abix-/skills/benchmark

Steps

1. Run benchmarks: Run k3sc cargo-lock bench --bench system_bench 2>&1 | tee /c/code/endless/rust/bench_results.txt and wait for completion.

2. Read results: Read rust/bench_results.txt and extract the timing data for each system at each entity count.

3. Read docs/performance.md: Find the ## Benchmark History section.

4. Append new entry: Add a new dated entry to the Benchmark History section with a markdown table showing all system timings. Format:

### YYYY-MM-DD — <commit hash (short)>

| System | 1K | 50K |
|--------|-----|-----|
| decision | Xus | Xus |
| damage | Xus | Xus |
| healing | Xus | Xus |
| attack | Xus | Xus |

Combined 50K: X.Xms (X.X% of 16ms budget)

5. Summarize: Print a short analysis comparing to the previous entry (if any) — which systems improved, regressed, or stayed the same. Include the Factorio-style budget math.

Rules

• Always write raw output to rust/bench_results.txt (overwritten each run)
• Round timings to nearest whole µs in the table
• Use the median value (middle of the [low median high] range) from Criterion output
• If a previous benchmark entry exists, note regressions >5% or improvements >5%
• Don't modify anything in performance.md outside the Benchmark History section
• If $ARGUMENTS is provided, pass it through to cargo bench (e.g. -- decision to filter)