Put rare PaintLayer fields into PaintLayerRareData

Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper			36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

This CL reduces sizeof(PaintLayer) from 344 bytes to 208 bytes on LP 64
systems.

BUG=581124
Committed: https://crrev.com/299d5beba4af85652d96a2b2da1733ebcdec2c5e
Cr-Commit-Position: refs/heads/master@{#372141}

[Xianzhu, 2016-01-26 00:28:21]

Description was changed from

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 461 in http://ct.skia.org for details):
blockSelectionGapsBounds: 0.00%
transform: 0.00%
reflectionInfo: 0.00%
enclosingPaginationLayer: 0.01%
subpixelAccumulation: 0.72%
groupdMapping: 0.43%
compositedLayerMapping: 1.52%
compositingReasons: 2.00%
offsetForInFlowPosition: 2.62%
potentialCompsotingReasonsFromStyle: 3.87%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

BUG=581124
==========

to

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 461 in http://ct.skia.org for details):
blockSelectionGapsBounds: 0.00%
transform: 0.00%
reflectionInfo: 0.00%
enclosingPaginationLayer: 0.01%
subpixelAccumulation: 0.72%
groupdMapping: 0.43%
compositedLayerMapping: 1.52%
compositingReasons: 2.00%
offsetForInFlowPosition: 2.62%
potentialCompsotingReasonsFromStyle: 3.87% (which is likely to cover most
of the above)

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

BUG=581124
==========

[Xianzhu, 2016-01-26 20:33:25]

Description was changed from

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 461 in http://ct.skia.org for details):
blockSelectionGapsBounds: 0.00%
transform: 0.00%
reflectionInfo: 0.00%
enclosingPaginationLayer: 0.01%
subpixelAccumulation: 0.72%
groupdMapping: 0.43%
compositedLayerMapping: 1.52%
compositingReasons: 2.00%
offsetForInFlowPosition: 2.62%
potentialCompsotingReasonsFromStyle: 3.87% (which is likely to cover most
of the above)

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

BUG=581124
==========

to

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper (count)		36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

BUG=581124
==========

[Xianzhu, 2016-01-27 18:22:01]

Description was changed from

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper (count)		36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

BUG=581124
==========

to

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper			36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

This CL reduces sizeof(PaintLayer) from 344 bytes to 208 bytes on LP 64
systems.

BUG=581124
==========

[Xianzhu, 2016-01-27 19:47:06]

wangxianzhu@chromium.org changed reviewers:
+ chrishtr@chromium.org, pdr@chromium.org

[Xianzhu, 2016-01-27 19:47:07]

Cluster telemetry result (Run 486 http://ct.skia.org) shows that the patch
slightly improves record_time (by 0.39%~1.6%). It even improves rasterize_time
by 0.23%. Running another cycle to confirm that.

[chrishtr, 2016-01-27 23:55:57]

Please post rasterize_and_record_micro timings from Cluster Telemetry on this
patch, excited
to see the results. This looks awesome.

https://codereview.chromium.org/1636563003/diff/120001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/paint/PaintLayer.h (right):

https://codereview.chromium.org/1636563003/diff/120001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/paint/PaintLayer.h:119:
OwnPtr<CompositedLayerMapping> compositedLayerMapping;
Briefly document these while you are at it.

https://codereview.chromium.org/1636563003/diff/120001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/paint/PaintLayer.h:230: PaintLayerType
layerType() const { return static_cast<PaintLayerType>(m_layerType); }
If this is just used in the ASSERT on line 231, just inline the static cast.

[Xianzhu, 2016-01-28 02:54:24]

https://codereview.chromium.org/1636563003/diff/120001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/paint/PaintLayer.h (right):

https://codereview.chromium.org/1636563003/diff/120001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/paint/PaintLayer.h:119:
OwnPtr<CompositedLayerMapping> compositedLayerMapping;
On 2016/01/27 23:55:57, chrishtr wrote:
> Briefly document these while you are at it.

Done.

https://codereview.chromium.org/1636563003/diff/120001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/paint/PaintLayer.h:230: PaintLayerType
layerType() const { return static_cast<PaintLayerType>(m_layerType); }
On 2016/01/27 23:55:56, chrishtr wrote:
> If this is just used in the ASSERT on line 231, just inline the static cast.

Done.

[chrishtr, 2016-01-28 16:45:08]

Any update on cluster telemetry?

[Xianzhu, 2016-01-28 16:55:24]

On 2016/01/28 16:45:08, chrishtr wrote:
> Any update on cluster telemetry?

Repeated additional twice on cluster telemetry. There seem some noise in the
result, but in general the CL seems to gain performance.

First run (run 486):
record_time_sk_null_canvas (ms)	0	0	0%	0%
rasterize_time (ms)	34106.314	34028.836	-0.227%	-0.227%
record_time (ms)	2067.185	2059.231	-0.385%	-0.386%
record_time_subsequence_caching_disabled (ms)	6394.631	6337.994	-0.886%	-0.890%
record_time_caching_disabled (ms)	13671.997	13526.607	-1.063%	-1.069%
record_time_painting_disabled (ms)	7449.227	7347.241	-1.369%	-1.379%
record_time_construction_disabled (ms)	6614.360	6506.418	-1.632%	-1.645%

Second run (run 487):
record_time_subsequence_caching_disabled (ms)	5898.708	5974.873	1.291%	1.283%
record_time_painting_disabled (ms)	6936.174	6966.180	0.433%	0.432%
rasterize_time (ms)	32316.639	32338.954	0.069%	0.069%
record_time_sk_null_canvas (ms)	0	0	0%	0%
record_time (ms)	1983.525	1983.097	-0.022%	-0.022%
record_time_construction_disabled (ms)	6143.964	6142.149	-0.030%	-0.030%
record_time_caching_disabled (ms)	12625.646	12414.430	-1.673%	-1.687%

Third run (run 488):
record_time_construction_disabled (ms)	6290.177	6321.200	0.493%	0.492%
record_time (ms)	2042.495	2051.082	0.420%	0.420%
record_time_caching_disabled (ms)	12754.881	12784.633	0.233%	0.233%
record_time_sk_null_canvas (ms)	0	0	0%	0%
rasterize_time (ms)	32352.361	32333.042	-0.060%	-0.060%
record_time_painting_disabled (ms)	7138.374	7094.605	-0.613%	-0.615%
record_time_subsequence_caching_disabled (ms)	6191.890	6124.590	-1.087%	-1.093%

[chrishtr, 2016-01-28 18:04:29]

The CQ bit was checked by chrishtr@chromium.org

[chrishtr, 2016-01-28 18:04:29]

lgtm

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h
(right):

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:75:
//     the layer is squashed;
s/layer/PaintLayer/

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:76:
// - Otherwise the layer doesn't own or directly reference any
CompositedLayerMapping.
s/layer/PaintLayer/

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/paint/PaintLayer.cpp (right):

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/paint/PaintLayer.cpp:100: struct
SameSizeAsPaintLayer : DisplayItemClient {
Isn't there already one of these?

[commit-bot: I haz the power, 2016-01-28 18:05:28]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1636563003/160001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1636563003/160001

[chrishtr, 2016-01-28 18:15:56]

The CQ bit was unchecked by chrishtr@chromium.org

[chrishtr, 2016-01-28 18:16:05]

(removed commit since I had a couple of small comments)

[Xianzhu, 2016-01-28 18:25:19]

The CQ bit was checked by wangxianzhu@chromium.org

[Xianzhu, 2016-01-28 18:25:19]

The patchset sent to the CQ was uploaded after l-g-t-m from
chrishtr@chromium.org
Link to the patchset: https://codereview.chromium.org/1636563003/#ps180001
(title: " ")

[commit-bot: I haz the power, 2016-01-28 18:25:29]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1636563003/180001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1636563003/180001

[Xianzhu, 2016-01-28 18:30:30]

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h
(right):

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:75:
//     the layer is squashed;
On 2016/01/28 18:04:29, chrishtr wrote:
> s/layer/PaintLayer/

Done.

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:76:
// - Otherwise the layer doesn't own or directly reference any
CompositedLayerMapping.
On 2016/01/28 18:04:29, chrishtr wrote:
> s/layer/PaintLayer/

Done.

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/paint/PaintLayer.cpp (right):

https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/paint/PaintLayer.cpp:100: struct
SameSizeAsPaintLayer : DisplayItemClient {
On 2016/01/28 18:04:29, chrishtr wrote:
> Isn't there already one of these?

Here we have several ways count the vtable pointer caused by the inheritance:
1. inheritance like the above;
2. declare a virtual destructor here;
3. add a pointer here.
1 seems the most robust.

[chrishtr, 2016-01-28 18:33:52]

On 2016/01/28 at 18:30:30, wangxianzhu wrote:
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> File
third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h
(right):
> 
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:75:
//     the layer is squashed;
> On 2016/01/28 18:04:29, chrishtr wrote:
> > s/layer/PaintLayer/
> 
> Done.
> 
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:76:
// - Otherwise the layer doesn't own or directly reference any
CompositedLayerMapping.
> On 2016/01/28 18:04:29, chrishtr wrote:
> > s/layer/PaintLayer/
> 
> Done.
> 
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> File third_party/WebKit/Source/core/paint/PaintLayer.cpp (right):
> 
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> third_party/WebKit/Source/core/paint/PaintLayer.cpp:100: struct
SameSizeAsPaintLayer : DisplayItemClient {
> On 2016/01/28 18:04:29, chrishtr wrote:
> > Isn't there already one of these?
> 
> Here we have several ways count the vtable pointer caused by the inheritance:
> 1. inheritance like the above;
> 2. declare a virtual destructor here;
> 3. add a pointer here.
> 1 seems the most robust.

What I meant was that the diff looked weird w.r.t.
https://codereview.chromium.org/1616193002. I expected an edit to the existing
test, not a new
one with no sign of the other. Maybe just not synced?

[Xianzhu, 2016-01-28 18:44:53]

On 2016/01/28 18:33:52, chrishtr wrote:
> On 2016/01/28 at 18:30:30, wangxianzhu wrote:
> >
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> > File
> third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h
> (right):
> > 
> >
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> >
third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:75:
> //     the layer is squashed;
> > On 2016/01/28 18:04:29, chrishtr wrote:
> > > s/layer/PaintLayer/
> > 
> > Done.
> > 
> >
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> >
third_party/WebKit/Source/core/layout/compositing/CompositedLayerMapping.h:76:
> // - Otherwise the layer doesn't own or directly reference any
> CompositedLayerMapping.
> > On 2016/01/28 18:04:29, chrishtr wrote:
> > > s/layer/PaintLayer/
> > 
> > Done.
> > 
> >
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> > File third_party/WebKit/Source/core/paint/PaintLayer.cpp (right):
> > 
> >
>
https://codereview.chromium.org/1636563003/diff/160001/third_party/WebKit/Sou...
> > third_party/WebKit/Source/core/paint/PaintLayer.cpp:100: struct
> SameSizeAsPaintLayer : DisplayItemClient {
> > On 2016/01/28 18:04:29, chrishtr wrote:
> > > Isn't there already one of these?
> > 
> > Here we have several ways count the vtable pointer caused by the
inheritance:
> > 1. inheritance like the above;
> > 2. declare a virtual destructor here;
> > 3. add a pointer here.
> > 1 seems the most robust.
> 
> What I meant was that the diff looked weird w.r.t.
> https://codereview.chromium.org/1616193002. I expected an edit to the existing
> test, not a new
> one with no sign of the other. Maybe just not synced?

That one was reverted because it broke oilpan builds.

Later I noticed that the OwnPtrWillBecomePersist (for m_scrollableArea) actually
contains 2 pointers when oilpan is enabled. There is another WIP CL to lazily
create PaintLayer::m_scrollableArea.

[chrishtr, 2016-01-28 18:56:02]

lgtm

[commit-bot: I haz the power, 2016-01-28 19:36:34]

Description was changed from

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper			36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

This CL reduces sizeof(PaintLayer) from 344 bytes to 208 bytes on LP 64
systems.

BUG=581124
==========

to

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper			36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

This CL reduces sizeof(PaintLayer) from 344 bytes to 208 bytes on LP 64
systems.

BUG=581124
==========

[commit-bot: I haz the power, 2016-01-28 19:36:35]

Committed patchset #10 (id:180001)

[commit-bot: I haz the power, 2016-01-28 19:38:46]

Description was changed from

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper			36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

This CL reduces sizeof(PaintLayer) from 344 bytes to 208 bytes on LP 64
systems.

BUG=581124
==========

to

==========
Put rare PaintLayer fields into PaintLayerRareData

Rare rates of the fields, based on top 10k sites
(See run 475 in http://ct.skia.org for details):

stackingNode		100.00%
ancestorDependentCompositingInputs 82.17%
clipper			36.25%
staticBlockPosition	11.00%
staticInlinePosition	7.49%
(The following fields are put into PaintLayerRareData)
potentialCompositingReasonsFromStyle 7.13%
  (^this is likely to cover most of the following)
compositingReasons	5.77%
compositedLayerMapping	3.32%
subpixelAccumulation	2.86%
offsetForInFlowPosition	2.70%
grouppedMapping		2.35%
transform		2.02%
enclosingPaginationLayer 0.01%
reflectionInfo (count)	0.00%
blockSelectionGapsBounds 0.00%

AncestorDependentCompositingInputs fields:
clippedAbsoluteBoundingBox 69.73%
clippingContainer 	46.57%
(The following fields are put into RareAncestorDependentCompositingInputs)
transformAncestor 	6.14%
opacityAncestor 	4.73%
ancestorScrollingLayer  4.26%
scrollParent 		2.76%
nearestFixedPositionLayer 2.31%
filterAncestor 		0.02%

Rare rate of a field is defined as: number of PaintLayers with the field
ever assigned non-default value / total number of PaintLayers.

This CL reduces sizeof(PaintLayer) from 344 bytes to 208 bytes on LP 64
systems.

BUG=581124

Committed: https://crrev.com/299d5beba4af85652d96a2b2da1733ebcdec2c5e
Cr-Commit-Position: refs/heads/master@{#372141}
==========

[commit-bot: I haz the power, 2016-01-28 19:38:47]

Patchset 10 (id:??) landed as
https://crrev.com/299d5beba4af85652d96a2b2da1733ebcdec2c5e
Cr-Commit-Position: refs/heads/master@{#372141}

[haraken, 2016-01-29 08:29:04]

I'm not 100% sure but it looks like this CL caused leaks in the following tests:

fast/css/getComputedStyle/computed-style-with-zoom.html
cssom/cssvalue-comparison.html

https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Linux%20Leak/b...