optimize entry cleanup

[SF-Zhou]

Optimize the entry cleanup process after EntryByVal/Ref is dropped. In practice, cleanup is only necessary when the reference count reaches zero.

Q: Why not check if the value is None inside EntryByVal/Ref::Drop to avoid redundant try_remove_entry calls?

We must decrement the reference count only after calling state.mutex.unlock(). If we were to decrement the reference count first and it hit zero, a concurrent map.remove(key) might assume the entry is no longer in use and deallocate the state. This would cause the subsequent state.mutex.unlock() in the current thread to fail.

Furthermore, once state.mutex.unlock() is executed, the value within the state could be immediately modified by another thread. Therefore, the current thread can no longer rely on the state of the value to determine whether try_remove_entry is required.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (82e7c3e) to head (153f836).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff            @@
##              main       #24   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            3         3           
  Lines          892       871   -21     
=========================================
- Hits           892       871   -21     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Copilot]

Pull request overview

This PR optimizes the entry cleanup mechanism in LockMap by refactoring how reference counting and entry removal are handled. The key change moves the reference count decrement from inside the shard lock to outside, improving concurrency by reducing lock hold time.

Key changes:

• Strengthened atomic memory ordering (Relaxed → AcqRel/Acquire) to ensure correctness with the new cleanup timing
• Refactored unlock() into try_remove_entry() that checks conditions before removal rather than unconditionally decrementing inside the lock
• Moved the fetch_sub operation from inside the shard lock to the Drop implementations, reducing critical section duration

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.