All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=833291

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=cc9d237fe1f455ea13f6e66084a31c9ce9dd9a407b38705b4a4fea0045899149


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

chromium-rel-mac11-air
chromium-rel-mac11-pro
chromium-rel-mac12

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/169bfa46c40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/169bfa46c40000

Reduce noise of memory metric. by ulan@chromium.org
https://chromium.googlesource.com/catapult/+/64d08119f342961a5dd64049a03a50d36585899c

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

Issue 833270 has been merged into this issue.

This seems to affect mobiles too.
It's surprising to me since I thought catapult was not included in mobile builds. Please double check it.

toyoshim@, it is a memory metric change. It affects all platforms.

Please not that this is not a real regression. The way how we measure memory is changing.

The metrics that regressed on mobile are mostly private_footprint_size_avg and private_dirty_size_avg. The regression is about 500KB (V8 page size).
V8 unmaps unused pages concurrently. We are now sampling memory usage immediately after GC, so there might be not enough time for unmapping. It makes sense to wait for the concurrent unmapper in GCs caused by telemetry. I will add such logic in V8.

Ah, I see, that all makes sense.

The following revision refers to this bug:
  https://chromium.googlesource.com/v8/v8.git/+/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9

commit 10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9
Author: Ulan Degenbaev <ulan@chromium.org>
Date: Thu Apr 19 16:13:04 2018

[heap] Do eager unmapping in CollectAllAvailableGarbage.

The memory metric samples memory usage immediately after forcing GC via
LowMemoryNotification. This makes the metric sensitive to the unmapper
tasks timing.

This patch forces eager unmapping in CollectAllAvailableGarbage.

It also forces eager unmapping of non-regular chunks at the beginning
of Mark-Compact to avoid accumulation of non-regular chunks.

Bug:  chromium:833291 ,  chromium:826384 
Change-Id: Iddf02cd4ab8613385d033899d29525fe6ee47fdd
Reviewed-on: https://chromium-review.googlesource.com/1017102
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52696}
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/src/heap/heap.cc
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/src/heap/mark-compact.cc
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/src/heap/spaces.cc
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/src/heap/spaces.h
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/src/isolate.cc
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/test/cctest/BUILD.gn
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/test/cctest/heap/test-page-promotion.cc
[modify] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/test/cctest/heap/test-spaces.cc
[add] https://crrev.com/10fce9c80a6557c7e2f60bf24b61118eaa8ce0b9/test/cctest/heap/test-unmapper.cc

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14f02199c40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/14f02199c40000

Only show default poster when the controls attr is set by steimel@chromium.org
https://chromium.googlesource.com/chromium/src/+/92619d7e8b26975354187ea6e81ac98b49da6953

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

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14e0e3a9c40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/14e0e3a9c40000

Only show default poster when the controls attr is set by steimel@chromium.org
https://chromium.googlesource.com/chromium/src/+/92619d7e8b26975354187ea6e81ac98b49da6953

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

My CL in comment #8 fixed the noise of the private footprint metric. 

There are alerts are about GPU and compositor memory, which are unrelated to my CLs.

steimel@, is it expected that "Only show default poster when the controls attr is set" increases GPU and compositor memory usage?

Assigning to you to check and fix (or close as won't fix).

steimel, beccahughes: can you clarify the answers to ulan's questions in #13?

That CL should have a positive effect on performance so I will close as WontFix.