Mark thread-local variables as initial-exec

[edolstra]

Motivation

As suggested by @xokdvium, use the initial-exec model for thread-local storage. This should be faster, at least until NixOS/nixpkgs#518519 lands. It didn't show a clear performance improvement in my testing, e.g. on a nix search nixpkgs:

maximum RSS:        median = 5037324.0000  mean = 5040103.2000  stddev =  19835.2237  min = 5005472.0000  max = 5080404.0000  [not rejected, p=0.10552, Δ=-6886.93333±12125.09181]
soft page faults:   median = 1264925.0000  mean = 1265636.1333  stddev =   5001.5549  min = 1258135.0000  max = 1275674.0000  [not rejected, p=0.10466, Δ=-1707.91111±2999.43676]
system CPU time:    median =     40.9134  mean =     40.8374  stddev =      0.6536  min =     39.3454  max =     41.9917  [not rejected, p=0.49896, Δ=-0.09703±0.41150]
user CPU time:      median =     40.0625  mean =     39.7700  stddev =      0.8387  min =     38.0676  max =     40.8150  [rejected, p=0.00000, Δ=-1.03734±0.54708]
elapsed time:       median =     13.2556  mean =     12.7748  stddev =      0.7929  min =     11.4628  max =     13.5534  [not rejected, p=0.83082, Δ=-0.03600±0.48379]

however it shouldn't hurt either.

Context

Summary by CodeRabbit

• Performance
  • Adjusted thread-local storage model across evaluation, parallel execution, logging, and signal handling to improve multi-threaded evaluation and responsiveness with no change to user-facing behavior.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 6e7c73f4-7c01-4b1b-b8c6-d0231325daf1

📥 Commits

Reviewing files that changed from the base of the PR and between 4c15336 and 3a91635.

📒 Files selected for processing (8)

• src/libexpr/eval.cc
• src/libexpr/include/nix/expr/eval-inline.hh
• src/libexpr/include/nix/expr/eval.hh
• src/libexpr/include/nix/expr/parallel-eval.hh
• src/libexpr/parallel-eval.cc
• src/libutil/logging.cc
• src/libutil/unix/include/nix/util/signals-impl.hh
• src/libutil/unix/signals.cc

✅ Files skipped from review due to trivial changes (3)

• src/libexpr/include/nix/expr/parallel-eval.hh
• src/libutil/logging.cc
• src/libutil/unix/include/nix/util/signals-impl.hh

🚧 Files skipped from review as they are similar to previous changes (4)

• src/libexpr/include/nix/expr/eval-inline.hh
• src/libexpr/eval.cc
• src/libexpr/include/nix/expr/eval.hh
• src/libexpr/parallel-eval.cc

---

📝 Walkthrough

Walkthrough

This PR annotates fourteen existing thread_local variables across evaluator, allocator caches, parallel executor, logging, and signal handling with [[gnu::tls_model("initial-exec")]]. No runtime logic or control flow is changed.

Changes

Thread-Local Storage Model Optimization

Layer / File(s)	Summary
EvalState thread-local annotations src/libexpr/include/nix/expr/eval.hh, src/libexpr/eval.cc	EvalState::callDepth and EvalState::evalContext declarations and definitions gain [[gnu::tls_model("initial-exec")]].
Memory allocation cache thread-locals src/libexpr/include/nix/expr/eval-inline.hh	GC allocation caches (valueAllocCache, env1AllocCache) and extern thread_local uint32_t myEvalThreadId are annotated with [[gnu::tls_model("initial-exec")]].
Executor parallel evaluation thread-locals src/libexpr/include/nix/expr/parallel-eval.hh, src/libexpr/parallel-eval.cc	Executor::amWorkerThread, spawn-local std::random_device/std::uniform_int_distribution, and myEvalThreadId receive the initial-exec TLS model attribute.
Utility subsystem thread-locals src/libutil/logging.cc, src/libutil/unix/include/nix/util/signals-impl.hh, src/libutil/unix/signals.cc	curActivity and interruptCheck declarations/definitions are annotated with [[gnu::tls_model("initial-exec")]].

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐰 I tuck each thread-local in its nest,

"initial-exec" gives them speed, not rest,
No logic hopped, no pathways rewired,
Just snappier hops where values are squired.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the main change: adding the initial-exec TLS model attribute to thread-local variables across multiple files.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch initial-exec-tls

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

🚀 Deployed on https://6a1daa63ab46434a634d78aa--determinate-nix.netlify.app