On Mon, Sep 21, 2009 at 10:25 AM, Erik Kay <[email protected]> wrote:
> I'm not on the devtools team, but I'll take a swing at these.
>
> On Sun, Sep 20, 2009 at 1:12 AM, Andy <[email protected]> wrote:
>> The profiling feature looks very cool.  I'm profiling some expensive
>> JavaScript, but most of the profile is attributed to "(program)" --
>> 80-90%.  What does this mean?
>
> I believe that means that it's something in the renderer process other
> than javascript code.  So it could be layout, image decoding, DOM
> manipulation, etc.

I see.  But it would be a lot more useful to attribute those
activities to the function that is calling them, no?  Or have a little
more breakdown other than the generic (program) category.  It is not
just one page where I see this -- basically any page I do is showing
up ~90% (program), which leads me scratching my head at what code to
try to optimize.

I think possibly the profiler is mostly useful for really CPU bound
code (like raytracing in JS or something).  Most JS apps aren't like
that, so the output isn't so useful for all those cases.  I see >95%
"(program)" and then 0.2% - 0.4% of 50 other things.  If you look at
some random profiles of say Gmail you can see what I mean.

> It's still a work in progress, so no docs yet.  The quick answer on
> how it works is that it's a sampling profiler.  This means it doesn't
> account for 100% of code run (it may miss some very fast/short
> snippets of code), but that it's very lightweight, and does a pretty
> good job of identifying hotspots in your code.

OK, that helps.  But my problem is precisely to identify a general
hotspot, not do anything really detailed, and it isn't really useful.
It's great to have any JS profiler at all though, so I look forward to
improvements.

Andy

--~--~---------~--~----~------------~-------~--~----~
Chromium Discussion mailing list: [email protected] 
View archives, change email options, or unsubscribe: 
    http://groups.google.com/group/chromium-discuss
-~----------~----~----~----~------~----~------~--~---

Reply via email to