diff options
| author | blasty <blasty@local> | 2026-08-07 03:13:22 +0200 |
|---|---|---|
| committer | blasty <blasty@local> | 2026-08-07 03:13:22 +0200 |
| commit | 4bc162ddc82dda13ca527b686190a8b723fdadd2 (patch) | |
| tree | 6ef9ad3062cc3d83ad1b0a68b4eab4e0f6e479ee /.auto/bench.py | |
| parent | RE-BASELINE (v4 bench). Adding a second repetition on the big target exposed ... (diff) | |
| download | ida-tui-4bc162ddc82dda13ca527b686190a8b723fdadd2.tar.gz ida-tui-4bc162ddc82dda13ca527b686190a8b723fdadd2.tar.xz ida-tui-4bc162ddc82dda13ca527b686190a8b723fdadd2.zip | |
bench: poll landings every 2ms, not 10 -- the wait interval was measurement overhead
Diffstat (limited to '.auto/bench.py')
| -rw-r--r-- | .auto/bench.py | 9 |
1 files changed, 8 insertions, 1 deletions
diff --git a/.auto/bench.py b/.auto/bench.py index d286c14..5d135f7 100644 --- a/.auto/bench.py +++ b/.auto/bench.py @@ -78,7 +78,14 @@ def fail(msg): FAILS.append(PREFIX + msg) -async def _wait(pilot, pred, t=60.0, step=0.01): +#: Poll interval for every "has it landed yet" wait. The phases time the wait, +#: so the interval is measurement overhead: at the 10ms this used to use, a +#: graph open that really took 30ms was charged up to 40, and the graph phase +#: (24 timed waits) carried ~12% of pure quantisation. +_STEP = 0.002 + + +async def _wait(pilot, pred, t=60.0, step=_STEP): return await wait_for(pred, pilot.pause, t, step) |