perf-benchmarklisted

# Performance Benchmark Pipeline ## Overview Run a deterministic Linux performance diagnosis. The pipeline profiles an explicit benchmark target or binary, scores a 7-dimension rubric, writes raw artifacts, and can emit shared-schema PERF findings plus an append-only trend ledger. Use this skill when you need evidence for algorithmic scaling, wall-time stability, CPU efficiency, cache behavior, branch prediction, memory profile, or ASM-level review. ## Command ```bash python scripts/perf_benchmark_pipeline.py \ --root /path/to/repo \ --out-dir /tmp/perf-bench \ --target "python -m benchmark_entrypoint {SIZE}" \ --sizes 1000,4000,16000 \ --tier fast \ --max-cv 5.0 \ --findings-out /tmp/perf-findings.json \ --baseline-ledger /tmp/perf-ledger.jsonl ``` Use `--binary ./program` instead of `--target` for standalone binaries. Use `--target` or `--binary` for non-pytest repos. Pytest benchmark autodiscovery is a convenience for Python repos. Multi-size explicit targets must include `{SIZE}`. ## Key Flags - `--root`: repository root under analysis. - `--out-dir`: output directory for reports and raw profiler artifacts. - `--target`: command template to benchmark; include `{SIZE}` with `--sizes`. - `--binary`: standalone binary entrypoint. - `--source-prefix`: project source filter; repeat for multiple prefixes. - `--tier`: `fast`, `medium`, `deep`, or `asm`. - `--sizes`: comma-separated input sizes for scaling checks. - `--expected-complexity`: expected growth