New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 665424 link

Starred by 1 user

Issue metadata

Status: WontFix
Owner: ----
Closed: Nov 7
Cc:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 3
Type: Bug



Sign in to add a comment

Many small regressions overwhelm memory improvements

Project Member Reported by perezju@chromium.org, Nov 15 2016

Issue description

On background usage here:
https://chrome-health.googleplex.com/health-plan/android-chrome/memory/nexus5/?from_commit=428740&to_commit=431875

Or chrome perf: https://chromeperf.appspot.com/report?sid=299508ce2a28f86977fc0a89ee4b11c396b6abd6543aaec28f0ed77a4dbc63cf&start_rev=1477945585&end_rev=1479176906


There are two nice improvements in pss, native heap and private dirty. The first one is issue 663698, the second is issue 663700.

However lots of small regressions in the range http://test-results.appspot.com/revision_range?start=429267&end=431748 appear to completely overwhelm those improvements, particularly for overall pss.

Not sure what is the best course of action here.
 

Comment 1 by picksi@chromium.org, Nov 15 2016

Is it expected that private dirty and native heap (from the background measurement graph) move in parallel? Is one a subset of the other? I'm hoping they are mutually exclusive and their parallel movement is a false reading that is down to an infra-structure issue...
In my dashboard-looking experience, all of {pss, private dirty, native heap} tend to move up/down together, and native heap is usually the cleanest (less noisy) of all three.
> Is it expected that private dirty and native heap
Yes they are directly related. private_dirty is a superset of native_heap (% rare negligible exceptions).

>I'm hoping they are mutually exclusive and their parallel movement is a false reading that is down to an infra-structure issue...
No. But on the other side they show the same regressions and are not dual-regressions.
Labels: Performance-Memory
Status: WontFix (was: Untriaged)
This regression has been open for half a year. It's not very actionable and the regression has been in all Chrome user's hands for months.

Sign in to add a comment