Run a Finch trial where we measure the impact of bypassing passthrough touch event queue filtering |
||
Issue description(Adding jwd@ because he's the randomly assigned Metrics team reviewer) A few weeks ago, I added an UMA histogram to measure the fraction of touch input events that are filtered at the passthrough touch event queue. https://uma.googleplex.com/p/chrome/histograms/?endDate=20190102&dayCount=7&histograms=Event.Touch.FilteredAtPassthroughQueue&fixupData=true&showMax=true&filters=platform%2Ceq%2CA%2Cchannel%2Ceq%2C1%2Cisofficial%2Ceq%2CTrue&implicitFilters=isofficial shows some preliminary Canary data: about 25% of touch input events are filtered there. The Speed Metrics team has two goals this quarter that would be made easier if passthrough touch event queue filtering were disabled: 1. Ensure that the first input is tracked for EventTiming even when it's filtered 2. Count all meaningful input events for EventTiming, even when those events are filtered Removing one of the two places that these inputs could be filtered would make these tasks easier, but is only realistic if the performance impact is minimal. This bug tracks the experiment to assess the performance impact of bypassing the passthrough touch event filter.
,
Jan 8
,
Jan 9
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/667a7dc5847073f1bf2eb016bb2def4a6d5658b3 commit 667a7dc5847073f1bf2eb016bb2def4a6d5658b3 Author: Charlie Andrews <charliea@chromium.org> Date: Wed Jan 09 16:02:12 2019 Add a feature to skip passthrough touch event queue filtering The only filter that's still applied is the timeout filter. nzolghadr@ and I agreed offline that it probably makes sense to move this filter to the top of FilterBeforeForwarding() to apply it unconditionally (previously it wasn't applied to touch scroll started events). This flag will be used to run a Finch trial to assess the performance impact of disabling filtering here. Previously added UMA metrics suggest that only 25% of inputs are filtered here. I'll write a doc with more details about this Finch trial, including what metrics we might want to look at, soon. Bug: 919843 Change-Id: I74d649fda6bb72f706acfec773c219204fbd1f72 Reviewed-on: https://chromium-review.googlesource.com/c/1400958 Reviewed-by: Camille Lamy <clamy@chromium.org> Reviewed-by: Navid Zolghadr <nzolghadr@chromium.org> Commit-Queue: Charlie Andrews <charliea@chromium.org> Cr-Commit-Position: refs/heads/master@{#621159} [modify] https://crrev.com/667a7dc5847073f1bf2eb016bb2def4a6d5658b3/content/browser/renderer_host/input/passthrough_touch_event_queue.cc [modify] https://crrev.com/667a7dc5847073f1bf2eb016bb2def4a6d5658b3/content/browser/renderer_host/input/passthrough_touch_event_queue.h [modify] https://crrev.com/667a7dc5847073f1bf2eb016bb2def4a6d5658b3/content/browser/renderer_host/input/passthrough_touch_event_queue_unittest.cc [modify] https://crrev.com/667a7dc5847073f1bf2eb016bb2def4a6d5658b3/content/public/common/content_features.cc [modify] https://crrev.com/667a7dc5847073f1bf2eb016bb2def4a6d5658b3/content/public/common/content_features.h |
||
►
Sign in to add a comment |
||
Comment 1 by charliea@chromium.org
, Jan 8