Add a setting to disable threaded scrolling

Add a setting to disable threaded scrolling

Add a web setting for disabling threaded scrolling. The
ScrollingCoordinator enforces this by tagging all layers as main-thread
scroll targets.

Such a setting is useful for local testing and potentially for perf
instrumentation to track and improve main-thread driven scrolling. Note
also that this is but a temporary solution as other synchronous
scrolling triggers are explored.

The Chromium flag to trigger this path is added in
https://codereview.chromium.org/571963003/.

BUG=414323, 410974
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=182611

[jdduke (slow), 2014-09-15 16:55:53]

jdduke@chromium.org changed reviewers:
+ rbyers@chromium.org

[jdduke (slow), 2014-09-15 16:55:54]

rbyers@: WDYT? My first plan was to have this rule enforced in Chromium's
InputEventFilter or InputHandlerProxy, then I thought of moving it to LTHI, then
I realized we should probably make it as similar in practice as we can to the
eventual "easter egg" approach.

[Rick Byers, 2014-09-15 21:00:05]

I don't have a strong preference for where we do this.  I agree leaving it up to
blink is probably preferable (although note that we might ultimately want a
per-layer knob instead of a page-wide one).

https://codereview.chromium.org/564043003/diff/1/Source/core/frame/Settings.in
File Source/core/frame/Settings.in (right):

https://codereview.chromium.org/564043003/diff/1/Source/core/frame/Settings.i...
Source/core/frame/Settings.in:72: acceleratedScrollingEnabled initial=true,
invalidate=Style
I've been a little unclear on what exactly we use the term "accelerated" to
mean.  I think it sometimes means "uses the compositor to do transforms rather
than repainting" and other times means "runs on the impl thread".  If it's not
too much hassle, I'd prefer a less ambiguous term here - eg.
"threadedScrollingEnabled".  If my understanding is wrong and our use of
"accelerated" is unambiguous, then this is fine as is. 

Also please a comment describing when this setting should be used, ideally
linking to a bug that will track eventually removing it (maybe just
crbug.com/410974). 
See
https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/984HMZyFQNo/AV...
for discussion.

[jdduke (slow), 2014-09-15 21:02:07]

https://codereview.chromium.org/564043003/diff/1/Source/core/frame/Settings.in
File Source/core/frame/Settings.in (right):

https://codereview.chromium.org/564043003/diff/1/Source/core/frame/Settings.i...
Source/core/frame/Settings.in:72: acceleratedScrollingEnabled initial=true,
invalidate=Style
On 2014/09/15 21:00:05, Rick Byers wrote:
> I've been a little unclear on what exactly we use the term "accelerated" to
> mean.  I think it sometimes means "uses the compositor to do transforms rather
> than repainting" and other times means "runs on the impl thread".  If it's not
> too much hassle, I'd prefer a less ambiguous term here - eg.
> "threadedScrollingEnabled".  If my understanding is wrong and our use of
> "accelerated" is unambiguous, then this is fine as is. 
> 
> Also please a comment describing when this setting should be used, ideally
> linking to a bug that will track eventually removing it (maybe just
> crbug.com/410974). 
> See
>
https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/984HMZyFQNo/AV...
> for discussion.

Yeah, I use "threaded" in the Chromium side, I was just patterning after this
verbiage (they also use threadedCompositingEnabled for this setting in
Chromium).

[jdduke (slow), 2014-09-16 22:36:05]

OK, terminology updated.

[Rick Byers, 2014-09-17 13:17:51]

lgtm

[jdduke (slow), 2014-09-17 15:24:31]

jdduke@chromium.org changed reviewers:
+ abarth@chromium.org

[jdduke (slow), 2014-09-17 15:24:31]

abarth@: Owner review for Source/web and public/web? Thanks.

[jdduke (slow), 2014-09-23 16:03:25]

jdduke@chromium.org changed reviewers:
+ jamesr@chromium.org
- abarth@chromium.org

[jdduke (slow), 2014-09-23 16:03:26]

jamesr@: Could you take a look at Source/web and public/web? Thanks.

[jamesr, 2014-09-23 23:34:46]

Why don't you just disable threaded scroll handling in cc?

[jamesr, 2014-09-23 23:35:14]

Hmm, are you trying to benchmark avoiding the cost of computing scroll rects?

In that case lgtm

[jdduke (slow), 2014-09-24 16:22:32]

The CQ bit was checked by jdduke@chromium.org

[commit-bot: I haz the power, 2014-09-24 16:23:49]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/564043003/20001

[commit-bot: I haz the power, 2014-09-24 16:28:24]

Committed patchset #2 (id:20001) as 182611