rsp=0。P5e 的“rsp 只计被采纳设障 pull”是历史计数规则,八对 admitted 的历史观察不再是当前允许集合 |
|
||||
| `SEG_REPLY` 2 MiB 单槽 payload cap | → **P9**(契约 §10.2.3 改判,ID-47);stream 链路上 reply 本就是消息,cap 变成 P6.5 nd 的 `maxReplyBytes` 窗口 |
|
||||
| `GetCaps` 的两个 blobref | 目前不骑 record;一旦运输,必须有 server→client carrier rule,不能套 `SEG_STAGE` |
|
||||
| PACK-PBO readback | P5 曾具名拒绝;`70df57bb`(2026-09-20)起 **client 侧已实现**(bounce 紧凑 reply + 按 PACK 步长逐行 `UploadSubData`,`EmitTables.cpp`;`glGetTexImage` 同形且按 `MaxReplyBytes` 分块);剩余只是 fire-and-forget(`MarkReadPixelsPackBuffer` 全仓无生产调用者)→ **P9**(ID-57) |
|
||||
| ABI fingerprint | **已由 P6.5 wf / nd 落地(`fe28bdb5`)**:`wireFingerprint` 布局摘要 + `buildFingerprint` 拆分(只在 `Dial==Fork` 或 `MOBILEGL_IPC_REQUIRE_SAME_BUILD=1` 下比较)、`LinkTerms` 四个尺寸由 server 陈述、`DynamicBackendParameters` 定宽;`CONTRACT-P6.md:666` 仍把定宽记在 P7 名下,待改 |
|
||||
| 默认 32 MiB staging | 目标负载单次 128 MiB 上传装不进默认 stage;**内容侧分块已落地**(buffer 范围走查 `1e7c372e`、纹理整宽 slab `9469d48e`,预算 `MGPipeStageChunkBytes()` = 默认 segment/4 = 8 MiB),故普查 / Redmi 的显式 256 MiB profile 不再是这两条路径的必需;未接入分片的 record 类型与专用 carrier 仍归 P8(开放问题 11) |
|
||||
| P5b 的 inproc 依赖 | 历史的 scoped client binding、mip registry 查询与 FBO death mailbox 已经 P5c/P5f 改为记录、句柄和控制/事件路径;P6 核验并装配已有载体,不再重做这些迁移 |
|
||||
| 未迁移与仿真路径 | **class-C 已归零到两个**(2026-09-22 核):`EmitTables.cpp` 的 `MGR_UNMIGRATED_*` 列表全空,只剩手写的 `SetSwapInterval`(P10)与按裁定无行的 `DeleteTransformFeedback`(P9);契约 §11 / SPAWN-PLAN §8 的"511 class-C entries"是 P5 joint 树上的旧普查(`notes/p6/census-classC.md`),不是当前数;~~client vertex arrays~~(已实现:`beba0256` 起 wire 复制到 owned buffer 并保留 baseVertex/baseInstance/drawID,本轮补齐 monolith 臂 `62bfe461..e258a822`)、multi-draw client indices、RGB 三通道 CPU mip、renderbuffer copy endpoint、未绑定 named clear、`texture-remint-pull` 仿真保留具名拒绝 → P8 / P9;Magma split compute / image 的“82 个错答”是 P5b 历史普查计数,不是当前失败清单;P5f 后续已补应用 buffer consumers 与 Android 旋转 blit,Magma inproc 已能运行所测 MC 世界;后续 run-ahead 已完成([报告](../p5f/magma-runahead.md));XFB、部分不对齐范围、placeholder/native-format 及进一步性能工作仍归 P7([报告](../p5f/magma-inproc-fix.md)) | **P5e 补 → CI 收尾更正**:client vertex arrays 曾是 run-ahead 下的具名拒绝(ID-82)与预期红车道 `integration-clientarrays-split`(ID-134);`beba0256` 起已实现,`347bb90c` 已把该车道翻转为**硬绿门**并纳入主 GPU/split 标签,本轮补齐 monolith 臂(`62bfe461..e258a822`);staging 仍归 P8
|
||||
| E2 wire 内容控制 | draw-drop 是 OpenRA 的有效控制(758 draws,SSIM 0.000036);clear-drop 被全屏 overdraw 掩盖,不能作控制 |
|
||||
| max record bytes | 实测 reduced / OpenRA `maxrec=784 B`,默认 cap 4 MiB;只代表所测负载,后续索引 / indirect 尾仍须记录 |
|
||||
| 树外脚本 | 普查跑器、`wsl_p5_gate.sh`、Redmi 定频行都在 `~/w7/notes/`,git merge 不会传播 |
|
||||
| 临时 CI trigger | 合入 dev 前删除 `test.yml` / `apk.yml` 的 `feat/disaggregated` TEMPORARY trigger |
|
||||
|
||||
## 本目录
|
||||
|
||||
| 文件 | 内容 |
|
||||
|---|---|
|
||||
| [`BRIEF-P5B.md`](BRIEF-P5B.md) | BRIEF-P5B — verb migration under inproc, Minecraft-first |
|
||||
| [`PACKAGE-PREAMBLE.md`](PACKAGE-PREAMBLE.md) | P5b 包通用前言 — 每个包开工前必读 |
|
||||
| [`p5b-results/`](p5b-results/) | 17 个文件:各包(d1 / i1 / t2 / f1 / r1 / r2 / blit / mip / sync)的报告 |
|
||||
@@ -0,0 +1,25 @@
|
||||
# P5b named blit migration
|
||||
|
||||
Commit `31b26a440de4d52c1afb746ed934247da8b8b145`, branch `p5b/blit-codex`, base `9606466a`; independent tree `/home/swung/w7/p5b-blit-codex`.
|
||||
|
||||
The client emit table now lowers BlitNamedFramebuffer through a scoped client-shadow read/draw binding override. It revalidates the existing classified fields, emits the original two handles/rectangles/mask/filter in the existing MGPBlit row, waits for server apply, then restores both bindings. BindingSlot versions make the following ordinary verb republish restored bindings. The server keeps using its existing bound backend entry point; no driver call runs on the client, and no pointer or extra side channel was added. Name zero uses the existing default-framebuffer handle. CONTRACT-P5B §6 item 13 records the integrator-approved ruling and its retirement when the lockstep barrier/process-shared access disappears.
|
||||
|
||||
Only split production sources changed. Pull G1 was not remeasured locally; the integrator owns the merged-head four-flavor gate. Slot partition is now A=2/B=49/C=20. Named clears are unchanged.
|
||||
|
||||
## Directed validation
|
||||
|
||||
- Split configure and full default ALL build succeeded at -j8; CTestTestfile.cmake exists. Logs `/home/swung/w7/p5b-blit-codex-{config-split,build-split,build-driver}.log`.
|
||||
- Four new split cases all executed and passed: two cases × DirectGLES/DirectVulkan. They check separately unbound source/destination, preserved public read/draw names, an immediate ordinary clear before another bind, following ordinary blit, and default FBO as source/destination.
|
||||
- RemoteEmitTable: 15/15 pass. Combined targeted CTest: 23 selected = 19 executed pass + four deliberately skipped monolith registrations of the split-only scenarios. Those skips are not counted as split execution. Log `p5b-blit-codex-targeted.log`.
|
||||
- Dynamic retrace: exactly four selected, stock `/home/swung/w7/retrace_gate.py`, `MOBILEGL_TRANSPORT=inproc`, **default SEG_STAGE=32 MiB**. No adb, LFS fetch, repeated full gate or adversarial review.
|
||||
|
||||
| Trace | Backend | Result / next first blocker |
|
||||
|---|---|---|
|
||||
| improved-transparency-minecraft-26.3 | DirectVulkan | pass, SSIM 0.999914064 (threshold 0.995), zero private Fatal |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-in-world | DirectVulkan | pass, SSIM 0.997324199 (threshold 0.99), zero private Fatal |
|
||||
| improved-transparency-minecraft-26.3 | DirectGLES | named blit crossed; Fatal{RingOverrun, "SEG_STAGE"}: 134217728-byte blob exceeds 33554432-byte segment |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-in-world | DirectGLES | named blit crossed; Fatal{UnmigratedEmulation, "generate-mipmap-storage"} |
|
||||
|
||||
Trace evidence: `/home/swung/w7/p5b-blit-codex-traces.log`, `/home/swung/w7/retrace-out/p5b-blit-codex/summary.json`, and per-case/backend `output/mobilegl.log`, `result.json`, images and outer logs below that output root. The trace fixtures were copied from hydrated pipe files for the run and restored to their original tracked pointers before commit.
|
||||
|
||||
The integrator owns a separate larger-stage improved-transparency run; it is not claimed here. The follow-up mip-storage fix is a separate commit/report. `git diff --check` clean, committer Swung0x48, no trailers, submodule links off at commit.
|
||||
@@ -0,0 +1,278 @@
|
||||
# c0b — the P5b contract package, v1
|
||||
|
||||
Branch `p5b/c0b`, forked from `feat/disaggregated @ c9878d69`, merged `feat/disaggregated @
|
||||
f5614995` (v1 round 3) before finishing. Worktree `~/w7/p5b-c0b`. Nothing pushed; nothing merged
|
||||
into the integration branch.
|
||||
|
||||
```
|
||||
4dd4027e [Feat] (MG_Pipe, P5b): append the five P5b verb rows at opcodes 72..76, arm draw_vbo
|
||||
with the kDrawIsIndirect second tail, give resource_copy_region its two GL names and
|
||||
the clear discriminants one spelling, and stamp every P5b verb boundary
|
||||
a1641130 [Feat] (MG_Remote, P5b): decode every P5b row to a WireVerbSink method, stub each
|
||||
ServerVerbSink body with a named UnmigratedVerb Fatal, and partition the emit table
|
||||
class-C list per migration package
|
||||
02972f70 [Test] (MG_Test, P5b): pin the 76-row catalogue, the 18 stamp rows and the 77 verify
|
||||
payloads, and round-trip each P5b row once through the codec
|
||||
538c3c19 [Docs] (MG_Remote): CONTRACT-P5B.md - the wire records, sinks, refusals and rulings
|
||||
for the 25 measured class-C slots
|
||||
f5676103 [Merge] (P5b, c0b): merge feat/disaggregated at f5614995 (v1 round 3, ID-62/ID-67)
|
||||
into p5b/c0b
|
||||
```
|
||||
|
||||
Range **`4dd4027e..f5676103`** (four single-line commits plus the merge). The working tree is
|
||||
clean; the worktree's submodule symlinks are left OFF (git-safe).
|
||||
|
||||
Notes written beside it: `~/w7/notes/p5b/PACKAGE-PREAMBLE.md`, `~/w7/notes/p5b/BRIEF-P5B.md`.
|
||||
The contract: `MobileGL/MG_Remote/CONTRACT-P5B.md` (beside `CONTRACT-P5.md`, which stays
|
||||
authoritative for everything it covers).
|
||||
|
||||
---
|
||||
|
||||
## 1. What landed
|
||||
|
||||
| # | item | state |
|
||||
|---|---|---|
|
||||
| 1 | `CONTRACT-P5B.md`: every one of the 25 slots - wire record, class, sink signature, reply slot (none), thread rule, refusal shape, monolith (untouched) - grouped d1 / i1 / t2 / f1; table 0 additions; table 2 and 3 under P5b; knobs (none); 12 rulings with overturn conditions; the 71-slot partition; ownership with two pre-granted edits | **done** |
|
||||
| 2 | `PipeCalls.def`: **five rows appended at opcodes 72..76** (`BindShaderImage`, `PatchParameter`, `BindStreamOutput`, `SetStorageBlockBinding` `kHasBlob`, `CopyFramebufferToTexture`), documented count 71 → 76; no opcode moved | **done** |
|
||||
| 3 | `MGPipeTypes.h`: the five PODs (40 / 8 / 16 / 40 / 48 B); `kDrawIsIndirect`; `MGPCopyRegion` 64 → 72 (two GL names); `MGPMipPlan::Target` ruled the GL enum; the clear discriminants `kMGPipeClearKind*` / `kMGPipeClearValueClass*` (one spelling; the two hand-minted copies in `EmitTables.h` and `PipeApplier.h` now alias them) | **done** |
|
||||
| 4 | `PipeFields.def`: five field lists + `MGPCopyRegion`'s two names; verify payloads 72 → 77. `FieldOwnership.def`: six stamp rows (`ResourceCopyRegion → CopyImageSubData` and the five new ops), 12 → 18. Generated tables regenerated; `gen_pipe.py --check` / `--self-test` (9/9), `gen_pipe_field_ownership.py --check` / `--self-test` (15/15), `gen_pipe_dirty_surface.py --check` all green | **done** |
|
||||
| 5 | Codec: `MGPW_FOR_EACH_CALL` + name/size tables for the five rows; `BlobSlotsFor` for the name blob; `draw_vbo`'s layout takes `kDrawIsIndirect` (exclusive with the span, `NumDraws == 0`, both `Fatal{ProtocolCorruption}` on both sides); `WireRecordLayout::SecondTailIsHostSpans` so the encoder's span honesty pass cannot read an indirect block as spans; arms for the six pre-existing rows that returned false (`LaunchGrid`, `MemoryBarrier`, the four stream-output rows), for `ResourceCopyRegion` and `GenerateMipmap`, and for the five new rows (the name blob is bounded at 4096 bytes, copied and re-terminated, rule C) | **done** |
|
||||
| 6 | `WireVerbSink`: `OnDrawVbo` gains `const MGPDrawIndirect*`; 13 new methods. `ServerVerbSink`: 13 overrides, every body `Fatal{UnmigratedVerb, "<GL slot>"}` by the slot's own name ("(server sink)" in the message text); `OnDrawVbo` declines the indirect block, the span, `NumDraws > 1` and the instanced arms by name instead of by log line | **done** |
|
||||
| 7 | `EmitTables.cpp`: class C partitioned into `MGR_UNMIGRATED_{D1,I1,T2,F1,TAIL}_SLOTS` with per-package `kEmittedSlots*` and ownership `static_assert`s 19 / 7 / 7 / 11 / 20 = 64; `EmitTables.h` includes `MGPipeTypes.h` for the aliases. **No slot flipped**: `UnmigratedSlotCount() == 64`, `ImplementedVerbCount() == 5` as `RemoteClientTest` pins | **done** |
|
||||
| 8 | Tests: `PipeCatalogueTest` (kCtxVerb 18, opcodes 72..76 by value, `kOpCount` 77, the P5b flags/sizes, the clear constants, verify payloads 77); `FieldOwnershipTest` (18 stamp rows, six by name); `PipeWireCodecTest` (13 new cases: 11 round trips - one per row, the four stream-output rows in one - and two forged-record Fatals for the draw flag exclusion and the indirect-with-ranges refusal) | **done** |
|
||||
| 9 | Notes: `PACKAGE-PREAMBLE.md` (numbers on this head, the ID-66 process rules, the `setsid` trap), `BRIEF-P5B.md` (goal, four packages with slots / rows / models, per-package gate, file ownership, the loop) | **done** |
|
||||
| 10 | Tools: `~/w7/p5b-c0b-{setup,submodlinks,gate,accept}.sh`, `~/w7/p5b-c0b-census.{sh,py}` (the P6 lane runner over any tree: `p5b-c0b-census.sh <tree> <outdir>`) | **done** |
|
||||
|
||||
---
|
||||
|
||||
## 2. Acceptance numbers
|
||||
|
||||
All on `~/w7/p5b-c0b @ f5676103` (the merged head), `GLIBC_TUNABLES=glibc.malloc.tcache_count=0`.
|
||||
Log: `~/w7/p5b-c0b-gate-run3.log`; the census: `~/w7/p5b-c0b-census-logs/results.json` (per entry:
|
||||
status, rc, first `Fatal{…}`, command, working directory, its own `.log`/`.out`).
|
||||
|
||||
### 2.1 Builds — four for four
|
||||
|
||||
`build-linux` (pull), `build-push`, `build-verify`, `build-split`: **configure rc=0, build rc=0,
|
||||
all four**, `CTestTestfile.cmake` confirmed in each.
|
||||
|
||||
### 2.2 G1 — did not move
|
||||
|
||||
```
|
||||
nm --defined-only build-linux/libMobileGL.so | grep -ic MG_Remote -> 0
|
||||
nm --defined-only build-split/libMobileGL.so | grep -ic MG_Remote -> 627
|
||||
symbol-report vs ~/w7/p5-before-libMobileGL.so (.READY 4595163c), --threshold 0:
|
||||
.text 10806611 -> 10806611 (+0, +0.000%)
|
||||
27814 -> 27814 defined symbols: 0 added, 0 removed, 0 resized, 0 renamed
|
||||
```
|
||||
|
||||
Held because everything c0b added is either a header-only type, a `.def` row whose generated
|
||||
tables the pull build never odr-uses (the baseline `.so` defines no `gMGPipe*` symbol), or an
|
||||
`MG_Remote` file compiled only under `MOBILEGL_BUILD_DISAGGREGATED`.
|
||||
|
||||
### 2.3 Generators and purity
|
||||
|
||||
```
|
||||
gen_pipe.py --check : generated files are up to date
|
||||
gen_pipe.py --self-test : 9 negative-control trips, positive control OK
|
||||
gen_pipe_field_ownership.py --check : generated file is up to date
|
||||
gen_pipe_field_ownership.py --self-test : 15 negative-control trips, positive control OK
|
||||
gen_pipe_dirty_surface.py --check : green
|
||||
check_include_closure.py : 4 probes, 0 skipped, 0 problem(s)
|
||||
check_doc_citations.py CONTRACT-P5B.md : 36 citations, 0 problem(s)
|
||||
```
|
||||
|
||||
### 2.4 Test lanes
|
||||
|
||||
| lane | result |
|
||||
|---|---|
|
||||
| `ctest -L unit` build-linux / push / verify | **1817 / 1817** each (the head's baseline) |
|
||||
| `ctest -L unit` build-split | **2084 / 2084** = the head's 2071 + c0b's 13 codec cases; no known red (c1g landed before c0b forked) |
|
||||
| `ctest -L integration-split` build-split, `MOBILEGL_TRANSPORT=inproc` | **22 / 22** (2 skipped by design, as in ID-66's quick gate) |
|
||||
| `ctest -L integration-gpu` build-push, `MOBILEGL_TRANSPORT=monolith` | **1128 / 1128** |
|
||||
| G2 pull-vs-push `ctest -N` names, every label | **0 diff lines** |
|
||||
| G14 build-linux names vs `~/w7/p5-before-ctest-names.txt` | **0 removed**, 43 added (c0b's 13 + v1 round 3's 30 `ServerLoopTest` cases) |
|
||||
| G5 `p3a_untouched_regions.sh ff2994d9 HEAD` | rc=0, byte-identical |
|
||||
| G5 `p4a_untouched_regions.sh ff2994d9 HEAD` | rc=0, byte-identical |
|
||||
|
||||
### 2.5 The inproc census on this head
|
||||
|
||||
`~/w7/p5b-c0b-census.sh ~/w7/p5b-c0b ~/w7/p5b-c0b-census-logs` (the P6 runner over this tree;
|
||||
1,149 entries, eight workers, private logs):
|
||||
|
||||
| status | P6 census (`e61d0012`) | c0b head (`f5676103`) |
|
||||
|---|---:|---:|
|
||||
| passed | 426 | **432** |
|
||||
| skipped | 185 | **185** |
|
||||
| subprocess-aborted | 511 | **505** |
|
||||
| ordinary failed | 27 | **27** |
|
||||
|
||||
**The verb census is unchanged: 505 `UnmigratedVerb` aborts, the same 505, the same per-slot
|
||||
table** (`BindImageTexture` 138, `BeginTransformFeedback` 95, `DrawElements` 56, `DispatchCompute`
|
||||
49, `PatchParameteri` 43, `CopyImageSubData` 30, `MemoryBarrier` 18, `MultiDrawElementsBaseVertex`
|
||||
18, `ClearBufferiv` 12, `ClearBufferuiv` 8, `ClearBufferfv` 6, `DrawArraysInstancedBaseInstance` 4,
|
||||
…). The six entries that moved are exactly the six the P6 census listed as
|
||||
`Fatal{StageSnapshotTooNarrow, "respecify_whole_store"}` (two `LargeArenaAdoptionScenario`
|
||||
pairs, `MapPersistentRoundtrips.LargeArenaAdoptionScenario.AnAdoptionCostsExactlyOneMap`,
|
||||
`ResourceSubsystemControlScenario.ClearingTheP3aBits`), now PASSED - v1 round 3's M-3 fix
|
||||
(`1646049a`, "aborted six LargeArenaAdoption / ResourceSubsystemControl joint entries by name"),
|
||||
which landed on `feat/disaggregated` while c0b ran and is in the merge. Entry names identical
|
||||
between the two runs; no other entry changed status. **432 / 185 / 505 / 27 is therefore the
|
||||
baseline the four packages diff against**, and every abort in it is a verb abort.
|
||||
|
||||
---
|
||||
|
||||
## 3. The 25 rows and their carriers, in one table
|
||||
|
||||
`R` = an existing row; `N` = appended by c0b. Every payload is an inline POD in `SEG_CMD`; the
|
||||
two variable parts are named. No row owns a reply slot.
|
||||
|
||||
| # | GL slot (census) | pkg | row (op) | R/N | fixed payload | var part / carrier | sink | stamp |
|
||||
|---|---|---|---|---|---|---|---|---|
|
||||
| 1 | `DrawElements` (56 + every MC trace) | d1 | `draw_vbo` (59) | R | `MGPDrawInfo` 56 B | `MGPDrawRange[NumDraws]` in `SEG_CMD`; client-side indices: `MGHostSpan` tail naming a `SEG_STAGE` run | `OnDrawVbo` | `DrawArrays` |
|
||||
| 2 | `DrawElementsInstancedBaseVertex` (trace) | d1 | 59 | R | `InstanceCount`, `IndexBias` | as 1 | `OnDrawVbo` | `DrawArrays` |
|
||||
| 3 | `MultiDrawElementsBaseVertex` (18) | d1 | 59 | R | `NumDraws = drawcount` | `MGPDrawRange[drawcount]` | `OnDrawVbo` | `DrawArrays` |
|
||||
| 4 | `DrawArraysInstancedBaseInstance` (4) | d1 | 59 | R | `IndexSize 0`, `InstanceCount`, `StartInstance` | one range | `OnDrawVbo` | `DrawArrays` |
|
||||
| 5 | `DrawElementsBaseVertex` (4) | d1 | 59 | R | `IndexBias` | one range | `OnDrawVbo` | `DrawArrays` |
|
||||
| 6 | `MultiDrawArrays` (2) | d1 | 59 | R | `IndexSize 0`, `NumDraws` | ranges | `OnDrawVbo` | `DrawArrays` |
|
||||
| 7 | `MultiDrawElements` (2) | d1 | 59 | R | `NumDraws` | ranges | `OnDrawVbo` | `DrawArrays` |
|
||||
| 8 | `DrawArraysInstanced` (2) | d1 | 59 | R | `InstanceCount` | one range | `OnDrawVbo` | `DrawArrays` |
|
||||
| 9 | `MultiDrawElementsIndirect` (2) | d1 | 59 | R (tail N) | `Flags = kDrawIsIndirect`, `NumDraws 0` | `MGPDrawIndirect` (40 B) as the second tail, in `SEG_CMD` | `OnDrawVbo(…, indirect)` | `DrawArrays` |
|
||||
| 10 | `MultiDrawArraysIndirectCount` (2) | d1 | 59 | R (tail N) | as 9 + `ParameterBuffer`, `ParameterOffset`, `DrawCount = maxdrawcount` | `MGPDrawIndirect` | `OnDrawVbo` | `DrawArrays` |
|
||||
| 11 | `BindImageTexture` (138) | i1 | `bind_shader_image` (72) | N | `MGPImageBind` 40 B (GL args verbatim + handle) | — | `OnBindShaderImage` | `BindImageTexture` |
|
||||
| 12 | `DispatchCompute` (49) | i1 | `launch_grid` (60) | R | `MGPGridInfo` 48 B | — | `OnLaunchGrid` | `DispatchCompute` |
|
||||
| 13 | `CopyImageSubData` (30) | i1 | `resource_copy_region` (53) | R (72 B) | `MGPCopyRegion` + `SrcGlName/DstGlName` | — | `OnResourceCopyRegion` | `CopyImageSubData` (new stamp row) |
|
||||
| 14 | `MemoryBarrier` (18) | i1 | `memory_barrier` (61) | R | `MGPMemoryBarrier` 8 B | — | `OnMemoryBarrier` | `MemoryBarrier` |
|
||||
| 15 | `ShaderStorageBlockBinding` (4) | i1 | `set_storage_block_binding` (75) | N, `kHasBlob` | `MGPStorageBlockBinding` 40 B | the block name, `SEG_STAGE`, `Size = strlen + 1` | `OnSetStorageBlockBinding(…, name)` | `ShaderStorageBlockBinding` |
|
||||
| 16 | `BeginTransformFeedback` (95) | t2 | `begin_stream_output` (62) | R | `MGPStreamOutputBegin` 8 B | — | `OnBeginStreamOutput` | `BeginTransformFeedback` |
|
||||
| 17 | `PatchParameteri` (43) | t2 | `patch_parameter` (73) | N | `MGPPatchParameter` 8 B | — | `OnPatchParameter` | `PatchParameteri` |
|
||||
| 18 | `BindTransformFeedback` (2) | t2 | `bind_stream_output` (74) | N | `MGPStreamOutputBind` 16 B | — | `OnBindStreamOutput` | `BindTransformFeedback` |
|
||||
| 19 | `ClearBufferiv` (12) | f1 | `clear` (57) | R | `MGPClear` 48 B, `Kind/ValueClass` per the one spelling | — | `OnClear` (live) | `Clear` |
|
||||
| 20 | `ClearBufferuiv` (8) | f1 | 57 | R | as 19 | — | `OnClear` (live) | `Clear` |
|
||||
| 21 | `ClearBufferfv` (6) | f1 | 57 | R | as 19 | — | `OnClear` (live) | `Clear` |
|
||||
| 22 | `ClearBufferfi` (1) | f1 | 57 | R | as 19 | — | `OnClear` (live) | `Clear` |
|
||||
| 23 | `ClearNamedFramebufferfv` (2) | f1 | 57 | R | `Fbo` = the named handle (Named record precedes) | — | `OnClear`; unbound → refused by name | `Clear` |
|
||||
| 24 | `CopyTexImage2D` (3) | f1 | `copy_framebuffer_to_texture` (76) | N | `MGPCopyFromFramebuffer` 48 B | — | `OnCopyFramebufferToTexture` | `CopyTexImage2D` |
|
||||
| 25 | `GenerateMipmap` (2) | f1 | `generate_mipmap` (54) | R | `MGPMipPlan` 16 B, `Target` = GL enum | — | `OnGenerateMipmap` | `GenerateMipmap` |
|
||||
|
||||
Companions that share a row and whose slots the same package owns (unmeasured, no gate):
|
||||
`DrawElementsInstanced*` ×3, `DrawRangeElements*` ×2, `*Indirect` ×4 (d1);
|
||||
`DispatchComputeIndirect`, `MemoryBarrierByRegion` (i1); `End/Pause/ResumeTransformFeedback`
|
||||
(t2; `DeleteTransformFeedback` has no row); `ClearNamedFramebufferfi/iv/uiv`,
|
||||
`CopyTexSubImage2D` (f1).
|
||||
|
||||
---
|
||||
|
||||
## 4. Rulings I made, each with what would overturn it
|
||||
|
||||
The full list with evidence is CONTRACT-P5B.md §6; the ones a package will trip over:
|
||||
|
||||
1. **No P5b slot gains an `MGPipeApply*`; all 25 reach `WireVerbSink`.** The applier's 37
|
||||
entry points are the object/state families; a verb is a backend CALL, and the census's
|
||||
correction (the P5 five already go to the sink) is the pattern. Consequence: R-17's generated
|
||||
routing is untouched, `PipeCatalogueTest` still reads 33 installed + 4 escapes, and a
|
||||
migration is two files per slot. *Overturned by* a verb that has to be replayable from the
|
||||
monolith side through a table (P13's recorder).
|
||||
|
||||
2. **Rule D: a verb crosses AS THE CALL.** GL enums and GL names verbatim beside the handle;
|
||||
the sink reproduces the backend call; the backend keeps reading its verb class's
|
||||
BARRIER-PULLED fields. This is why nothing on the monolith path changes and why `rsp` grows -
|
||||
the phase's measurement, not a regression. *Overturned by* nothing in P5b; retiring the pulls
|
||||
is P3b/P4b/P7/P8's per CONTRACT-P5 table 2.
|
||||
|
||||
3. **Four of the five appended rows exist because the backend does work AT THE CALL** (Espryt:
|
||||
`SyncImageTextureBinding` at `DirectGLES.cpp:9154`, `glPatchParameteri` at `:8760`, the XFB
|
||||
object rebind at `:1390`, the driver-program rebinding at `:9200`) that no validate-time set
|
||||
record reproduces; the fifth (`copy_framebuffer_to_texture`) has a source no resource row can
|
||||
name. *Overturned by* the backend moving that work into its draw prep - G5-pinned regions,
|
||||
so not P5b's.
|
||||
|
||||
4. **`resource_copy_region` = `glCopyImageSubData` only**, and the framebuffer-sourced copies
|
||||
get row 76 - which also settles the stamp row `FieldOwnership.def` had left to "the phase
|
||||
that emits it".
|
||||
|
||||
5. **`kCapNeedsHostIndexBytes` stays 0 while the user-index span is ARMED.** The cap is the
|
||||
server asking for VBO-backed index bytes (P8's mirror); the span is the client handing over
|
||||
bytes only it has (the P8 resolve-on-client rule applied to the one case that needs it).
|
||||
CONTRACT-P5 table 0 conflated them because P5 produced neither. *Overturned by* a client
|
||||
index array larger than a `SEG_STAGE` run at 32 MiB (R-10's `maxrec=` would say).
|
||||
|
||||
6. **Renderbuffer copy endpoints, unbound DSA clears, and multi-draws with client-side indices
|
||||
are refused BY NAME** (ID-57's shape) - no barrier-pulled forward can serve them and none
|
||||
was measured. *Overturned by* a measured workload hitting one.
|
||||
|
||||
7. **The `copy-image-shadow-mirror` site is skipped under split** (behind
|
||||
`#if MOBILEGL_BUILD_DISAGGREGATED`, i1's edit); the two mipmap emulation sites stay Fatal
|
||||
(one may be re-derived from the descriptor by f1; the CPU fallback needs P9's
|
||||
`OnTextureWriteback`). *Overturned by* a measured `GetTexImage` of a copy destination before
|
||||
P9 - which is the wave-3 `GetTexImage` Fatal anyway.
|
||||
|
||||
8. **`GetProgramForDispatch` FATAL → BARRIER_PULLED "P7 (Magma), P8 (Espryt)"** (i1's one-row
|
||||
edit; granted). **`kCapBackendOwnsXfbCapture` (1<<9)** for the `EndTransformFeedback` probe
|
||||
at `GL_Drawing.cpp:1290` (t2's edit; granted). The only two c0b-file edits a package makes.
|
||||
|
||||
9. **The emit-table class-C list is partitioned per package** with five ownership
|
||||
`static_assert`s, so four parallel landings edit four disjoint regions and a package that
|
||||
touches another's list breaks the other's line. Arithmetic; nothing overturns it.
|
||||
|
||||
10. **`MGPCopyRegion` 64 → 72.** No producer, no consumer, no wire history; the names are the
|
||||
pull keys the P5b sink needs beside the handles P7 will use. *Overturned by* P7's
|
||||
handle-keyed endpoints, when the two fields become padding (and stay in the layout).
|
||||
|
||||
11. **Client-side vertex arrays are not a P5b question**: served on the server through the
|
||||
barrier-pulled `GetBoundVertexArray` (CONTRACT-P5 table 2 "P8"), correct under inproc by
|
||||
the barrier and the shared address space, retired by P8's `HostResolve.cpp`. *Overturned
|
||||
by* P6's spawn - which ID-68 puts after P5b.
|
||||
|
||||
12. **Companions ride for free but are not gated.** Nine draw variants, two compute/barrier
|
||||
variants, three XFB span controls and four clear/copy variants share a measured row; each
|
||||
package owns their slots (the ownership counts include them) and may flip them in the same
|
||||
change, but the census gate names only the 25.
|
||||
|
||||
---
|
||||
|
||||
## 5. What I believe the brief or the census got wrong or incomplete
|
||||
|
||||
1. **The brief's `IVerbSink` is `MG_Remote::Wire::WireVerbSink`** (declared in the codec header
|
||||
so the codec does not depend on the session), and its server implementation is
|
||||
`Server::ServerVerbSink`. Nothing else in the tree is named `IVerbSink`.
|
||||
2. **The census's "reuse `MGPipeApplySetShaderImages` / `MGPipeApplySetPatchState`" for
|
||||
`BindImageTexture` / `PatchParameteri` does not work**: both appliers write the working block
|
||||
at VALIDATE time and the backend slots do their work AT THE CALL (§4.3). Those two slots
|
||||
needed rows of their own, which is where two of the five appended rows come from. The
|
||||
census's "add/finish `MGPipeApplyLaunchGrid` / `MGPipeApplyDrawVbo` / `MGPipeApplyClear`"
|
||||
is likewise not the shape: the census's own correction (verbs reach the sink) is.
|
||||
3. **`ShaderStorageBlockBinding` was documented as "folded into the reflection archive"**
|
||||
(`PipeCalls.def` footer, `BackendObject.h`). The archive carries the bindings as LINKED;
|
||||
`glShaderStorageBlockBinding` moves one AFTER link and both backends apply it to an
|
||||
already-built program. It needed a row (75); the footer is corrected in place.
|
||||
4. **The 25 are 25 only if `DrawElementsInstancedBaseVertex` counts** (trace-only, 0 lane
|
||||
entries). The brief's d1 list has 10 with it; the census's integration table has 24 non-zero
|
||||
rows. Both are right; the contract lists all 25 by name.
|
||||
5. **`FieldOwnership.def`'s stamp map covered the pre-existing rows but not
|
||||
`ResourceCopyRegion`**, by design ("the phase that emits it adds its row"). Left to a
|
||||
package, an omitted stamp row is the ONE silent failure the table has (the record applies
|
||||
under the previous verb's serial). All six rows are landed here, not left to the packages.
|
||||
6. **`feat/disaggregated` moved twice while c0b ran** (c1g at `c9878d69` before I forked; v1
|
||||
round 3 at `f5614995` after). Merged; one four-line overlap in `PipeApplier.h`, no conflict.
|
||||
|
||||
---
|
||||
|
||||
## 6. Debts, for the packages and the phases after
|
||||
|
||||
- **`rsp` per Minecraft frame is unmeasured** until d1 renders one; each package reports it
|
||||
for its scenarios (PACKAGE-PREAMBLE §"五件事" 4).
|
||||
- **Renderbuffer `glCopyImageSubData` endpoints** need a renderbuffer forward or P7's handle
|
||||
keying (§4.6). **Unbound DSA clears** need Espryt's handle-keyed
|
||||
`SyncAndBindFramebufferObject` sibling (P7 / P3b-P4b, or f1 behind the option). **Multi-draw
|
||||
with client-side indices** needs P8's `HostResolve.cpp`.
|
||||
- **`DeleteTransformFeedback` has no row**: the driver object leaks on the server until P9's
|
||||
XFB namespace work; a bind of name 0 is what the backend does on delete of the bound one.
|
||||
- **`MGPGridInfo::Block*` is 0 in P5b**; P7 Magma fills it from the reflection archive if it
|
||||
needs it.
|
||||
- **The `Fatal` NAME a pull prints under a P5b draw is `DrawArrays`** for all twenty draw verbs
|
||||
(one stamp row, one kDraw mask). The wire's `UnmigratedVerb` names are exact; the poison's
|
||||
`@DrawArrays` is not, and d1's reports should say so once rather than be surprised.
|
||||
- **P5's `.text` baseline moved** between c0 (10806323) and this head (10806611) with the
|
||||
landed P5 packages; `p5-before-libMobileGL.so` (`.READY` `4595163c`) is the comparand and
|
||||
the report's G1 is against it.
|
||||
@@ -0,0 +1,87 @@
|
||||
# P5b Codex APK / Redmi runner handoff
|
||||
|
||||
2026-09-16. Dedicated scripts; original campaign scripts and evidence remain intact.
|
||||
|
||||
## Build state
|
||||
|
||||
Final source head: `82683d4a83b5c8f4f7c375a7a92ebee9a1854646`, including both closing-review fixes. The first real build started at 2026-09-16T14:12:42-04:00 in phase `p5bcodex1` and was interrupted by an entire WSL restart at approximately 14:14 (retained exec session exited 1; uptime reset to 0 minutes; all native/Gradle processes disappeared). Its output directory, build log and native cache are preserved. Pull reached `mergeTraceReleaseJniLibFolders`; no signed APK/proof was produced.
|
||||
|
||||
State: waiting for the root task's memory/concurrency coordination. The next authorized build will use **p5bcodex2**, the same exact source head and dedicated APK tree, and a fresh output directory. The commands below name that next phase; they must not be run until its three APK proofs exist.
|
||||
|
||||
Build tree: `/home/swung/w7/p5b-apk-codex`; output: `/home/swung/w7/notes/p5b/apk/p5bcodex2`.
|
||||
|
||||
The wrapper `notes/tools/p5b_codex_build_apks.sh` requires `--head` with the full SHA and a clean dedicated checkout. It shares the old `wsl_build_p5_apks.lock`, refuses an existing output directory, retains each arm's build log and proofs, and builds pull/push/split sequentially. `P5B_APK_JOBS=4` is the default. A generated Gradle init script also sets CMake compile/link job pools, because AGP invokes Ninja directly. No tracked source build file is modified.
|
||||
|
||||
In WSL Arch, after the root task has moved the clean APK tree to the final source commit:
|
||||
|
||||
```bash
|
||||
HEAD_SHA=82683d4a83b5c8f4f7c375a7a92ebee9a1854646
|
||||
test "$(git -C /home/swung/w7/p5b-apk-codex rev-parse HEAD)" = "$HEAD_SHA"
|
||||
P5B_APK_JOBS=4 bash /home/swung/w7/notes/tools/p5b_codex_build_apks.sh --head "$HEAD_SHA" --phase p5bcodex2 /home/swung/w7/p5b-apk-codex
|
||||
```
|
||||
|
||||
Prerequisites: local hydrated `~/w7/pipe` dependencies, JDK 21 at `/usr/lib/jvm/java-21-openjdk`, SDK at `~/android-sdk`, Gradle 8.10.2 and its offline dependencies. The native build must be optimized `RelWithDebInfo`, arm64 only. Each proof checks source SHA, embedded Git stamp, package/signing/ABI, push/remote symbol identities, and byte-for-byte equality between the packaged library and stripped build library. The three outputs are `trace-pull.apk`, `trace-push.apk`, and `trace-split.apk`; splitctl reuses the split APK with monolith runtime.
|
||||
|
||||
## Windows Git Bash setup and prerequisites
|
||||
|
||||
Run in Windows **Git Bash**, not WSL, since the existing Android runner uses Windows Python/adb. The root task owns all device operations and synchronization of the clean Windows source tree.
|
||||
|
||||
```bash
|
||||
TREE=/c/Users/geekerwan/AndroidStudioProjects/FoldCraftLauncher/MobileGL-disagg
|
||||
TOOLS=//wsl.localhost/Arch/home/swung/w7/notes/tools
|
||||
APKS=//wsl.localhost/Arch/home/swung/w7/notes/p5b/apk/p5bcodex2
|
||||
FIXTURES=//wsl.localhost/Arch/home/swung/w7/pipe/tools/trace_replay/fixtures
|
||||
OUT_ROOT=/c/Users/geekerwan/AndroidStudioProjects/FoldCraftLauncher/MobileGL/.trace-work/p5b-redmi
|
||||
PHASE=p5bcodex2
|
||||
cd "$TREE"
|
||||
command -v bash git python adb
|
||||
python --version
|
||||
HEAD_SHA=$(python -c 'import json,sys; print(json.load(open(sys.argv[1]))["source_head"])' "$APKS/trace-pull.proof.json")
|
||||
test "$(git rev-parse HEAD)" = "$HEAD_SHA"
|
||||
git diff --quiet --ignore-submodules=all HEAD --
|
||||
for arm in pull push split; do
|
||||
test -s "$APKS/trace-$arm.apk" && test -s "$APKS/trace-$arm.proof.json" || exit 1
|
||||
done
|
||||
python "$TOOLS/p5b_codex_runner_selftest.py"
|
||||
```
|
||||
|
||||
All four fixture archives and their manifest goldens must already be hydrated. The adapter validates their existence/content and refuses Git LFS pointer files; it never fetches LFS. Case names:
|
||||
|
||||
- `improved-transparency-minecraft-26.3`
|
||||
- `minecraft-1.21.4-in-world`
|
||||
- `minecraft-1.21.4-fabric-sodium-in-world`
|
||||
- `minecraft-1.21.4-fabric-iris-bsl-in-world`
|
||||
|
||||
Dry runs perform no device operations, locking, builds, or file writes:
|
||||
|
||||
```bash
|
||||
bash "$TOOLS/p5b_codex_redmi.sh" --head "$HEAD_SHA" --phase "$PHASE" --mode correctness --tree "$TREE" --out-root "$OUT_ROOT" --dry-run
|
||||
bash "$TOOLS/p5b_codex_redmi.sh" --head "$HEAD_SHA" --phase "$PHASE" --mode benchmark --tree "$TREE" --out-root "$OUT_ROOT" --dry-run
|
||||
```
|
||||
|
||||
Expect `DRY_GROUPS=8` and `DRY_GROUPS=32`. The target is fixed at Redmi serial `2f7cbe2e`, which must support adb and `su` after reboot. The existing pin script must be readable at `//wsl.localhost/Arch/home/swung/w7/notes/p2/devices/pin_device.sh`. The session uses the existing campaign lock at `/c/Users/geekerwan/AppData/Local/Temp/claude/C--Users-geekerwan-AndroidStudioProjects-FoldCraftLauncher-MobileGL/8ab6b3cb-c783-4826-9fd7-39ff3498af5b/scratchpad/w7/locks/2f7cbe2e`; it waits at most 900 seconds and never breaks another owner's lock. No other campaign should operate that device concurrently.
|
||||
|
||||
## Device commands (root task only)
|
||||
|
||||
```bash
|
||||
bash "$TOOLS/p5b_codex_redmi.sh" --head "$HEAD_SHA" --phase "$PHASE" --mode correctness --tree "$TREE" --out-root "$OUT_ROOT"
|
||||
```
|
||||
|
||||
Correctness must finish with 8/8 groups and `session-exit-code.txt` equal to 0. Each case/backend requires an actual fresh PNG, the manifest target call and SSIM threshold (including improved-transparency 0.995), successful replay, and private-log proof of inproc, apply thread, and positive wire records. A skip is failure. Then:
|
||||
|
||||
```bash
|
||||
python "$TOOLS/p5b_codex_matrix.py" require-correctness "$OUT_ROOT/$PHASE/2f7cbe2e/correctness" --head "$HEAD_SHA" --phase "$PHASE"
|
||||
bash "$TOOLS/p5b_codex_redmi.sh" --head "$HEAD_SHA" --phase "$PHASE" --mode benchmark --tree "$TREE" --out-root "$OUT_ROOT"
|
||||
```
|
||||
|
||||
Benchmark is four arms × two backends × four cases = 32 groups, with three repeats each and the existing nofinish/last-200-frame protocol. The reducer retains the previous lowest-mean-repeat selection and reports p50/p99, CPU, mapped views and available memory evidence. No timing delta is a pass/fail threshold. Failed replay, missing evidence, short samples or pin drift invalidate a group and exclude it from paired metrics.
|
||||
|
||||
All arms explicitly use `MOBILEGL_IPC_STAGE_MB=256`, the authorized P5b exit profile for the 128 MiB improved-transparency upload. This leaves the production 32 MiB default unchanged. Stage mapping is 256 MiB per role view / 512 MiB summed inproc views; this is not an inferred RSS increase. Other controls are `MOBILEGL_IPC_VERB_BARRIER=1`, stats enabled, and transport inproc for split / monolith for pull, push and splitctl. rd12 is excluded from this matrix.
|
||||
|
||||
The session reboots, fixes fan level 2, checks temperature below 40000 mC, pins/checks before and after each group, and cleans up unpin/fan0/stayon-off before releasing its own lock. Each retry uses a new attempt directory; rerunning a completed mode requires a new `--out-root` or phase. Do not remove old evidence to rerun.
|
||||
|
||||
Outputs are under `$OUT_ROOT/$PHASE/2f7cbe2e/{correctness,benchmark}`: `matrix.json`, `matrix.md`, cleanup exit, per-group attempt/runner exit and pin evidence; benchmark also produces `ab-tables.md`. Per-repeat originals include result JSON, private `mobilegl.log`, screenshots and benchmark JSON. Missing/failed evidence cannot produce a green matrix.
|
||||
|
||||
## Runner checks already performed
|
||||
|
||||
Both shell scripts pass `bash -n`; Python modules compile. Real Windows Git Bash dry runs produced exactly 8/32 groups, and the Windows adapter imports the existing Windows runner correctly. Nine artifact-input selftests passed on Windows and WSL (skip, SSIM, missing/truncated PNG, split proof, wrong splitctl runtime, stale directories, short benchmark). Logs: `~/w7/p5b-codex-redmi-dry-{correctness,benchmark}.log` and `~/w7/p5b-codex-runner-selftest.log`.
|
||||
@@ -0,0 +1,124 @@
|
||||
# d1 Codex handoff completion — 2026-09-16
|
||||
|
||||
Source d1 commits: `e5649396..fbe17d06` (base `37fc4fdb`); preserved from the clean Windows
|
||||
worktree through a local bundle. Integrated with f1/i1/t2/r2 at `9606466a`.
|
||||
No d1 implementation rewrite was needed. Existing nineteen slots emit draw_vbo; Linux
|
||||
validation completes the missing evidence. Windows STATIC trace intersections are explicitly
|
||||
not execution evidence, and their SetSwapInterval/FenceSync predictions do not describe the
|
||||
actual retracer path.
|
||||
|
||||
## Validation
|
||||
|
||||
- Linux split build succeeded (clang, Release, ccache, -j8).
|
||||
- Split unit: 2098/2098, set -e script progressed to integration; no Windows timeout debt.
|
||||
- integration-split: 54/54, including 32 d1 additions; two pre-existing map-arm skips.
|
||||
- No repeated full gate or review. Integrator runs four-lane quick checks on the combined head.
|
||||
- Dynamic Minecraft census: 77 backend cases, 39 fixtures (iterationRP DirectVulkan only).
|
||||
All entries reached rendering or a non-draw first blocker. Results below are real inproc
|
||||
execution, not static predictions. Source tree was merged while the census ran, but the
|
||||
d1 binary was held unchanged until every entry completed.
|
||||
|
||||
Binary SHA256: `aeee247d77f56622fcb622fbdb7ad4a636d2f21268a858edc287572a6091df6a`.
|
||||
|
||||
Outcome counts: {'UnmigratedVerb, "BlitNamedFramebuffer"': 31, 'PASS': 28, 'UnmigratedVerb, "DispatchCompute"': 2, 'UnmigratedVerb, "GenerateMipmap"': 7, 'UnmigratedVerb, "BindImageTexture"': 1, 'RingOverrun, "SEG_STAGE"': 2, 'UnmigratedVerb, "MemoryBarrier"': 2, 'UnmigratedVerb, "ClearBufferfv"': 2, 'InitialBytesNotCarried, "resource_respecify"': 1, 'BarrierTimeout, "Present"': 1}
|
||||
|
||||
Evidence: `/home/swung/w7/retrace-out/p5b-d1-codex/summary.json`; each `actual_private_log`
|
||||
points to its isolated retrace output directory. Console summaries are in
|
||||
`/home/swung/w7/p5b-d1-codex-trace-logs/`.
|
||||
|
||||
## First blocker by fixture and backend
|
||||
|
||||
| Fixture | Backend | Result / first blocker | SSIM |
|
||||
|---|---|---|---|
|
||||
| improved-transparency-minecraft-26.3 | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| improved-transparency-minecraft-26.3 | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.17-main-menu-854 | DirectGLES | PASS | 0.999962 |
|
||||
| minecraft-1.17-main-menu-854 | DirectVulkan | PASS | 0.999961 |
|
||||
| minecraft-1.21.1-neoforge-create-indirect-in-world | DirectGLES | UnmigratedVerb, "DispatchCompute" | — |
|
||||
| minecraft-1.21.1-neoforge-create-indirect-in-world | DirectVulkan | UnmigratedVerb, "DispatchCompute" | — |
|
||||
| minecraft-1.21.1-neoforge-create-instancing-in-world | DirectGLES | PASS | 0.999958 |
|
||||
| minecraft-1.21.1-neoforge-create-instancing-in-world | DirectVulkan | PASS | 0.999952 |
|
||||
| minecraft-1.21.11-main-menu | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.11-main-menu | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-common-mods-in-world | DirectGLES | UnmigratedVerb, "GenerateMipmap" | — |
|
||||
| minecraft-1.21.4-fabric-common-mods-in-world | DirectVulkan | UnmigratedVerb, "GenerateMipmap" | — |
|
||||
| minecraft-1.21.4-fabric-common-mods-inventory | DirectGLES | UnmigratedVerb, "GenerateMipmap" | — |
|
||||
| minecraft-1.21.4-fabric-common-mods-inventory | DirectVulkan | UnmigratedVerb, "GenerateMipmap" | — |
|
||||
| minecraft-1.21.4-fabric-iris-bliss-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-bliss-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-esc-menu-854 | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-esc-menu-854 | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-chocapic-v6-lite-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-chocapic-v6-lite-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-reimagined-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-reimagined-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-unbound-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-unbound-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-derivative-main-d24.4.14-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-derivative-main-d24.4.14-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-iterationrp-in-world | DirectVulkan | UnmigratedVerb, "BindImageTexture" | — |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-in-world | DirectGLES | RingOverrun, "SEG_STAGE" | — |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-nodsa-in-world | DirectGLES | RingOverrun, "SEG_STAGE" | — |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-nodsa-in-world | DirectVulkan | UnmigratedVerb, "GenerateMipmap" | — |
|
||||
| minecraft-1.21.4-fabric-iris-makeup-ultrafast-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-makeup-ultrafast-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-mellow-in-world | DirectGLES | UnmigratedVerb, "MemoryBarrier" | — |
|
||||
| minecraft-1.21.4-fabric-iris-mellow-in-world | DirectVulkan | UnmigratedVerb, "MemoryBarrier" | — |
|
||||
| minecraft-1.21.4-fabric-iris-nostalgia-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-nostalgia-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.1-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.1-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.3b-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.3b-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-sundial-lite-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-sundial-lite-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-super-duper-vanilla-in-world | DirectGLES | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-iris-super-duper-vanilla-in-world | DirectVulkan | UnmigratedVerb, "BlitNamedFramebuffer" | — |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world-normal-world | DirectGLES | PASS | 0.999979 |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world-normal-world | DirectVulkan | PASS | 0.999979 |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world | DirectGLES | PASS | 0.999984 |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world | DirectVulkan | PASS | 0.999984 |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory-normal-world | DirectGLES | UnmigratedVerb, "ClearBufferfv" | — |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory-normal-world | DirectVulkan | UnmigratedVerb, "ClearBufferfv" | — |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory | DirectGLES | UnmigratedVerb, "GenerateMipmap" | — |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory | DirectVulkan | UnmigratedVerb, "GenerateMipmap" | — |
|
||||
| minecraft-1.21.4-fabric-rei-inventory-normal-world | DirectGLES | PASS | 0.999988 |
|
||||
| minecraft-1.21.4-fabric-rei-inventory-normal-world | DirectVulkan | PASS | 0.999988 |
|
||||
| minecraft-1.21.4-fabric-rei-inventory | DirectGLES | PASS | 0.99999 |
|
||||
| minecraft-1.21.4-fabric-rei-inventory | DirectVulkan | PASS | 0.99999 |
|
||||
| minecraft-1.21.4-fabric-sodium-in-world | DirectGLES | PASS | 0.999986 |
|
||||
| minecraft-1.21.4-fabric-sodium-in-world | DirectVulkan | PASS | 0.999986 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world-normal-world | DirectGLES | PASS | 0.999992 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world-normal-world | DirectVulkan | PASS | 0.999992 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world | DirectGLES | PASS | 0.999999 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world | DirectVulkan | PASS | 0.999999 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world-normal-world | DirectGLES | PASS | 0.999995 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world-normal-world | DirectVulkan | PASS | 0.999995 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world | DirectGLES | PASS | 1.0 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world | DirectVulkan | PASS | 1.0 |
|
||||
| minecraft-1.21.4-in-world | DirectGLES | PASS | 0.999976 |
|
||||
| minecraft-1.21.4-in-world | DirectVulkan | PASS | 0.999976 |
|
||||
| minecraft-1.21.4-main-menu | DirectGLES | PASS | 0.997009 |
|
||||
| minecraft-1.21.4-main-menu | DirectVulkan | PASS | 0.997009 |
|
||||
| minecraft-1.21.4-rd12-odinlite-in-world | DirectGLES | InitialBytesNotCarried, "resource_respecify" | — |
|
||||
| minecraft-1.21.4-rd12-odinlite-in-world | DirectVulkan | BarrierTimeout, "Present" | — |
|
||||
| minecraft-1.21.4-startup | DirectGLES | PASS | 1.0 |
|
||||
| minecraft-1.21.4-startup | DirectVulkan | PASS | 1.0 |
|
||||
|
||||
## Rulings and follow-up
|
||||
|
||||
Plain base-vertex-zero indexed draws dispatch to DrawElements, rather than the optional
|
||||
DrawElementsBaseVertex backend entry, preserving the monolith path on ES 3.1/ANGLE. This is
|
||||
the d1 implementation's documented correction to CONTRACT-P5B's original sink pseudocode.
|
||||
Client-index multi-draw remains the named `+CLIENT_INDICES` refusal pending P8; it is not a
|
||||
Minecraft trace blocker. GuiBatch was not in the standalone d1 split registration because
|
||||
its next MemoryBarrier belongs to i1.
|
||||
|
||||
Vanilla and sodium target traces already render (both backends); iris-bsl and
|
||||
improved-transparency now stop at BlitNamedFramebuffer. The integrator owns that migration.
|
||||
Sync migration is a separate next commit and contract appendix. rd12 DirectGLES hits
|
||||
InitialBytesNotCarried/resource_respecify; DirectVulkan times out at Present. Neither result
|
||||
is the historical dev scudo crash.
|
||||
@@ -0,0 +1,26 @@
|
||||
# P5 / P5b docs close-out, Codex
|
||||
|
||||
Date: 2026-09-16. Branch `p5b/docs-codex`, independent worktree `/home/swung/w7/p5b-docs-codex`, base `2deca994`.
|
||||
Commit: `dab238938eb55eec993d90ce1f5a9b753d8712bb` — `[Docs] (Disaggregated): record P5 final gate stop and scoped control evidence`.
|
||||
Only `docs/Disaggregated/{README,ROADMAP,MEASUREMENTS}.md` changed: 102 insertions / 32 deletions. No source, integration worktree, main progress note, device, build or test changes. No repeated tests or adversarial review.
|
||||
|
||||
## What changed
|
||||
|
||||
- README and ROADMAP now distinguish landed P5 code, the historical reduced IPC-frame milestone, the stopped final-head gate, and the P5b real-workload exit. P5.5 is renamed P5b. P6 remains after P5b.
|
||||
- MEASUREMENTS §27 retains historical joint readings and adds x2, v1-r3 and r2 corrections. Old 511 aborts are 505 class-C plus six P5 StageSnapshotTooNarrow defects fixed by v1-r3. The 27 wrong answers are 22 texture readbacks, three queries, one inspection and one FBO/RBO ReadPixels lifetime path; the last pixel causality remains unisolated.
|
||||
- §28 replaces maxrec pending with measured 784 B SetVertexAttribDefaults on reduced/OpenRA, the default 4 MiB cap and the SmallRing ledger. Old x2 ringwaits was reclaim evidence, not an actual wait; r2's 1/8 MiB wait/wrap controls are separately attributed. Existing rsp is a lower bound and inproc RSS remains one shared-process measurement.
|
||||
- §30 records Redmi A/B runner `eec0e836`, APK source `e61d0012`, 40 groups / 27 completed / 13 failed; all ten split groups abort before benchmark. Five concise rows preserve useful frame-time readings. No barrier-tax estimate is invented; full tables and per-thread data remain in ab-v1.md.
|
||||
- New §31 records `37fc4fdb` final gate: builds/G1/unit/reduced ran, E1 own diagnostics passed, then E3(a) 4 pixel failures + 2 designed skips was correctly rejected. Part 2 and Part 4 were not reached. Package, historical joint and APK evidence is not relabeled final merged-head proof.
|
||||
- r2 corrections #5–13 and package acceptance are summarized, including exact library restore, four own E3 pixel/private-log reds, all-skip rejection, REQUIRE_GPU, fresh/nonempty census, production G5 pins and the current-name c1f 15/15 mutation run. No new G1 measurement is claimed for r2. ROADMAP names r1's four remaining corrections, non-verb rsp, illegal reply status and tree-external runner propagation.
|
||||
|
||||
## Evidence read
|
||||
|
||||
Under `notes/p5/p5-results`: joint-v1.md, x2-v1.md, x2m-v1.md, j0-v3.md, v1-v3.md, ab-v1.md, ap-v1.md, apkci-v1.md, p5-close-codex-review.md. Under `notes/p5b/p5b-results`: r2-v1.md. Also `/home/swung/w7/p5-final-gate.log` (741 lines at documentation time), which pins its own source HEAD on line 1 and ends with E3(a)'s selected-skip failure.
|
||||
|
||||
Final-gate log SHA-256: `51d15bd11d06851fa9f7934eeac1a162f8db45c8b81dc345dd76e77701c7531c`.
|
||||
|
||||
## Validation and handoff
|
||||
|
||||
`git diff --check` clean; UTF-8/replacement-character and final-newline checks on all three edited files passed. Changes were checked against the cited report tables and the final gate's stopping point. Committer identity before commit: `Swung0x48 <swung0x48@outlook.com>`; single-line commit without trailers; final tree clean. No binary testing was needed for this documentation-only change.
|
||||
|
||||
Integrator can cherry-pick `dab23893` onto the evolving merge head. This report does not claim r1 is currently fixed or a later integration gate is complete; those need their own report/head. Existing §26 historical numbers and architecture text are retained. The main progress MD is owned by the integrator and was not edited.
|
||||
@@ -0,0 +1,18 @@
|
||||
# P5b final documentation draft — Codex
|
||||
|
||||
State: prepared, uncommitted; waiting for DEVICE_DONE + final sourcehead and final host/trace counts.
|
||||
Base: 348d22a4b9c15cc571af840dedc4b4edc8c53dbb.
|
||||
Worktree: /home/swung/w7/p5b-docs-final-codex, branch p5b/docs-final-codex.
|
||||
Files: docs/Disaggregated/README.md, ROADMAP.md, MEASUREMENTS.md only.
|
||||
|
||||
The draft records migrated slots A=2/B=54/C=15; separates package measurements, the fixed-head
|
||||
one-time host gate, census, phase review/targeted repairs, and final device identity/results.
|
||||
Host JUnit selected/passed/skipped counts are recorded individually; its live-runner shell
|
||||
anomaly is retained. Integration census is 1267 selected: 811 passed,203 skipped,62 aborted,
|
||||
191 failed; baseline 432 passes all retained, no removed old names. Trace results remain pending.
|
||||
The real-workload profile is stage 256 MiB across traces/device arms; production default 32 MiB
|
||||
and large-single-blob boundary are explicit. P5b remains in progress, P6 remains after its exit.
|
||||
|
||||
The shared joint report's CODEX CENSUS markers are untouched. Only CODEX HOST DOCS was appended.
|
||||
No builds, tests, new reviews, adb calls or integration tree changes were made for this task.
|
||||
git diff --check passed on the draft. No commit until the integrator's final evidence signal.
|
||||
@@ -0,0 +1,144 @@
|
||||
# f1-v1 — clear family, framebuffer copies, mipmaps
|
||||
|
||||
Base: `f5676103` (c0b); branch `p5b/f1`, tree `/home/swung/w7/p5b-f1`.
|
||||
|
||||
## Implementation and proof
|
||||
|
||||
All eleven slots move C → B. The client calls `BeforeReadOnlyVerb()` before `EmitAndWait`; frontend framebuffer/texture publications remain ahead of the verb. No opcode, carrier, generator, backend or monolith implementation changed. Copy/mipmap records dispatch on the apply thread through the existing codec arms. Rule D remains: GL arguments travel verbatim beside resource handles, and backend barrier-pulled reads remain.
|
||||
|
||||
| Slot | What crosses | Proof |
|
||||
|---|---|---|
|
||||
| `ClearBufferfi` | Null bound-draw handle, DepthStencil kind, depth/stencil values, drawbuffer index | `RemoteF1.ClearBufferfiFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearBufferfiPixels` |
|
||||
| `ClearBufferfv` | Null bound-draw handle, kind, Float/Int/Uint value class, drawbuffer index, exact color bits (or scalar depth/stencil) | `RemoteF1.ClearBufferfvFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearBufferfvPixels` |
|
||||
| `ClearBufferiv` | Null bound-draw handle, kind, Float/Int/Uint value class, drawbuffer index, exact color bits (or scalar depth/stencil) | `RemoteF1.ClearBufferivFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearBufferivPixels` |
|
||||
| `ClearBufferuiv` | Null bound-draw handle, kind, Float/Int/Uint value class, drawbuffer index, exact color bits (or scalar depth/stencil) | `RemoteF1.ClearBufferuivFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearBufferuivPixels` |
|
||||
| `ClearNamedFramebufferfi` | Named framebuffer handle, DepthStencil kind, depth/stencil values, drawbuffer index | `RemoteF1.ClearNamedFramebufferfiFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearNamedFramebufferfiPixels` |
|
||||
| `ClearNamedFramebufferfv` | Named framebuffer handle, kind, Float/Int/Uint value class, drawbuffer index, exact color bits (or scalar depth/stencil) | `RemoteF1.ClearNamedFramebufferfvFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearNamedFramebufferfvPixels` |
|
||||
| `ClearNamedFramebufferiv` | Named framebuffer handle, kind, Float/Int/Uint value class, drawbuffer index, exact color bits (or scalar depth/stencil) | `RemoteF1.ClearNamedFramebufferivFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearNamedFramebufferivPixels` |
|
||||
| `ClearNamedFramebufferuiv` | Named framebuffer handle, kind, Float/Int/Uint value class, drawbuffer index, exact color bits (or scalar depth/stencil) | `RemoteF1.ClearNamedFramebufferuivFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.ClearNamedFramebufferuivPixels` |
|
||||
| `CopyTexImage2D` | Bound destination texture handle, GL target, level, source rectangle, image format or sub-image offsets and discriminator | `RemoteF1.CopyTexImage2DFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.CopyTexImage2DPixels` |
|
||||
| `CopyTexSubImage2D` | Bound destination texture handle, GL target, level, source rectangle, image format or sub-image offsets and discriminator | `RemoteF1.CopyTexSubImage2DFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.CopyTexSubImage2DPixels` |
|
||||
| `GenerateMipmap` | Bound texture handle, GL target, base level and level count | `RemoteF1.GenerateMipmapFieldsCross`; `DirectGLES.Split.F1.F1WireScenario.GenerateMipmapPixels` |
|
||||
|
||||
The pixel cases require a live split session and an advancing emitted ordinal. Integer and float clears are read back from matching attachment types; fi checks both packed depth and stencil; the named variants clear the bound named framebuffer. Copies read back the destination, and mip generation reads level 2 after clearing only level 0. Four additional `RemoteF1.UnboundNamed{fv,iv,uiv,fi}RefusesByName` cases assert each exact refusal string.
|
||||
|
||||
## Executed negative controls
|
||||
|
||||
`/home/swung/w7/p5b-f1-red.log`: **26/26 fail**, with their own failure strings; all mutations reverted before the final gate. Each new test carries a matching “Red once” source comment.
|
||||
|
||||
- `ClearBufferfiFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearBufferfi.fields`.
|
||||
- `ClearBufferfvFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearBufferfv.fields`.
|
||||
- `ClearBufferivFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearBufferiv.fields`.
|
||||
- `ClearBufferuivFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearBufferuiv.fields`.
|
||||
- `ClearNamedFramebufferfiFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferfi.fields`.
|
||||
- `ClearNamedFramebufferfvFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferfv.fields`.
|
||||
- `ClearNamedFramebufferivFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferiv.fields`.
|
||||
- `ClearNamedFramebufferuivFieldsCross`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferuiv.fields`.
|
||||
- `CopyTexImage2DFieldsCross`: red once by incrementing the emitted copy level; asserted `F1.CopyTexImage2D.fields`.
|
||||
- `CopyTexSubImage2DFieldsCross`: red once by incrementing the emitted copy level; asserted `F1.CopyTexSubImage2D.fields`.
|
||||
- `GenerateMipmapFieldsCross`: red once by incrementing the emitted base level; asserted `F1.GenerateMipmap.fields`.
|
||||
- `ClearBufferfiPixels`: red once by zeroing the clear values; asserted `F1.ClearBufferfi.pixels`.
|
||||
- `ClearBufferfvPixels`: red once by zeroing the clear values; asserted `F1.ClearBufferfv.pixels`.
|
||||
- `ClearBufferivPixels`: red once by zeroing the clear values; asserted `F1.ClearBufferiv.pixels`.
|
||||
- `ClearBufferuivPixels`: red once by zeroing the clear values; asserted `F1.ClearBufferuiv.pixels`.
|
||||
- `ClearNamedFramebufferfiPixels`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferfi.pixels`.
|
||||
- `ClearNamedFramebufferfvPixels`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferfv.pixels`.
|
||||
- `ClearNamedFramebufferivPixels`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferiv.pixels`.
|
||||
- `ClearNamedFramebufferuivPixels`: red once by zeroing the clear values; asserted `F1.ClearNamedFramebufferuiv.pixels`.
|
||||
- `CopyTexImage2DPixels`: red once by omitting the copy sink call; asserted `F1.CopyTexImage2D.pixels`.
|
||||
- `CopyTexSubImage2DPixels`: red once by omitting the copy sink call; asserted `F1.CopyTexSubImage2D.pixels`.
|
||||
- `GenerateMipmapPixels`: red once by omitting the mipmap sink call; asserted `F1.GenerateMipmap.pixels`.
|
||||
- Each of the four unbound named-clear controls: red once by disabling the handle refusal; the required exact `Fatal{UnmigratedVerb, "ClearNamedFramebuffer<suffix>+UNBOUND"}` disappears.
|
||||
|
||||
## Census: original 1,149-entry registry
|
||||
|
||||
Command: `bash ~/w7/p5b-c0b-census.sh ~/w7/p5b-f1 ~/w7/p5b-f1-census-logs`. Evidence: that directory’s `results.json`, per-entry `.out`/`.log`; summary `/home/swung/w7/p5b-f1-census.log`. Compared by name with `/home/swung/w7/p5b-c0b-census-logs/results.json`.
|
||||
|
||||
Baseline **432 passed / 185 skipped / 505 aborted / 27 failed** → **460 passed / 185 skipped / 473 aborted / 31 failed**. All **34 bare f1 first-slot aborts → 0**. Exactly those 34 entries change: 28 pass, 2 reach the qualified unbound refusal, 4 reach assertions. No other original entry changes status or normalized Fatal.
|
||||
|
||||
| Entry | Previous first slot | New first blocker/result |
|
||||
|---|---|---|
|
||||
| `DirectGLES.DepthStencilReadbackMatrixScenario.CopyTexImageFromADepthAttachmentSurvivesAReadBack` | `CopyTexImage2D` | FAIL: 3072/3072 copied depth values wrong, first word `0x0` (`DepthStencilReadbackMatrixScenario.cpp:568`) |
|
||||
| `DirectGLES.ForcedDepthStencilEmulation.DepthStencilReadbackMatrixScenario.CopyTexImageFromADepthAttachmentSurvivesAReadBack` | `CopyTexImage2D` | FAIL: 3072/3072 copied depth values wrong, first word `0x0` (`DepthStencilReadbackMatrixScenario.cpp:568`) |
|
||||
| `DirectGLES.IntegerBorderColorScenario.ANegativeBorderOnAnUnsignedFormatClampsToTheFormatsMaximum` | `ClearBufferiv` | PASS |
|
||||
| `DirectGLES.IntegerBorderColorScenario.ASignedIntegerBorderIsClampedToTheFormatsRepresentableRange` | `ClearBufferiv` | PASS |
|
||||
| `DirectGLES.IntegerBorderColorScenario.AnOversizedUnsignedBorderIsClampedToTheFormatsMaximum` | `ClearBufferiv` | PASS |
|
||||
| `DirectGLES.IntegerBorderColorScenario.InsideTexelsAreUnaffectedByTheBorderColour` | `ClearBufferiv` | PASS |
|
||||
| `DirectGLES.IntegerBorderColorScenario.SamplerParameterIivBorderColourSurvivesToAnIntegerSampler` | `ClearBufferiv` | PASS |
|
||||
| `DirectGLES.IntegerBorderColorScenario.TexParameterIivBorderColourSurvivesToAnIntegerSampler` | `ClearBufferiv` | PASS |
|
||||
| `DirectGLES.P4aFinalFixScenario.AFramebufferDeletedAfterADsaClearLeavesItsAttachmentIntact` | `ClearNamedFramebufferfv` | `Fatal{UnmigratedVerb, "ClearNamedFramebufferfv+UNBOUND"}` |
|
||||
| `DirectGLES.P4aFinalFixScenario.GenerateMipmapAcrossAnUnrelatedDraw` | `GenerateMipmap` | PASS |
|
||||
| `DirectGLES.P4aFinalFixScenario.GenerateMipmapImmediately` | `GenerateMipmap` | PASS |
|
||||
| `DirectGLES.SampledSetStalenessScenario.AQueuedClearIsMaterialisedWhenAFilterChangeCompletesTheTexture` | `ClearBufferfv` | PASS |
|
||||
| `DirectGLES.SampledSetStalenessScenario.AQueuedClearIsMaterialisedWhenASamplerObjectCompletesTheTexture` | `ClearBufferfv` | PASS |
|
||||
| `DirectGLES.SwizzleAccessRoutineScenario.EveryAccessRoutineFetchesTheSameTexelUnderTheIdentitySwizzle` | `ClearBufferuiv` | PASS |
|
||||
| `DirectGLES.SwizzleAccessRoutineScenario.EveryAccessRoutineSeesAReversedSwizzle` | `ClearBufferuiv` | PASS |
|
||||
| `DirectGLES.SwizzleAccessRoutineScenario.EveryAccessRoutineSeesConstantSwizzleSources` | `ClearBufferuiv` | PASS |
|
||||
| `DirectGLES.SwizzleAccessRoutineScenario.RepeatedProgramBuildsKeepFetchingTheSameTexel` | `ClearBufferuiv` | PASS |
|
||||
| `DirectGLES.TextureParamsWithoutASamplerViewScenario.AReadAttachmentOnlyTexturesDepthStencilModeReachesTheDriver` | `ClearBufferfi` | FAIL: applied `DepthStencilMode=0`, expected `GL_STENCIL_INDEX (6401)` (`TextureParamsWithoutASamplerViewScenario.cpp:437`) |
|
||||
| `DirectGLES.ViewportArrayScenario.EachViewportIndexUsesItsOwnDepthRange` | `ClearBufferfv` | PASS |
|
||||
| `DirectVulkan.DepthStencilReadbackMatrixScenario.CopyTexImageFromADepthAttachmentSurvivesAReadBack` | `CopyTexImage2D` | FAIL: 3072/3072 copied depth values wrong, first word `0x0` (`DepthStencilReadbackMatrixScenario.cpp:568`) |
|
||||
| `DirectVulkan.IntegerBorderColorScenario.ANegativeBorderOnAnUnsignedFormatClampsToTheFormatsMaximum` | `ClearBufferiv` | PASS |
|
||||
| `DirectVulkan.IntegerBorderColorScenario.ASignedIntegerBorderIsClampedToTheFormatsRepresentableRange` | `ClearBufferiv` | PASS |
|
||||
| `DirectVulkan.IntegerBorderColorScenario.AnOversizedUnsignedBorderIsClampedToTheFormatsMaximum` | `ClearBufferiv` | PASS |
|
||||
| `DirectVulkan.IntegerBorderColorScenario.InsideTexelsAreUnaffectedByTheBorderColour` | `ClearBufferiv` | PASS |
|
||||
| `DirectVulkan.IntegerBorderColorScenario.SamplerParameterIivBorderColourSurvivesToAnIntegerSampler` | `ClearBufferiv` | PASS |
|
||||
| `DirectVulkan.IntegerBorderColorScenario.TexParameterIivBorderColourSurvivesToAnIntegerSampler` | `ClearBufferiv` | PASS |
|
||||
| `DirectVulkan.P4aFinalFixScenario.AFramebufferDeletedAfterADsaClearLeavesItsAttachmentIntact` | `ClearNamedFramebufferfv` | `Fatal{UnmigratedVerb, "ClearNamedFramebufferfv+UNBOUND"}` |
|
||||
| `DirectVulkan.SampledSetStalenessScenario.AQueuedClearIsMaterialisedWhenAFilterChangeCompletesTheTexture` | `ClearBufferfv` | PASS |
|
||||
| `DirectVulkan.SampledSetStalenessScenario.AQueuedClearIsMaterialisedWhenASamplerObjectCompletesTheTexture` | `ClearBufferfv` | PASS |
|
||||
| `DirectVulkan.SwizzleAccessRoutineScenario.EveryAccessRoutineFetchesTheSameTexelUnderTheIdentitySwizzle` | `ClearBufferuiv` | PASS |
|
||||
| `DirectVulkan.SwizzleAccessRoutineScenario.EveryAccessRoutineSeesAReversedSwizzle` | `ClearBufferuiv` | PASS |
|
||||
| `DirectVulkan.SwizzleAccessRoutineScenario.EveryAccessRoutineSeesConstantSwizzleSources` | `ClearBufferuiv` | PASS |
|
||||
| `DirectVulkan.SwizzleAccessRoutineScenario.RepeatedProgramBuildsKeepFetchingTheSameTexel` | `ClearBufferuiv` | PASS |
|
||||
| `DirectVulkan.ViewportArrayScenario.EachViewportIndexUsesItsOwnDepthRange` | `ClearBufferfv` | PASS |
|
||||
|
||||
## Rulings and overturn conditions
|
||||
|
||||
1. **Unbound named clears remain named refusals.** The census’s DSA-delete scenario is actually unbound (it binds framebuffer 0 before the call), on both backends. The bound named case crosses and renders in the new pixel cases. No server rebind workaround: Espryt’s barrier-pulled binding would undo it. Overturn with an integrator-authorized handle-keyed backend binding path (P7/P3b/P4b); the contract’s optional stretch is not required to report this next blocker.
|
||||
2. **Depth-copy wrong-answer failures are next-round work.** All three measured variants now execute the copy and reach `CopyTexImageFromADepthAttachmentSurvivesAReadBack`; all 3072 values read as zero. Color image/sub-image copy pixel controls pass. No claim that depth copies render correctly; fixes outside the granted emitter/sink boundary are requested from the integrator. Overturn by a depth-copy/readback fix making those same assertions pass.
|
||||
3. **Read-attachment-only depth/stencil inspection remains a next blocker.** `AReadAttachmentOnlyTexturesDepthStencilModeReachesTheDriver` reaches its inspection assertion; it is no longer stopped at ClearBufferfi. The focused fi test verifies both actual depth and stencil values. Overturn with a fix to the applied-mode observation/state path, validated against that scenario.
|
||||
4. **No mip-storage re-derivation was needed.** Both measured mip cases pass, as does a generated-level pixel control. Existing `generate-mipmap-storage` and `generate-mipmap-cpu-fallback` refusals remain intact. Overturn if a measured format reaches them; only the descriptor-derived storage site is granted, CPU writeback still needs P8/P9.
|
||||
5. **No `+RENDERBUFFER` refusal is added to framebuffer-sourced copies.** The source is the read framebuffer, which may legitimately have a renderbuffer attachment. The contract’s explicit resource-endpoint `CopyImageSubData+RENDERBUFFER` refusal belongs to i1, not f1. Overturn only by a contract revision for these two slots.
|
||||
6. **Existing codec arms suffice.** Clear (57), generate_mipmap (54), framebuffer copy (76) already decode their PODs and invoke the corresponding sinks; live peer cases prove fields survive that path. No codec layout changes were necessary. Overturn with a newly measured payload shape requiring a contract change.
|
||||
7. **Partition arithmetic is additive.** f1 moves 11 slots; counts on this base are A=2/B=16/C=53. The old global `C==64` assertion becomes `B+C==69`, retaining all per-owner asserts. Integrator must add the other packages’ counts in the shared RemoteClient partition test.
|
||||
|
||||
## Expanded census and newly exposed companion blockers
|
||||
|
||||
After registering the split-only source, the normal split-build backend registries also discover 22 controls. `/home/swung/w7/p5b-f1-final-census-logs/results.json` records **478 passed / 185 skipped / 477 aborted / 31 failed** over 1,171 entries. Restricted to the original 1,149 names, the result remains **460/185/473/31**. The 22 added entries contribute 18 passes and four named refusals.
|
||||
|
||||
The four `DirectVulkan.F1WireScenario.ClearNamedFramebuffer{fv,iv,uiv,fi}Pixels` entries hit their corresponding `+UNBOUND` refusal **even though the application has bound that framebuffer**: the server applier has no matching BoundFramebuffer handle at this point. The exact contract predicate prevents a silent clear of an unproven binding. This is a newly measured publication/capability-path gap; it is not evidence that those applications used unbound DSA. Vulkan advertises consumer mask `0x1f7f` (resource subsystem 0x80 absent), while PipeFill has dependency gates for texture/framebuffer publications. DirectGLES passes all eleven focused controls; Vulkan passes the other seven. **Integrator request:** resolve the bound-framebuffer publication/capability dependency for Magma or revise the contract before relaxing the sink predicate. Overturn with these four existing controls passing while the four deliberately unbound unit controls still refuse exactly. No emitter bypass of the R-8 mirror gate was introduced.
|
||||
|
||||
## Gate and landing
|
||||
|
||||
f1 implementation commit `7ffa4998`; merged the moved integrator head `37fc4fdb` (P5 x2 wire-ledger/control changes) as `fe5ac006`, without conflicts. Commit identity checked: `Swung0x48 <swung0x48@outlook.com>`. No attribution trailers. The merge only imports integrator-owned changes; f1’s own diff is the five assigned source/test files.
|
||||
|
||||
Pre-merge quick gate: builds 4/4; unit **1817 × 3 / split 2099** (2084 + 15); integration-split **33/33**, two expected skips (22 + 11); push integration-gpu **1128/1128**. G2 all-label pull/push names: **0 diff**. G14: **0 removed / 43 added** vs the P5 name baseline, matching c0b. G1: **0/0/0/0**, `.text 10806611 → 10806611 (+0)`, 27814 defined symbols unchanged. Generators green. Both p3a/p4a untouched-region checks against `ff2994d9 HEAD` green after the merge.
|
||||
|
||||
The first integration-split pass exposed missing private log paths for the eleven new entries; the f1-only registration now supplies a unique absolute path for each. Final rerun: **33/33** including `SplitLogPaths.PrivateAndDistinct`.
|
||||
|
||||
Post-merge gate on **fe5ac006**: builds **4/4**; unit **1817/1817 × 3**, split **2101/2101** (2084 + f1’s 15 + x2’s 2); integration-split **33/33**, two expected skips; push integration-gpu **1128/1128**. G2 **0 diff**, G14 **0 removed / 43 added**; G1 **0/0/0/0, `.text +0`**; generators and both purity checks green.
|
||||
|
||||
Post-merge census: `/home/swung/w7/p5b-f1-merged-census-logs/results.json`, **478/185/477/31** over 1,171 entries, **zero status or normalized-Fatal differences** from the pre-merge expanded census. Original-name projection remains **460/185/473/31**.
|
||||
|
||||
Final integrator head check: `feat/disaggregated=37fc4fdb` is an ancestor of `fe5ac006`. Worktree clean, submodule links off.
|
||||
|
||||
## Retained evidence and tools
|
||||
|
||||
All paths below are under `/home/swung/w7/`:
|
||||
|
||||
- `p5b-f1-{census-logs,final-census-logs,merged-census-logs}/`: manifests, per-entry outputs/logs and results for the original, expanded and post-merge censuses, respectively.
|
||||
- `p5b-f1-census.log`, `p5b-f1-final-census.log`, `p5b-f1-merged-census.log`: census summaries.
|
||||
- `p5b-f1-gate.log`, `p5b-f1-merged-gate.log`: pre/post-merge builds, generators, G1 and all unit lanes.
|
||||
- `p5b-f1-accept.log` (initial private-log registration failure, superseded), `p5b-f1-isplit-final.log` (fixed), `p5b-f1-merged-accept.log` (all green).
|
||||
- `p5b-f1-igpu-push.log`, `p5b-f1-isplit.log`: final post-merge integration output.
|
||||
- `p5b-f1-unit-linux.log`, `p5b-f1-unit-push.log`, `p5b-f1-unit-verify.log`, `p5b-f1-unit-split.log`: final per-lane unit output.
|
||||
- `p5b-f1-build-linux.log`, `p5b-f1-build-push.log`, `p5b-f1-build-verify.log`, `p5b-f1-build-split.log`: final build output.
|
||||
- `p5b-f1-symbols.md`, `p5b-f1-p3a-purity.log`, `p5b-f1-p4a-purity.log`.
|
||||
- `p5b-f1-names-linux.txt`, `p5b-f1-names-push.txt`, `p5b-f1-names-split.txt`: sorted all-label registrations.
|
||||
- `p5b-f1-red.log`: every new control fails its own diagnostic after the intentional production mutations; all restored.
|
||||
- `p5b-f1-final-client.log`: 23 field/partition/refusal controls green after moving the old unmigrated control to the GetTexImage tail.
|
||||
- `p5b-f1-submodlinks.sh`, `notes/tools/p5b-f1-gate.sh`, `notes/tools/p5b-f1-accept.sh`: reproducible local helpers.
|
||||
|
||||
Uncited temporary configure/intermediate-build/test logs were deleted individually. No devices, LFS downloads or other package worktrees were modified.
|
||||
|
||||
@@ -0,0 +1,299 @@
|
||||
# i1 — image / compute / barrier / copy-image / storage block, v1
|
||||
|
||||
Branch `p5b/i1`, forked from `feat/disaggregated @ f5676103` (c0b's head). Worktree
|
||||
`~/w7/p5b-i1`. Nothing pushed; nothing merged into the integration branch.
|
||||
|
||||
```
|
||||
734287ac [Feat] (MG_Remote, P5b): emit i1 seven image/compute/barrier/copy-image/storage-block
|
||||
slots as class B and give each ServerVerbSink body the backend call the contract names
|
||||
b82c63a9 [Feat] (MG_Pipe, DirectGLES, P5b): move GetProgramForDispatch FATAL to BARRIER_PULLED
|
||||
now that compute crosses, and skip the copy-image shadow mirror under a non-monolith
|
||||
transport
|
||||
a72a96d8 [Test] (MG_Test, MG_IntegrationTest, P5b): pin i1 seven class-B slots and the served
|
||||
compute program, and arm 24 inproc lane entries whose results only exist if the records
|
||||
crossed
|
||||
e352e9a6 [Merge] (P5b, i1): merge feat/disaggregated at 37fc4fdb (x2 via x2m, ID-65) into p5b/i1
|
||||
```
|
||||
|
||||
Range **`734287ac..e352e9a6`** (three single-line commits plus the merge). `feat/disaggregated`
|
||||
moved to `37fc4fdb` while i1 ran (x2 via x2m); merged, no conflicts, although x2m also edits
|
||||
`EmitTables.cpp` (+157) — the merge is textually clean and the whole gate was re-run on the
|
||||
merged head. Working tree clean; the worktree's submodule symlinks are left OFF (git-safe).
|
||||
|
||||
---
|
||||
|
||||
## 1. What landed, per slot
|
||||
|
||||
Seven slots moved class C → B: the five the census measured plus the two companions that share
|
||||
their rows. `kEmittedSlotsI1 = 0 → 7`, `MGR_UNMIGRATED_I1_SLOTS` is now the empty macro, and
|
||||
`UnmigratedSlotCount()` is 64 → 57 / `ImplementedVerbCount()` 5 → 12.
|
||||
|
||||
| slot (census entries) | row (op) | emitter | sink body | proven by |
|
||||
|---|---|---|---|---|
|
||||
| `BindImageTexture` (138) | `bind_shader_image` (72) | `EmitBindImageTexture`: GL args verbatim, `Res` from the unit's `ImageTextureBinding`, `Access` the GL token | `OnBindShaderImage` → `GL.BindImageTexture(Unit, GlName, Level, Layered, Layer, Access, Format)` | 138 → 0 aborts; 13 `UnboundImageDescriptorScenario` lane entries |
|
||||
| `DispatchCompute` (49) | `launch_grid` (60) | `EmitDispatchCompute`, b1's hook order kept verbatim from the stub (`PushPersistentMapsBeforeVerb(); MarkGpuWritesForDispatch();` then the record) | `OnLaunchGrid` → `GL.DispatchCompute(GridX,Y,Z)` | 49 → 0; 3 `AtomicCounterScenario` lane entries |
|
||||
| `DispatchComputeIndirect` (companion) | `launch_grid` (60) | `IsIndirect = 1`, `IndirectOffset = indirect`, `IndirectBuffer` = the bound `GL_DISPATCH_INDIRECT_BUFFER` handle | `OnLaunchGrid` → `GL.DispatchComputeIndirect(IndirectOffset)` | unit: the two slots are not one thunk |
|
||||
| `MemoryBarrier` (18) | `memory_barrier` (61) | `EmitMemoryBarrier`, bits verbatim, no client-side lowering | `OnMemoryBarrier` → `GL.MemoryBarrier(Bits)` | 18 → 0 |
|
||||
| `MemoryBarrierByRegion` (companion) | `memory_barrier` (61) | `ByRegion = 1` | `OnMemoryBarrier` → `GL.MemoryBarrierByRegion(Bits)` | unit |
|
||||
| `CopyImageSubData` (30) | `resource_copy_region` (53) | `EmitCopyImageSubData`: handles + the two GL names + the targets verbatim; a renderbuffer endpoint refused by name before emission | `OnResourceCopyRegion` rebuilds two `CopyImageEndpoint`s through the barrier-pulled `GetTextureObject(name)` and calls `GL.CopyImageSubData(...)` | 30 → 0; 6 `CopyImageLayeredScenario` lane entries |
|
||||
| `ShaderStorageBlockBinding` (4) | `set_storage_block_binding` (75), `kHasBlob` | `EmitShaderStorageBlockBinding`: `GlName`, `Binding`, `ShaderCso`, the block name staged `strlen + 1` | `OnSetStorageBlockBinding` → `GL.ShaderStorageBlockBinding(GlName, name, Binding)` | 4 → 0; the 2 `ProgramPipelineScenario` storage-block lane entries |
|
||||
|
||||
Beside them, the two edits the contract names:
|
||||
|
||||
- **`FieldOwnership.def`, the one row §6.9 grants i1**: `GetProgramForDispatch` FATAL →
|
||||
`BARRIER_PULLED, "P7 (Magma), P8 (Espryt)"`. Without it every migrated dispatch aborts on the
|
||||
backend's own `PrepareForCompute` (`DirectGLES.cpp:5779`, `VulkanRenderer.cpp:7327`). The
|
||||
FATAL block is two rows now, not three; tables regenerated with
|
||||
`gen_pipe_field_ownership.py`.
|
||||
- **`DirectGLES.cpp`, i1's one backend edit** (§2 i1, ruling §6.7): the
|
||||
`copy-image-shadow-mirror` site returns early under a non-monolith transport, behind
|
||||
`#if MOBILEGL_BUILD_DISAGGREGATED` so the pull build's code does not move (G1 `.text +0`),
|
||||
and armed on the TRANSPORT and not the build so build-split's monolith lanes still mirror.
|
||||
The pre-existing `#if MOBILEGL_PIPE_PUSH` `MGPipeUnmigratedEmulation` call is left where it
|
||||
was, so a push-only build is byte-for-byte unchanged.
|
||||
|
||||
Tests: `RemoteEmitTable.TheSevenI1SlotsAreClassBAndAreNotTheFatalThunk`,
|
||||
`RemoteEmitTable.EachI1SlotReachesRequireSessionUnderItsOwnName` (five death tests, each
|
||||
asserting `Fatal{NoClientSession, "<slot>"}` and the ABSENCE of `Fatal{UnmigratedVerb`),
|
||||
`FieldOwnershipTest.TheComputeProgramIsServedUnderADispatchStampFromP5bOn`, the retargeted
|
||||
`FieldOwnershipTest.AFatalClassReadAbortsEvenWithoutStrictErrors` (its old exemplar was the
|
||||
field i1 moved), and 24 `integration-split` entries. Every one ships its "red once by doing X"
|
||||
line.
|
||||
|
||||
---
|
||||
|
||||
## 2. Acceptance numbers
|
||||
|
||||
All on `~/w7/p5b-i1 @ e352e9a6`, `GLIBC_TUNABLES=glibc.malloc.tcache_count=0`.
|
||||
Logs `~/w7/p5b-i1-*.log`; census `~/w7/p5b-i1-census-logs/results.json`.
|
||||
|
||||
### 2.1 Builds — four for four
|
||||
`build-linux` (pull), `build-push`, `build-verify`, `build-split`: configure rc=0, build rc=0,
|
||||
`CTestTestfile.cmake` confirmed in each.
|
||||
|
||||
### 2.2 G1 — did not move
|
||||
```
|
||||
nm --defined-only build-linux/libMobileGL.so | grep -ic MG_Remote -> 0
|
||||
symbol-report vs ~/w7/p5-before-libMobileGL.so, --threshold 0:
|
||||
.text 10806611 -> 10806611 (+0, +0.000%)
|
||||
27814 -> 27814 defined symbols: 0 added, 0 removed, 0 resized, 0 renamed
|
||||
```
|
||||
Everything i1 adds is either inside `MG_Remote` (compiled only under
|
||||
`MOBILEGL_BUILD_DISAGGREGATED`), a `.def` row whose generated table the pull build never
|
||||
odr-uses, or the one backend site behind `#if MOBILEGL_BUILD_DISAGGREGATED`.
|
||||
|
||||
### 2.3 Generators and purity
|
||||
`gen_pipe.py --check` / `--self-test` (9 trips), `gen_pipe_field_ownership.py --check` /
|
||||
`--self-test` (15 trips, 63 fields, 7 sticky forwards, 9 refusals), `gen_pipe_dirty_surface.py
|
||||
--check`, `check_include_closure.py` (4 probes, 0 problems): all green.
|
||||
|
||||
### 2.4 Test lanes
|
||||
|
||||
| lane | c0b baseline | i1 |
|
||||
|---|---|---|
|
||||
| `ctest -L unit` build-linux / push / verify | 1817 / 1817 each | **1817 / 1817** each, unchanged |
|
||||
| `ctest -L unit` build-split | 2084 / 2084 | **2089 / 2089** (+3 i1 cases, +2 from the x2m merge) |
|
||||
| `ctest -L integration-split` build-split, inproc | 22 / 22 | **46 / 46** (+24 i1 lane entries) |
|
||||
| `ctest -L integration-gpu` build-push, monolith | 1128 / 1128 | **1128 / 1128** |
|
||||
| G2 pull-vs-push `ctest -N`, every label | 0 diff | **0 diff** |
|
||||
| G14 build-linux names vs `p5-before-ctest-names.txt` | 0 removed, 43 added | **0 removed, 43 added** — identical to c0b; i1 adds NOTHING to the pull build |
|
||||
| G5 `p3a_untouched_regions.sh ff2994d9 HEAD` | rc=0 | **rc=0** |
|
||||
| G5 `p4a_untouched_regions.sh ff2994d9 HEAD` | rc=0 | **rc=0** |
|
||||
|
||||
### 2.5 The inproc census
|
||||
|
||||
`~/w7/p5b-c0b-census.sh ~/w7/p5b-i1 ~/w7/p5b-i1-census-logs`. The manifest GREW by 24 — i1's
|
||||
lane entries carry `integration-gpu` as well as `integration-split`, and the census enumerates
|
||||
that label — so the diff below is over the **1149 entries c0b's baseline and this run share**,
|
||||
and the 24 new ones are reported separately (all 24 passed).
|
||||
|
||||
| status | c0b baseline (`f5676103`) | i1, common 1149 | i1, all 1173 |
|
||||
|---|---:|---:|---:|
|
||||
| passed | 432 | **503** | 527 |
|
||||
| skipped | 185 | **195** | 195 |
|
||||
| subprocess-aborted | 505 | **335** | 335 |
|
||||
| ordinary failed | 27 | **116** | 116 |
|
||||
|
||||
**Every transition came out of `aborted`. There is not one regression:**
|
||||
|
||||
```
|
||||
passed -> passed 432 aborted -> aborted 335 skipped -> skipped 185
|
||||
failed -> failed 27 aborted -> failed 89
|
||||
aborted -> passed 71 aborted -> skipped 10
|
||||
```
|
||||
|
||||
No entry that passed on c0b's head does anything else on this one, and no entry that failed
|
||||
fails differently.
|
||||
|
||||
**The census was run twice, before and after the x2m merge, and the two runs are identical** —
|
||||
same totals, same transition matrix, same per-slot table. So nothing below is x2m's, and every
|
||||
number is attributable to i1.
|
||||
|
||||
---
|
||||
|
||||
## 3. The gate: my five slots, and where their 239 entries went
|
||||
|
||||
**All five counts are zero, and so are both companions:**
|
||||
|
||||
| slot | c0b | i1 |
|
||||
|---|---:|---:|
|
||||
| `BindImageTexture` | 138 | **0** |
|
||||
| `DispatchCompute` | 49 | **0** |
|
||||
| `CopyImageSubData` | 30 | **0** |
|
||||
| `MemoryBarrier` | 18 | **0** |
|
||||
| `ShaderStorageBlockBinding` | 4 | **0** |
|
||||
| `DispatchComputeIndirect`, `MemoryBarrierByRegion` | 0 | **0** |
|
||||
|
||||
239 entries, not the brief's 235 — the brief's number omits the four
|
||||
`ShaderStorageBlockBinding` entries. Where each went:
|
||||
|
||||
| was |
|
||||
|---|
|
||||
| **`BindImageTexture` 138** → 30 passed · 10 skipped · 45 `Fatal{UnmigratedEmulation, "texture-remint-pull"}` · 53 wrong-answer failures (52 of them DirectVulkan) |
|
||||
| **`DispatchCompute` 49** → 23 passed · 1 `texture-remint-pull` · 25 wrong-answer failures (all DirectVulkan) |
|
||||
| **`CopyImageSubData` 30** → 16 passed · 9 wrong-answer failures · 3 `Fatal{UnmigratedVerb, "ClearBufferfv"}` (f1's) · 2 `Fatal{UnmigratedVerb, "CopyImageSubData+RENDERBUFFER"}` (i1's own named refusal) |
|
||||
| **`MemoryBarrier` 18** → 12 `Fatal{UnmigratedVerb, "DrawElementsBaseVertex"}` · 6 `Fatal{UnmigratedVerb, "DrawElements"}` (both d1's) |
|
||||
| **`ShaderStorageBlockBinding` 4** → 2 passed · 2 wrong-answer failures (DirectVulkan) |
|
||||
|
||||
### The next first blockers, by name
|
||||
|
||||
1. **`Fatal{UnmigratedEmulation, "texture-remint-pull"}` — 46 entries, all DirectGLES.**
|
||||
`Managers.cpp:5634`. Exactly the bounded loss contract §6.7 predicted for the mirror skip.
|
||||
By scenario: `ImageTargetKindScenario` 22, `NonCoreImageFormatScenario` 9,
|
||||
`ImageFormatQualifierScenario` 4, `FormatlessImageBakeScenario` 3,
|
||||
`ImageLoadStoreSsoScenario` 2, `LayeredTextureReadbackScenario` 2, and one each of
|
||||
`AdvertisedLimitsScenario`, `ImageSizeAfterRespecScenario`, `P4aFinalFixScenario`,
|
||||
`TextureParamsWithoutASamplerViewScenario`.
|
||||
2. **d1's draw slots — 18 entries**, the whole of `GuiBatchScenario` on both backends:
|
||||
`DrawElementsBaseVertex` 12, `DrawElements` 6. `GuiBatchScenario` was `MemoryBarrier`'s only
|
||||
scenario, and it is now entirely d1's.
|
||||
3. **f1's `ClearBufferfv` — 3 entries**, the three
|
||||
`CopyImagePacked16Scenario.RenderbufferLandsInArrayMipLevelIntact` variants.
|
||||
4. **i1's own named refusal — 2 entries**:
|
||||
`{DirectGLES,DirectVulkan}.PackedWordReadbackScenario.ACopyThroughARenderbufferReachesAnRgb9E5Destination`
|
||||
reach `Fatal{UnmigratedVerb, "CopyImageSubData+RENDERBUFFER"}`. The refusal is MEASURED now,
|
||||
which is what §6.6 asked for.
|
||||
5. **89 wrong-answer failures** — §4.2 and §4.3 below. 82 DirectVulkan, 7 DirectGLES.
|
||||
|
||||
---
|
||||
|
||||
## 4. Rulings I made, each with what would overturn it
|
||||
|
||||
1. **The handles are LOOKED UP, never minted.** `MGPImageBind::Res`, `MGPCopyRegion::Src/Dst`,
|
||||
`MGPStorageBlockBinding::ShaderCso` and `MGPGridInfo::IndirectBuffer` come from
|
||||
`MGPipeSlots().FindByLifetimeId(...)`, not `Acquire(...)`. Every one of these sinks
|
||||
dispatches on the GL NAME — that is what rule D carries it for — so the handle is for P7.
|
||||
`Acquire` would MINT a slot at a call site that emits no create record, handing the server an
|
||||
identity it has never seen; a null field is the true statement "no handle published yet".
|
||||
*Overturned by* P7 dispatching on `Res` instead of `GlName`, which needs the minting form
|
||||
plus an ordering guarantee this row does not have.
|
||||
2. **Only the two dispatch slots carry a pre-verb hook.** b1's hooks describe "the work the
|
||||
record is about to start". A bind, a barrier, an image copy and a block rebinding start no
|
||||
shader and read no mapped buffer; the dispatch that later reads the image is the verb that
|
||||
carries both hooks. Running them on all seven would push the same maps several times per
|
||||
dispatch and inflate b1's per-row counters. *Overturned by* a measured workload whose
|
||||
coherent-map write is read by a copy or a barrier with no dispatch between.
|
||||
3. **The renderbuffer endpoint is refused on BOTH sides of the wire.** The client refuses before
|
||||
emission (ID-57's shape, so the census greps it where the other client refusals are) and the
|
||||
sink refuses a record that names `GL_RENDERBUFFER` in either target. *Overturned by* a
|
||||
renderbuffer sticky forward or P7's handle keying.
|
||||
4. **A copy whose GL name no longer resolves DECLINES, it does not Fatal.** `OnResourceCopyRegion`
|
||||
logs once and returns false — the monolith's own answer in its own words
|
||||
(`DirectGLES.cpp:9067`, "declining the copy"); the frontend validator is what keeps it
|
||||
unreachable and what reports the `INVALID_VALUE` the application is owed. *Overturned by* a
|
||||
measured entry that reaches it, which would mean a pull went stale under the barrier.
|
||||
5. **The atomic-counter barrier lowering stays in the backend.** `OnMemoryBarrier` forwards
|
||||
`Bits` untouched. Espryt's "counter bit implies storage bit" is a statement about the driver
|
||||
(`DirectGLES.cpp:8837`); repeating it client-side would make the split arm and the monolith
|
||||
arm two different calls. *Overturned by* nothing in P5b.
|
||||
|
||||
### 4.2 The ruling I was asked to make, and the measurement that changes its JUSTIFICATION
|
||||
|
||||
Contract §6.7 rules the mirror skipped under split and says **"what the skip loses is bounded by
|
||||
two Fatals"**. I implemented the skip. The bound is **half right, and I measured both halves**:
|
||||
|
||||
- The `texture-remint-pull` half is real: 46 entries reach it, by name (§3.1).
|
||||
- **The `glGetTexImage` half does not hold, and the skip is not what breaks those entries
|
||||
anyway.** I ran the experiment: with the skip compiled out (`if (false && …)`, rebuilt,
|
||||
re-run) the five DirectGLES packed-readback entries —
|
||||
`CopyImagePacked16Scenario.{Array,Flat}…` ×2, `WidenedPacked16` ×2,
|
||||
`PackedWordReadbackScenario.ACopiedRgb9E5Word…` — **fail identically**. So i1's mirror skip
|
||||
costs the lane exactly the 46 named aborts and nothing else.
|
||||
- What actually breaks those five is `glGetTexImage` under split: `GetTexImage_State`
|
||||
(`GL_Texture.cpp:5331`) rejects with `GL_INVALID_OPERATION` and returns BEFORE
|
||||
`GetTexImage_Backend`, so the class-C `Fatal{UnmigratedVerb, "GetTexImage"}` never fires and
|
||||
the caller keeps its zero-filled buffer. The scenario reads the zeros as texels.
|
||||
**A class-C slot that silently answers nothing instead of refusing by name is the one shape
|
||||
ID-47 says a picture comparison cannot see**, and it is worth the integrator's attention
|
||||
independently of i1 — it is wave-3 / P9 territory.
|
||||
|
||||
*What would overturn the skip itself:* nothing i1 measured. If the integrator wants the honest
|
||||
alternative anyway, it is one line — delete the `#if MOBILEGL_BUILD_DISAGGREGATED` early return
|
||||
and those 46 entries go back to aborting `Fatal{UnmigratedEmulation, "copy-image-shadow-mirror"}`
|
||||
by name instead of proceeding. That trades 46 named aborts for 46 entries that get further.
|
||||
|
||||
### 4.3 The 89 wrong-answer failures, and why i1 does not believe they are i1's
|
||||
|
||||
**82 of 89 are DirectVulkan (Magma); 7 are DirectGLES (Espryt).** The load-bearing observation
|
||||
is that **the same scenarios pass on Espryt and fail on Magma from the same wire records** —
|
||||
`DoublePrecisionScenario` 7/7 pass on DirectGLES and 7/7 fail on DirectVulkan, and so do
|
||||
`AtomicCounterScenario`, `SsboArrayLengthScenario`, `UniformInitializerScenario` and
|
||||
`ProgramPipelineScenario`. The record is backend-agnostic; only the backend differs.
|
||||
|
||||
The Magma shape is uniform: the compute dispatch runs, no server-side error is logged, and the
|
||||
SSBO the shader writes reads back all zeros
|
||||
(`DirectVulkan.DoublePrecisionScenario.ADoubleUniformKeepsItsDeclaredInitializer`: `values[0]`
|
||||
is 0, 2.5 expected). Magma registers a real `DispatchCompute` slot, so the sink is calling it;
|
||||
what is missing is inside `VulkanRenderer::DispatchCompute`'s own state resolution under split.
|
||||
**Magma's split compute/image path is the next round's package, and it is the single biggest
|
||||
item i1 hands on.**
|
||||
|
||||
The 7 DirectGLES failures: 5 are the `glGetTexImage` shape of §4.2; the other two are
|
||||
`TextureParamsWithoutASamplerViewScenario.ACopyImageEndpointOnlyTexturesParamsReachTheDriver`
|
||||
(a white-box G9 case whose own message diagnoses it: "the record reached the server and the
|
||||
server has not pushed it" — Espryt defers texture-param application to the first sampler view,
|
||||
and a copy-image-only destination never gets one) and one
|
||||
`ResourceSubsystemOff.LargeArenaAdoptionScenario…` entry that the runner classes `failed` while
|
||||
its own output says `SKIPPED` ("runs only in its own lane") — a census-classifier artifact, not
|
||||
a product failure.
|
||||
|
||||
---
|
||||
|
||||
## 5. What I believe the brief or the contract got wrong or left open
|
||||
|
||||
1. **The brief says 235 lane entries; it is 239.** 138 + 49 + 30 + 18 + 4. The four
|
||||
`ShaderStorageBlockBinding` entries are missing from the brief's arithmetic.
|
||||
2. **Contract §6.7's "bounded by two Fatals" does not hold for the `glGetTexImage` half** —
|
||||
§4.2, with the experiment that shows it and the site that causes it.
|
||||
3. **The census manifest is not a fixed 1149.** It is `ctest -L integration-gpu` on the tree
|
||||
under test, and every package that adds a split lane entry (which also carries
|
||||
`integration-gpu`) grows it. The four packages will each grow it; the diff has to be taken
|
||||
over the intersection, and c0b's 432/185/505/27 is a baseline for the COMMON entries only.
|
||||
4. **A `DirectGLES.Split.` lane entry must emit a record** (`ScenarioFixture.h:86`), so a
|
||||
whole-scenario `TEST_FILTER` cannot be used for a scenario that has pure state cases:
|
||||
`ProgramPipelineScenario` has nine of them and they are red for having nothing to say. i1's
|
||||
block filters to `*StorageBlock*`. A package adding lane entries also has to extend
|
||||
`Harness/SplitLogPaths.cmake.in`, or `SplitLogPaths.PrivateAndDistinct` fails for the whole
|
||||
lane — that file's loop is a hard-coded list of TEST_LIST variables and every package needs a
|
||||
line in it.
|
||||
5. **`FieldOwnershipTest` runs only in the split unit lane.** The three other unit lanes stay at
|
||||
1817 with an i1 case added to that file; worth knowing before someone expects 1818.
|
||||
|
||||
---
|
||||
|
||||
## 6. Debts, and what the next round starts on
|
||||
|
||||
- **Magma's split compute/image path** (§4.3) — 82 entries. Biggest single item.
|
||||
- **`glGetTexImage` silently answers nothing under split** (§4.2) instead of raising its class-C
|
||||
Fatal. Wave 3 / P9 owns the slot; the SILENCE is the part that should not wait for it.
|
||||
- **`texture-remint-pull`** — 46 entries; ROADMAP P8's client-side mirror retires it.
|
||||
- **`GetProgramForDispatch` is now a counted pull** (`rsp`), retired by P7 (Magma) / P8 (Espryt)
|
||||
with `GetProgramForDraw`.
|
||||
- **`MGPGridInfo::Block*` stay 0** and `IndirectBuffer` travels unread — P7 Magma fills the
|
||||
first from the reflection archive if it needs it.
|
||||
- **`MemoryBarrier` has no green lane case of its own**: `GuiBatchScenario` is its only census
|
||||
scenario and it is d1's `DrawElements` now. `AtomicCounterScenario` covers the row in the
|
||||
lane; a first-class one has to wait for d1.
|
||||
- **i1's slots are zero, so the lane's remaining 335 aborts are d1's, t2's and f1's**, led by
|
||||
`BeginTransformFeedback` 95, `DrawElements` 62, `PatchParameteri` 43, `DrawElementsBaseVertex`
|
||||
16, `MultiDrawElementsBaseVertex` 18, `ClearBufferiv` 12, `ClearBufferfv` 9.
|
||||
@@ -0,0 +1,9 @@
|
||||
# P5b user-index span extent fix
|
||||
|
||||
- Commit: `0b67568e` (`p5b/indexspan-codex`), parent `348d22a4`.
|
||||
- Tree: `/home/swung/w7/p5b-d1-codex`.
|
||||
- Finding: `Offset + Size` staying within SEG_STAGE did not prove that the draw's declared index count fit inside that run. A span at the final two bytes with `Size=2`, `IndexSize=2`, `Count=3` could reach a backend read beyond the segment.
|
||||
- Change: one shared gate in the encoder, decoder and server sink requires exactly one direct indexed range, index width 1/2/4, client-array `Start=0`, and `Uint64(Count) * IndexSize <= Size`. Existing segment bounds/honesty gates remain in place.
|
||||
- Validation: `PipeWireCodecTest.*Draw*:PipeWireCodecTest.UserIndex*` **9/9 passed**. The two-byte reproducer dies at all three entry points. Additional negative cases cover arrays, invalid width, zero/multiple ranges, nonzero Start and 32-bit multiplication wrap. An exact-fit two-byte span at segment end passes encoding, decoding and sink validation.
|
||||
- Build: `cmake --build build-split -j 8 --target PipeWireCodecTest`, completed. Logs: `/home/swung/w7/p5b-indexspan-build.log`, `/home/swung/w7/p5b-indexspan-test.log`.
|
||||
- No additional full gate was run. The concurrent final census froze `348d22a4`; its results are not evidence for this later commit.
|
||||
@@ -0,0 +1,175 @@
|
||||
# P5b joint Codex evidence — 2026-09-16
|
||||
|
||||
<!-- BEGIN CODEX CENSUS -->
|
||||
## Inproc census — merged head, lane default 32 MiB / traces explicit 256 MiB
|
||||
|
||||
Head: `348d22a4b9c15cc571af840dedc4b4edc8c53dbb`. Evidence directory: `/home/swung/w7/p5b-final-census-348d22a4`.
|
||||
|
||||
Frozen libMobileGL SHA256: `50fecb248368a4de02d7f51d4229c4ebac555b3bea18d3a22398c1247d2a28c4`. `identity.json` records the head, all frozen executables/scripts, hashes and explicit profiles.
|
||||
|
||||
The single complete integration census is the final gate’s inproc debt arm. No second full
|
||||
inproc ctest run was used for these numbers. Each case has its own fresh log; static fixture
|
||||
intersections are not execution evidence. Missing fixtures/results, zero cases and timeouts
|
||||
are not passes.
|
||||
|
||||
### Integration lane — stage 32 MiB
|
||||
|
||||
Final **1267** exact manifest entries: **62 aborted / 191 failed / 811 passed / 203 skipped**.
|
||||
|
||||
Baseline c0b is 432 passed / 185 skipped / 505 aborted / 27 failed (1149 entries).
|
||||
Exact-name join: 1149 retained, 118 added, 0 removed. Previously passing entries now failed/aborted: 0.
|
||||
|
||||
| c0b status | Final status | Matched entries |
|
||||
|---|---|---:|
|
||||
| aborted | aborted | 58 |
|
||||
| aborted | failed | 164 |
|
||||
| aborted | passed | 265 |
|
||||
| aborted | skipped | 18 |
|
||||
| failed | failed | 26 |
|
||||
| failed | passed | 1 |
|
||||
| passed | passed | 432 |
|
||||
| skipped | skipped | 185 |
|
||||
|
||||
All exact names, including additions/removals, are in `lane-transitions.md` and
|
||||
`lane-transitions.json` beside `identity.json`. Final first-blocker distribution:
|
||||
|
||||
| Outcome / first blocker | Entries |
|
||||
|---|---:|
|
||||
| passed | 811 |
|
||||
| skipped | 203 |
|
||||
| failed | 191 |
|
||||
| UnmigratedEmulation, "texture-remint-pull" | 46 |
|
||||
| UnmigratedVerb, "CopyImageSubData+RENDERBUFFER" | 5 |
|
||||
| UnmigratedVerb, "ClearNamedFramebufferfv+UNBOUND" | 3 |
|
||||
| UnmigratedVerb, "DeleteTransformFeedback" | 2 |
|
||||
| UnmigratedVerb, "MultiDrawElementsBaseVertex+CLIENT_INDICES" | 2 |
|
||||
| UnimplementedWritebackWait | 1 |
|
||||
| UnmigratedVerb, "ClearNamedFramebufferfi+UNBOUND" | 1 |
|
||||
| UnmigratedVerb, "ClearNamedFramebufferiv+UNBOUND" | 1 |
|
||||
| UnmigratedVerb, "ClearNamedFramebufferuiv+UNBOUND" | 1 |
|
||||
|
||||
### Complete trace census — stage 256 MiB
|
||||
|
||||
**79** backend entries: **6 aborted / 1 failed / 72 passed**. This inventory includes OpenRA and honors each fixture’s declared ci_backends.
|
||||
|
||||
Production staging remains 32 MiB. The 256 MiB profile is authorized for real workloads
|
||||
with a 128 MiB upload; a profile-assisted success is not evidence that default-capacity debt
|
||||
was fixed. Standalone d1’s stage32 baseline had 28 passes / 49 failures over 77 Minecraft
|
||||
entries, including SEG_STAGE failures on GLES iterationt and iterationt-nodsa. r1’s retirement
|
||||
wait fix and a larger capacity address different conditions and are not merged in attribution.
|
||||
|
||||
| Fixture | Backend | Final outcome / first blocker | SSIM |
|
||||
|---|---|---|---|
|
||||
| OpenRA | DirectGLES | passed | 1.0 |
|
||||
| OpenRA | DirectVulkan | passed | 1.0 |
|
||||
| improved-transparency-minecraft-26.3 | DirectGLES | passed | 1.0 |
|
||||
| improved-transparency-minecraft-26.3 | DirectVulkan | passed | 0.999914064 |
|
||||
| minecraft-1.17-main-menu-854 | DirectGLES | passed | 0.9999616 |
|
||||
| minecraft-1.17-main-menu-854 | DirectVulkan | passed | 0.99996113 |
|
||||
| minecraft-1.21.1-neoforge-create-indirect-in-world | DirectGLES | UnmigratedEmulation, "texture-remint-pull" | — |
|
||||
| minecraft-1.21.1-neoforge-create-indirect-in-world | DirectVulkan | failed | — |
|
||||
| minecraft-1.21.1-neoforge-create-instancing-in-world | DirectGLES | passed | 0.999957572 |
|
||||
| minecraft-1.21.1-neoforge-create-instancing-in-world | DirectVulkan | passed | 0.99995247 |
|
||||
| minecraft-1.21.11-main-menu | DirectGLES | passed | 0.993474922 |
|
||||
| minecraft-1.21.11-main-menu | DirectVulkan | passed | 0.99347492 |
|
||||
| minecraft-1.21.4-fabric-common-mods-in-world | DirectGLES | passed | 0.999947917 |
|
||||
| minecraft-1.21.4-fabric-common-mods-in-world | DirectVulkan | passed | 0.999947917 |
|
||||
| minecraft-1.21.4-fabric-common-mods-inventory | DirectGLES | passed | 0.999996158 |
|
||||
| minecraft-1.21.4-fabric-common-mods-inventory | DirectVulkan | passed | 0.999996158 |
|
||||
| minecraft-1.21.4-fabric-iris-bliss-in-world | DirectGLES | passed | 0.998057535 |
|
||||
| minecraft-1.21.4-fabric-iris-bliss-in-world | DirectVulkan | passed | 0.997986272 |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-esc-menu-854 | DirectGLES | InitialBytesNotCarried, "resource_respecify" | — |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-esc-menu-854 | DirectVulkan | passed | 0.998402027 |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-in-world | DirectGLES | passed | 0.997496023 |
|
||||
| minecraft-1.21.4-fabric-iris-bsl-in-world | DirectVulkan | passed | 0.997324199 |
|
||||
| minecraft-1.21.4-fabric-iris-chocapic-v6-lite-in-world | DirectGLES | passed | 0.999992296 |
|
||||
| minecraft-1.21.4-fabric-iris-chocapic-v6-lite-in-world | DirectVulkan | passed | 0.99575325 |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-reimagined-in-world | DirectGLES | passed | 0.999503181 |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-reimagined-in-world | DirectVulkan | passed | 0.99929651 |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-unbound-in-world | DirectGLES | passed | 0.999478525 |
|
||||
| minecraft-1.21.4-fabric-iris-complementary-unbound-in-world | DirectVulkan | passed | 0.999244084 |
|
||||
| minecraft-1.21.4-fabric-iris-derivative-main-d24.4.14-in-world | DirectGLES | UnmigratedEmulation, "texture-remint-pull" | — |
|
||||
| minecraft-1.21.4-fabric-iris-derivative-main-d24.4.14-in-world | DirectVulkan | passed | 0.996658574 |
|
||||
| minecraft-1.21.4-fabric-iris-iterationrp-in-world | DirectVulkan | passed | 0.995833421 |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-in-world | DirectGLES | passed | 0.997752215 |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-in-world | DirectVulkan | passed | 0.999198504 |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-nodsa-in-world | DirectGLES | passed | 0.995919679 |
|
||||
| minecraft-1.21.4-fabric-iris-iterationt-nodsa-in-world | DirectVulkan | passed | 0.99864077 |
|
||||
| minecraft-1.21.4-fabric-iris-makeup-ultrafast-in-world | DirectGLES | passed | 0.99542552 |
|
||||
| minecraft-1.21.4-fabric-iris-makeup-ultrafast-in-world | DirectVulkan | passed | 0.996629722 |
|
||||
| minecraft-1.21.4-fabric-iris-mellow-in-world | DirectGLES | passed | 0.999822891 |
|
||||
| minecraft-1.21.4-fabric-iris-mellow-in-world | DirectVulkan | passed | 0.999312266 |
|
||||
| minecraft-1.21.4-fabric-iris-nostalgia-in-world | DirectGLES | passed | 0.999988313 |
|
||||
| minecraft-1.21.4-fabric-iris-nostalgia-in-world | DirectVulkan | passed | 0.999983572 |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.1-in-world | DirectGLES | passed | 0.999135212 |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.1-in-world | DirectVulkan | passed | 0.998781357 |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.3b-in-world | DirectGLES | UnmigratedEmulation, "texture-remint-pull" | — |
|
||||
| minecraft-1.21.4-fabric-iris-photon-v1.3b-in-world | DirectVulkan | passed | 0.998529629 |
|
||||
| minecraft-1.21.4-fabric-iris-sundial-lite-in-world | DirectGLES | passed | 0.996300657 |
|
||||
| minecraft-1.21.4-fabric-iris-sundial-lite-in-world | DirectVulkan | passed | 0.996284254 |
|
||||
| minecraft-1.21.4-fabric-iris-super-duper-vanilla-in-world | DirectGLES | passed | 0.999643966 |
|
||||
| minecraft-1.21.4-fabric-iris-super-duper-vanilla-in-world | DirectVulkan | passed | 0.999439052 |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world-normal-world | DirectGLES | passed | 0.99997899 |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world-normal-world | DirectVulkan | passed | 0.99997899 |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world | DirectGLES | passed | 0.999984122 |
|
||||
| minecraft-1.21.4-fabric-journeymap-in-world | DirectVulkan | passed | 0.999984122 |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory-normal-world | DirectGLES | passed | 0.999997315 |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory-normal-world | DirectVulkan | passed | 0.999997292 |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory | DirectGLES | passed | 0.999996645 |
|
||||
| minecraft-1.21.4-fabric-modernui-inventory | DirectVulkan | passed | 0.999996645 |
|
||||
| minecraft-1.21.4-fabric-rei-inventory-normal-world | DirectGLES | passed | 0.99998801 |
|
||||
| minecraft-1.21.4-fabric-rei-inventory-normal-world | DirectVulkan | passed | 0.99998801 |
|
||||
| minecraft-1.21.4-fabric-rei-inventory | DirectGLES | passed | 0.999989912 |
|
||||
| minecraft-1.21.4-fabric-rei-inventory | DirectVulkan | passed | 0.999989912 |
|
||||
| minecraft-1.21.4-fabric-sodium-in-world | DirectGLES | passed | 0.999985623 |
|
||||
| minecraft-1.21.4-fabric-sodium-in-world | DirectVulkan | passed | 0.999986092 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world-normal-world | DirectGLES | passed | 0.999991832 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world-normal-world | DirectVulkan | passed | 0.999991832 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world | DirectGLES | passed | 0.999999456 |
|
||||
| minecraft-1.21.4-fabric-xaero-minimap-in-world | DirectVulkan | passed | 0.999999456 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world-normal-world | DirectGLES | passed | 0.999995376 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world-normal-world | DirectVulkan | passed | 0.999995376 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world | DirectGLES | passed | 0.999999823 |
|
||||
| minecraft-1.21.4-fabric-xaero-world-map-in-world | DirectVulkan | passed | 0.999999823 |
|
||||
| minecraft-1.21.4-in-world | DirectGLES | passed | 0.999976222 |
|
||||
| minecraft-1.21.4-in-world | DirectVulkan | passed | 0.999976105 |
|
||||
| minecraft-1.21.4-main-menu | DirectGLES | passed | 0.997008676 |
|
||||
| minecraft-1.21.4-main-menu | DirectVulkan | passed | 0.99700863 |
|
||||
| minecraft-1.21.4-rd12-odinlite-in-world | DirectGLES | InitialBytesNotCarried, "resource_respecify" | — |
|
||||
| minecraft-1.21.4-rd12-odinlite-in-world | DirectVulkan | BarrierTimeout, "Present" | — |
|
||||
| minecraft-1.21.4-startup | DirectGLES | passed | 0.999999662 |
|
||||
| minecraft-1.21.4-startup | DirectVulkan | passed | 0.999999662 |
|
||||
|
||||
`trace-transitions.json` maps all names to standalone d1’s first blocker and carries both
|
||||
profile sizes. Every result links its actual private log and result.json in the isolated
|
||||
trace output directory. A new first blocker counts as progress, never as rendered success.
|
||||
|
||||
<!-- END CODEX CENSUS -->
|
||||
|
||||
<!-- BEGIN CODEX HOST DOCS -->
|
||||
## Host gate and documentation — pending final replay/device evidence
|
||||
|
||||
The one host gate is pinned to `348d22a4b9c15cc571af840dedc4b4edc8c53dbb`.
|
||||
Authoritative directory: `/home/swung/w7/p5b-final-host-348d22a4/`;
|
||||
`head.txt`, `libraries.sha256`, `status.tsv`, complete per-lane JUnit,
|
||||
`summary-counts.json` and `summary-counts.md` distinguish selection from actual passes/skips.
|
||||
The completed split lane is 107 selected = 105 passed + 2 skipped, zero failures.
|
||||
The host replay runner is still in flight; no complete verdict is asserted here.
|
||||
|
||||
A runner execution anomaly is retained separately: appending an archive command while Bash
|
||||
was reading the live runner produced one `-test-dir: command not found` line (original session
|
||||
87762; outer shell lacked set-e). It is not classified as a required-test failure. The verify
|
||||
and audit lanes have complete independent JUnit and recorded step rc=0; E2 then ran normally.
|
||||
The E1/E3 private logs were already copied successfully to `negative-private-logs` manually.
|
||||
Final required-step inventory, JUnit counts and replay counts must show no omitted step;
|
||||
a final shell exit alone is not the proof. No repeated full gate is requested.
|
||||
|
||||
Three tree documents are prepared in independent `/home/swung/w7/p5b-docs-final-codex`, branch
|
||||
`p5b/docs-final-codex`, based on `348d22a4`. They separate package evidence (§32), the one host
|
||||
gate and combined census (§33), the one phase review plus targeted repairs (§34), and the final
|
||||
Redmi exit/four-arm measurements (§35). Default stage remains 32 MiB; real-workload trace and
|
||||
all four device arms explicitly use a 256 MiB profile. No final device/source identity or
|
||||
successful P5b exit is filled in before the integrator sends DEVICE_DONE and sourcehead.
|
||||
|
||||
Documentation commit: pending final host/census/targeted-closure/device evidence.
|
||||
<!-- END CODEX HOST DOCS -->
|
||||
@@ -0,0 +1,53 @@
|
||||
# P5b GLES mip storage follow-up — Codex v1
|
||||
|
||||
Date: 2026-09-16. Branch `p5b/blit-codex`; parent `31b26a440de4d52c1afb746ed934247da8b8b145`.
|
||||
Commit: `4d1219cdadd1c1ecb3ebbd25ff44bc702ce42ac0`.
|
||||
|
||||
## Result
|
||||
|
||||
`minecraft-1.21.4-fabric-iris-bsl-in-world.DirectGLES` crosses the previous
|
||||
`Fatal{UnmigratedEmulation, "generate-mipmap-storage"}` and renders in inproc:
|
||||
SSIM **0.997496**, threshold 0.99, runner exit 0. Actual selection **1 / 79**.
|
||||
The private `output/mobilegl.log` contains **zero Fatal records**.
|
||||
|
||||
## Implementation and authority
|
||||
|
||||
CONTRACT-P5B §2 f1 expressly permits deriving storage growth from the applier descriptor.
|
||||
The frontend `EnsureGeneratedMipmapStorageAllocated` already allocates and publishes the
|
||||
complete chain before the GenerateMipmap verb. In a disaggregated build under a real transport,
|
||||
the GLES helper now reads the live applier resource descriptor and verifies `Levels` covers the
|
||||
chain derived from `Width/Height/Depth/Target`. Array layers are excluded from shrinking axes.
|
||||
The backend then proceeds to GPU generation without allocating or dirtying client level shadows.
|
||||
Missing/insufficient descriptors retain the named storage refusal.
|
||||
|
||||
Identity lookup reuses the existing backend registry `HandleOf`: it reads the barrier-held
|
||||
object lifetime id and `FindByLifetimeId`, with a registry memo update on hits. It neither mints
|
||||
a slot nor mutates the client allocator. **This identity lookup remains P5b barrier debt**; this
|
||||
change does not claim process separation. No new client mip-level, texel-data, or MappedData read
|
||||
is introduced. The RGB16F/RGB32F CPU fallback refusal remains intact, pending writeback work.
|
||||
The new code is behind `MOBILEGL_BUILD_DISAGGREGATED`, with monolith falling through unchanged.
|
||||
|
||||
## Directed validation
|
||||
|
||||
- Split ALL incremental build completed, max `-j8`; `p5b-mip-codex-build.log`.
|
||||
- Three selected DirectGLES split mip tests passed (0 skipped): existing RGBA8 control, plus
|
||||
new mutable base-only R11F_G11F_B10F and DEPTH_COMPONENT32F controls reading generated mip 2.
|
||||
Both new tests assert wire emission, framebuffer completeness, GL_NO_ERROR and actual pixels.
|
||||
`p5b-mip-codex-tests.log`; all three private logs contain zero Fatal records.
|
||||
- Dynamic trace library: `/home/swung/w7/p5b-blit-codex/build-split/libMobileGL.so`.
|
||||
SHA-256 `e932dd21542f29a0cf1a4337a1204b68c8c2e25034bfa71aa7d858dfa9d10ad5`.
|
||||
- Trace scripts and hydrated fixtures came from `/home/swung/w7/p5b-integrate-codex`; only the
|
||||
explicit output directory was written. This was a library-under-test run, not evidence that
|
||||
the integration tree had already built this change.
|
||||
- Output `/home/swung/w7/retrace-out/p5b-mip-codex/summary.json`;
|
||||
runner log `/home/swung/w7/p5b-mip-codex-trace.log`;
|
||||
private log `/home/swung/w7/retrace-out/p5b-mip-codex/minecraft-1.21.4-fabric-iris-bsl-in-world/DirectGLES/output/mobilegl.log`.
|
||||
- Default SEG_STAGE **32 MiB**. No stage-size override here; parent handles the separate
|
||||
improved-transparency 128 MiB blob with its 256 MiB-stage run.
|
||||
- `git diff --check` passed. No full gate, whole census, device run or adversarial review repeated.
|
||||
|
||||
## Remaining
|
||||
|
||||
Parent integrates this commit and verifies combined Android/device behavior. Vulkan results
|
||||
from the named-blit package remain separate prior evidence; they were not rerun for this
|
||||
GLES-only backend fix. RGB three-channel CPU mip fallback is still named debt.
|
||||
@@ -0,0 +1,195 @@
|
||||
# P5b phase-close Codex review
|
||||
|
||||
Date: 2026-09-16. Reviewed tree: `/home/swung/w7/p5b-integrate-codex`.
|
||||
Pinned target: **348d22a4b9c15cc571af840dedc4b4edc8c53dbb**; diff base: **37fc4fdb**.
|
||||
HEAD was rechecked at the end and still matched. This is the phase's one closing review,
|
||||
not an additional implementation-round review.
|
||||
|
||||
**Verdict: 0 blocker, 2 major, 1 minor.** The two new major findings concern the now-live
|
||||
user-index carrier and the now-live blocking sync call. The minor is the previously named
|
||||
ReadPixels reply-status gap, which remains present. Findings below are source-confirmed;
|
||||
no new runtime reproduction, mutation, build, broad test, or device operation was performed.
|
||||
They do not claim that the four selected Minecraft targets exercise these edge cases.
|
||||
|
||||
## Findings
|
||||
|
||||
### 1. A user-index span can declare fewer bytes than the draw consumes — major
|
||||
|
||||
Locations:
|
||||
|
||||
- `MobileGL/MG_Remote/Wire/PipeWireCodec.cpp:1049`: encoder host-span check.
|
||||
- `MobileGL/MG_Remote/Wire/PipeWireCodec.cpp:1966`: decoder host-span check.
|
||||
- `MobileGL/MG_Remote/Server/PipeApplier.cpp:555`: the sink resolves the span.
|
||||
- `MobileGL/MG_Remote/Server/PipeApplier.cpp:593`: the ordinary indexed arm hands the
|
||||
resolved pointer and independent range count to the backend; the other indexed arms do
|
||||
the same at lines 582, 586, 590, 598, 602, 607 and 610.
|
||||
|
||||
`CheckHostSpanIsHonest` proves that **the declared** `Offset + Size` lies in a registered
|
||||
segment. Neither codec direction compares `Size` with `MGPDrawRange::Count * IndexSize`.
|
||||
The sink casts `range.Count` to GLsizei and passes it on without that check. Once the pointer
|
||||
is resolved, `span.Size` no longer constrains the backend's reads.
|
||||
|
||||
Concrete corrupt-record trigger: one direct indexed draw, `IndexSize=2`, `NumDraws=1`,
|
||||
`Flags=kDrawHasUserIndices`, range `{Start=0, Count=3, IndexBias=0}`, and a null-Ptr span
|
||||
`{Seg=SEG_STAGE, Offset=segment_capacity-2, Size=2}`. The existing layout, tail-size and
|
||||
four honesty checks all accept this shape. The sink then asks the backend to consume six
|
||||
index bytes from the last two bytes of the segment. Even away from the segment boundary,
|
||||
this can consume bytes from an adjacent staged allocation rather than the declared run.
|
||||
|
||||
This is a newly armed P5b carrier-boundary defect, not the documented P8 host-index mirror
|
||||
deferral. The ordinary emitter currently stages `count * indexSize`, so a healthy emitter
|
||||
does not itself produce the mismatch; the missing defense is against corrupted/inconsistent
|
||||
wire input, which the codec expressly promises to reject before handing the sink validated
|
||||
arguments. No driver read outside memory was deliberately executed during this review.
|
||||
|
||||
Suggested bounded fix: share a DrawVbo user-index validation helper between encode and
|
||||
decode. For the supported one-range client-index shape, validate the index width, range
|
||||
shape and stage carrier, and require the declared run to cover the count-derived extent
|
||||
(the producer's exact-size contract permits equality). Use checked/widened arithmetic.
|
||||
Keep the already documented multi-draw client-index refusal. A focused short-span negative
|
||||
case and its exact-span positive counterpart are sufficient evidence for this finding.
|
||||
|
||||
### 2. A legal long ClientWaitSync is cut short by the 30-second transport watchdog — major
|
||||
|
||||
Locations:
|
||||
|
||||
- `MobileGL/MG_Remote/Client/EmitTables.cpp:1483`: `EmitClientWaitSync` preserves the
|
||||
application's Uint64 timeout, then calls `ReadFenceReply` at line 1471.
|
||||
- `MobileGL/MG_Remote/Server/PipeApplier.cpp:127`: `OnFenceWait` forwards that timeout
|
||||
unchanged to the native backend at line 137.
|
||||
- `MobileGL/MG_Remote/Client/ClientSession.cpp:126`: fixed `kBarrierTimeoutMs = 30000`.
|
||||
- `MobileGL/MG_Remote/Client/ClientSession.cpp:779`: all reply-owning records use that
|
||||
fixed timeout for `WaitForApplied`; line 797 aborts on expiry.
|
||||
- `MobileGL/MG_Backend/DirectGLES/DirectGLES.cpp:12114`: the backend synchronously waits
|
||||
in native `glClientWaitSync`. DirectVulkan's wait entry at `DirectVulkan.cpp:1045` is
|
||||
also reached through the same client deadline.
|
||||
|
||||
Concrete trigger: an unsignaled fence and
|
||||
`glClientWaitSync(sync, GL_SYNC_FLUSH_COMMANDS_BIT, 60000000000ULL)`, while valid queued
|
||||
work keeps the native wait outstanding beyond 30 seconds. The server is still carrying out
|
||||
the application's permitted 60-second wait, so it cannot post the result or advance
|
||||
appliedSeq yet. At 30 seconds the client aborts with `Fatal{BarrierTimeout,"FenceWait"}`
|
||||
instead of returning the eventual native wait result. A timeout longer than 30 seconds is
|
||||
not rejected by frontend GL validation and CONTRACT-P5B's sync appendix preserves the
|
||||
64-bit timeout. This is different from a wedged ordinary verb: the call explicitly authorizes
|
||||
the server to remain inside it for that interval.
|
||||
|
||||
Suggested minimal fix: derive the **applied/reply** wait budget for `FenceWait` from its
|
||||
declared timeout, plus the existing transport grace period, leaving command-capacity waits
|
||||
and all other opcodes unchanged. Round nanoseconds upward, handle overflow and the doorbell's
|
||||
Uint32-millisecond / wait-forever sentinel deliberately, and retain shutdown wakeup. A budget
|
||||
helper can be tested with 0, a short wait, 60 seconds and UINT64_MAX without a 30-second sleep.
|
||||
Alternatively split the server's native wait into bounded pieces while preserving the full
|
||||
client-visible timeout and final GL result; merely clamping the native timeout changes GL
|
||||
semantics and does not close this finding. Update the timeout diagnostic to report the actual
|
||||
selected budget. `FenceWaitServer` is a server-side GPU ordering call and should not inherit
|
||||
the application's GL_TIMEOUT_IGNORED as a CPU deadline by accident.
|
||||
|
||||
### 3. ReadPixels still accepts an unknown reply status when the byte count matches — minor
|
||||
|
||||
Locations: `MobileGL/MG_Remote/Client/EmitTables.cpp:863`, used at lines 969 and 982;
|
||||
the stricter exposed predicate is at line 1792.
|
||||
|
||||
`RequireReadbackReplyComplete` rejects ERROR and DECLINED individually, then checks only
|
||||
the byte count. An exact-sized reply with status 3 (or another non-OK value) passes and is
|
||||
returned/scattered as pixels. `ReadbackReplyIsComplete` correctly requires status OK, but
|
||||
the production helper does not invoke that predicate. Reply-slot reading does not reject
|
||||
unknown status values earlier.
|
||||
|
||||
This was already named in the P5 close review's unassigned-debt section and r2's follow-up
|
||||
table. It is not a newly observed normal server output: the production server currently posts
|
||||
the expected statuses. Close it by requiring OK after the named ERROR/DECLINED diagnostics,
|
||||
and drive the actual live readback return path with an exact-sized status-3 answer. Testing
|
||||
only the exposed stricter predicate would miss the current disconnect again.
|
||||
|
||||
## Previous P5 findings 1–13
|
||||
|
||||
The inspected integration contains the intended r1/r2 repairs. No additional defect in those
|
||||
repairs was established during this review. This is source review plus inspection of the
|
||||
package evidence references, not a claim that their runtime controls were rerun here.
|
||||
|
||||
| Prior finding | Reviewed disposition at this target |
|
||||
|---|---|
|
||||
| 1 coherent-map server producer re-entry | Backend sync early-outs on the apply role; producer guards precede map reads / serial mutations. r1's client-only suppression report addresses the earlier shared-knob blind spot. |
|
||||
| 2 frontend framebuffer death | The callback synchronously uses the apply control mailbox before twin destruction/cache edits; reentrant server calls stay on the apply thread. P6 still owns separate-process lifetime transport. |
|
||||
| 3 PACK_SWAP_BYTES | Client applies component/packed-word swaps on both tight and scatter paths, with the mixed float/depth-stencil type split into two 32-bit words. |
|
||||
| 4 false RingOverrun on delayed retirement | Staging waits after immediate reclaim; command retry retains both draw tails and accounts for wrap padding. Individually oversized allocations still refuse. |
|
||||
| 5 census prevents CI controls | Broad inproc status is recorded as debt; hard reduced gates remain; negative controls have `!cancelled()`. |
|
||||
| 6 pull control poisons following draw-drop | Frozen library backup/EXIT restoration exists; both controls check library symbols, including nm failure. |
|
||||
| 7 intentional E3 skips and combined diagnostic | E3 selects the four pixel cases; assertions and private logs are checked per selected entry. |
|
||||
| 8 reclaim count mislabeled as wait | Immediate reclamation does not increment waits; outstanding-retirement blocking does. SmallRing asserts an actual wait as well as wrap. |
|
||||
| 9 all-skip false green | Baseline failure/nonzero exit/all-skip rejects; CI has the runtime-only nonempty-success tally. |
|
||||
| 10 optional EGL unit proof | Split unit CI explicitly requires GPU. |
|
||||
| 11 census zero-run / stale log | Updated external runners clear private logs before launch, reject zero executed gtests, and retain process-status distinctions; these external tools still must accompany the integration. |
|
||||
| 12 pin self-test bypasses production selection | Both revised self-tests call production `extract_baseline`, rather than directly testing the pin helper alone. |
|
||||
| 13 obsolete c1f test name | The runner selects the current different-tuple case and checks the identical-tuple companion in baseline/restoration. |
|
||||
|
||||
## Scope and limits
|
||||
|
||||
Read CONTRACT-P5/P5B, the P5b brief, prior P5 closing review, and the r1/r2/d1/i1/t2/f1,
|
||||
named-blit, mip and sync reports; inspected the affected emit/codec/sink/lifetime paths and
|
||||
CI/control changes. Also inspected the new Redmi runner/reducer's exact-head prerequisites,
|
||||
fresh attempt paths, fixed serial, stage profile and completeness checks. No new general
|
||||
runner false-green finding was established.
|
||||
|
||||
The 256 MiB stage profile is explicit and applies to the four A/B arms; the production
|
||||
32 MiB default and its large-upload boundary are recorded limits. The named-blit scoped
|
||||
binding override and mip descriptor's existing identity lookup remain intentional inproc
|
||||
barrier dependencies. They are not claimed as spawn-safe. Client vertex arrays, resource
|
||||
identity pulls, P7/P8/P9 readback/query work, rd12, and the known deferred shapes remain their
|
||||
recorded debts, not newly relabeled blockers in this review.
|
||||
|
||||
In particular, i1's old comment claiming the missing copy shadow is always bounded by a
|
||||
GetTexImage Fatal must be read with the recorded frontend-shadow fallback debt: the frontend
|
||||
can still select its local GetTexImage fallback. The established texture readback failures
|
||||
already cover that limitation; this review does not count the same debt as a new finding.
|
||||
|
||||
The final host gate, merged census, APK build and Redmi runs were underway concurrently.
|
||||
Their incompleteness at review time is not a finding, and no successful phase exit or final
|
||||
binary identity is inferred here. The integrator must append their actual pinned results
|
||||
and the targeted closure evidence for findings 1–3 before claiming P5b complete. This report
|
||||
does not request a second full review or a repeated full gate after focused repairs.
|
||||
|
||||
## Integrator addendum, 2026-09-16 — pinned results and closure evidence
|
||||
|
||||
All concurrently-running evidence named above has since completed; nothing below reruns a
|
||||
gate or a review, it appends the pinned outcomes the report asked for.
|
||||
|
||||
- **Host closing gate (`348d22a4`, `~/w7/p5b-final-host-348d22a4/`)**: the WSL restart left
|
||||
31/32 steps at rc=0; the missing `retrace-verify` arm was resumed against the SHA-256
|
||||
pinned verify library (`verify-recovery-r1/`), 75 remaining entries serial, exit 0.
|
||||
Combined: **retrace-verify 79/79**, retrace-push 79/79, OpenRA 2/2; all unit/gpu/split/
|
||||
verify/audit lanes 0 failed. Step-by-step reconciliation in `reconciliation-20260916.md`;
|
||||
`exit-code.txt=0`, `complete=true`, `summary-counts.{json,md}` regenerated.
|
||||
- **Merged census (`348d22a4`, `~/w7/p5b-final-census-348d22a4/`)**: lane **1267 selected =
|
||||
811 passed / 203 skipped / 62 aborted / 191 failed**, zero regressions against the c0b
|
||||
baseline (1149 retained, 118 added, 0 removed). Trace census **79 = 72 passed / 6 aborted
|
||||
/ 1 failed** after a three-round resume (resume1: create-indirect DirectVulkan >60 GiB
|
||||
RSS guard-killed; resume2: same case guard-killed again = the 1 failed, five cases
|
||||
deferred by the memory guard; resume3 on a fresh instance: 4 passed / 1 aborted).
|
||||
Non-passing first blockers: rd12 GLES `InitialBytesNotCarried/resource_respecify`, rd12
|
||||
Vulkan `BarrierTimeout/Present`, iris-photon GLES / iris-derivative GLES /
|
||||
create-indirect GLES `UnmigratedEmulation/texture-remint-pull`, iris-bsl-esc-menu GLES
|
||||
`InitialBytesNotCarried/resource_respecify`, create-indirect Vulkan memory-guard kill.
|
||||
Census block injected into `joint-codex-v1.md`; `identity.json`/`counts.json`/
|
||||
`trace-transitions.json` beside it.
|
||||
- **Findings 1–3 closure**: fixes `a021e3cc` (finding 1) and `82683d4a` (findings 2–3) keep
|
||||
their package evidence (`indexspan-codex-v1.md` 9/9, `wait-codex-v1.md` 7/7). The two
|
||||
`PipeWireCodecTest.UserIndexSpan*` failures in the pre-`7cb29d46` quickgate were the
|
||||
test log-capture defect fixed by `7cb29d46`; re-evidenced green on `c77831e0`
|
||||
(3/3). Final device source: **`82683d4a83b5c8f4f7c375a7a92ebee9a1854646`** (APK phase
|
||||
`p5bcodex2`: pull/push/split signed APKs, provenance and proofs in
|
||||
`~/w7/notes/p5b/apk/p5bcodex2/`).
|
||||
- **Redmi exit (device 2f7cbe2e, stage 256 MiB explicit profile)**: correctness matrix
|
||||
**8/8** (4 traces x 2 backends, inproc split arm, per-group pin evidence and inproc/
|
||||
apply-thread/wire-record proofs; `MOBILEGL_IPC_STAGE_MB=256`). Benchmark four-arm A/B:
|
||||
**24/32 groups green with validated 200-frame tails** (all groups of
|
||||
improved-transparency-26.3, 1.21.4-in-world, 1.21.4-sodium-in-world; first barrier-tax
|
||||
measurement, see `ab-tables.md`). The 8 `iris-bsl-in-world` groups failed validation on a
|
||||
fixture limit, not a regression: the trace yields only 123 benchmark frames, below the
|
||||
runner's 200-frame-tail rule, on every arm including pull. Their retained 123-frame
|
||||
series were reduced separately as a labeled supplement (p99 is shader-compile dominated
|
||||
on all arms, as already recorded in the P2 baseline). Note for future campaigns: the
|
||||
device no longer holds the vendor GPU ceiling observed 2026-09-11; the deterministic pin
|
||||
is now 1100 MHz (pin_device.sh updated 2026-09-16), so these numbers are not
|
||||
clock-comparable with a campaign pinned at 1050 MHz.
|
||||
@@ -0,0 +1,76 @@
|
||||
# P5b r1 — server ownership, readback byte order, retirement back-pressure
|
||||
|
||||
Completed 2026-09-16. Branch `p5b/r1-codex`, isolated tree
|
||||
`/home/swung/w7/p5b-r1-codex`, commit
|
||||
`db8b3479520862cd47c546e96506b7332158abcf`.
|
||||
Base is r2 `d50183cb`, so merge r2 first or merge this branch including its ancestry.
|
||||
This implements ID-73 findings 1–4; it is not another adversarial review or a phase gate.
|
||||
|
||||
The inherited Windows tree `/mnt/c/Users/geekerwan/p5b-r1` was at `37fc4fdb`
|
||||
with three uncommitted files: `PersistentMapTracker.cpp`, `.h`, and `BufferObject.cpp`.
|
||||
Their complete file contents were copied into `/home/swung/w7/p5b-r1-codex-backup/`
|
||||
before continuing in this new tree. The old tree and any old Claude process were left alone.
|
||||
|
||||
## Changes and evidence
|
||||
|
||||
| Finding | Implementation | Focused evidence |
|
||||
|---|---|---|
|
||||
| 1: coherent map bypass | Under split, retained backend `SyncPersistentMappedRange` calls return on the apply thread. `PushAllMembers`, `PushBlocksFor`, and the final `PushMappedSpanBlock` producer refuse server-role entry with `Fatal{RoleViolation,...}`, before reading map bytes or modifying frontend content serials. Server map-state teardown keeps membership valid but does not emit client live-write state. | Real GPU store readback test now holds a coherently mapped frontend object with bytes A and a server shadow with bytes B; Ensure uploads B. A live client session also asserts the producer's named apply-thread refusal. Separately, removing only client pre-verb push calls makes both coherent-pointer pixel cases fail while backend sync sites remain intact; restoring the calls makes both pass. |
|
||||
| 2: framebuffer death on client | GLES' shared frontend-death callback forwards synchronously through the existing apply-thread control mailbox before any twin destruction or binding-cache changes. This covers every kind in that callback; framebuffer still has no invented wire opcode. | Native `glDeleteFramebuffers` interception observes one deletion on the apply thread and verifies that the native draw binding becomes zero. The previously failing `DirectGLES.P4aFinalFixScenario.ARenderbufferAndItsFramebufferDeletedAfterAClearLeaveTheNextDrawIntact` now passes. |
|
||||
| 3: PACK_SWAP_BYTES | After validating the tight reply, swap each requested component/packed word on the client. Apply to both direct and bounce paths, before scatter; byte types remain unchanged. `FLOAT_32_UNSIGNED_INT_24_8_REV` swaps two separate 32-bit words. | Live emit-table peer covers ushort, float, packed uint, mixed depth/stencil, and byte data with swap off/on and tight/scattered layouts; sentinel gaps are preserved. Actual GL test reads a non-symmetric ushort from RGBA16F with tight and skipped/strided pack destinations: both monolith and inproc pass. RGBA8 was deliberately avoided because expanded ushort components repeat the same byte and cannot expose a missing swap. |
|
||||
| 4: applied before retired | Preserve r2's staging allocator wait and truthful wait counter unchanged. `ClientSession::EmitAndWait` now treats command-ring capacity exhaustion as back-pressure after the encoder's individual-record size check. It waits for enough retired bytes including the wrap pad, then retries the same record without changing its sequence. Shutdown/timeout gets a named `RetirementWaitFailed` instead of a silent drop or indefinite wait. | Two live-session cases hold the server between applied publication and retirement using r2's scheduling-only hook. The staging case fills 3/4 of 1 MiB twice and observes one wait. The command case positions a legal 393 KiB draw record so its successor needs a wrap pad, observes the producer parked, releases retirement, and verifies the next ordinal is exactly prior+1. |
|
||||
|
||||
The death callback uses the existing inproc control mailbox, not a new wire lifetime record.
|
||||
P6's separate address space still needs the already-planned lifetime/control transport work.
|
||||
Staging's r2 five-second shutdown-aware bound and once-per-blocked-allocation counter are intact.
|
||||
The command wait uses the session's existing `kBarrierTimeoutMs` (30 seconds); individually
|
||||
oversized records still fail at `PipeWireEncoder::EncodeRecord`.
|
||||
|
||||
## Executed acceptance
|
||||
|
||||
All paths below are under `/home/swung/w7/`.
|
||||
|
||||
- Split build completed, `CTestTestfile.cmake` verified by the build driver. Logs:
|
||||
`p5b-r1-codex-config-split.log`, `p5b-r1-codex-build-split.log`,
|
||||
`p5b-r1-codex-build-fixes.log`. Build concurrency was at most 8.
|
||||
- Final targeted run: **10/10 passed, zero skips**, `MOBILEGL_TRANSPORT=inproc`,
|
||||
`MOBILEGL_ITEST_REQUIRE_GPU=1`, `p5b-r1-codex-final-targeted.log`.
|
||||
Six unit entries plus four integration entries: the six ownership/wait/swap controls,
|
||||
actual ushort swap pixels, FBO/RBO deletion pixels, and two coherent-pointer draw cases.
|
||||
- Actual GL ushort swap: **1/1 monolith and 1/1 inproc**, zero skips,
|
||||
`p5b-r1-codex-swap-{monolith,inproc}.log`.
|
||||
- Client-only push suppression: **two own pixel failures** (CTest exit 8), plus the existing
|
||||
map-arm metadata case's one intentional skip. The skip is not counted as a red.
|
||||
`p5b-r1-codex-coherent-red.log`; driver `p5b-r1-codex-coherent-control.sh`.
|
||||
It restores the source through an EXIT trap and rebuilds the restored integration binary.
|
||||
The restored two pixel cases are included in the final 10/10 run.
|
||||
- `git diff --check` passed. Final worktree clean, submodule links off. Commit is a single
|
||||
line without trailers, author/committer `Swung0x48 <swung0x48@outlook.com>`.
|
||||
|
||||
The primary negative control removes **only** `PushPersistentMapsBeforeVerb()` from the
|
||||
client pre-verb hooks. It does not set the block-size-zero knob, alter the backend's retained
|
||||
sync sites, or remove the new server-role guard. It therefore proves client omission is no
|
||||
longer repaired by server producer re-entry.
|
||||
|
||||
No new all-lane census, four-flavor unit run, G1 binary comparison, full integration lane,
|
||||
device operation, push, or phase review was run here: the integrator explicitly owns the
|
||||
combined quick gate. All production changes outside MG_Remote are inside
|
||||
`#if MOBILEGL_BUILD_DISAGGREGATED`, so pull code is structurally unaffected; this is not
|
||||
reported as a new measured G1 result. No `.def`, generated catalogue, contract, or opcode changed.
|
||||
|
||||
## Integration notes
|
||||
|
||||
1. Keep r2's StageAllocate implementation, its session doorbell hookup, telemetry semantics,
|
||||
and before-retire hook. This commit changes command wait policy in ClientSession only;
|
||||
it does not replace r2's codec implementation.
|
||||
2. Five new unit entries; the existing staged-shadow unit is strengthened to use a persistent
|
||||
coherent mapping. One new ordinary integration scenario, registered for both backends;
|
||||
`integration-split` remains the same 22 selected entries. The new scenario is
|
||||
`PixelStoreSweepScenario.ReadPixelsSwapsUnsignedShortComponentsAndPreservesPackGaps`.
|
||||
3. The known 27 inproc ordinary failures should no longer be relabeled wholesale as unchanged:
|
||||
the GLES FBO/RBO deletion case is now green on this isolated tree. The merged census must
|
||||
determine the new aggregate; no count was extrapolated from this single case.
|
||||
|
||||
Tools retained: `p5b-r1-codex-submodlinks.sh`, `p5b-r1-codex-build.sh`,
|
||||
`p5b-r1-codex-coherent-control.sh`; external backup directory above. No integration worktree,
|
||||
other package source/build directory, adb device, or remote ref was modified.
|
||||
@@ -0,0 +1,119 @@
|
||||
# P5b r2 — CI, controls and runner fixes
|
||||
|
||||
Range: `37fc4fdb..d50183cb`, isolated worktree `/home/swung/w7/p5b-r2`, branch `p5b/r2`.
|
||||
`6fa6a925` contains the CI/control fixes; `d50183cb` isolates finding 8's allocator/telemetry
|
||||
and harness seam for the r1 merge. `feat/disaggregated` remained `37fc4fdb` at completion.
|
||||
This is ID-73's implementation round, not another adversarial review.
|
||||
|
||||
## Findings and executed red-once evidence
|
||||
|
||||
| Finding | Change | Executed red-once / own diagnostic |
|
||||
|---|---|---|
|
||||
| 5 | Broad inproc CTest status is recorded, its JUnit uploaded and its pass/skip/abort/fail counts written to the job summary. Monolith, reduced split and controls remain hard gates; E1/E3 also run under `!cancelled()`. No temporary branch-trigger additions. | Executed the actual YAML shell body with an indebted CTest exit 8: fixed body records aborted=1 and reaches `CONTROLS_REACHED`. Removing only `|| inproc_rc=$?` exits 8 before that marker. `p5b-r2-probes.log`: `F5 RED: fail-fast restored -> controls not reached (exit 8)`; transcript `p5b-r2-probes/5-census-step.log`. |
|
||||
| 6 | Pull-library control saves/restores the exact frozen split library through an EXIT trap; both retrace controls assert `nm` identity, including nm failure. Draw-drop receives the frozen path explicitly from CI. Output restoration also runs on failure/interruption. Smoke checks compare the restored library bytes. | Real CTest OpenRA baseline, pull control, then draw-drop ran in workflow order. SHA256 unchanged after pull control. Preplacing the pull library at that path makes draw-drop fail before replay: `draw-drop control requires a split library: MG_Remote=0`. `p5b-r2-retrace.log`, `p5b-r2-retrace/{baseline,pull-control,drop-control,wrong-library-red}.log`. |
|
||||
| 7 | E3(a) selects the two pixel cases explicitly in default/SmallRing lanes, excluding `TheMapLandsInTheArmItsLaneDeclares` with the by-design reason in the script. Pixel evidence is required in EACH selected JUnit testcase, not the combined transcript. Added E3-specific pre-flight-skip smoke case. | Restoring the broad selection on the live build fails: `SplitLogPaths FAILED: E3(a) control: the knob killed the pre-flight, not the entry - 2 selected entries skipped` (`p5b-r2-e3-red.log`). Restored selection: four own pixel failures plus four own private disabled-push diagnostics. E3 pre-flight skip remains rejected in smoke. |
|
||||
| 8 | `ringwaits` now counts an allocation blocked on outstanding retirement after immediate reclaim failed, once per allocation. The live encoder receives the session's shutdown-aware doorbell and waits/retries outstanding staging marks; unsubmitted exhaustion and individually oversized blobs still refuse. SmallRing drives two ordinary 768 KiB GL uploads with a scheduling-only before-retire hook, then asserts both wrap and wait. New unit case proves lazy reclamation alone is zero waits. | `p5b-r2-ring-red.log`: 1 MiB arm green; raising command AND staging rings to 8 MiB fails on both `NEVER WENT ROUND` and `producer NEVER WAITED`. Logs `p5b-r2-ring-{1,8}.{out,log}`. At 1 MiB: maxrec 784, cap 524288, cmdbytes 1315256, ringwraps 1, ringpads 0, ringwaits 1, emitseq 24328. |
|
||||
| 9 | All-skip control baseline is a hard failure; baseline CTest errors cannot arm controls. CI and local baseline invoke shared `junit_tally.py --require-split-ran`, filtering out the metadata test so it cannot arm an empty runtime lane. | Actual control with all-skipping stub: `split baseline FAILED: every DirectGLES.Split. entry SKIPPED; the split implementation did not execute`. `p5b-r2-probes/9-all-skip.log`; smoke now expects failure. |
|
||||
| 10 | `MOBILEGL_ITEST_REQUIRE_GPU=1` on the split unit step and local split unit invocation. Local gate now propagates that unit exit code rather than swallowing it in a summary substitution. | Real `ServerLoopEglTest.MakeCurrentBindsNativelyOncePerContextAndNeverForwardsAClientRelease`, invalid EGL vendor: optional arm SKIPPED/exit 0; required arm fails with `a failure, not a skip`. `p5b-r2-probes/10-{optional,required}.log`. |
|
||||
| 11 | Both external census Python runners truncate each private log BEFORE launch, count actual gtest RUN records, reject zero-run success as `harness-error`, distinguish segfault/abort, and do not let skip text override failure. Timeout bytes and launch errors handled. The shell runner preserves its output directory and rewrites this run's artifacts instead of recursive deletion. | Both exact runners executed against a nonexistent gtest selection: `HARNESS ERROR: zero executed tests`. Both preseeded with stale FATAL evidence, then launched a nonexistent executable: rc=127, current private log empty, fatal=None. `p5b-r2-probes.log` and its `r2-{zero,stale}-*` artifacts. Real census reproduced 432/185/505/27. |
|
||||
| 12 | Both pin self-tests call production `extract_baseline` using real historical refs with drifted pinned bodies (p3a ff2994d9; p4a 37da3c3a); their negative arm also uses production baseline extraction. They no longer directly call the pin helper. | Removed ONLY the production `apply_pinned_shas` call, leaving helper intact. Both self-tests failed with `PIN CONTROL FAILED: a baseline that carried a DIFFERENT sha for FlushPendingRangesFrom...`. `p5b-r2-p3a-red.log`, `p5b-r2-p4a-red.log`; restored positives `p5b-r2-{p3a,p4a}.log`. |
|
||||
| 13 | c1f mutation now targets ID-67's different-tuple adoption case. Both different-tuple and identical-tuple cases must pass in baseline/restored runs; the adoption mutation is expected to redden only different-tuple. | Full shipped mutation runner: 15 named REDs, restored GREEN, `FAILED_CONTROLS=[]` in `p5b-r2-c1f-full.log`; focused `codex12-repeat-skip: RED build=0 run=1` in `p5b-r2-c1f.log`. The obsolete name is no longer selected. |
|
||||
|
||||
All evidence paths above are under `/home/swung/w7/`. External driver sources are
|
||||
`notes/tools/p5b-r2-{setup,build,gate}.sh`, `notes/tools/p5b-r2-{probes,retrace,ring-red}.py`;
|
||||
the owned link toggle is `p5b-r2-submodlinks.sh`. `p5b-r2-venv` supplies PyYAML for YAML probes.
|
||||
The real retrace uses existing host replay tooling and hydrated fixtures read from pipe; no
|
||||
LFS fetch, adb access, or write to another worktree/build directory occurred.
|
||||
|
||||
## Integration seam and external files
|
||||
|
||||
Finding 8 necessarily meets r1 finding 4 at `PipeWireEncoder::StageAllocate`.
|
||||
This branch adds the smallest live staging wait needed to make a truthful wait counter and
|
||||
SmallRing gate execute: `ClientSession` supplies its own doorbell; immediate reclaim is not
|
||||
counted; only pending published marks are waited on, with a five-second shutdown-aware bound.
|
||||
It does not fix command-ring exhaustion or claim closure of r1's full finding 4.
|
||||
Integrator must preserve r1's retirement policy and these counter semantics when merging,
|
||||
plus the before-retire scheduling hook used by the GL-driven SmallRing gate. The hook does
|
||||
not publish watermarks, allocate bytes or modify counters. All of this is split-only.
|
||||
|
||||
Changes outside git, applied in place: `/home/swung/w7/p5b-c0b-census.{sh,py}`,
|
||||
`/home/swung/w7/notes/tools/p6_census_lane.py`, and minimal split unit/baseline/control
|
||||
exit propagation plus ID-65 wording in `/home/swung/w7/notes/tools/wsl_p5_gate.sh`.
|
||||
These must travel with the integration tooling; a git merge alone cannot install them elsewhere.
|
||||
|
||||
## Corrections for the docs close-out (docs deliberately not edited)
|
||||
|
||||
1. x2's “all three obligations closed” was package-local evidence, not final workflow proof.
|
||||
Old SmallRing asserted wrap only, and old ringwaits counted reclaim attempts. This report
|
||||
supplies the combined control sequence and actual wait assertion.
|
||||
2. The unmapped E5 staged-shadow control does not prove coherent-map wire-only correctness.
|
||||
r1 finding 1 owns server re-entry; old `pmap` figures cannot establish exclusive transport.
|
||||
3. `SlotCaps.h`'s “frontend CPU readback, which is exact” is too broad: 22 retained texture
|
||||
readback failures contradict it. A phase waiver is not correctness.
|
||||
4. “none is v1's” and describing all 27 wrong answers as texture readback are unsupported.
|
||||
The record separates 22 texture readbacks, three queries, one inspection and one FBO/RBO
|
||||
lifetime path. The latter uses ReadPixels; its full pixel causality was not isolated.
|
||||
5. Old 426/185/511/27 comprises 505 class-C aborts AND six P5 overbroad
|
||||
StageSnapshotTooNarrow refusals. Those six were fixed by v1 r3, not waived to P8/P11.
|
||||
The reproduced review-head census is 432/185/505/27, all 505 UnmigratedVerb.
|
||||
6. MEASUREMENTS max-record/census/A-B “pending” entries need the later measured reports.
|
||||
Minecraft inproc aborts belong to ID-66/68's P5b exit work; device barrier cost remains
|
||||
unmeasured, not inferred from splitctl or aborted benchmarks.
|
||||
7. Historical joint G1/G2/G14, unit/retrace counts, RSS and controls remain historical and
|
||||
revision-qualified. CTest totals include skips. E3(a)'s old “6 red” was 4 red/2 skip;
|
||||
ID-42/audit had documented skips; rsp=35 is a lower bound; inproc RSS is shared-process,
|
||||
not two independent peaks or an N-frame slope. No old summary proves a final merged workflow.
|
||||
8. Final-head full-gate/APK evidence was unavailable to the review (the gate log was then
|
||||
empty; the APK report used e61d0012). The docs close-out must cite completed, pinned new
|
||||
evidence rather than relabel package or joint gates. c1f's historical 15/15 used the old
|
||||
case name; this report supplies a reproducible current-name run.
|
||||
|
||||
## ROADMAP-shaped rows for previously unnamed debts
|
||||
|
||||
| Phase / proposed owner | Concrete gap | Retirement gate / status |
|
||||
|---|---|---|
|
||||
| P5b r1 / buffer + server | Coherent-map server re-entry into the client producer, including frontend serial/aggregate writes and direct adapter bypass | Suppress only client pre-draw push; coherent mapped draws must expose missing transported bytes; producer must refuse server-role access. Assigned r1, not claimed fixed here. |
|
||||
| P5b r1 / readback | Client multibyte PACK_SWAP_BYTES conversion after tight server readback | One-pixel non-symmetric UNSIGNED_SHORT read matches monolith with swap toggled; r1 finding 3. |
|
||||
| P5b r1 + r2 / transport | Applied-before-retired scheduling and allocation wait/retry policy | Delayed-retirement uploads wait without false RingOverrun; command ring and shutdown policy covered by r1; r2 adds actual staging-wait telemetry and GL SmallRing proof. |
|
||||
| P5b r2 / CI controls | Combined control lifecycle and own-entry red selection | Restore exact frozen split library, run pull then draw-drop; E3 skips cannot satisfy pixel control; broad census cannot suppress reduced gates. Closed by this package's executed controls. |
|
||||
| P7/P8 / PipeInputs owner (assignment proposed) | Sticky/non-verb forwards bypass rsp accounting and strict checks when no verb is stamped | Count and classify every non-verb frontend read; deleting one classification must produce its own named gate failure. Existing rsp is a lower bound until then. |
|
||||
| P5b follow-up / client reply protocol (assignment proposed) | Exact-sized ReadPixels reply status 3 is accepted though only 0/1/2 are valid | Adversarial live reply sink posts status 3 with correct extent; client must reject by named protocol diagnostic. Production emission of status 3 remains unverified. |
|
||||
|
||||
Previously named context-loss restaging, PACK-PBO, oversized readback, GetCaps/ABI sizing,
|
||||
texture readback, query and inspection debts are not relabeled as new debts here.
|
||||
|
||||
## Final acceptance
|
||||
|
||||
| Gate | Result on the final source |
|
||||
|---|---|
|
||||
| Split unit, REQUIRE_GPU=1 | **2087 selected, zero failures** (2086 baseline + lazy-reclaim unit case; pre-existing non-EGL skips remain) |
|
||||
| inproc integration-split, REQUIRE_GPU=1 | **22 selected, zero failures**, including the two documented map-arm body skips; metadata is excluded from the runtime baseline tally |
|
||||
| E1 / E3(a), exact CI script | **14 own barrier reds / 4 own pixel+private-log reds**, script exit 0 |
|
||||
| OpenRA E2, real CTest | Baseline **SSIM 1.0**; sequential pull-library and draw-drop controls exit 0; draw-drop **0.000036**, **758** drops |
|
||||
| Smoke / smoke redcheck | **16 core + 12 private-log cases**, zero failures; `P5_T1_CONTROL_SMOKE_REDCHECK_OK` |
|
||||
| Full c1f mutation runner | **15/15 RED**, every restored baseline green, `FAILED_CONTROLS=[]` |
|
||||
| Census on this tree | **432 passed / 185 skipped / 505 aborted / 27 failed**, 1149 total; all 505 aborts carry UnmigratedVerb, zero segfault / zero harness-error |
|
||||
| Push integration-gpu | **1128 selected, zero failures** |
|
||||
| G5 pin self-tests | Both green; deleting each production pin-selection call fails on its own PIN CONTROL FAILED message |
|
||||
| YAML | `test.yml` and `apk.yml` parse successfully; actual changed census shell body executed by the red-once probe |
|
||||
| G1 | **Unaffected by construction: no pull-reachable production code changed.** MG_Remote implementation/header changes are split-only; other C++ changes are test harnesses/tests. No new pull binary comparison is claimed; inherited 0/0/0/0 and `.text +0` are not relabeled as a new measurement. |
|
||||
|
||||
Final gate driver exits 0 with `UNIT_OK SPLIT_OK CONTROLS_OK SMOKE_OK CENSUS_OK PUSH_OK`
|
||||
in `p5b-r2-final-gate.log`. Authoritative test transcripts: `p5b-r2-{unit,split,push,controls,
|
||||
smoke,redcheck,census}.log`, census per-entry evidence `p5b-r2-census-logs/results.json` and
|
||||
its adjacent fresh private logs/output, and `p5b-r2-controls/` (JUnit/control evidence).
|
||||
The intentional E3 broad-selection run also retains `p5b-r2-controls-red/`.
|
||||
|
||||
Retained configure/build evidence: `p5b-r2-config-{split,push}.log`,
|
||||
`p5b-r2-build-{split,push,wait,wait-final,push-final}.log`. `p5b-r2-wait.log` is an earlier
|
||||
focused SmallRing run; the later two-arm `p5b-r2-ring-red.log` is the final wait control.
|
||||
The initial focused c1f attempt hit a CMake regeneration timeout scanning the inherited Windows
|
||||
PATH; it was stopped before source mutation, and the cited successful runs use
|
||||
`PATH=/usr/bin:/bin`. No failed build was accepted as a mutation red.
|
||||
|
||||
Commit driver: `notes/tools/p5b-r2-commit.sh`; identity and exact SHAs in
|
||||
`p5b-r2-commit.log`. Both commits are one-line messages with no trailers, committer
|
||||
`Swung0x48 <swung0x48@outlook.com>`. `git diff --check` clean; final worktree clean,
|
||||
submodule links **off**. No integration-head merge was needed.
|
||||
Final clean-tree, links-off, unchanged integration head and per-abort attribution check:
|
||||
`notes/tools/p5b-r2-final-check.py`, transcript `p5b-r2-final-check.log`.
|
||||
@@ -0,0 +1,53 @@
|
||||
# P5b sync migration — 2026-09-16
|
||||
|
||||
Commit: `08fd9e9ca1888d795654fe133968526411be334c`, parent `9606466a`.
|
||||
Worktree: `/home/swung/w7/p5b-d1-codex`. Integrator-approved contract is CONTRACT-P5B §9.
|
||||
|
||||
## Result
|
||||
|
||||
Five existing opcodes now serve FenceSync, ClientWaitSync, GetSyncStatus, WaitSync and
|
||||
DeleteSync. The client allocates Fence-kind handles and retains local opaque proxies; the
|
||||
wire carries only `{slot, gen}`. The apply thread owns native fences, validates generations,
|
||||
preserves timeout/failed wait results, and deletes native orphan fences before backend detach.
|
||||
Frontend orphan cleanup after session shutdown releases only proxies. New slots legitimately
|
||||
start at generation zero; recycled generations must increase, and stale/destroyed handles
|
||||
are protocol corruption. No host pointer crosses the wire.
|
||||
|
||||
MGPFenceWait is 24 bytes (was 16), with Flags/Pad0 appended; ClientWaitSync retains its flush
|
||||
flag. FenceStatus and FenceWait use their existing reply slots with a 4-byte answer. All five
|
||||
opcodes have matching verb stamps. Field declarations and generated stamp tables agree.
|
||||
The class counts after this commit are A=2/B=53/C=16 (before integrator's named-blit migration).
|
||||
|
||||
The frontend FenceSync guard previously returned nullptr for every split session through
|
||||
MGL_BACKEND_SLOT_PTR_LOCAL. This explains why the dynamic d1 census had no FenceSync first
|
||||
blocker despite many trace calls. The guard now reads the class-B emitter pointer; in a pull
|
||||
build this is the exact expression the macro previously produced. A backend with no native
|
||||
fence slot, or a fence creation that returns nullptr, retains GL_Sync.cpp's old monolith
|
||||
always-signaled fallback **on the server**. Real native wait answers are never replaced.
|
||||
|
||||
## Validation
|
||||
|
||||
- Split build successful, clang Release, maximum -j8.
|
||||
- Related unit selection: 40/40 (RemoteEmitTable/FieldOwnership/FenceWireRoundTrip at that
|
||||
point). Final FenceWireRoundTrip selection after adding the decline control: 4/4.
|
||||
- Real GL `DirectGLES.Split.SyncWireScenario` plus private-log gate: 2/2.
|
||||
- The GL scenario asserts all five sync records crossed, waits for GPU completion, validates
|
||||
rendered pixels and leaves one orphan to exercise session teardown.
|
||||
- Wire controls preserve timeout (rather than silently returning signaled), flush flags and
|
||||
64-bit timeout; reject stale reused handles; verify native deletion and null-native fallback;
|
||||
missing consumers produce DECLINED with no manufactured answer.
|
||||
- Representative Minecraft retraces: 4/4. Sodium DirectGLES/DirectVulkan SSIM 0.999986;
|
||||
vanilla in-world both 0.999976. Same values as standalone d1.
|
||||
- gen_pipe.py --check and gen_pipe_field_ownership.py --check pass; git diff --check passes.
|
||||
- No full-stage gate or review repeated. G1 and remaining quick lanes belong to the integrator.
|
||||
|
||||
Logs: `/home/swung/w7/p5b-sync-{unit,unit-final,integration}.log`;
|
||||
trace summary `/home/swung/w7/retrace-out/p5b-sync-targets/summary.json`.
|
||||
|
||||
## Integration notes
|
||||
|
||||
The four A/B trace targets' immediate missing verb was named framebuffer blit, owned by the
|
||||
integrator; sync completes real fence semantics rather than removing that particular blocker.
|
||||
Cherry-picking over the integrator's blit change may need B/C count arithmetic combined and
|
||||
both new ServerVerbSink declarations retained. r1 was deliberately not merged here.
|
||||
No backend implementation changes, no adb, no push.
|
||||
@@ -0,0 +1,200 @@
|
||||
# P5b t2 Codex takeover report
|
||||
|
||||
Date: 2026-09-16. Code: `p5b/t2-codex@fc5b52e4`; clean independent worktree `/home/swung/w7/p5b-t2-codex`. Original `/home/swung/w7/p5b-t2` had no uncommitted non-submodule changes. No source changes were needed during takeover; the already committed implementation and completed gates below are reused. No new all-lane run, adversarial review, adb operation or push was performed.
|
||||
|
||||
## Implementation and commits
|
||||
|
||||
- `c240557f`: capture ownership cap, server publication and client probe.
|
||||
- `dd293203`: six client emitters and six server sinks for Begin/End/Pause/ResumeTransformFeedback, BindTransformFeedback and PatchParameteri.
|
||||
- `c831ab03`: wire ordering/accounting and six-slot tests; DirectGLES inproc tessellation and buffer-reuse lanes.
|
||||
- `fc5b52e4`: merge P5 final integration head `37fc4fdb`.
|
||||
|
||||
Records carry the GL arguments and the contract accounting/lifetime fields. The pre-verb read-only hook stays before every control record. Vulkan null backend slots decline exactly as the monolith guards did; XFB capture ownership uses `kCapBackendOwnsXfbCapture` from the server table. Existing EndTransformFeedback capture-target marking and buffer visibility rules remain. DeleteTransformFeedback deliberately remains named-unmigrated (no P5b record). `set_stream_output_targets` remains deferred under the barrier, as CONTRACT-P5B §2 requires. No contract deviation or backend-source change.
|
||||
|
||||
## Reused completed package evidence
|
||||
|
||||
| Check | Result | Evidence under `/home/swung/w7` |
|
||||
|---|---|---|
|
||||
| Four builds and unit lanes | 1817 / 1817 / 1817 / 2092, all green | `p5b-t2-postmerge.log`, `p5b-t2-unit-*.log` |
|
||||
| G1 pull equivalence | 27,814 symbols; 0 added/removed/resized/renamed; .text 10,806,611 (+0) | `p5b-t2-symbols.md`, `p5b-t2-postmerge.log` |
|
||||
| Code generators / ownership checks / purity | green | `p5b-t2-postmerge.log` |
|
||||
| Push integration-gpu | 1128/1128, monolith | `p5b-t2-igpu-push.log` |
|
||||
| Split integration-split | 33/33 nonfailure, including 2 existing skips; 11 new t2 entries pass | `p5b-t2-isplit.log` |
|
||||
| G2 and G14 | pull-vs-push zero name differences; baseline removed 0, added 43 | `p5b-t2-postmerge.log`, `p5b-t2-names-*.txt` |
|
||||
| G5 | both unchanged-region checks passed | `p5b-t2-p3a_untouched_regions.sh.log`, `p5b-t2-p4a_untouched_regions.sh.log` |
|
||||
|
||||
All four original build directories still contain `CTestTestfile.cmake`. These are package results at the t2 head, not evidence of a final merged-head gate. The older `p5b-t2-accept-run1.log` includes a missing census-script invocation and stale 1149-name output; it is not used as successful census evidence. The completed postmerge log and current 1160-name results are the authoritative reused outputs.
|
||||
|
||||
## Census and next blockers
|
||||
|
||||
| Cohort | Passed | Skipped | Aborted | Failed |
|
||||
|---|---:|---:|---:|---:|
|
||||
| c0b baseline, 1149 names | 432 | 185 | 505 | 27 |
|
||||
| t2 on the same 1149 names | 490 | 191 | 390 | 78 |
|
||||
| t2 complete manifest, 1160 names | 501 | 191 | 390 | 78 |
|
||||
|
||||
All 432 baseline passing cases remain passing; all 185 prior skips and 27 prior failures retain their statuses. The 51 additional failures were previously aborted. They are newly exposed wrong answers, not cases that regressed from passing. Six further previously aborted cases reach designed skips. The 11 newly added inproc t2 names all pass.
|
||||
|
||||
The measured t2 first-blocker cohort is **140**, not 138: BeginTransformFeedback 95 + PatchParameteri 43 + BindTransformFeedback 2. All six migrated slot names have **zero remaining UnmigratedVerb aborts**. Outcomes for the 140 cases: 58 passed, 6 skipped, 51 failed, 18 DrawArraysInstanced, 4 DrawArraysInstancedBaseInstance, 2 DeleteTransformFeedback, 1 UnimplementedWritebackWait.
|
||||
|
||||
The 51 new wrong answers are all DirectVulkan: 47 capture/readback assertions observe the prefill/poison value or zero; four PrimitivesGeneratedNoXfb cases report counts 0 vs 1 or 1 vs 3. The capture visibility debt is consistent with the current DirectVulkan resource consumer mask withholding the resource family, so GPU-written resident capture data does not reach the client shadow through resource_readback. Existing probe logs confirm active capture state on both arms and STRICT_ERRORS first-refuses GetProgramForDraw@DrawArrays. This is scoped as the newly reached P7/P9 resource/readback and P7 query work, not claimed as a proven root-cause fix for every assertion. A later resource-consumer/readback implementation and green captured-byte assertions would retire it.
|
||||
|
||||
DeleteTransformFeedback remains the contract-defined next blocker until the XFB namespace/lifetime carrier lands. The single UnimplementedWritebackWait is reported by name below; it is not converted to a skip. No trace census was rerun: d1 owns the Minecraft trace gate.
|
||||
|
||||
### Every affected scenario and its next result
|
||||
|
||||
| Scenario | Previous first blocker | Current status / first observed result |
|
||||
|---|---|---|
|
||||
| `DirectGLES.GeometryDrawModeScenario.PointsInGeometryProgramRejectsNonPointModesOnFeedbackDraws` | `BindTransformFeedback` | **aborted**: {UnmigratedVerb, "DeleteTransformFeedback"} |
|
||||
| `DirectGLES.IoBlockNameCollisionScenario.DistinctlyNamedBlocksCarryThePayloadThroughFiveStages` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.IoBlockNameCollisionScenario.OneBlockNameInBothDirectionsStillCarriesThePayload` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotion.PointSizeDemotionScenario.ACaptureSurvivesAStageThatOnlyReadsThePointSize` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotion.PointSizeDemotionScenario.AGeometryOnlyChainCarriesTheVertexValue` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotion.PointSizeDemotionScenario.TheDemotionIsActuallyArmedWhenTheEnvironmentPinsItOn` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotion.PointSizeDemotionScenario.TheEvaluationStageOwnsTheCaptureWithoutAGeometryStage` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotion.PointSizeDemotionScenario.TheValueSurvivesTheFiveStageChainIntoTheCapture` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotionScenario.AGeometryOnlyChainCarriesTheVertexValue` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotionScenario.TheEvaluationStageOwnsTheCaptureWithoutAGeometryStage` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.PointSizeDemotionScenario.TheValueSurvivesTheFiveStageChainIntoTheCapture` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.RelinkStageSetScenario.RelinkingToAddATessEvalStageRunsTheNewExecutable` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationDrawModeScenario.PatchesRejectedWithoutATessellationEvaluationStage` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationDrawModeScenario.TessellationProgramRejectsNonPatchModes` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationXfbCaptureScenario.CapturesFromASeparableEvaluationProgramInAPipelineObject` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationXfbCaptureScenario.CapturesGlPointSizeByNameFromTheGeometryStage` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationXfbCaptureScenario.CapturesGlPositionAndGlPointSizeByNameFromTheEvaluationStage` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationXfbCaptureScenario.CapturesGlPositionByNameFromTheEvaluationStage` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationXfbCaptureScenario.MapsTheCaptureBufferAfterEachOfTwoPatchDraws` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationXfbCaptureScenario.TheEvaluationStageSeesGlPointSizeAcrossItsPatch` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.TessellationXfbCaptureScenario.TheEvaluationStageSeesTheUserPerVertexBlockOfItsPatch` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnlocatedIoBlockScenario.BlocksCarryTheirPayloadThroughFiveStages` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnlocatedIoBlockScenario.BlocksNamedInBothDirectionsStillMeetWithoutLocations` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnlocatedIoBlocks.UnlocatedIoBlockScenario.BlocksCarryTheirPayloadThroughFiveStages` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnlocatedIoBlocks.UnlocatedIoBlockScenario.BlocksNamedInBothDirectionsStillMeetWithoutLocations` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnlocatedIoBlocks.UnlocatedIoBlockScenario.TheEmulationIsActuallyArmedWhenTheEnvironmentPinsItOn` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnwrittenPositionOutputScenario.ARedeclaredButUnwrittenPositionStillDraws` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnwrittenPositionOutputScenario.AShaderWithNoPositionBlockStillDraws` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnwrittenPositionOutputScenario.AWrittenRedeclaredPositionStillDraws` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.UnwrittenPositionOutputScenario.CapturingAWrittenPositionStillDraws` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.VertexArrayEnableDisableScenario.EitherHalfOfTheAttributesInTurn` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.VertexArrayEnableDisableScenario.TheEvenHalfAlone` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.VertexArrayEnableDisableScenario.TheOddHalfAlone` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.VertexAttribBindingScenario.BaseInstanceLeavesPerVertexArraysWhereTheyWere` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstancedBaseInstance"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.BaseInstanceMovesTheInstancedArraysStartElement` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstancedBaseInstance"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.BindingDivisorAppliesToEveryAttributeOnThePoint` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.BindingStrideZeroRepeatsOneElement` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.DoubleArrayIsFetchedAtFloat32Precision` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.EveryInputArrayElementGetsItsOwnCurrentValue` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.VertexAttribBindingScenario.FormatAndBindingFeedTheDraw` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.FormatBeforeBufferStillResolves` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.InputArrayCaptureProgramFeedsTheDraw` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.VertexAttribBindingScenario.LongDoubleArrayIsFetchedAtFloat32Precision` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.NormalizedIsIgnoredForDoubleArrays` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.PointerStrideZeroStaysTightlyPacked` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.VertexAttribBindingScenario.RelativeOffsetComposesWithBindingOffset` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectGLES.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceEnables` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadLines8` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadPoints1` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbAfterClipDistanceScenario.SkipComponentsCaptureAlone` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbAfterClipDistanceScenario.SkipComponentsCaptureSurvivesEveryClipWorkloadStopPoint` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbCaptureBufferReuseScenario.ARespecificationMayChangeTheCaptureBufferSize` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbCaptureBufferReuseScenario.EverySpanIntoABufferObjectOfItsOwn` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbCaptureBufferReuseScenario.EverySpanIntoOneImmutableStorageBuffer` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbCaptureBufferReuseScenario.EverySpanIntoOneRespecifiedBufferObject` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbPrimitiveQueryScenario.ACaptureThatFitsReportsEveryPrimitiveOnBothTargets` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbPrimitiveQueryScenario.AnOverflowingVertexOnlyCaptureStopsWritingAtTheBufferCapacity` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbRepeatedCaptureScenario.ACaptureFromAFragmentlessProgramRecords` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbRepeatedCaptureScenario.ACaptureFromAPatchesDrawRecords` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbRepeatedCaptureScenario.ACaptureFromAnAdjacencyDrawRecords` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbRepeatedCaptureScenario.ACaptureListBeginningWithGlNextBufferSparesTheEarlierBuffer` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbRepeatedCaptureScenario.ASecondGeometryCaptureAfterADeqpStateResetStillRecords` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectGLES.XfbRepeatedCaptureScenario.ASpanThatNeverDrawsLeavesTheCaptureBufferAlone` | `BeginTransformFeedback` | **aborted**: {UnimplementedWritebackWait} - a ClientSession is active and buffer 2 still has an outstanding GPU write after its readback was emitted. The wait is ClientSession::EmitAndWait's (R-3: the reply slot id IS the record seq); P5 package b1 landed the third state and s1/c1 own the wait itself. |
|
||||
| `DirectGLES.XfbRepeatedCaptureScenario.AVertexOnlyAdjacencyCaptureRecords` | `BeginTransformFeedback` | **skipped**: DirectGLES cannot forward a geometry-shader-less adjacency draw: ES 3.2 10.1 supports the adjacency primitive types only with a geometry stage. The frontend no longer rejects the draw (checked above), which is the defect this covers. / |
|
||||
| `DirectVulkan.AsyncCompile.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceEnables` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:526: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceEnables (25 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (25 ms total) |
|
||||
| `DirectVulkan.AsyncCompile.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadLines8` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:540: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadLines8 (353 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (353 ms total) |
|
||||
| `DirectVulkan.AsyncCompile.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadPoints1` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:552: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadPoints1 (95 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (95 ms total) |
|
||||
| `DirectVulkan.AsyncCompile.XfbAfterClipDistanceScenario.SkipComponentsCaptureAlone` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:511: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAlone (32 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (32 ms total) |
|
||||
| `DirectVulkan.AsyncCompile.XfbAfterClipDistanceScenario.SkipComponentsCaptureSurvivesEveryClipWorkloadStopPoint` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:575: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / (stop point: primitive 0, redeclaration 0, clip count 1) / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:575: Failure |
|
||||
| `DirectVulkan.GeometryDrawModeScenario.PointsInGeometryProgramRejectsNonPointModesOnFeedbackDraws` | `BindTransformFeedback` | **aborted**: {UnmigratedVerb, "DeleteTransformFeedback"} |
|
||||
| `DirectVulkan.IoBlockNameCollisionScenario.DistinctlyNamedBlocksCarryThePayloadThroughFiveStages` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.IoBlockNameCollisionScenario.OneBlockNameInBothDirectionsStillCarriesThePayload` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.PointSizeDemotion.PointSizeDemotionScenario.ACaptureSurvivesAStageThatOnlyReadsThePointSize` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:457: Failure / Value of: ComponentIs(captured, 0, 7.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 7)) / Expected: true / the unrelated varying captured alongside gl_PointSize did not survive; the point-size capture took the whole set with it / / [ FAILED ] PointSizeDemotionScenario.ACaptureSurvivesAStageThatOnlyReadsThePointSize (5 ms) |
|
||||
| `DirectVulkan.PointSizeDemotion.PointSizeDemotionScenario.AGeometryOnlyChainCarriesTheVertexValue` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:412: Failure / Value of: ComponentIs(captured, 0, 8.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 8)) / Expected: true / / [ FAILED ] PointSizeDemotionScenario.AGeometryOnlyChainCarriesTheVertexValue (6 ms) / [----------] 1 test from PointSizeDemotionScenario (6 ms total) |
|
||||
| `DirectVulkan.PointSizeDemotion.PointSizeDemotionScenario.TheDemotionIsActuallyArmedWhenTheEnvironmentPinsItOn` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:507: Failure / Value of: ComponentIs(captured, 0, 8.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 8)) / Expected: true / the pinned-on lane did not even capture correctly / / [ FAILED ] PointSizeDemotionScenario.TheDemotionIsActuallyArmedWhenTheEnvironmentPinsItOn (6 ms) |
|
||||
| `DirectVulkan.PointSizeDemotion.PointSizeDemotionScenario.TheEvaluationStageOwnsTheCaptureWithoutAGeometryStage` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:385: Failure / Value of: ComponentIs(captured, 0, 12.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 12)) / Expected: true / / [ FAILED ] PointSizeDemotionScenario.TheEvaluationStageOwnsTheCaptureWithoutAGeometryStage (7 ms) / [----------] 1 test from PointSizeDemotionScenario (7 ms total) |
|
||||
| `DirectVulkan.PointSizeDemotion.PointSizeDemotionScenario.TheValueSurvivesTheFiveStageChainIntoTheCapture` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:361: Failure / Value of: ComponentIs(captured, 0, 24.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 24)) / Expected: true / / [ FAILED ] PointSizeDemotionScenario.TheValueSurvivesTheFiveStageChainIntoTheCapture (9 ms) / [----------] 1 test from PointSizeDemotionScenario (9 ms total) |
|
||||
| `DirectVulkan.PointSizeDemotionScenario.ACaptureSurvivesAStageThatOnlyReadsThePointSize` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:457: Failure / Value of: ComponentIs(captured, 0, 7.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 7)) / Expected: true / the unrelated varying captured alongside gl_PointSize did not survive; the point-size capture took the whole set with it / / [ FAILED ] PointSizeDemotionScenario.ACaptureSurvivesAStageThatOnlyReadsThePointSize (7 ms) |
|
||||
| `DirectVulkan.PointSizeDemotionScenario.AGeometryOnlyChainCarriesTheVertexValue` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:412: Failure / Value of: ComponentIs(captured, 0, 8.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 8)) / Expected: true / / [ FAILED ] PointSizeDemotionScenario.AGeometryOnlyChainCarriesTheVertexValue (7 ms) / [----------] 1 test from PointSizeDemotionScenario (7 ms total) |
|
||||
| `DirectVulkan.PointSizeDemotionScenario.TheEvaluationStageOwnsTheCaptureWithoutAGeometryStage` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:385: Failure / Value of: ComponentIs(captured, 0, 12.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 12)) / Expected: true / / [ FAILED ] PointSizeDemotionScenario.TheEvaluationStageOwnsTheCaptureWithoutAGeometryStage (7 ms) / [----------] 1 test from PointSizeDemotionScenario (7 ms total) |
|
||||
| `DirectVulkan.PointSizeDemotionScenario.TheValueSurvivesTheFiveStageChainIntoTheCapture` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PointSizeDemotionScenario.cpp:361: Failure / Value of: ComponentIs(captured, 0, 24.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 24)) / Expected: true / / [ FAILED ] PointSizeDemotionScenario.TheValueSurvivesTheFiveStageChainIntoTheCapture (8 ms) / [----------] 1 test from PointSizeDemotionScenario (8 ms total) |
|
||||
| `DirectVulkan.PrimGenReroute.PrimitivesGeneratedNoXfbScenario.APausedSpanCountsACpuPricedDrawExactlyOnce` | `BeginTransformFeedback` | **skipped**: no counter this backend can reach answers a capture-less draw on this stack, so the paused half of the comparison proves nothing; the bring-up probe measures the same hole and the POST row reports it / |
|
||||
| `DirectVulkan.PrimGenReroute.PrimitivesGeneratedNoXfbScenario.APausedSpanCountsATessellatedPatchExactlyOnce` | `PatchParameteri` | **skipped**: no counter this backend can reach answers a capture-less patch draw on this stack, so the paused half proves nothing; the bring-up probe measures the same hole and the POST row reports it / |
|
||||
| `DirectVulkan.PrimGenReroute.PrimitivesGeneratedNoXfbScenario.ASpanMixingActiveAndInactiveDrawsAccumulatesBoth` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PrimitivesGeneratedNoXfbScenario.cpp:395: Failure / Expected equality of these values: / generated / Which is: 1 / 3u / Which is: 3 / one triangle before the span, one inside it, one after |
|
||||
| `DirectVulkan.PrimGenReroute.PrimitivesGeneratedNoXfbScenario.CountsATessellatedPatchWithNoCaptureSpan` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PrimitivesGeneratedNoXfbScenario.cpp:356: Failure / Expected equality of these values: / generated / Which is: 0 / 1u / Which is: 1 / a triangles-domain patch with every level 1 tessellates to exactly one triangle, and GL_PRIMITIVES_GENERATED must say so with no capture active |
|
||||
| `DirectVulkan.PrimitivesGeneratedNoXfbScenario.APausedSpanCountsACpuPricedDrawExactlyOnce` | `BeginTransformFeedback` | **skipped**: no counter this backend can reach answers a capture-less draw on this stack, so the paused half of the comparison proves nothing; the bring-up probe measures the same hole and the POST row reports it / |
|
||||
| `DirectVulkan.PrimitivesGeneratedNoXfbScenario.APausedSpanCountsATessellatedPatchExactlyOnce` | `PatchParameteri` | **skipped**: no counter this backend can reach answers a capture-less patch draw on this stack, so the paused half proves nothing; the bring-up probe measures the same hole and the POST row reports it / |
|
||||
| `DirectVulkan.PrimitivesGeneratedNoXfbScenario.ASpanMixingActiveAndInactiveDrawsAccumulatesBoth` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PrimitivesGeneratedNoXfbScenario.cpp:395: Failure / Expected equality of these values: / generated / Which is: 1 / 3u / Which is: 3 / one triangle before the span, one inside it, one after |
|
||||
| `DirectVulkan.PrimitivesGeneratedNoXfbScenario.CountsATessellatedPatchWithNoCaptureSpan` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/PrimitivesGeneratedNoXfbScenario.cpp:356: Failure / Expected equality of these values: / generated / Which is: 0 / 1u / Which is: 1 / a triangles-domain patch with every level 1 tessellates to exactly one triangle, and GL_PRIMITIVES_GENERATED must say so with no capture active |
|
||||
| `DirectVulkan.RelinkStageSetScenario.RelinkingToAddATessEvalStageRunsTheNewExecutable` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.TessellationDrawModeScenario.PatchesRejectedWithoutATessellationEvaluationStage` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.TessellationDrawModeScenario.TessellationProgramRejectsNonPatchModes` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.TessellationXfbCaptureScenario.CapturesFromASeparableEvaluationProgramInAPipelineObject` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:883: Failure / Value of: ComponentIs(captured, 0, 11.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 11)) / Expected: true / / ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:884: Failure / Value of: ComponentIs(captured, 1, 12.0f) |
|
||||
| `DirectVulkan.TessellationXfbCaptureScenario.CapturesGlPointSizeByNameFromTheGeometryStage` | `BeginTransformFeedback` | **skipped**: this backend cannot capture from a geometry stage at all, with or without gl_PointSize (DirectVulkan, Magma (MobileGL Core) (4.6.0)) / |
|
||||
| `DirectVulkan.TessellationXfbCaptureScenario.CapturesGlPositionAndGlPointSizeByNameFromTheEvaluationStage` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:395: Failure / Value of: ComponentIs(captured, 0, 11.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 11)) / Expected: true / / ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:396: Failure / Value of: ComponentIs(captured, 1, 12.0f) |
|
||||
| `DirectVulkan.TessellationXfbCaptureScenario.CapturesGlPositionByNameFromTheEvaluationStage` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:370: Failure / Value of: ComponentIs(captured, 0, 11.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 11)) / Expected: true / / ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:371: Failure / Value of: ComponentIs(captured, 1, 12.0f) |
|
||||
| `DirectVulkan.TessellationXfbCaptureScenario.MapsTheCaptureBufferAfterEachOfTwoPatchDraws` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:769: Failure / Value of: ComponentIs(captured, 0, 11.0f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 11)) / Expected: true / iteration 0 / / ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:770: Failure |
|
||||
| `DirectVulkan.TessellationXfbCaptureScenario.TheEvaluationStageSeesGlPointSizeAcrossItsPatch` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:558: Failure / Value of: ComponentIs(captured, static_cast<std::size_t>(c), referencePosition[c], 1e-2f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 1984)) / Expected: true / te_position.0 (gl_in[].gl_Position) / / ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:558: Failure |
|
||||
| `DirectVulkan.TessellationXfbCaptureScenario.TheEvaluationStageSeesTheUserPerVertexBlockOfItsPatch` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:558: Failure / Value of: ComponentIs(captured, static_cast<std::size_t>(c), referencePosition[c], 1e-2f) / Actual: false (component 0 still holds the poison value - the capture never reached these bytes (expected 1984)) / Expected: true / te_position.0 (gl_in[].gl_Position) / / ../MobileGL/MG_IntegrationTest/Scenarios/TessellationXfbCaptureScenario.cpp:558: Failure |
|
||||
| `DirectVulkan.UnlocatedIoBlockScenario.BlocksCarryTheirPayloadThroughFiveStages` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.UnlocatedIoBlockScenario.BlocksNamedInBothDirectionsStillMeetWithoutLocations` | `PatchParameteri` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.UnwrittenPositionOutputScenario.ARedeclaredButUnwrittenPositionStillDraws` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/UnwrittenPositionOutputScenario.cpp:208: Failure / Expected equality of these values: / readback[userVaryingOffset + i] / Which is: -1 / static_cast<float>(i + 1) / Which is: 1 / redeclared, never written: captured float 0 came back as -1; the pre-fill value means the draw never produced a vertex, which is what an invalid shader module looks like from out here |
|
||||
| `DirectVulkan.UnwrittenPositionOutputScenario.AShaderWithNoPositionBlockStillDraws` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/UnwrittenPositionOutputScenario.cpp:208: Failure / Expected equality of these values: / readback[userVaryingOffset + i] / Which is: -1 / static_cast<float>(i + 1) / Which is: 1 / no gl_PerVertex block: captured float 0 came back as -1; the pre-fill value means the draw never produced a vertex, which is what an invalid shader module looks like from out here |
|
||||
| `DirectVulkan.UnwrittenPositionOutputScenario.AWrittenRedeclaredPositionStillDraws` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/UnwrittenPositionOutputScenario.cpp:208: Failure / Expected equality of these values: / readback[userVaryingOffset + i] / Which is: -1 / static_cast<float>(i + 1) / Which is: 1 / redeclared and written: captured float 0 came back as -1; the pre-fill value means the draw never produced a vertex, which is what an invalid shader module looks like from out here |
|
||||
| `DirectVulkan.UnwrittenPositionOutputScenario.CapturingAWrittenPositionStillDraws` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/UnwrittenPositionOutputScenario.cpp:208: Failure / Expected equality of these values: / readback[userVaryingOffset + i] / Which is: -1 / static_cast<float>(i + 1) / Which is: 1 / capturing a written gl_Position: captured float 0 came back as -1; the pre-fill value means the draw never produced a vertex, which is what an invalid shader module looks like from out here |
|
||||
| `DirectVulkan.UnwrittenPositionOutputScenario.CapturingAnUnwrittenPositionStillDraws` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/UnwrittenPositionOutputScenario.cpp:208: Failure / Expected equality of these values: / readback[userVaryingOffset + i] / Which is: -1 / static_cast<float>(i + 1) / Which is: 1 / capturing an unwritten gl_Position: captured float 0 came back as -1; the pre-fill value means the draw never produced a vertex, which is what an invalid shader module looks like from out here |
|
||||
| `DirectVulkan.VertexArrayEnableDisableScenario.EitherHalfOfTheAttributesInTurn` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/VertexArrayEnableDisableScenario.cpp:235: Failure / Expected equality of these values: / DrawAndRead(0) / Which is: 0 / ExpectedSum(0) / Which is: 240 / even attributes |
|
||||
| `DirectVulkan.VertexArrayEnableDisableScenario.TheEvenHalfAlone` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/VertexArrayEnableDisableScenario.cpp:249: Failure / Expected equality of these values: / DrawAndRead(0) / Which is: 0 / ExpectedSum(0) / Which is: 240 / |
|
||||
| `DirectVulkan.VertexArrayEnableDisableScenario.TheOddHalfAlone` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/VertexArrayEnableDisableScenario.cpp:259: Failure / Expected equality of these values: / DrawAndRead(1) / Which is: 0 / ExpectedSum(1) / Which is: 256 / |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.BaseInstanceLeavesPerVertexArraysWhereTheyWere` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstancedBaseInstance"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.BaseInstanceMovesTheInstancedArraysStartElement` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstancedBaseInstance"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.BindingDivisorAppliesToEveryAttributeOnThePoint` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.BindingStrideZeroRepeatsOneElement` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.DoubleArrayIsFetchedAtFloat32Precision` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.EveryInputArrayElementGetsItsOwnCurrentValue` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/VertexAttribBindingScenario.cpp:666: Failure / Expected equality of these values: / data[element * 4 + 0] / Which is: -1234 / base / Which is: 1 / element 0 |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.FormatAndBindingFeedTheDraw` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.FormatBeforeBufferStillResolves` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.InputArrayCaptureProgramFeedsTheDraw` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/VertexAttribBindingScenario.cpp:590: Failure / Expected equality of these values: / data[0] / Which is: -1234 / 0.0f / Which is: 0 / |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.LongDoubleArrayIsFetchedAtFloat32Precision` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.NormalizedIsIgnoredForDoubleArrays` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.PointerStrideZeroStaysTightlyPacked` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.VertexAttribBindingScenario.RelativeOffsetComposesWithBindingOffset` | `BeginTransformFeedback` | **aborted**: {UnmigratedVerb, "DrawArraysInstanced"} |
|
||||
| `DirectVulkan.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceEnables` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:526: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceEnables (26 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (26 ms total) |
|
||||
| `DirectVulkan.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadLines8` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:540: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadLines8 (251 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (251 ms total) |
|
||||
| `DirectVulkan.XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadPoints1` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:552: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAfterClipDistanceWorkloadPoints1 (66 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (66 ms total) |
|
||||
| `DirectVulkan.XfbAfterClipDistanceScenario.SkipComponentsCaptureAlone` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:511: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / / [ FAILED ] XfbAfterClipDistanceScenario.SkipComponentsCaptureAlone (23 ms) / [----------] 1 test from XfbAfterClipDistanceScenario (23 ms total) |
|
||||
| `DirectVulkan.XfbAfterClipDistanceScenario.SkipComponentsCaptureSurvivesEveryClipWorkloadStopPoint` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:575: Failure / Value of: CheckSkipComponents(captured) / Actual: false (capture mismatch at index 1: got -2, expected 1 (the whole buffer is still the pre-fill: NOTHING was captured)) / Expected: true / (stop point: primitive 0, redeclaration 0, clip count 1) / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbAfterClipDistanceScenario.cpp:575: Failure |
|
||||
| `DirectVulkan.XfbCaptureBufferReuseScenario.ARespecificationMayChangeTheCaptureBufferSize` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:265: Failure / Value of: CapturedIs(readback.data(), value) / Actual: false (component 0 is -1234, expected 10 (the capture never reached these bytes)) / Expected: true / span 0 / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:265: Failure |
|
||||
| `DirectVulkan.XfbCaptureBufferReuseScenario.EverySpanIntoABufferObjectOfItsOwn` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:199: Failure / Value of: CapturedIs(readback, value) / Actual: false (component 0 is -1234, expected 10 (the capture never reached these bytes)) / Expected: true / span 0 / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:199: Failure |
|
||||
| `DirectVulkan.XfbCaptureBufferReuseScenario.EverySpanIntoOneImmutableStorageBuffer` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:308: Failure / Value of: CapturedIs(readback, value) / Actual: false (component 0 is -1234, expected 10 (the capture never reached these bytes)) / Expected: true / span 0 / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:308: Failure |
|
||||
| `DirectVulkan.XfbCaptureBufferReuseScenario.EverySpanIntoOneRespecifiedBufferObject` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:232: Failure / Value of: CapturedIs(readback, value) / Actual: false (component 0 is -1234, expected 10 (the capture never reached these bytes)) / Expected: true / span 0 / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbCaptureBufferReuseScenario.cpp:232: Failure |
|
||||
| `DirectVulkan.XfbPrimitiveQueryScenario.ACaptureThatFitsReportsEveryPrimitiveOnBothTargets` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbPrimitiveQueryScenario.cpp:230: Failure / Value of: CapturedVertexIs(readback.data() + static_cast<std::size_t>(point) * kFloatsPerVertex, point) / Actual: false (point 0 component 0 is -1234, expected 0 (the capture never reached these bytes)) / Expected: true / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbPrimitiveQueryScenario.cpp:230: Failure / Value of: CapturedVertexIs(readback.data() + static_cast<std::size_t>(point) * kFloatsPerVertex, point) |
|
||||
| `DirectVulkan.XfbPrimitiveQueryScenario.AnOverflowingVertexOnlyCaptureStopsWritingAtTheBufferCapacity` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbPrimitiveQueryScenario.cpp:260: Failure / Value of: CapturedVertexIs(readback.data() + static_cast<std::size_t>(point) * kFloatsPerVertex, point) / Actual: false (point 0 component 0 is -1234, expected 0 (the capture never reached these bytes)) / Expected: true / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbPrimitiveQueryScenario.cpp:260: Failure / Value of: CapturedVertexIs(readback.data() + static_cast<std::size_t>(point) * kFloatsPerVertex, point) |
|
||||
| `DirectVulkan.XfbRepeatedCaptureScenario.ACaptureFromAFragmentlessProgramRecords` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbRepeatedCaptureScenario.cpp:434: Failure / Value of: CapturedIs(captured, expected) / Actual: false (component 0 is -987654, expected 10 (the capture never reached these bytes)) / Expected: true / / [ FAILED ] XfbRepeatedCaptureScenario.ACaptureFromAFragmentlessProgramRecords (47 ms) / [----------] 1 test from XfbRepeatedCaptureScenario (47 ms total) |
|
||||
| `DirectVulkan.XfbRepeatedCaptureScenario.ACaptureFromAPatchesDrawRecords` | `PatchParameteri` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbRepeatedCaptureScenario.cpp:412: Failure / Value of: CapturedIs(captured, expected) / Actual: false (component 0 is -987654, expected 10 (the capture never reached these bytes)) / Expected: true / / [ FAILED ] XfbRepeatedCaptureScenario.ACaptureFromAPatchesDrawRecords (54 ms) / [----------] 1 test from XfbRepeatedCaptureScenario (54 ms total) |
|
||||
| `DirectVulkan.XfbRepeatedCaptureScenario.ACaptureFromAnAdjacencyDrawRecords` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbRepeatedCaptureScenario.cpp:459: Failure / Value of: CapturedIs(captured, expected) / Actual: false (component 0 is -987654, expected 11 (the capture never reached these bytes)) / Expected: true / / [ FAILED ] XfbRepeatedCaptureScenario.ACaptureFromAnAdjacencyDrawRecords (45 ms) / [----------] 1 test from XfbRepeatedCaptureScenario (45 ms total) |
|
||||
| `DirectVulkan.XfbRepeatedCaptureScenario.ACaptureListBeginningWithGlNextBufferSparesTheEarlierBuffer` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbRepeatedCaptureScenario.cpp:551: Failure / Value of: CapturedIs(afterA, spanAExpected) / Actual: false (component 0 is -987654, expected 10 (the capture never reached these bytes)) / Expected: true / the setup span itself did not capture / / [ FAILED ] XfbRepeatedCaptureScenario.ACaptureListBeginningWithGlNextBufferSparesTheEarlierBuffer (47 ms) |
|
||||
| `DirectVulkan.XfbRepeatedCaptureScenario.ASecondGeometryCaptureAfterADeqpStateResetStillRecords` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbRepeatedCaptureScenario.cpp:377: Failure / Value of: CapturedIs(captured, expected) / Actual: false (component 0 is -987654, expected 10 (the capture never reached these bytes)) / Expected: true / capture 0 of 3 in this process / / ../MobileGL/MG_IntegrationTest/Scenarios/XfbRepeatedCaptureScenario.cpp:377: Failure |
|
||||
| `DirectVulkan.XfbRepeatedCaptureScenario.ASpanThatNeverDrawsLeavesTheCaptureBufferAlone` | `BeginTransformFeedback` | **passed**: render/capture assertions passed |
|
||||
| `DirectVulkan.XfbRepeatedCaptureScenario.AVertexOnlyAdjacencyCaptureRecords` | `BeginTransformFeedback` | **failed**: ../MobileGL/MG_IntegrationTest/Scenarios/XfbRepeatedCaptureScenario.cpp:502: Failure / Value of: CapturedIs(captured, expected) / Actual: false (component 0 is -987654, expected 11 (the capture never reached these bytes)) / Expected: true / / [ FAILED ] XfbRepeatedCaptureScenario.AVertexOnlyAdjacencyCaptureRecords (55 ms) / [----------] 1 test from XfbRepeatedCaptureScenario (55 ms total) |
|
||||
|
||||
## Source evidence identity
|
||||
|
||||
- `p5b-c0b-census-logs/results.json`: SHA-256 `af55b6c5cbf58d6cc37f71eaaf1edb05a3cb04c8d3ab83f57f414979753075ba`, modified 2026-09-16T15:20:13.293362+00:00.
|
||||
- `p5b-t2-census-logs/results.json`: SHA-256 `cbdaaca2412675860189c6337a5b6177ca18ea59a8a01d4a28387187b40a0809`, modified 2026-09-16T16:30:20.426480+00:00.
|
||||
- `p5b-t2-census-logs/manifest.json`: SHA-256 `e0b8e97f3cf0cdfd0db6a8a1c2232cb4a49190843a1ad16c2f67c0a5971d11e5`, modified 2026-09-16T16:29:28.172948+00:00.
|
||||
- `p5b-t2-postmerge.log`: SHA-256 `3dae25bbe13407442805d58a5d9988f6a94fc35e28a6e0b86430f6da8e906c8a`, modified 2026-09-16T16:30:20.448940+00:00.
|
||||
|
||||
The JSON entries retain the command, working directory, private `.log` and `.out` per name. Additional prior probes: `p5b-t2-probe-{strict,vk-mono,vk-split,gles-mono,mono}.log`.
|
||||
|
||||
## Integrator notes
|
||||
|
||||
Merge the existing t2 code tip `fc5b52e4` (or branch `p5b/t2-codex`); no new patch is required. Resolve cumulative class-B/class-C pins with the other packages (t2 alone adds 6 implemented and subtracts 6 unmigrated slots). Keep the two added split discovery lists in SplitLogPaths alongside other packages. The new cap initialization in MG_Backend/Init.cpp must compose with later capability bits rather than overwrite them.
|
||||
@@ -0,0 +1,34 @@
|
||||
# P5b closing-review fixes: FenceWait and ReadPixels status
|
||||
|
||||
2026-09-16. Branch `p5b/wait-codex`, dedicated tree `/home/swung/w7/p5b-wait-codex`.
|
||||
Base: `348d22a4b9c15cc571af840dedc4b4edc8c53dbb`.
|
||||
Commit: `6388035f809861fd9b45cd1ba67c458a14b8b836`.
|
||||
|
||||
## Changes
|
||||
|
||||
Closing review Major #2: `FenceWait` applied/reply wait now allows `ceil(TimeoutNs / 1,000,000) + 30,000` milliseconds. Quotient/remainder rounding avoids overflow at `UINT64_MAX`. The finite 64-bit budget is consumed in transport chunks of at most `UINT32_MAX - 1`, avoiding both narrowing and the transport's forever sentinel, and avoiding overflow from one enormous chrono deadline. Shutdown still returns the original `ShutDown` result immediately; finite exhaustion still reports `Fatal{BarrierTimeout}` with the full effective budget. Only CPU `FenceWait` expands its budget. All other verb barriers and command-space retirement retain 30 seconds; `FenceWaitServer` with `GL_TIMEOUT_IGNORED` remains an ordinary bounded CPU submission.
|
||||
|
||||
Closing review Minor #3: both tight and scattered ReadPixels paths now abort `Fatal{ReplyStatusInvalid, "ReadPixels"}` for unknown reply status. Existing ERROR/DECLINED diagnostics remain intact. Correctly sized pixels with status 3 can no longer return as successful readback.
|
||||
|
||||
## Directed verification
|
||||
|
||||
Built only `RemoteClientTest` in this dedicated split/inproc tree with Clang Release and `-j4`. Selected **7**, passed **7**, failed **0**, skipped **0** (80 ms runtime).
|
||||
|
||||
New runtime controls use an installed test peer and a wrapper of the real transport doorbell to observe the deadline actually selected by `EmitAndWait`; no 30/60-second sleep or production test switch is needed. Coverage:
|
||||
|
||||
- `FenceWait` timeouts 0 ns, 1 ns, 60 seconds and `UINT64_MAX`: requested timeout is respected and huge waits use a finite transport chunk.
|
||||
- `FenceWaitServer` + `GL_TIMEOUT_IGNORED`: ordinary 30-second budget.
|
||||
- Huge `FenceWait` interrupted by a dead doorbell: DECLINED, no reply, prompt return.
|
||||
- Status 3 with a complete readback payload: named abort on both tight and bounce paths.
|
||||
- Existing complete readback, short tight reply, ERROR reply and short bounce reply controls remain green.
|
||||
|
||||
Command:
|
||||
|
||||
```bash
|
||||
build-split/MobileGL/MG_Test/Wire/RemoteClientTest --gtest_filter='RemoteClientControls.FenceWaitBudgetHonorsGlTimeoutAndFiniteTransportChunks:RemoteClientControls.LongFenceWaitBudgetStillWakesOnShutdown:RemoteClientControls.UnknownReadPixelsReplyStatusRefusesTightAndBounce:RemoteClientControls.ReadPixelsPutsTheTightExtentOnTheWire:RemoteClientControls.ShortOkReadPixelsReplyRefusesByName:RemoteClientControls.ErrorReadPixelsReplyRefusesByName:RemoteClientControls.BounceReadPixelsShortReplyRefusesBeforeScatter'
|
||||
```
|
||||
|
||||
Evidence: `/home/swung/w7/p5b-wait-{configure,build,focused}.log`.
|
||||
`git diff --check` passed. Commit tree is clean; dependency links are off. Old r1 tree/commit and all existing evidence remain intact. No repeated phase gate, additional adversarial review, device operation or push was performed.
|
||||
|
||||
The host full gate remains pinned to `348d22a4`; this directed result is explicitly a later review-fix delta. APKs must be built from the final integrated head containing both closing-review fixes.
|
||||
Reference in New Issue
Block a user