Benchmarks and validation¶
These records describe numerical parity and performance under the stated case, runtime and resource conditions. Repeated measurements report the median and maximum absolute deviation; single runs are labelled separately. Results are engineering evidence for those conditions rather than general performance guarantees.
| Record | Scope |
|---|---|
| Reconstruction | Native OpenFOAM comparison, command-line overhead and cross-distribution compatibility |
| Dataset export and reduction | Direct collated decoding, memory use and filtering kernels |
Keep case size, field selection, runtime, worker count and timing boundaries attached to each comparison. Preserve source cases and raw measurements outside the source tree. See Verification for checks after implementation changes.