Unify benchmark scripts with improved robustness

[jserv]

This consolidates separate benchmark scripts into a single unified 'tests/bench.py' with:

• Registration pattern for extensible benchmark definitions
• Adaptive run count: single run if >5min, time budget of 10min max
• Thread-safe progress indicator with spinner animation
• Proper subprocess management with explicit kill on timeout
• Parallel execution support (--parallel N)
• Robust error handling with preserved logs on failure
• time.monotonic() for accurate elapsed time measurement

It also fixes LOG_TRACE -> LOG_WARN to avoid excessive logging overhead during benchmark runs.

---

Summary by cubic

Unified all benchmark scripts into a single tests/bench.py runner with parallel execution and adaptive timing, and reduced logging noise to speed up runs. CI now calls the unified runner and writes a single JSON output.

• New Features

  • Extensible @register_benchmark pattern.
  • Adaptive runs with a 10-minute cap; single run if a pass takes >5 minutes.
  • Thread-safe progress spinner; quiet mode for CI.
  • Safe subprocess timeouts with explicit kill; logs preserved on failure.
  • Parallel execution (--parallel N) with ordered output; JSON saved to benchmark_output.json.

• Refactors

  • Removed tests/dhrystone.sh, tests/coremark.py, and tests/bench-aggregator.py.
  • Updated workflow to run: python3 tests/bench.py --json --quiet.
  • Switched log level from LOG_TRACE to LOG_WARN to reduce overhead.
  • Standardized timing with time.monotonic().

^{Written for commit 352a2af. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 6 files

[jserv]

Benchmarks

Details

Benchmark suite	Current: f4aff45	Previous: d8129e5	Ratio
Dhrystone	1635 Average DMIPS over 10 runs	1615 Average DMIPS over 10 runs	0.99
Coremark	965.643 Average iterations/sec over 10 runs	961.802 Average iterations/sec over 10 runs	1.00

This comment was automatically generated by workflow using github-action-benchmark.