144863 - startup (ts) slowed down by 1.58%-2.64% (or 40ms - 52ms)

Status: RESOLVED WONTFIX

(Core :: XPCOM, defect)

Description

[Cathleen]

checkin caused slowdown

Comment 1

[Cathleen]

-> jaggernaut

http://bonsai.mozilla.org/cvsquery.cgi?module=MozillaTinderboxAll&date=explicit&mindate=1021432680&maxdate=1021434359

tinderbox ts:
-------------------
comet 1513 -> 1536  ( +23ms or +1.52% )
luna  3294 -> 3317  ( +23ms or +0.7% )

Comment 2

[jag (Peter Annema)]

Well, I don't know how much faster we can make this code. scc has a slightly
faster algorithm which we can look into.

scc, any other suggestions?

Comment 3

[jag (Peter Annema)]

After my second checkin for that bug:

comet 1537 -> 1553  ( +16ms or +1.04% )
luna  3297 -> 3326  ( +29ms or +0.88% )

Dunno what made luna go back down in between (doesn't seem to have helped
comet), I've compensated for this below (*):

Total regression times:

comet 1513 -> 1553  ( +40ms or +2.64% )
luna  3294 -> 3346* ( +52ms or +1.58% )

Comment 4

[Scott Collins]

see http://bugzilla.mozilla.org/show_bug.cgi?id=139649#c35 for one theory about
the slowdown, suggested profiling steps and solution.

Comment 5

[Cathleen]

this is also suspecious on slowing down page-load (tp) on moz 1.0 branch.
added 12ms or 0.5% slowdown on facedown tinderbox.

build log's cvs update

U mozilla/string/public/MANIFEST
U mozilla/string/public/Makefile.in
U mozilla/string/public/makefile.win
cvs server: mozilla/string/public/nsAPromiseString.h is no longer in the repository
U mozilla/string/public/nsAString.h
U mozilla/string/public/nsDependentConcatenation.h
U mozilla/string/public/nsDependentSubstring.h
U mozilla/string/public/nsPromiseFlatString.h
U mozilla/string/public/nsSlidingString.h
U mozilla/string/public/nsStringFwd.h
U mozilla/string/public/nsStringTraits.h
U mozilla/string/src/nsAString.cpp
U mozilla/string/src/nsDependentConcatenation.cpp
U mozilla/string/src/nsPromiseFlatString.cpp

http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=MozillaBranchTinderboxAll&branch=MOZILLA_1_0_0_BRANCH&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=5%2F14%2F2002+20%3A00%3A00&maxdate=5%2F14%2F2002+21%3A00%3A00&cvsroot=%2Fcvsroot

Comment 6

[Cathleen]

jag any updates on this bug?

Comment 7

[Cathleen]

Attachment: graph of startup regression on tinderbox comet

Comment 8

[jag (Peter Annema)]

I have a few ideas that might speed things up, I'll see if I can work on that
this week.

Comment 9

[Henrik Lynggaard Hansen]

This bug has not been touched for nearly 2 years. In most cases, that 
means it has "slipped through the net".

I don't think this is relevant anymore since there has been alot of changes and
speedups, in the last two year. Could someone please take a moment to 
add a comment to the bug with current status, and/or close it.

Comment 10

[Chris McAfee]

Agreed, I think we should probably punt here unless jag wants
to hand this off to someone for a look.

Comment 11

[Ryan VanderMeulen [:RyanVM]]

jag, are you still interested or should this bug be WONTFIXED out of its misery?

Comment 12

[Wayne Mery (:wsmwk)]

(In reply to comment #11)
>  should this bug be WONTFIXED out of its misery?