commit 1125e092045aa91b4b90b439a3811a8f16a57cf6
author Miguel Casas-Sanchez <mcasas@chromium.org> Fri Apr 12 17:08:47 2019
committer Commit Bot <commit-bot@chromium.org> Fri Apr 12 17:08:47 2019
tree ddf5040a16bf4afbbf1f30b52d6e8bb7bb960164
parent 4c14665cb7d88f915ed61b7601660d2973b967ef

Teach WebRTC display perf test about VideoFrameSubmitter

WebRtcVideoDisplayPerfBrowserTest is a perf test that receives and
displays a remote peer connection, collects a series of intermediate
event timestamps, associates them with each other, and produces an
estimation of the decoded-video-to-raster delay.  Since it only
tracks Display::DrawAndSwap inside the viz process, it misses things
like the BufferQueue, which is sizeable. This CL extends the test to
add the collection and calculation of the whole mentioned delay,
including any time spent in the Display Compositor (a.k.a. viz
compositor).

Local runs indeed prove that the new "Post-decode-to-raster latency"
is always > than ToT's "Total Latency Mean" (e.g. [3]).

The test was introduced as part of [1] with a sample dashboard in [2].

[1] https://goo.gl/PKb69u
[2] https://chromeperf.appspot.com/report?sid=3f56e4a6575b0839e4f272eb72b49d7664a869ff23f0cce5528d34d6c6e1cf2d
[3] https://pastebin.com/raw/JcLSNtuT

Bug: 950183
Change-Id: I3b6525811859ddb339ec448ff000c87a55931d9e
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1565008
Reviewed-by: Frank Liberato <liberato@chromium.org>
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#650348}