Regression in multi_widget_construction_perf__e2e_summary benchmarls

[kf6gpe]

See https://flutter-flutter-perf.skia.org/e/?begin=1602102124&end=1603169996&keys=Xa6ab3890c3982db78aadaec0616a0db6&requestType=0&xbaroffset=22068.

Was previously bimodal, with the predominant mode lower. Now it's clearly unimodal, with the higher value. Probably worth investigating, not a P0.

This affects all of
average_frame_rasterizer_time_millis90th_percentile_frame_build_time_millis, 90th_percentile_frame_rasterizer_time_millis, and average_frame_build_time_millis.

Maybe commit #68424? It started to climb on the previous commit, but that doesn't seem relevant.

[zanderso]

#67594 Looks like it could be related as well. @CareF @liyuqian

[CareF]

#67594 introduces #64780 to all macrobenchmark for consistency, including multi_widget_construction_perf__e2e_summary. That might be the reason but I don't have a local MOTO G4 to really see the time distribution to confirm.

[liyuqian]

Confirmed that #67594 caused this regression because the added delay changed the frames that we measured and their frame time distributions. Specifically, later frames are slower to build than earlier frames as shown in the screenshot below. I'm closing this issue as there's no actual performance regression.

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. If you are still experiencing a similar issue, please open a new bug, including the output of flutter doctor -v and a minimal reproduction of the issue.