Download CSV · Full records · Campaign manifest
Only boundaries supported by each software count toward completion. Solid color = successful, stripes = failed or skipped, gray = pending. Each planned repetition is one case. Percentages count recorded cases, not elapsed time.
Uses the completion scope above. Counts qualified, published references on the benchmark frequency grid, once per physical case across software and repetitions. Missing includes cases not yet calculated or not qualified. Percentages measure coverage, not elapsed time.
X = failed prediction, in the software's color, at its recorded time to failure. Each failed repetition is shown separately. Failures without a positive recorded prediction duration, including timeouts, remain in the coverage counts.
Lines connect adjacent sampled frequencies only. Gaps include pending, failed and unsupported cases. Timings show medians of successful repetitions. A single repetition has no variability estimate. Timeout durations are not successful solve times. ΔTS uses references matched to the same physical inputs. Local references retain their manuscript spatial convergence checks. Independently qualified references also require Fourier checks. Convergence evidence remains in the downloadable records. This file refreshes every minute. Downloaded artifacts are snapshots.
Memory is the sampled peak additional resident RAM during one prediction, above its warmed-up baseline, including child processes. Setup and warmup are excluded. Reusing allocated memory can produce zero additional RAM. Short-lived peaks between samples may be missed. Values are medians across successful repetitions, not cumulative allocated bytes.
| Software | Boundary | Completion | Success | Other recorded | Pending | Total solve time (s) | Average solve time (s) | RMSE ΔTS (dB) | Predictions with reference |
|---|
Total and average solve time include successful timed predictions only.
RMSE uses the median ΔTS across successful repetitions of each prediction, then √mean(ΔTS²) across predictions with valid comparisons. Predictions with reference counts the frequency and incidence combinations that have a matching BEM/FEM-BEM reference. Repetitions count once. No matched reference means successful predictions exist but their exact reference cases are missing. No successful prediction means there is no completed successful result in this selection.