I have tried to look a bit more into this, but without finding anything to
explain the matter. First of all I looked at all the perf-bots:

*XP Release v8 latest*
http://build.chromium.org/buildbot/perf/xp-release-v8-latest/sunspider/report.html?history=150&rev=54000
*50969 - 51216 +30% on SunSpider** and much more variance*

*XP Release Single Core*
http://build.chromium.org/buildbot/perf/xp-release-single-core/sunspider/report.html?history=150&rev=54000
*50901 - 52449 +5% on SunSpider and much more variance*

*XP Release Dual Core*
http://build.chromium.org/buildbot/perf/xp-release-dual-core/sunspider/report.html?history=450&rev=52000
*Nothing*

*Vista Release Dual Core*
http://build.chromium.org/buildbot/perf/vista-release-dual-core/sunspider/report.html?history=1000&rev=53000
*Nothing*

*Vista Release Single Core*
http://build.chromium.org/buildbot/perf/vista-release-single-core/sunspider/report.html?history=250&rev=53000
*Nothing*

*Linux Release Hardy*
http://build.chromium.org/buildbot/perf/linux-release-hardy/sunspider/report.html?history=1000&rev=53000
*Nothing*

*Mac Release 10.5*
http://build.chromium.org/buildbot/perf/mac-release-10.5/sunspider/report.html?history=1000&rev=53000
*Nothing*

Besides the regression on SunSpider the variance overall is also much
larger, and also startup time looks to be hurt.

The strange thing is that it is only *XP Release v8 Latest* which show a
regression, as v8 trunk is updated bi-weekly, so it might be the bot that
needs to be clobbered. I tried to checkout revisions 50969 and 51216
locally, but could not measure any regressions locally (Z600 with Vista).

Nicolas: I cant find v8-latest on any waterfall, so therefore I cannot
clobber it. Where should I look for it?

How the regression with *XP Release Single Core* is releated is also not
clear. Looking through the changes from 50969 to 51216 the following two has
to do with startup and timers.

  http://src.chromium.org/viewvc/chrome?view=rev&revision=51006
  http://src.chromium.org/viewvc/chrome?view=rev&revision=51102

I tried to revert in a local build, but could also not measure any
difference.

Regards,
Søren

On Wed, Jul 28, 2010 at 10:44, Anton Muhin <[email protected]> wrote:

> From the right address.
>
> On Wed, Jul 28, 2010 at 12:41 PM, Anton Muhin <[email protected]> wrote:
> > The strange thing is v8 has its own perf monitoring infrastructure
> > which didn't notice any regression.  And v8 gets pushed to trunk---the
> > version which all other builders should use---quite regularly (~two
> > times a week).
> >
> > Technically v8 perf monitoring is on Linux while build bots are on
> Windows.
> >
> > Anyway, I'll check if I can repro it on my win box.
> >
> > yours,
> > anton.
> >
> > On Wed, Jul 28, 2010 at 12:51 AM, Peter Kasting <[email protected]>
> wrote:
> >> +CC v8-dev
> >> V8 team, are you aware of the issue below?
> >> PK
> >>
> >> On Tue, Jul 27, 2010 at 1:48 PM, Carlos Pizano <[email protected]>
> wrote:
> >>>
> >>> just surprised that is 4 weeks old and nobody complaining
> >>>
> >>> On Jul 27, 1:37 pm, Ojan Vafai <[email protected]> wrote:
> >>> > Looks like it's only a regression on the v8-latest bots. So, it seems
> >>> > like a
> >>> > v8 regression that hasn't been rolled into the Chromium tree.
> >>> >
> >>> > On Tue, Jul 27, 2010 at 1:15 PM, Carlos Pizano <[email protected]>
> wrote:
> >>> >
> >>> > > >
> http://build.chromium.org/buildbot/perf/xp-release-v8-latest/sunspider/report.html?history=150&rev=-1
> >>> >
> >>> > > Looking at the chromium-dev traffic I don't see any mention of
> this.
> >>> > > Nor a bug.
> >>
> >> --
> >> Chromium Developers mailing list: [email protected]
> >> View archives, change email options, or unsubscribe:
> >> http://groups.google.com/a/chromium.org/group/chromium-dev
> >>
> >
>
> --
> v8-dev mailing list
> [email protected]
> http://groups.google.com/group/v8-dev
>

-- 
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev

Reply via email to