All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=916138

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=ae06e91ee6d8e93f9583f379c24578ec09685427b31a793895f7db3b6e6b12d3


Bot(s) for this bug's original alert(s):

linux-perf

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/120a56da140000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/120a56da140000

Set VizDisplayCompositor feature enabled on desktop platforms. by kylechar@chromium.org
https://chromium.googlesource.com/chromium/src/+/f1ef4405b08f701d2bdae9494e0cdf92506de98e
CastV2Performance_gpu_60fps: 0.1472 → 0.1739 (+0.02673)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Looks like we just shifted time from capture_duration to send_to_renderer. There is a corresponding improvement in capture_duration and no increase in total_latency.

https://chromeperf.appspot.com/report?sid=ab63bcad6dc5b249e1e498e0091b1ce0cdeb29d2ec43e5fb1fea1761562d302c

It seems like capture end event is probably getting recorded earlier, likely because something has moved from browser to GPU process? This doesn't seem like a problem though.

 Issue 917268  has been merged into this issue.