NM Dev benchmark explorer

Large-matrix comparisons: 2000 × 2000 · 1 thread(s) · 2000 × 2000 · 4 thread(s)

2000 × 2000 · 4 requested compute thread(s). Four libraries; multiplication, LU, Cholesky, QR and solve. Two JVM forks; two 100 ms warmups and three 200 ms measurements per fork. One JMH worker. Each fork validates 2000-square outputs before timing. Native worker counts are not independently audited. Timing tables and evidence.

Compare saved measurements from 2026-10-18-2000-4thread. Hover, tap or keyboard-focus a point for timing, uncertainty and backend details.

Lower values mean less time. Both modes use a logarithmic horizontal axis. Ratios are ratios of measured means, without confidence intervals for the ratios. Results describe the recorded shared VM and backend configurations, not universal performance rankings.

Chart settings

Selections are saved in this browser. A shared link takes priority over saved settings. Hiding the reference library does not change the ratios.

Visible libraries
Visible operations

Results tables · Methodology · Versions and backends · Reproduce

SVG downloads contain the selected chart, labels, units and run identifier. CSV exports include original millisecond timings, confidence limits and the selected reference ratios. All calculations and exports use the saved measurements; no benchmark is run in your browser.