All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=817235

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=428d1e45d2d0099e5c5ce1e03732eae669eb2a852154111eb6a8a276089f7663


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

android-nexus5

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14ae1ce8440000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/14ae1ce8440000

Revert "Revert "Separate creation time & flush time behavior in GrDrawOpAtlas (take 2)"" by robertphillips@google.com
https://skia.googlesource.com/skia/+/934c3d0c548d254a0d986963ee751caf0ffcb500

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

This recovered.

https://skia-review.googlesource.com/c/skia/+/110221 (Revert "Revert "Separate creation time & flush time behavior in GrDrawOpAtlas (take 2)") relanded: https://skia-review.googlesource.com/c/skia/+/109749 (Separate creation time & flush time behavior in GrDrawOpAtlas (take 2)) which had an unfortunate texture over-allocation bug.

It was reverted in: https://skia-review.googlesource.com/c/skia/+/111080 (Revert "Revert "Revert "Separate creation time & flush time behavior in GrDrawOpAtlas (take 2)""") which corresponds to the perf recovery on the Chrome bots.

This bug was fixed (and the CL relanded in) in: https://skia-review.googlesource.com/c/skia/+/111100 (Separate creation time & flush time behavior in GrDrawOpAtlas (take 3))

This CL rolled into Chrome on 3/5/18 at r540824 and displayed no perf regressions.

Note that:

the buggy CL landed on 2/27/18 at r539339.
the revert landed on 2/28/18 at 539929.
the fixed version landed on on 3/5/18 at r540824.