Repository navigation
Establish a repeatable editor performance baseline #14
Description
Establish a manual, paired performance reference for the focused Notepad2-mod fork, using Python/pywinauto and psutil outside the application.
Protocol: published x64 v4.2.25.1029-rc.1 vs v4.2.25.1040-rc.1 in Original/Dark; fixed viewport; generated 128 KiB, 25 MiB and 1 MiB single-line plain-text files; isolated INIs; seven measured runs after per-mode/workload warmup; rotating mode order. Keep raw samples, medians, spread, executable/workload hashes and machine metadata.
Measure loading, first repaint, literal search, edit/undo, scroll/repaint, private/working-set memory and sampled idle CPU. Explicitly exclude claims about cold-start, syntax-heavy files, physical input latency, or a universal regression threshold. Never introduce app telemetry/background work or change user preferences.
Tooling is committed on perf/editor-baseline; baseline measurement is running. Results and reproduction instructions will be linked from the PR. This is a reference dataset, not a mandate to optimize or upgrade Scintilla.
Activity
itlezy commentedon Nov 9, 2026
Completed in #16, merged as 5b49158. Manual Python/pywinauto benchmark tooling and hash-locked optional test dependencies are available on master; no application runtime code changed. All four CI configurations passed before merge.
Raw samples and summaries: https://github.com/itlezy/notepad2-mod/blob/master/tests/baselines/published-1029-vs-1040.json
Reproduction command: https://github.com/itlezy/notepad2-mod/blob/master/Readme.md#performance-baseline
The seven-run published x64 comparison found close loading/private-memory results. The 1 MiB single-line scroll/repaint difference remains open in #15; these measurements do not establish a cause or performance parity.