Commit graph

2 commits

Author SHA1 Message Date
Antoine
1e9b1d458b
GPU benchmark: isolate trials, time the card, say which map drew
- A trial now ends before the next starts: the clock stops, the frame
  callback is cancelled (and a queued one turned away), a day still in the
  route worker is waited for (or dropped), the clock's leftovers are cleared,
  and every trial starts from the same random numbers.
- The renderers take turns going first, scenario by scenario; ?reps=N runs
  each N times and reports medians.
- draw() is labelled as CPU submission time; the card's own time comes from
  EXT_disjoint_timer_query_webgl2 where available (read frames later, never
  waited for, disjoint results dropped); a run drawn by the other map is
  flagged.
- npm run perf --browser reads its sync pixel back from the surface the map
  was drawn on (the WebGL canvas waits for the card) and says which map each
  version drew with.
2026-11-04 22:37:13 +01:00
Frederic Jung
c76dd4d556
Benchmark in real Chromium too, and only call a row slower when the samples say so
npm run perf -- --browser times the same workloads in Chromium as a desktop
and as a phone (390x844 at 3x, CPU slowed 4x), one browser per version. Each
run collects garbage first and ends by forcing layout and the map canvas, so
deferred work counts. The jsdom DOM is about 30x slower than a browser's and
has no layout, so the interface rows need this.

Shared machines swing single runs by 30% on identical code, CPU time
included. A row now counts as slower or faster only when it moved 15% and a
Mann-Whitney U test gives p < 0.01 (15 runs a side by default). Identical
code: no flags in three full runs; a routing slowdown planted in a copy is
flagged in jsdom and on the phone.

The panel workload renders both selections every run instead of alternating,
which split its samples in two. CI drives the runner's Chrome through
playwright-core, no browser download.
2026-10-26 12:04:57 +01:00