# Benchmark Results

This document records a reproducible local measurement of the native CLI
benchmark. The counters printed by the benchmark are deterministic; the
wall-clock values are specific to the machine, tool cache, and current system
load.

## Environment

- Date: 2026-08-24 03:14:15 +08:00
- Operating system: Windows (processor description unavailable in the current
  restricted PowerShell environment)
- MoonBit: `moon 0.1.20260819 (fc2a4ee 2026-08-19)`
- Moon compiler: `moonc v0.10.9+6e6c44045 (2026-08-19)`

## Workload

Command-level measurement:

```text
pwsh -NoProfile -File scripts/benchmark.ps1 -Runs 5
```

The fixed workload runs 1000 native 10 ms cycles on the standard station,
visits 255 route slots per cycle, and performs 10000 hostile-route bitset
checks. The benchmark verifies 1 processed route event, 10000 hostile
rejections, and zero fatal safety violations.

## Raw local output

```text
run=1 elapsed_ms=156.193
run=2 elapsed_ms=85.382
run=3 elapsed_ms=85.831
run=4 elapsed_ms=75.904
run=5 elapsed_ms=78.091
min_ms=75.904 max_ms=156.193 average_ms=96.280
result=workload=standard-route-1000-cycles-bitmask-10000 cycles=1000 events=1 active_slots_scanned=255000 bitmask_checks=10000 hostile_rejections=10000 fatal_violations=0 verification_passed=true
```

These timings include the `moon run` command and native executable startup.
They are reference data for this local environment, not a cross-platform
performance gate. CI checks the deterministic counters and safety result; it
does not compare command-level wall-clock timings across runners.