# Native benchmark — 2026-08-24

This record was generated locally from the repository benchmark package. It is
a parser throughput measurement, not a claim about every ELF/DWARF input.

## Command

```bash
moon run benchmarks --target native --release
```

## Workload

- Operation 1: `DebugImage::parse`
- Operation 1 input: deterministic 52-byte ELF32 little-endian header with no
  section payloads
- Operation 1 iterations per run: 10,000
- Operation 2: `SymbolTableIndex::build` plus address lookup
- Operation 2 input: 1,024 deterministic function symbols
- Operation 2 iterations per run: 1,000
- Every iteration completed successfully in all recorded runs.

## Environment

- OS: Windows NT 10.0.26200.0
- Moon: 0.1.20260819
- Moon compiler: v0.10.9+6e6c44045
- CPU model: not available because the local WMI query was denied by the
  operating-system permission profile

## Three local runs

| Run | Parse elapsed | Parse operations/second | Index elapsed | Index operations/second |
| ---: | ---: | ---: | ---: | ---: |
| 1 | 10 ms | 1,000,000 | 12 ms | 83,333 |
| 2 | 12 ms | 833,333 | 11 ms | 90,909 |
| 3 | 15 ms | 666,666 | 17 ms | 58,823 |

Median parse result: 12 ms for 10,000 successful parses, or 833,333
operations/second. Median index result: 12 ms for 1,000 builds and lookups, or
83,333 operations/second. These workloads are intentionally deterministic and
small; re-run the command when comparing changes because timings are
machine- and load-dependent.