cc: Block the main thread for texture layers during impl-side painting

cc: Block the main thread for texture layers during impl-side painting

This patch adds an additional completion event to thread proxy.  Right now, it checks for the lack of a pending tree after drawing (which is currently where tree activation is triggered).

R=nduca@chromium.org
BUG=164993
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=177218

[enne (OOO), 2012-12-29 00:44:00]

If a scheduling event is added for tree activation, then maybe the proxy can be
notified when that happens and unblock the main thread then.

The only bad part about this patch is that it makes WebGL Aquarium go from 60FPS
to a janky ~40ish on my Z600.  I suspect this is because even though the tree
could be immediately activated, it isn't until the next draw, and so the main
thread is blocked until vsync.

[nduca, 2013-01-05 04:56:22]

https://codereview.chromium.org/11710004/diff/1/cc/thread_proxy.cc
File cc/thread_proxy.cc (right):

https://codereview.chromium.org/11710004/diff/1/cc/thread_proxy.cc#newcode724
cc/thread_proxy.cc:724: m_commitCompletionEventOnImplThread->signal();
how about making not signaling this in impl side painting mode and then instead
if we activate the pending tree in scheduledActionDraw we signal this then?

[enne (OOO), 2013-01-05 06:25:45]

https://codereview.chromium.org/11710004/diff/1/cc/thread_proxy.cc
File cc/thread_proxy.cc (right):

https://codereview.chromium.org/11710004/diff/1/cc/thread_proxy.cc#newcode724
cc/thread_proxy.cc:724: m_commitCompletionEventOnImplThread->signal();
On 2013/01/05 04:56:22, nduca wrote:
> how about making not signaling this in impl side painting mode and then
instead
> if we activate the pending tree in scheduledActionDraw we signal this then?

You mean conditionally not signaling this here based on blocksPendingCommit and
just communicating that boolean rather than the mutex?

[nduca, 2013-01-05 06:34:02]

Sort of. Basically, we let the commit finish but leave the completion event
still set if we had a blocksCommit bit. Then anytime we activate the tree and
there m_pendingCompletionEvent, signal it then.... if we had done a commit and
there wans't a blocksCommit, m_pendingCompletionEvent would be null so we'd not
need to do antyhing...

[enne (OOO), 2013-01-08 02:00:51]

Done.  PTAL.

[nduca, 2013-01-08 04:49:27]

LGTM

https://codereview.chromium.org/11710004/diff/7001/cc/thread_proxy.cc
File cc/thread_proxy.cc (right):

https://codereview.chromium.org/11710004/diff/7001/cc/thread_proxy.cc#newcode726
cc/thread_proxy.cc:726: TRACE_EVENT0("cc",
"ThreadProxy::scheduledActionCommit_HoldCommit");
You can just write HoldCommitUntilActivate and use TRACE_EVENT_INSTANT0 ---
this'll get issued be inside a ScheduldActionCommit, so it'll be pretty
straightfowrad to see who did it.

[commit-bot: I haz the power, 2013-01-08 05:12:32]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/enne@chromium.org/11710004/3002

[commit-bot: I haz the power, 2013-01-08 07:13:47]

Retried try job too often on win_rel for step(s) browser_tests

[commit-bot: I haz the power, 2013-01-08 08:01:05]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/enne@chromium.org/11710004/3002

[commit-bot: I haz the power, 2013-01-08 12:20:33]

The commit queue went berserk retrying too often for a
seemingly flaky test. Builder is android_dbg, revision is HEAD

[commit-bot: I haz the power, 2013-01-08 15:13:18]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/enne@chromium.org/11710004/3002

[commit-bot: I haz the power, 2013-01-08 18:41:56]

Retried try job too often on android_dbg for step(s) build

[commit-bot: I haz the power, 2013-01-08 22:19:05]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/enne@chromium.org/11710004/3002

[commit-bot: I haz the power, 2013-01-09 00:39:23]

Commit queue patch verification failed without an error message.
Something went wrong, probably a crash, a hickup or simply
the monkeys went out for dinner.
Please email commit-bot@chromium.org with the CL url.

[commit-bot: I haz the power, 2013-01-09 18:02:26]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/enne@chromium.org/11710004/3002

[commit-bot: I haz the power, 2013-01-09 22:41:27]

The commit queue went berserk retrying too often for a
seemingly flaky test. Builder is android_dbg, revision is HEAD

[commit-bot: I haz the power, 2013-01-16 18:07:00]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/enne@chromium.org/11710004/3002

[commit-bot: I haz the power, 2013-01-16 21:12:42]

Change committed as 177218