The SPIN model `wake_workers_no_std` must follow the new loop order

未關閉 適合新手
#804 0 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視

維護者通常 2 天內回覆

還沒有人認領這個 Issue。

評估

難度
2/5
預估耗時
1-3 小時
新手友好度
75/100
Issue 類型
缺陷
描述清晰度
描述清楚
活躍度
活躍
技術堆疊
rust

研究方向

閱讀 specification/awkernel_async_lib/src/task/wake_workers_no_std/wake_workers.pml 中的 primary_main,並將其迴圈順序與 kernel/src/lib.rs:109-114 比較。更新模型,讓它依序執行 wake_tasks、wake_workers,然後執行 sleep;當模型與實作一致,且受影響的已提交結果檔案已重新產生時,工作即告完成。issue 指出,run.sh 可能會將受記憶體限制而未完成的搜尋錯誤地回報為成功,因此重新產生結果時,請檢查是否出現搜尋未完成警告。

由索引模型根據 Issue 內容生成。

描述

File: specification/awkernel_async_lib/src/task/wake_workers_no_std/wake_workers.pml
(primary_main), related to kernel/src/lib.rs:109-114.

The model still has the old order:

wake_tasks(time_to_wait)
sleep(0, time_to_wait)
wake_workers(0)

The loop has three steps, so wake_tasks; wake_workers; sleep is a different
cyclic order, and the 32 committed result files no longer describe the
implementation. Please update the model in this PR or in a follow-up:

wake_tasks(time_to_wait)
wake_workers(0)
sleep(0, time_to_wait)

I ran this reordered model locally. All 16 configurations of
CONCURRENT_WORK_CONSERVATION and EVENTUALLY_EXECUTE
(TIMER_EDGE/TIMER_LEVEL x IRQ_POS 0-3) completed a full state-space
(TIMER_EDGE/TIMER_LEVEL x IRQ_POS 0-3) completed a full state-space
search with errors: 0. I did not run the 16 CPU_WAKING_TO_ACTIVE_CPU0/1
configurations.

Notes for regenerating the results:

  • The state space grows by a few percent, and -DMEMLIM=55000 in the Makefile
    is no longer enough. EVENTUALLY_EXECUTE needs 50.7-52.8 GB for
    IRQ_POS 1-3 and more for IRQ_POS=0. EVENTUALLY_EXECUTE /
    TIMER_LEVEL / IRQ_POS=0 needed 55.5 GB, and it only completed with a
    smaller hash table (./pan -a -n -w26) and -DMEMLIM=56500.
  • run.sh only checks for a trail file, so a run that stops at the memory
    limit (Warning: Search not completed) is recorded as true. It should also
    check for that warning.
  • The README lists kernel/src/main.rs as a target, but the loop is in
    kernel/src/lib.rs.

Other parts of the model need no change:

  • spawn_task does run_queue++ and then wake_up(..., 0, ...). The
    implementation keeps this order; the notification only moved from
    ArcWake::wake into the schedulers.
  • wake_tasks has no wake_up, which matches wake_cpu(0) being a no-op on
    CPU 0.
  • The model uses run_queue both as the queue and as the counter, so it
    already assumes an exact counter. This PR makes the implementation match it.
  • The std model (wake_workers) has a two-step loop
    (select_wait; wake_workers), so the reorder does not change it after the
    first iteration.
主要語言
Rust
星號
282
分支
12
平均合併
2 天 9 小時
30 天內合併 PR
28

環境準備

這個專案沒有提供開發容器、Dockerfile 或貢獻指南,環境需要你自己搭建:先看它的 README,通用步驟見我們的新手貢獻指南。

從這裡開始

  1. 先讀完整個 Issue,再讀專案的貢獻指南。
  2. 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
  3. Fork 儲存庫,在一個分支上完成修改。
  4. 送出 Pull Request,並在描述裡引用這個 Issue 編號。

tier4/awkernel 的其他 Issue

查看 tier4/awkernel 的全部 Issue

相似的 Issue

更多 Rust Issue

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。