1375470 - 4.16% Heap Unclassified (windows10-64-vm) regression on push 2b0d662801c24270a80187c9444dd829ee06be4e (Fri Jun 16 2017)

Status: NEW

(Core :: DOM: Core & HTML, defect, P3)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

We have detected an awsy regression from push:

https://hg.mozilla.org/releases/mozilla-beta/pushloghtml?changeset=2b0d662801c24270a80187c9444dd829ee06be4e

As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

  4%  Heap Unclassified summary windows10-64-vm opt      40,860,583.69 -> 42,560,716.24


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=7426

On the page above you can see an alert for each affected platform as well as a link to a graph showing the history of scores for this test. There is also a link to a treeherder page showing the jobs in a pushlog format.

To learn more about the regressing test(s), please see: https://developer.mozilla.org/en-US/docs/Mozilla/Performance/AWSY

Comment 1

[Joel Maher ( :jmaher ) (UTC -8)]

this showed up on mozilla-beta and we don't see any matching regression on trunk (firefox 56).  I am not sure why that is the case, but we are trying to document all regressions.

:erahm would know the most about what this specific regression means.

:smaug, As the patch author, can you help figure out why your patch from bug 1373067 is causing this AWSY regression?

Comment 2

[Olli Pettay [:smaug][[e-mail]]]

I can sort of understand why bug 1373067 could have caused some regression (collecting less often), but why beta only... no idea.
But, the initial patch had memory usage improvements
https://bugzilla.mozilla.org/show_bug.cgi?id=1367905#c52
Are we now close to those numbers?

Comment 3

[Joel Maher ( :jmaher ) (UTC -8)]

The improvements in bug 1367905 are Explicit Memory and Resident Memory, the issue we see on Beta is Heap Unclassified.

Looking over the data again, the values for win10 Heap are slightly higher on trunk:
https://treeherder.mozilla.org/perf.html#/graphs?timerange=7776000&series=%5Bmozilla-beta,fd986b27b1f1c0e10bef8d5ae050097f19686133,1,4%5D&series=%5Bautoland,fd986b27b1f1c0e10bef8d5ae050097f19686133,1,4%5D&series=%5Bmozilla-inbound,fd986b27b1f1c0e10bef8d5ae050097f19686133,1,4%5D&zoom=1496719904621.2766,1498156492000,37146068.10151861,47483146.753204

There was a regression June 13th- this change seems to be the same magnitude.  We tracked that change in bug 1373083 and what was the root cause is not uplifted to beta.

I wonder if there is something else going on here- maybe we hit certain limits in binary sizes, etc.

Comment 4

[Eric Rahm [:erahm]]

I'm concerned that our changes to the reported values from AWSY in bug 1372437 and bug 1372667 might be causing some confusion. We might want to uplift those.

Comment 5

[Andrew Overholt [:overholt]]

Is there anything left to do here?

Comment 6

[Joel Maher ( :jmaher ) (UTC -8)]

I am not sure, the values have continued to increase:
https://treeherder.mozilla.org/perf.html#/graphs?timerange=31536000&series=mozilla-beta,1486481,1,4&series=mozilla-beta,1533757,1,4&selected=%5Bmozilla-beta,fbe8096b6a3baec89600e46643641a6aac882aea,241181,313004915,4%5D

we have shipped this already, but we should either focus on the regressions or investigate this measurement from scratch.