Turn on enable begin frame scheduling by default

Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all platforms but mus
and Blimp.  This change means browser->renderer begin frame ticks
instead of sending vsync information and having a synthetic source on
the renderer side.  This feature was already on for Android so should
only be a real change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

This previously landed and regressed a number of metrics.  Some of these
have been investigated and fixed and the rest are WontFix.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/d020d554c9ffc0476d43f5364a47c522b64e3d9a
Cr-Commit-Position: refs/heads/master@{#408172}

[enne (OOO), 2016-05-02 20:00:16]

Description was changed from

==========
[wip] turn on begin frame scheduling everywhere
==========

to

==========
[wip] turn on begin frame scheduling everywhere
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_site_isolation
==========

[enne (OOO), 2016-05-02 20:00:17]

Description was changed from

==========
[wip] turn on begin frame scheduling everywhere
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_site_isolation
==========

to

==========
[wip] turn on begin frame scheduling everywhere
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[enne (OOO), 2016-05-04 20:05:16]

Description was changed from

==========
[wip] turn on begin frame scheduling everywhere
CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Turn on enable begin frame scheduling by default

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[enne (OOO), 2016-05-04 20:21:02]

Description was changed from

==========
Turn on enable begin frame scheduling by default

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all[*] platforms.
This was already on for Android (non-WebView) so should only be a real
change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performances regressions.

[*] WebView still uses a synchronous begin frame source.  MUS is not
hooked up to begin frame scheduling yet, but mojo:mash_session in an
"oxygen" build still works with this patch applied.  Blimp also doesn't
use begin frame scheduling and will eventually just be transitioned to a
synthetic begin frame source for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[enne (OOO), 2016-05-04 20:31:22]

Description was changed from

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all[*] platforms.
This was already on for Android (non-WebView) so should only be a real
change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performances regressions.

[*] WebView still uses a synchronous begin frame source.  MUS is not
hooked up to begin frame scheduling yet, but mojo:mash_session in an
"oxygen" build still works with this patch applied.  Blimp also doesn't
use begin frame scheduling and will eventually just be transitioned to a
synthetic begin frame source for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all[*] platforms.
This was already on for Android (non-WebView) so should only be a real
change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

[*] WebView still uses a synchronous begin frame source.  MUS is not
hooked up to begin frame scheduling yet, but mojo:mash_session in an
"oxygen" build still works with this patch applied.  Blimp also doesn't
use begin frame scheduling and will eventually just be transitioned to a
synthetic begin frame source for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[enne (OOO), 2016-05-04 20:35:54]

enne@chromium.org changed reviewers:
+ danakj@chromium.org

[enne (OOO), 2016-05-04 20:40:34]

enne@chromium.org changed reviewers:
+ boliu@chromium.org, piman@chromium.org, skyostil@chromium.org,
sunnyps@chromium.org
- danakj@chromium.org

[enne (OOO), 2016-05-04 20:44:03]

Description was changed from

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all[*] platforms.
This was already on for Android (non-WebView) so should only be a real
change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

[*] WebView still uses a synchronous begin frame source.  MUS is not
hooked up to begin frame scheduling yet, but mojo:mash_session in an
"oxygen" build still works with this patch applied.  Blimp also doesn't
use begin frame scheduling and will eventually just be transitioned to a
synthetic begin frame source for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling[1] for all[2] platforms.
This was already on for Android (non-WebView) so should only be a real
change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.


[1] In this case, "begin frame scheduling" means browser->renderer
begin frame ticks instead of sending vsync information and having
a synthetic source on the renderer side.

[2] WebView still uses a synchronous begin frame source.  MUS is not
hooked up to begin frame scheduling yet, but mojo:mash_session in an
"oxygen" build still works with this patch applied.  Blimp also doesn't
use begin frame scheduling and will eventually just be transitioned to a
synthetic begin frame source for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[enne (OOO), 2016-05-04 20:45:11]

boliu: android_webview/
piman: chrome/browser/chromeos/login/, content/
skyostil: content/browser/android/
sunnyps: fyi

[enne (OOO), 2016-05-04 20:47:35]

Description was changed from

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling[1] for all[2] platforms.
This was already on for Android (non-WebView) so should only be a real
change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.


[1] In this case, "begin frame scheduling" means browser->renderer
begin frame ticks instead of sending vsync information and having
a synthetic source on the renderer side.

[2] WebView still uses a synchronous begin frame source.  MUS is not
hooked up to begin frame scheduling yet, but mojo:mash_session in an
"oxygen" build still works with this patch applied.  Blimp also doesn't
use begin frame scheduling and will eventually just be transitioned to a
synthetic begin frame source for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling[1] for all[2] platforms.
This was already on for Android so should only be a real change
for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.


[1] In this case, "begin frame scheduling" means browser->renderer
begin frame ticks instead of sending vsync information and having
a synthetic source on the renderer side.

[2] MUS is not hooked up to begin frame scheduling yet, but
mojo:mash_session in an "oxygen" build still works with this patch
applied.  Blimp also doesn't use begin frame scheduling and will
eventually just be transitioned to a synthetic begin frame source
for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[boliu, 2016-05-04 20:48:58]

lgtm

[piman, 2016-05-04 22:38:50]

If this is a patch to tease out performance issues, should we keep a disable
flag?

[Sami, 2016-05-05 13:36:57]

lgtm \o/

We could turn the enable flag into a disable in this patch, but I'm not sure if
its worth the effort -- unified BeginFrame is where we want to be.

[enne (OOO), 2016-05-05 17:39:15]

On 2016/05/04 at 22:38:50, piman wrote:
> If this is a patch to tease out performance issues, should we keep a disable
flag?

I don't expect anything huge.  Telemetry smoke tests of smoothness and scheduler
tests looked fine when I ran them locally.  We're already doing this on Android.
 I expect regressions to be of the sort where the perfbot notices latency and we
want to bisect, rather than somebody in the field notcing that everything runs
at 1fps and needing a flag to check if it's my patch.

[enne (OOO), 2016-05-10 22:31:56]

piman: here's a patch with a disable flag.  Thanks for the suggestion.  :)

[piman, 2016-05-10 22:39:22]

LGTM, thanks!

[enne (OOO), 2016-05-11 19:51:43]

Hmm, looks like mac isn't happy with this.  I'll get my mac back into a workable
state and investigate.

[enne (OOO), 2016-05-24 00:26:31]

Ok, I'm fixing mac content shell issues in
https://codereview.chromium.org/2005013002, and so once that lands, I'll come
back and land this.

[enne (OOO), 2016-06-03 18:46:05]

On 2016/05/24 at 00:26:31, enne wrote:
> Ok, I'm fixing mac content shell issues in
https://codereview.chromium.org/2005013002, and so once that lands, I'll come
back and land this.

Thousands of years later, I have sorted through multiple mac issues.  I am
planning on landing this soon if the trybots don't have indigestion.

[enne (OOO), 2016-06-22 18:37:29]

Mac changes have rerelanded and this is finally green.  I'm going to give the
mac changes a day or so to bake and then will land this.

[enne (OOO), 2016-06-23 23:26:19]

The CQ bit was checked by enne@chromium.org

[enne (OOO), 2016-06-23 23:26:19]

The patchset sent to the CQ was uploaded after l-g-t-m from
skyostil@chromium.org, boliu@chromium.org, piman@chromium.org
Link to the patchset: https://codereview.chromium.org/1939253002/#ps200001
(title: "Rebase")

[commit-bot: I haz the power, 2016-06-23 23:26:52]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1939253002/200001

[commit-bot: I haz the power, 2016-06-24 02:41:57]

Committed patchset #11 (id:200001)

[commit-bot: I haz the power, 2016-06-24 02:43:21]

Description was changed from

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling[1] for all[2] platforms.
This was already on for Android so should only be a real change
for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.


[1] In this case, "begin frame scheduling" means browser->renderer
begin frame ticks instead of sending vsync information and having
a synthetic source on the renderer side.

[2] MUS is not hooked up to begin frame scheduling yet, but
mojo:mash_session in an "oxygen" build still works with this patch
applied.  Blimp also doesn't use begin frame scheduling and will
eventually just be transitioned to a synthetic begin frame source
for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling[1] for all[2] platforms.
This was already on for Android so should only be a real change
for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

[1] In this case, "begin frame scheduling" means browser->renderer
begin frame ticks instead of sending vsync information and having
a synthetic source on the renderer side.

[2] MUS is not hooked up to begin frame scheduling yet, but
mojo:mash_session in an "oxygen" build still works with this patch
applied.  Blimp also doesn't use begin frame scheduling and will
eventually just be transitioned to a synthetic begin frame source
for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/f2d7f5e1891703ec4384ededd80f896816921204
Cr-Commit-Position: refs/heads/master@{#401796}
==========

[commit-bot: I haz the power, 2016-06-24 02:43:24]

Patchset 11 (id:??) landed as
https://crrev.com/f2d7f5e1891703ec4384ededd80f896816921204
Cr-Commit-Position: refs/heads/master@{#401796}

[enne (OOO), 2016-06-27 18:23:46]

A revert of this CL (patchset #11 id:200001) has been created in
https://codereview.chromium.org/2100203002/ by enne@chromium.org.

The reason for reverting is: Causes smoothness regressions, other bugs

BUG=623174,623467,623490.

[enne (OOO), 2016-07-26 21:41:32]

Description was changed from

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling[1] for all[2] platforms.
This was already on for Android so should only be a real change
for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

[1] In this case, "begin frame scheduling" means browser->renderer
begin frame ticks instead of sending vsync information and having
a synthetic source on the renderer side.

[2] MUS is not hooked up to begin frame scheduling yet, but
mojo:mash_session in an "oxygen" build still works with this patch
applied.  Blimp also doesn't use begin frame scheduling and will
eventually just be transitioned to a synthetic begin frame source
for its engine half.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/f2d7f5e1891703ec4384ededd80f896816921204
Cr-Commit-Position: refs/heads/master@{#401796}
==========

to

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all platforms but mus
and Blimp.  This change means browser->renderer begin frame ticks
instead of sending vsync information and having a synthetic source on
the renderer side.  This feature was already on for Android so should
only be a real change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

This previously landed and regressed a number of metrics.  Some of these
have been investigated and fixed and the rest are WontFix.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

[enne (OOO), 2016-07-26 21:50:19]

Ok, now that https://codereview.chromium.org/2170053002 has landed, I think it's
time to give this another go.

I investigated all the regressions and put the findings in
https://docs.google.com/document/d/1aTaqW0yJj5C-9uTdkiQ7U6bCtBg6hdskzZcNOYrhT....

Summary: Real mac power issues (aka align display compositor with vsync oops)
were fixed.  A slight IO thread time regression for the extra IPC is WontFix. 
The biggest WontFix is cases where the browser compositor thread is pegged and
throughput is reduced because begin frames aren't sent.  (On the other hand,
latency is improved.)  It makes throughput worse for things like software
composited animometer focus benchmark where both renderer main thread and
browser main thread are both under a lot of load, though.  I'm inclined to
WontFix this as MUS tadpole will make it a lot better but am happy to discuss.

[enne (OOO), 2016-07-26 21:50:30]

The CQ bit was checked by enne@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-07-26 21:51:22]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-26 21:51:25]

Your CL relies on deprecated CQ feature(s):
 * Specifying master names in CQ_INCLUDE_TRYBOTS part of description without
"master." prefix is deprecated:
     tryserver.blink
   For more details, see http://crbug.com/617627.

[commit-bot: I haz the power, 2016-07-26 22:53:48]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-07-26 22:53:49]

Dry run: This issue passed the CQ dry run.

[piman, 2016-07-27 04:47:18]

Thanks a lot for the investigation!

LGTM.

[enne (OOO), 2016-07-27 17:38:47]

The CQ bit was checked by enne@chromium.org

[enne (OOO), 2016-07-27 17:38:48]

The patchset sent to the CQ was uploaded after l-g-t-m from
skyostil@chromium.org, boliu@chromium.org
Link to the patchset: https://codereview.chromium.org/1939253002/#ps240001
(title: "Rebase")

[commit-bot: I haz the power, 2016-07-27 17:39:14]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-07-27 17:39:17]

Your CL relies on deprecated CQ feature(s):
 * Specifying master names in CQ_INCLUDE_TRYBOTS part of description without
"master." prefix is deprecated:
     tryserver.blink
   For more details, see http://crbug.com/617627.

[commit-bot: I haz the power, 2016-07-27 17:43:39]

Committed patchset #13 (id:240001)

[commit-bot: I haz the power, 2016-07-27 17:45:43]

Description was changed from

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all platforms but mus
and Blimp.  This change means browser->renderer begin frame ticks
instead of sending vsync information and having a synthetic source on
the renderer side.  This feature was already on for Android so should
only be a real change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

This previously landed and regressed a number of metrics.  Some of these
have been investigated and fixed and the rest are WontFix.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel
==========

to

==========
Turn on enable begin frame scheduling by default

This turns on --enable-begin-frame-scheduling for all platforms but mus
and Blimp.  This change means browser->renderer begin frame ticks
instead of sending vsync information and having a synthetic source on
the renderer side.  This feature was already on for Android so should
only be a real change for desktop / ChromeOS platforms.

Lots of cleanup can follow from this like removing all commit vsync /
authoritative vsync / CompositorVSyncManager things, but this is a
smaller patch to suss out any performance regressions.

This previously landed and regressed a number of metrics.  Some of these
have been investigated and fixed and the rest are WontFix.

CQ_INCLUDE_TRYBOTS=tryserver.blink:linux_blink_rel

Committed: https://crrev.com/d020d554c9ffc0476d43f5364a47c522b64e3d9a
Cr-Commit-Position: refs/heads/master@{#408172}
==========

[commit-bot: I haz the power, 2016-07-27 17:45:45]

Patchset 13 (id:??) landed as
https://crrev.com/d020d554c9ffc0476d43f5364a47c522b64e3d9a
Cr-Commit-Position: refs/heads/master@{#408172}