Pass FROM_HERE to ObserverListThreadSafe::Notify to improve profile.

Pass FROM_HERE to ObserverListThreadSafe::Notify to improve profile.

ObserverListThreadSafe::Notify is the 4th most called function in a
profile generated on my host. This is because all tasks posted from it
are stamped with this callsite, providing very little information.

This patch forces callers to pass in their own tracking data so that
the profile will differentiate between each observer list holder.

TBR=garykac@chromium.org,rdsmith@chromium.org,stuartmorgan@chromium.org,gbillock@chromium.org

Committed: https://crrev.com/9a77a7236aeab865e82dcb0baef21ab6e039cd2b
Cr-Commit-Position: refs/heads/master@{#315372}

[Reilly Grant (use Gerrit), 2015-02-06 22:29:38]

reillyg@chromium.org changed reviewers:
+ thakis@chromium.org

[Reilly Grant (use Gerrit), 2015-02-06 23:24:58]

Patchset #1 (id:1) has been deleted

[Reilly Grant (use Gerrit), 2015-02-06 23:25:13]

Patchset #1 (id:1) has been deleted

[Reilly Grant (use Gerrit), 2015-02-06 23:26:04]

Nico, please review this change to the ObserverListThreadSafe::Notify interface.
I'm still running try jobs, there may be some platform-specific files I have
missed.

[Nico, 2015-02-07 00:28:03]

If this is the 4th-hottest function, are there any performance concerns about
creating tracked_objects::Location object at each of the calls?

[Reilly Grant (use Gerrit), 2015-02-07 01:01:34]

On 2015/02/07 00:28:03, Nico wrote:
> If this is the 4th-hottest function, are there any performance concerns about
> creating tracked_objects::Location object at each of the calls?

This should be cheaper than the existing code as only one
tracked_objects::Location object per callsite instead of one per invocation of
PostTask.

[Nico, 2015-02-07 02:03:20]

lgtm

On 2015/02/07 01:01:34, reillyg wrote:
> On 2015/02/07 00:28:03, Nico wrote:
> > If this is the 4th-hottest function, are there any performance concerns
about
> > creating tracked_objects::Location object at each of the calls?
> 
> This should be cheaper than the existing code as only one
> tracked_objects::Location object per callsite instead of one per invocation of
> PostTask.

Err right. (Why cheaper though? Before every Notify() call built one object like
this, and this is still true – shouldn't it be the same? (except that there's
another pointer argument, but that should be cheap))

[Reilly Grant (use Gerrit), 2015-02-07 02:08:21]

On 2015/02/07 02:03:20, Nico wrote:
> lgtm
> 
> On 2015/02/07 01:01:34, reillyg wrote:
> > On 2015/02/07 00:28:03, Nico wrote:
> > > If this is the 4th-hottest function, are there any performance concerns
> about
> > > creating tracked_objects::Location object at each of the calls?
> > 
> > This should be cheaper than the existing code as only one
> > tracked_objects::Location object per callsite instead of one per invocation
of
> > PostTask.
> 
> Err right. (Why cheaper though? Before every Notify() call built one object
like
> this, and this is still true – shouldn't it be the same? (except that there's
> another pointer argument, but that should be cheap))

Depends on how smart the compiler was about realizing that FROM_HERE was a
constant and hoisting it out of the loop.

[Reilly Grant (use Gerrit), 2015-02-07 02:30:06]

reillyg@chromium.org changed reviewers:
+ jrummell@chromium.org, nasko@chromium.org

[Reilly Grant (use Gerrit), 2015-02-07 02:31:25]

reillyg@chromium.org changed reviewers:
+ rdsmith@chromium.org

[Reilly Grant (use Gerrit), 2015-02-07 02:32:08]

Adding nasko@ for content/, jrummel@ for media/ and rdsmith@ for net/.

[Reilly Grant (use Gerrit), 2015-02-07 02:34:49]

reillyg@chromium.org changed reviewers:
+ garykac@chromium.org, kalman@chromium.org, stuartmorgan@chromium.org

[Reilly Grant (use Gerrit), 2015-02-07 02:35:52]

reillyg@chromium.org changed reviewers:
+ gbillock@chromium.org

[Reilly Grant (use Gerrit), 2015-02-07 02:37:22]

Adding stuartmorgan@ for components/password_manager/, gbillock@ for
components/storage_monitor, kalman@ for extensions/ and garykac@ for remoting/.

[jrummell, 2015-02-09 18:23:10]

media/ lgtm

[Nico, 2015-02-09 18:26:14]

(i think this is mechanical enough that you can tbr other owners)

[nasko, 2015-02-09 18:27:07]

content/ LGTM

[not at google - send to devlin, 2015-02-09 18:34:20]

tbr works for me, but lgtm anyway

[Reilly Grant (use Gerrit), 2015-02-09 18:36:22]

The CQ bit was checked by reillyg@chromium.org

[commit-bot: I haz the power, 2015-02-09 18:37:49]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/877993003/20001

[commit-bot: I haz the power, 2015-02-09 20:18:41]

Committed patchset #1 (id:20001)

[commit-bot: I haz the power, 2015-02-09 20:19:35]

Patchset 1 (id:??) landed as
https://crrev.com/9a77a7236aeab865e82dcb0baef21ab6e039cd2b
Cr-Commit-Position: refs/heads/master@{#315372}

[Randy Smith (Not in Mondays), 2015-02-16 14:21:20]

net/ LGTM.

[stuartmorgan, 2015-02-18 16:56:41]

password_manager LGTM