This page records development-time performance baselines for JU SeqWorkbench so that later renderer changes can be measured against a fixed reference point.
Scope: this is a JU-only development baseline. It is not a cross-software ranking or a claim that these timings represent every real-world workflow.
The current sequence renderer still uses a full-document QPlainTextEdit + QSyntaxHighlighter architecture. Before migrating or substantially changing that renderer, a fixed baseline was captured so that later optimization work can be evaluated with the same fixtures and metrics.
The public dataset intentionally contains only JU SeqWorkbench measurements. Internal observations involving other software are kept separate because the measurement methods are not directly comparable enough for a fair public benchmark.
post-safe-optimization / pre-renderer-migrationbeta-dev@96ae3f9QPlainTextEdit + QSyntaxHighlighter full-document architectureLower timing values are better. All timing values below are milliseconds.
| Fixture | Rows | Alignment columns | Total cells |
|---|---|---|---|
| 35 × 1,000 | 35 | 1,000 | 35,000 |
| 35 × 5,000 | 35 | 5,000 | 175,000 |
| 100 × 5,000 | 100 | 5,000 | 500,000 |
| 100 × 10,000 | 100 | 10,000 | 1,000,000 |
| 500 × 1,000 | 500 | 1,000 | 500,000 |
The internal frozen benchmark also tracks fixture hashes and additional provenance fields. The public-friendly CSV keeps the timing values and the metadata needed for later JU before/after comparisons without publishing the separate internal comparison master.
| Fixture | Refresh | Edit + refresh | Color ON | Dot ON |
|---|---|---|---|---|
| 35 × 1,000 (35k cells) | 167.7 ms | 135.9 ms | 659.5 ms | 138.9 ms |
| 35 × 5,000 (175k cells) | 674.1 ms | 618.8 ms | 1,581.6 ms | 328.9 ms |
| 100 × 5,000 (500k cells) | 772.5 ms | 781.0 ms | 4,775.9 ms | 916.6 ms |
| 100 × 10,000 (1.0M cells) | 1,452.3 ms | 1,731.6 ms | 9,545.6 ms | 1,912.6 ms |
| 500 × 1,000 (500k cells) | 213.0 ms | 185.2 ms | 3,182.4 ms | 301.9 ms |
| Fixture | H-scroll | V-scroll | Color OFF | 10-column highlight |
|---|---|---|---|---|
| 35 × 1,000 (35k cells) | 53.6 ms | 0.0 ms* | 97.8 ms | 84.0 ms |
| 35 × 5,000 (175k cells) | 108.6 ms | 0.0 ms* | 189.2 ms | 135.4 ms |
| 100 × 5,000 (500k cells) | 206.2 ms | 194.5 ms | 370.4 ms | 295.0 ms |
| 100 × 10,000 (1.0M cells) | 398.7 ms | 387.0 ms | 816.0 ms | 503.6 ms |
| 500 × 1,000 (500k cells) | 53.5 ms | 102.4 ms | 104.1 ms | 333.9 ms |
* The 35-row fixtures did not overflow vertically in the 800-pixel test window, so vertical scrolling was effectively a no-op and should not be interpreted as a meaningful scrolling result.
The largest bottleneck in this renderer is residue coloring with Color ON. Under the current full-document highlighting architecture, the 100 × 10,000 fixture reached a median of about 9.55 s for that metric. After reviewing this result, renderer migration and visible-range rendering were moved into the pre-alpha release gate rather than being left as an immediate post-alpha patch.
The two 500,000-cell fixtures also behave differently depending on whether the dataset is wider or taller. That is useful for future optimization work because total cell count alone does not fully describe renderer cost.
This baseline is deliberately published before the renderer is improved. Its purpose is to preserve an honest reference point for later before/after development logs rather than to present tuned marketing numbers.
The developer renderer prototype has moved to a visible-range viewport model: paint work is limited to the rows and columns currently visible instead of rebuilding the full alignment on every repaint. Prototype Color and Dot display paths are working, and geometry/visible-only checks have been exercised on the large synthetic fixtures. The new renderer is not yet the default production viewer; production integration, editing/clipboard/Undo-Redo parity, regression testing, and manual QA remain release gates.
Download/view the public JU pre-renderer baseline CSV
The CSV contains all 40 measurements: 5 fixtures × 8 metrics, with median/min/max values and the public comparison metadata.