SpinLockMutex small cleanup; improved and extended benchmark

[Reneg973]

Fixes # (issue)

Changes

• load() not required in try_lock()
• shortened loop in lock()
• fixed erroneous yield() in BM_ThreadYieldSpinLockThrashing
• added std::mutex to benchmark
• removed superfluous static as code is already in unnamed namespace

It seems that at least on Apple M1 (lClang+libc) the std::mutex outperforms SpinlockMutex with higher number of threads.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: lalitb / name: Lalit Kumar Bhasin (61f60a7)
• ✅ login: Reneg973 / name: Rene Greiner (0fac0b9, 5bfa70d)

[Copilot]

Pull Request Overview

This PR refactors the SpinLockMutex implementation and its benchmark suite to improve code clarity and performance testing. The changes simplify the locking logic and expand benchmark coverage to include std::mutex for performance comparison.

Key changes:

• Simplified try_lock() by removing redundant load() check before exchange operation
• Refactored lock() to use try_lock() internally for cleaner code
• Fixed erroneous yield logic in BM_ThreadYieldSpinLockThrashing benchmark
• Added std::mutex benchmark for performance comparison across different thread counts

Reviewed Changes

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

File	Description
api/include/opentelemetry/common/spin_lock_mutex.h	Simplified try_lock() and lock() implementations by removing redundant checks and restructuring loop logic
api/test/common/spinlock_benchmark.cc	Refactored benchmark template to use lambdas with captures, removed static qualifiers, fixed yield bug, and added std::mutex benchmark

[Copilot]

The spin-fast loop at line 101 is now unreachable. The while (!try_lock()) loop continuously retries try_lock() without ever entering the body to execute the spin-fast logic. The original structure with explicit if (!flag_.exchange(...)) return; allowed falling through to the spin logic when the lock failed. This breaks the intended backoff behavior.

[Copilot]

[nitpick] The alignas(8) specifier appears to be added without explanation. If alignment is needed for performance, consider adding a comment explaining why 8-byte alignment is chosen, or remove it if it was added unintentionally.

Suggested change

	alignas(8) std::atomic_flag mutex = ATOMIC_FLAG_INIT;
	std::atomic_flag mutex = ATOMIC_FLAG_INIT;

[lalitb]

This seems less efficient under contention. Calling try_lock() in every iteration causes repeated atomic exchanges and cache invalidations. The previous version avoided that by doing the initial exchange outside the spin loop.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.95%. Comparing base (70daba9) to head (5bfa70d).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3706      +/-   ##
==========================================
+ Coverage   89.90%   89.95%   +0.06%     
==========================================
  Files         225      225              
  Lines        7281     7101     -180     
==========================================
- Hits         6545     6387     -158     
+ Misses        736      714      -22     

Files with missing lines	Coverage Δ
api/include/opentelemetry/common/spin_lock_mutex.h	100.00% <100.00%> (ø)

... and 47 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[maxmarkin01]

How is this more efficient? load(std::memory_order_relaxed) would avoid having a memory fence and wait for cache sync. Having said that the whole idea of SpinLock in user mode seems dubious especially in high contention scenarios.

[Reneg973]

I'm confused why this CI/Format check fails! Any hint is welcome

[marcalff]

@Reneg973

Tip to format all the code (C++, CMake, Bazel):

[malff@malff-desktop ~]$ more ~/bin/otel_cf.sh 
#!/bin/bash
podman run --rm --privileged=true --volume ${PWD}:/otel otel/cpp_format_tools

See https://github.com/open-telemetry/cpp-build-tools/tree/main/cpp_format_tools