From aa8ef34641640698222d77c42f6aace6ffc2f9f6 Mon Sep 17 00:00:00 2001 From: Swung0x48 Date: Tue, 22 Sep 2026 20:46:34 -0400 Subject: [PATCH] =?UTF-8?q?[docs]=20(P7):=20magma-b3=20-=20=C2=A77.2=20rec?= =?UTF-8?q?ords=20that=20WaitForFrameSerial=20can=20answer=20true=20for=20?= =?UTF-8?q?a=20serial=20the=20clamped=20floor=20has=20not=20reached,=20for?= =?UTF-8?q?=20the=20integrator=20to=20copy=20into=20=C2=A712?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fable review of the B3 rework, note for CONTRACT-P7 §12. On the disaggregated build VulkanRenderer::WaitForFrameSerial waits the FIRST in-flight record with frameSerial >= serial, calls OnSubmitsCompletedUpTo and returns true. With the §3.1 clamp, a later record carrying the same serial (S2, the Present submission - the two-submissions-per-serial shape of §2.2) keeps the floor below `serial`, so the function returns true while IsFrameSerialComplete(serial) is false. The only caller, WaitForTimerQueryResult, re-checks IsTimerQueryResultReady, so today it degrades to a spurious timeout, not a wrong read. The fix belongs to B4 and is not made here: after OnSubmitsCompletedUpTo, return only if IsFrameSerialComplete(serial), else restart the scan (the call erases from the vector being iterated, so restart, never continue). --- docs/Disaggregated/notes/p7/magma-b3.md | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/docs/Disaggregated/notes/p7/magma-b3.md b/docs/Disaggregated/notes/p7/magma-b3.md index 2e3a20b7..325da8c3 100644 --- a/docs/Disaggregated/notes/p7/magma-b3.md +++ b/docs/Disaggregated/notes/p7/magma-b3.md @@ -512,6 +512,15 @@ Present（`:14094`）三处「等一条 fence、把它之前的都记成完成 「OnSubmitsCompletedUpTo calls NotifyFrameSerialComplete for every record it retires, so the completed-serial floor still advances correctly after one fence wait」——disagg 构建上 §3.1 之后 **不再逐记录通知**，地板只推到无在飞持有的序号。该函数是 pull 与 disagg 共用代码，注释归 B4 一起改。 +- **`WaitForFrameSerial` 在 disagg 构建上可能答 true 而序号并未完成（修正轮，fable 复审；写给集成者抄进 §12， + 本包不修）**：它等的是 `m_inFlightSubmits` 里**第一条** `frameSerial >= serial` 的记录， + `OnSubmitsCompletedUpTo(record.submitIndex)` 之后就 `return true`。有了 §3.1 的钳制，若**更晚**的一条记录 + （S2，Present）持有**同一个**序号——正是 §2.2 的「一个序号两次提交」——地板到不了 `serial`，于是函数返回 true + 而 `IsFrameSerialComplete(serial)` 为 false。唯一的调用者 `WaitForTimerQueryResult` 返回前重查 + `IsTimerQueryResultReady(record)`，所以今天它退化为一次**假超时**（结果这次拿不到、下次再收），不是错读。 + **修法归 B4**：`OnSubmitsCompletedUpTo` 之后只在 `IsFrameSerialComplete(serial)` 为真时返回，否则**从头重扫** + ——`OnSubmitsCompletedUpTo` 会从正在迭代的 vector 里 erase，所以是 restart，不能 continue； + 与 §7.1 的聚合等待与上面的注释一起做。 - **disaggregated 构建的 monolith 臂现在比 pull 构建更可靠（复审 (a)）**：monolith 的 `VkBufferResource` 路径 `OnSubData` 经 `IsResourceBusy`（`VkBufferManager.cpp:709`）读的是**同一个** `GetCompletedSerial()` 地板，所以 §3.1 的钳制在 disagg 构建里也保护了 monolith 臂的 mid-frame flush；