I originally assumed both would be x64, but I'm not sure if Chromium 
installed through Ubuntu's software center is 32 or 64 bit.  The shell was 
64 for sure.  Thanks for the additional parameter to check.

We picked these versions because that is what installed on a regular user 
machine, but on my Windows box it has 3.12.19.15, so maybe a newer version 
isn't such a bad test.  And as you say nobody there likely cares about the 
difference in this version.

The subtests that differ don't follow me across different CPUs (Ivy Bridge 
vs. Llano vs. ...), but there often seem to be a couple that are faster in 
the browser.  Originally we found this on an AMD Bobcat (E 350) CPU, and 
the two tests were ai-astar and audio-oscillator.  Ai-astar was about 25% 
faster and audio-oscillator I think was even higher.  I don't have the 
machines or data in front of me today to say more conclusively, but maybe 
this can spark something.  I think audio-oscillator might have been common 
to the lists.

On Saturday, October 20, 2012 5:42:04 AM UTC-7, Jakob Kummerow wrote:
>
> Hard to say without knowing more about what you did and what you observed. 
> E.g. if your browser was an x64 build and your command-line V8 was an ia32 
> build, that would explain all sorts of differences. Telling us *which* 
> tests exactly were faster in the browser would also add quite a bit of 
> relevant information.
>
> That said, V8 3.8 is ancient. I don't think anyone really cares about its 
> performance any more. Please repeat your tests with the current development 
> version (V8 trunk or bleeding_edge, Chromium M24 trunk or Chrome Dev 
> channel).
>
>
> On Fri, Oct 19, 2012 at 10:09 PM, Joe Millenbach 
> <[email protected]<javascript:>
> > wrote:
>
>> [Cross posting between Chromium and v8 as related to both]
>> I wondered if anyone knows why running some Kraken sub tests in a
>> browser go faster than if you run the same test from a command line
>> shell.  Like by 20-30% sometimes.  We expected the browser to only add
>> overhead, like it does for the web based SunSpider.
>>
>> I should note we don't have exactly the same version of shell and
>> browser v8 engine (shell is 3.8.9 and browser says 3.8.10).
>>
>> --
>> v8-users mailing list
>> [email protected] <javascript:>
>> http://groups.google.com/group/v8-users
>>
>
>

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

Reply via email to