385957 - 2%-3% Tp/Tp2 regression evening of 2007-06-25

Status: RESOLVED INCOMPLETE

(Core :: General, defect, P2)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

There was a 2%-3% Tp and Tp2 regression on both Linux and Windows on the Firefox tinderbox a little before midnight on 2007-06-25.  That is, on the tinderboxes:
Linux bl-bldlnx01 Dep fx-linux-tbox perf test
WINNT 5.1 bl-bldxp01 Dep fx-win32-tbox perf test

The first of these tinderboxes reports more often and has more stable numbers than the second, but this is the size regression that's a little hard to pick out as a particular cycle even though there was clearly a change.  The most likely window for the regression is one of the three cycles between 20:52 and 22:23 (MOZ_CO_DATE).

Comment 1

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Checkins in that window:
http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=PhoenixTinderbox&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2007-06-25+20%3A52&maxdate=2007-06-25+22%3A23&cvsroot=%2Fcvsroot

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

We don't seem to have any talos performance numbers for the trunk (only for the branch).

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I just tried backing out bug 384836 and bug 385686.

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(That said, given the quality of the numbers we have, I'm not even all that confident in the wide window I gave above.)

Comment 5

[Dietrich Ayala (:dietrich)]

Bug 327350 should not adversely affect Tp, as it's only run during initialization of the History service, which occurs once at application startup.

Comment 6

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Tp2 pretty clearly went down after backing out those two patches, so it was one of the 2, at least for the Linux numbers, so I'm reopening the tree.

Comment 7

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

So it actually wasn't clear that it went down after backing those out -- it probably was just one low cycle.  So it seems like it might be one of the other patches in the URL above, or maybe even some other patches at nearby times.

Comment 8

[Dietrich Ayala (:dietrich)]

backed out bug 327350

Comment 9

[Dietrich Ayala (:dietrich)]

(In reply to comment #8)
> backed out bug 327350
> 

i've reverted this backout until windows perf #s are back and there's less
noise.

Comment 10

[Damon Sicore (:damons)]

David, do you still think this might block?

Comment 11

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

It probably should, but we need to generate some more stable numbers to figure it out.

Comment 12

[Damon Sicore (:damons)]

Ok.  +'ing this.  Marking as P2 so we get to it first thing after Beta 2.  

Comment 13

[Mike Schroepfer]

Dietrich - can you either take this or find the right owner

Comment 14

[Shawn Wilsher :sdwilsh]

I'm going through and marking old performance regression bugs as INCOMPLETE that are likely too old to be valid or get any traction on them.

Please re-open if you have more information or can demonstrate the regression still exists.