All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=838454

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=49a13e0b628bf2ca958784ffafb76cd89ed458af73d69f7a95235aa86ef1744e


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

linux
mac
win

Looks like a size regression occurred with a v8 autoroll. Assigning to V8 folks for further triage...

https://chromium.googlesource.com/chromium/src/+/98eb4947531658548004a76ced588517a8a32998%5E%21/#F0

Hey Jakob, may it be your CL [1]?

[1] https://chromium-review.googlesource.com/c/v8/v8/+/1030550

+ perf sheriff

Yes, my CL in #3 will increase the size of libv8.so (and static chrome builds). It moves code from the snapshot into the binary. At runtime, this is a net win since binary-embedded code only exists once instead of once per isolate.

miu, please use Blink>JavaScript when sending bugs to V8 and Infra>Client>V8 when  V8 infra support is needed. For regressions, CCing current perf sheriff is also useful: http://shortn/_d3V8V0adV1 (sorry, internal only).