# Benchmark results

The benchmark is deterministic and can be rerun with:

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

Test environment: Windows 11 (`Microsoft Windows NT 10.0.26200.0`), AMD Ryzen 7 5800H with Radeon Graphics, MoonBit stable `0.1.20260814`, `moonc 0.10.8+8606a5800`, recorded 2026-08-18.

| Run | Scalar updates | Scalar elapsed | Scalar updates/s | Matrix updates | Matrix elapsed | Matrix updates/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 1 | 1,000,000 | 11 ms | 90,909,090.91 | 10,000 | 68 ms | 147,058.82 |
| 2 | 1,000,000 | 12 ms | 83,333,333.33 | 10,000 | 70 ms | 142,857.14 |
| 3 | 1,000,000 | 12 ms | 83,333,333.33 | 10,000 | 70 ms | 142,857.14 |

The scalar checksum was `512911.08916493575` in all three runs. The 4×2 matrix checksum was `9850.129406998216` in all three runs. These numbers are measurements from the release executable, not estimates or asymptotic claims; they will vary with CPU load and toolchain version.

## Follow-up local validation

After the production-source expansion, the same release command was run three more times locally on 2026-08-19. The checksums remained unchanged. Matrix timings were more variable because the sample is short and the machine was not isolated from other workload.

| Run | Scalar elapsed | Scalar updates/s | Matrix elapsed | Matrix updates/s |
| ---: | ---: | ---: | ---: | ---: |
| 1 | 15 ms | 66,666,666.67 | 195 ms | 51,282.05 |
| 2 | 15 ms | 66,666,666.67 | 170 ms | 58,823.53 |
| 3 | 14 ms | 71,428,571.43 | 143 ms | 69,930.07 |

Follow-up checksum: scalar `512911.08916493575`; 4×2 matrix `9850.129406998216`.