LatencyInfo: Better handling of coalesced LatencyInfos |
||
Issue descriptionLatencyTracker doesn’t really care whether a LatencyInfo is coalesced or not, except for expecting non-coalesced instances to arrive in ComputeEndToEndLatencyHistograms [1]. RenderWidgetHostLatencyTracker similarly doesn’t really care a whole lot: on an input-event ack, if a LatencyInfo is coalesced, then it terminates [2] the LatencyInfo (which triggers some trace-flow events), but nothing else from the LatencyInfo gets reported (early out in ComputeInputLatencyHistograms() [3]). Instead of carrying around coalesced latency-infos, and then doing nothing with them, we should just terminate the LatencyInfo immediately when it gets coalesced (so that the trace-flow events have an end), and stop carrying them around (e.g. not add a coalesced LatencyInfo to CompositorFrameMetadata etc.) [1] https://cs.chromium.org/chromium/src/ui/latency/latency_tracker.cc?type=cs&sq=package:chromium&g=0&l=159 [2] https://cs.chromium.org/chromium/src/content/browser/renderer_host/input/render_widget_host_latency_tracker.cc?type=cs&sq=package:chromium&g=0&l=216 [3] https://cs.chromium.org/chromium/src/content/browser/renderer_host/input/render_widget_host_latency_tracker.cc?type=cs&sq=package:chromium&g=0&l=54
,
Jan 15
Ping, any update for this bug?
,
Jan 15
|
||
►
Sign in to add a comment |
||
Comment 1 by npm@chromium.org
, Sep 25