Enable scavenging in RenderThread::IdleHandler.

Enable scavenging in RenderThread::IdleHandler.

[Mike Belshe, 2009-09-02 22:08:35]

http://codereview.chromium.org/172109/diff/1005/1012
File chrome/renderer/render_thread.cc (left):

http://codereview.chromium.org/172109/diff/1005/1012#oldcode468
Line 468: // as much as it can.  There is no point in continuing to call it.
I don't think you should remove this.

There are two important things going on here.

a) We dampen the rate at which we call the IdleNotification over time.
b) When the IdleNotification returns true, it means "I did a full GC Compact,
you can't squeeze me anymore, so leave me alone".

These two together make it so that we won't drain laptop batteries with idle tab
work.

[Mike Belshe, 2009-09-11 19:40:52]

LGTM - with questions addressed.

Ask Erik Corry or Darin Fisher for a second opinion on the backoff.

http://codereview.chromium.org/172109/diff/12070/13027
File chrome/renderer/render_thread.cc (left):

http://codereview.chromium.org/172109/diff/12070/13027#oldcode467
Line 467: // When V8::IdleNotification returns true, it means that it has
cleaned up
I still think we should leave the backoff in place.

http://codereview.chromium.org/172109/diff/12070/13021
File third_party/tcmalloc/tcmalloc.cc (right):

http://codereview.chromium.org/172109/diff/12070/13021#newcode621
Line 621: Static::pageheap()->Scavenge();
Do we need to lock the pageheap_lock() here?  I looked at PageHeap, and couldn't
conclude that we do?

[darin (slow to review), 2009-09-11 20:14:20]

RenderThread changes LGTM

[antonm, 2009-09-15 15:19:36]

Mike,

I incorporated slowing down of idle handlers invocation as per separate email
thread.  May you have another look?

Try bots are almost fine, the problem is with
http://build.chromium.org/buildbot/try-server/builders/win/builds/18399/steps...

You mentioned that there are some failures, is one above know or induced by my
patch?

[mbelshe, 2009-09-15 16:46:00]

That looks like a test error to me.
Mike


On Tue, Sep 15, 2009 at 8:19 AM, <antonm@chromium.org> wrote:

> Mike,
>
> I incorporated slowing down of idle handlers invocation as per separate
> email
> thread.  May you have another look?
>
> Try bots are almost fine, the problem is with
>
>
http://build.chromium.org/buildbot/try-server/builders/win/builds/18399/steps...
>
> You mentioned that there are some failures, is one above know or induced by
> my
> patch?
>
>
> http://codereview.chromium.org/172109
>

[Anton Muhin, 2009-09-15 17:15:00]

Good.  Thanks a lot for review, Mike!

yours,
anton.

On Tue, Sep 15, 2009 at 8:45 PM, Mike Belshe <mbelshe@google.com> wrote:
> That looks like a test error to me.
> Mike
>
> On Tue, Sep 15, 2009 at 8:19 AM, <antonm@chromium.org> wrote:
>>
>> Mike,
>>
>> I incorporated slowing down of idle handlers invocation as per separate
>> email
>> thread. =A0May you have another look?
>>
>> Try bots are almost fine, the problem is with
>>
>> http://build.chromium.org/buildbot/try-server/builders/win/builds/18399/=
steps/ui_tests/logs/stdio
>>
>> You mentioned that there are some failures, is one above know or induced
>> by my
>> patch?
>>
>> http://codereview.chromium.org/172109
>
>