Add approximate end-to-end Input Latency Metrics for all input types |
||||||||||||||||
Issue descriptionWe currently have Event.Latency.Renderer2, Event.Latency.RendererImpl and Event.Latency.Browser for each input type. These aren't very useful, as they don't really approximate the end-to-end latency well. Ideally we'd measure from the hardware input timestamp until a lower bound for when the input's visible response would have happened. If there's no visible response, we'd put that input in a separate bucket, using the time event handler processing finished (or would have started, if there is no event handler processing).
,
Dec 5 2016
,
Dec 12 2016
,
Dec 13 2016
,
Dec 13 2016
,
Feb 3 2017
,
Feb 3 2017
,
Mar 13 2017
,
Mar 15 2017
Is there any update? The implementation of idle time spell checker has been fully landed, so it will be great if we can start an experiment to see how much input latency can be reduced.
,
Mar 15 2017
I've got a patch locally that plumbs this through for keyboard, but I've been doing a bunch of associated cleanup instead of actually landing the patch. I'll land this within the next few weeks.
,
Mar 15 2017
Thanks a lot for the update!
,
Aug 15 2017
I'm not going to get to this any time soon.
,
Jan 4 2018
Parking this on Navid for now. Feel free to close it if you don't think this is worth tracking.
,
Mar 16 2018
,
Jun 27 2018
No updates in a while, should we close this?
,
Jun 27 2018
Tim, I don't fully understand what you had in mind. Did you think of a latency measurement for the events from the hardware timestamp until we call the js event handler (excluding the running time of handlers in js)?
,
Oct 1
I was thinking we'd include JS event handler time, and, if a frame was produced, time until the frame was produced.
,
Jan 15
Navid, do you have any update here?
,
Jan 15
Let's keep this as P3 and not something we would need immediately.
,
Jan 21
(2 days ago)
Leaving this on the input team for now.
,
Jan 21
(2 days ago)
,
Yesterday
(39 hours ago)
|
||||||||||||||||
►
Sign in to add a comment |
||||||||||||||||
Comment 1 by tdres...@chromium.org
, Oct 31 2016