mirror of
https://github.com/MobileGL-Dev/MobileGL
synced 2026-09-18 00:58:30 +09:00
[Docs] (Disaggregated): plan P5c - audit the inproc split's remaining direct cross-role memory access (59 sites at a79a0af6) and schedule its zeroing before P6 as texture staged shadow, SEG_EVENT producers, handle-keyed resolution, control records, residual-value records and role guards
This commit is contained in:
@@ -1,6 +1,6 @@
|
||||
# MGPipe:MobileGL 前后端拆分
|
||||
|
||||
> 状态:**P0–P5b 已收官**(2026-09-16,代码头 `82683d4a`)。目标负载(四条 Minecraft A/B trace)已在 Redmi Adreno 830 上以 `inproc`(独立 apply 线程)双后端渲染,barrier tax 首次实测。**下一个是 P6 spawn transport**。当前头、逐门数字与开放项见 [`CURRENT_STAGE_PROGRESS.md`](CURRENT_STAGE_PROGRESS.md)。
|
||||
> 状态:**P0–P5b 已收官**(2026-09-16,代码头 `82683d4a`)。目标负载(四条 Minecraft A/B trace)已在 Redmi Adreno 830 上以 `inproc`(独立 apply 线程)双后端渲染,barrier tax 首次实测。**下一个是 P5c**:审计(2026-09-17)显示 `inproc` 的两个角色仍有 59 处不经 wire 的直接内存访问(纹理纹素回读 client、反向通道裸指针、server 用 client 的 slot 分配器等),P5c 先把它们归零,**然后才是 P6 spawn transport**(计划见 `ROADMAP.md` "P5c 计划")。当前头、逐门数字与开放项见 [`CURRENT_STAGE_PROGRESS.md`](CURRENT_STAGE_PROGRESS.md)。
|
||||
>
|
||||
> 性能纪律(2026-09-08 起):逐线程 CPU 与 tracker 绝对 ns **对着 pull 臂基线记录**,不作阻塞门;专门的优化阶段排在路线图推完之后。
|
||||
|
||||
|
||||
Reference in New Issue
Block a user