Talos Regression :( Dromaeo (CSS) decrease 2.73% on WINNT 6.1 x64 Firefox

[Nobody]

Regression :( Dromaeo (CSS) decrease 2.73% on WINNT 6.1 x64 Firefox
-------------------------------------------------------------------
Previous: avg 3265.708 stddev 36.744 of 30 runs up to revision eb84780783ed
New : avg 3176.438 stddev 14.846 of 5 runs since revision f694183357ec
Change : -89.270 (2.73% / z=2.429)
Graph : http://mzl.la/tNtAc2

Changeset range: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=eb84780783ed&tochange=f694183357ec

Changesets:
* http://hg.mozilla.org/mozilla-central/rev/61094ae6da18
: David Zbarsky <dzba...@gmail.com> - Bug 660959 - Update link state less lazily r=bz
: http://bugzilla.mozilla.org/show_bug.cgi?id=660959

* http://hg.mozilla.org/mozilla-central/rev/03970d1fc919
: David Zbarsky <dzba...@gmail.com> - Bug 661746 - Part 1: Delay setting slow selector flags during selector matching to avoid mutating Elements. r=bzbarsky
: http://bugzilla.mozilla.org/show_bug.cgi?id=661746

* http://hg.mozilla.org/mozilla-central/rev/1cb3705557d9
: David Zbarsky <dzba...@gmail.com> - Bug 661746 - Part 2: Make selector matching work with const elements. r=bzbarsky
: http://bugzilla.mozilla.org/show_bug.cgi?id=661746

* http://hg.mozilla.org/mozilla-central/rev/61534874aed7
: David Zbarsky <dzba...@gmail.com> - Bug 661746 - Part 3: Use const selectors during selector matching. r=bzbarsky
: http://bugzilla.mozilla.org/show_bug.cgi?id=661746

* http://hg.mozilla.org/mozilla-central/rev/2f403e4c42c8
: David Zbarsky <dzba...@gmail.com> - Bug 661746 - Part 4: use const nsIDocument in selector matching. r=bzbarsky
: http://bugzilla.mozilla.org/show_bug.cgi?id=661746

* http://hg.mozilla.org/mozilla-central/rev/15cf68a3c027
: David Zbarsky <dzba...@gmail.com> - Bug 661746 - Part 5: Avoid creating atoms during selector matching. r=bzbarsky
: http://bugzilla.mozilla.org/show_bug.cgi?id=661746

* http://hg.mozilla.org/mozilla-central/rev/dc82c3485d1a
: Brian Birtles <bir...@gmail.com> - Bug 690994 - Check for self-dependent times when there are coincident zero-duration intervals; r=dholbert
: http://bugzilla.mozilla.org/show_bug.cgi?id=690994

* http://hg.mozilla.org/mozilla-central/rev/b1e2ee0bbdb5
: Brian Birtles <bir...@gmail.com> - Bug 697640 - Ignore self-dependent end instance times when determining if an open-ended interval is ok; r=dholbert
: http://bugzilla.mozilla.org/show_bug.cgi?id=697640

* http://hg.mozilla.org/mozilla-central/rev/677fe017e6a0
: Joe Drew <j...@drew.ca> - Bug 685322 - Create explicit conversion constructors for Rect and Point taking IntRect and IntPoint. r=roc
: http://bugzilla.mozilla.org/show_bug.cgi?id=685322

* http://hg.mozilla.org/mozilla-central/rev/e6705f49eaaa
: Joe Drew <j...@drew.ca> - Bug 685326 - Move Round{,In,Out} from gfxRect to BaseRect, and make them no-ops on IntRect. r=Bas
: http://bugzilla.mozilla.org/show_bug.cgi?id=685326

* http://hg.mozilla.org/mozilla-central/rev/6ae6d3beeaf4
: Joe Drew <j...@drew.ca> - Bug 685767 - Factor blurring out into its own class, and use it from gfxAlphaBoxBlur. r=mattwoodrow
: http://bugzilla.mozilla.org/show_bug.cgi?id=685767

* http://hg.mozilla.org/mozilla-central/rev/33547cc48103
: Robert O'Callahan <rob...@ocallahan.org> - Bug 701528. Fix component alpha shader so correct destination alpha values are produced. r=bas
: http://bugzilla.mozilla.org/show_bug.cgi?id=701528

* http://hg.mozilla.org/mozilla-central/rev/cb9611719bc7
: Makoto Kato <m_k...@ga2.so-net.ne.jp> - Bug 693227 - Remove nsIWindowsShellService.unreadMailCount. r=jimm
: http://bugzilla.mozilla.org/show_bug.cgi?id=693227

* http://hg.mozilla.org/mozilla-central/rev/09ad1943c19a
: Henri Sivonen <hsiv...@iki.fi> - Bug 693399 - r=bzbarsky.
: http://bugzilla.mozilla.org/show_bug.cgi?id=693399

* http://hg.mozilla.org/mozilla-central/rev/1b823c46084b
: Jonathan Watt <jw...@jwatt.org> - Bug 702283 - Remove dead code RegionMark (left behind by bug 614732 attachment 563617 (changeset 8ddd09649e13)) since now we use GetMarkBBoxContribution (which doesn't need to update the covered regions of a mark's content). r=longsonr.
: http://bugzilla.mozilla.org/show_bug.cgi?id=702283

* and 9 more

Bugs:
* http://bugzilla.mozilla.org/show_bug.cgi?id=701761
* http://bugzilla.mozilla.org/show_bug.cgi?id=702283
* http://bugzilla.mozilla.org/show_bug.cgi?id=685767
* http://bugzilla.mozilla.org/show_bug.cgi?id=697640
* http://bugzilla.mozilla.org/show_bug.cgi?id=701682
* http://bugzilla.mozilla.org/show_bug.cgi?id=685322
* http://bugzilla.mozilla.org/show_bug.cgi?id=693227
* http://bugzilla.mozilla.org/show_bug.cgi?id=690994
* http://bugzilla.mozilla.org/show_bug.cgi?id=685326
* http://bugzilla.mozilla.org/show_bug.cgi?id=701259
* http://bugzilla.mozilla.org/show_bug.cgi?id=661746
* http://bugzilla.mozilla.org/show_bug.cgi?id=698326
* http://bugzilla.mozilla.org/show_bug.cgi?id=693399
* http://bugzilla.mozilla.org/show_bug.cgi?id=701528
* http://bugzilla.mozilla.org/show_bug.cgi?id=697279
* http://bugzilla.mozilla.org/show_bug.cgi?id=700764
* http://bugzilla.mozilla.org/show_bug.cgi?id=697246
* http://bugzilla.mozilla.org/show_bug.cgi?id=660959

[Nobody]

Regression :( Tp5 MozAfterPaint (Memset) increase 5.28% on XP Firefox-Non-PGO
-----------------------------------------------------------------------------
Previous: avg 129648766.667 stddev 1402593.624 of 30 runs up to revision eb84780783ed
New : avg 136495600.000 stddev 732855.238 of 5 runs since revision f694183357ec
Change : +6846833.333 (5.28% / z=4.882)
Graph : http://mzl.la/ulIwoM

[Nobody]

Regression :( Tp5 MozAfterPaint (Private Bytes) increase 5.51% on XP Firefox-Non-PGO
------------------------------------------------------------------------------------
Previous: avg 123252466.667 stddev 1462051.080 of 30 runs up to revision eb84780783ed
New : avg 130041200.000 stddev 283209.640 of 5 runs since revision f694183357ec
Change : +6788733.333 (5.51% / z=4.643)
Graph : http://mzl.la/sTTvxn

[Nobody]

Regression :( Dromaeo (jslib) decrease 2.03% on XP Firefox-Non-PGO
------------------------------------------------------------------
Previous: avg 135.348 stddev 0.967 of 30 runs up to revision eb84780783ed
New : avg 132.603 stddev 0.351 of 5 runs since revision f694183357ec
Change : -2.745 (2.03% / z=2.840)
Graph : http://mzl.la/rvF1wa

[Nobody]

Regression :( Tp5 MozAfterPaint (Memset) increase 4.87% on XP Firefox
---------------------------------------------------------------------
Previous: avg 132363266.667 stddev 1174859.755 of 30 runs up to revision eb84780783ed
New : avg 138814600.000 stddev 209673.079 of 5 runs since revision 9ae1d4f44b8b
Change : +6451333.333 (4.87% / z=5.491)
Graph : http://mzl.la/u7Xt8x

Changeset range: http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=eb84780783ed&tochange=9ae1d4f44b8b

* and 12 more

Bugs:
* http://bugzilla.mozilla.org/show_bug.cgi?id=701761
* http://bugzilla.mozilla.org/show_bug.cgi?id=702283
* http://bugzilla.mozilla.org/show_bug.cgi?id=685767
* http://bugzilla.mozilla.org/show_bug.cgi?id=697640
* http://bugzilla.mozilla.org/show_bug.cgi?id=701682
* http://bugzilla.mozilla.org/show_bug.cgi?id=685322
* http://bugzilla.mozilla.org/show_bug.cgi?id=693227

* http://bugzilla.mozilla.org/show_bug.cgi?id=701996

* http://bugzilla.mozilla.org/show_bug.cgi?id=690994
* http://bugzilla.mozilla.org/show_bug.cgi?id=685326

* http://bugzilla.mozilla.org/show_bug.cgi?id=702029

* http://bugzilla.mozilla.org/show_bug.cgi?id=701259
* http://bugzilla.mozilla.org/show_bug.cgi?id=661746
* http://bugzilla.mozilla.org/show_bug.cgi?id=698326
* http://bugzilla.mozilla.org/show_bug.cgi?id=693399

* and 6 more

[Nobody]

Regression :( Tp5 MozAfterPaint (Private Bytes) increase 4.74% on XP Firefox
----------------------------------------------------------------------------
Previous: avg 126243400.000 stddev 1330430.177 of 30 runs up to revision eb84780783ed
New : avg 132231000.000 stddev 481950.205 of 5 runs since revision 9ae1d4f44b8b
Change : +5987600.000 (4.74% / z=4.500)
Graph : http://mzl.la/tkFaMs

[Armen Zambrano Gasparnian]

Has any of these being investigated?

[Marco Bonardo]

On 16/11/2011 15:24, Armen Zambrano Gasparnian wrote:
> Has any of these being investigated?

I don't think so.

Looking at the Tp5 private bytes test the range should be this one, that
includes multiple changes, none of which looks particularly memory
hunger. Maybe someone else may identify a possible culprit?

http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=15cf68a3c027&tochange=736d3db54c41

[Ed Morley]

After poking around in the graphs and trying to use compare-talos without success, I've just kicked off a number of try runs to see if that helps. For the ranges where bug 661746 had landed but not yet been backed out, I backed it out before sending to try since it breaks Win PGO builds.

From newest cset to old...

Built from: https://hg.mozilla.org/integration/mozilla-inbound/rev/f694183357ec
https://tbpl.mozilla.org/?tree=Try&rev=df9b4267a896

Built from: https://hg.mozilla.org/integration/mozilla-inbound/rev/920c5da54a5c
https://tbpl.mozilla.org/?tree=Try&rev=3527db5b4dbb

Built from: https://hg.mozilla.org/integration/mozilla-inbound/rev/add05f884131
https://tbpl.mozilla.org/?tree=Try&rev=79e09f2350ff

Built from: https://hg.mozilla.org/integration/mozilla-inbound/rev/09ad1943c19a
https://tbpl.mozilla.org/?tree=Try&rev=3d570cbf42aa

Built from: https://hg.mozilla.org/integration/mozilla-inbound/rev/e6705f49eaaa
https://tbpl.mozilla.org/?tree=Try&rev=ca9af4dfae90

[Ed Morley]

Backing out bug 685767 (http://hg.mozilla.org/mozilla-central/rev/b62e6ee5ba9b) appears to have fixed this:
http://graphs-new.mozilla.org/graph.html#tests=[[116,1,1],[124,1,1]]&sel=none&displayrange=7&datatype=running

In addition, users in bug 702942 have confirmed their issue no longer occurs.

[Armen Zambrano Gasparnian]

Thanks Ed! You're splendid :)