All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=865582

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=2dc559265f765727e7d100b3a7e67fd663f911525276e0727cc278d15d73f53e


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

linux-perf

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/1465eed1a40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/1465eed1a40000

Migrate AutofillDownloadManager to SimpleURLLoader by tonikitoo@igalia.com
https://chromium.googlesource.com/chromium/src/+/3328d5c434b366bf9c6ddd344867be2c7f209a5e
2.855 → 2.897 (+0.0423)

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

If you look at the graph over time, it appears to have historically had a bi-modal distribution (https://chromeperf.appspot.com/report?sid=748a299c03f94bea7f3966bc1afa32ad4d1ddbccdec7d46aed013ed9b460ff57&start_rev=513348&end_rev=576788), going back and forth between two values.

Given the unlikelihood of a change in the code to fetch autofill data having an impact on rasterization performance that's only visible on one Linux test suite, I'd tend to call this noise (Maybe it's based on binary size, or relative placement of two methods within the binary).

+benchmark owners for comment on noise.