Improve positioned/percentHeight descendant/container tracking

Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52% based on top 10k sites -- Run 402
   https://ct.skia.org/chromium_perf_runs/ )
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01-10_00-31-59
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01-10_11-52-52

More discussions: https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A

Committed: https://crrev.com/0ba5554bd01f1f3dd94f38375d02f234e683c9ca
Cr-Commit-Position: refs/heads/master@{#371388}

[Xianzhu, 2015-12-29 21:56:40]

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

[commit-bot: I haz the power, 2015-12-29 21:57:20]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1544423002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1544423002/60001

[commit-bot: I haz the power, 2015-12-29 23:06:12]

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

[commit-bot: I haz the power, 2015-12-29 23:06:12]

Dry run: This issue passed the CQ dry run.

[Xianzhu, 2016-01-04 19:45:35]

Description was changed from

==========
Remove positioned/percent-height descendant/container maps
==========

to

==========
Track positoned/percent-height descendant/container with raredata instead of
maps
==========

[Xianzhu, 2016-01-04 19:52:13]

Description was changed from

==========
Track positoned/percent-height descendant/container with raredata instead of
maps
==========

to

==========
Track positoned/percent-height descendant/container with raredata instead of
maps

This will improve layout performance. Blink_perf.layout tests show 1%~13% gain:

http://storage.googleapis.com/chromium-telemetry/html-results/results-2015-12...

http://storage.googleapis.com/chromium-telemetry/html-results/results-2015-12...

http://storage.googleapis.com/chromium-telemetry/html-results/results-2015-12...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

[Xianzhu, 2016-01-05 01:43:29]

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

[commit-bot: I haz the power, 2016-01-05 01:44:30]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1544423002/160001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1544423002/160001

[commit-bot: I haz the power, 2016-01-05 03:01:17]

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

[commit-bot: I haz the power, 2016-01-05 03:01:17]

Dry run: This issue passed the CQ dry run.

[Xianzhu, 2016-01-05 17:09:16]

wangxianzhu@chromium.org changed reviewers:
+ esprehn@chromium.org

[chrishtr, 2016-01-06 18:02:52]

chrishtr@chromium.org changed reviewers:
+ chrishtr@chromium.org, eae@chromium.org

[chrishtr, 2016-01-06 18:02:52]

Adding Emil for feedback and review also.

[esprehn, 2016-01-06 19:37:25]

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBox.h (right):

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBox.h:65: virtual
~LayoutBoxRareData() { }
we make this non-virtual in ElementRareData/NodeRareData and make it the
responsibility of the owner to destroy it at the correct type which saves 1 ptr
(the vtable) for every rare data.

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBox.h:911: m_rareData =
adoptPtr(createRareData());
adoptPtr should be at the creation site, not around a virtual call.

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBox.h:914: virtual
LayoutBoxRareData* createRareData() { return new LayoutBoxRareData; }
PassOwnPtr, you should never leak like this.

[Xianzhu, 2016-01-06 20:49:07]

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBox.h (right):

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBox.h:65: virtual
~LayoutBoxRareData() { }
On 2016/01/06 19:37:25, esprehn wrote:
> we make this non-virtual in ElementRareData/NodeRareData and make it the
> responsibility of the owner to destroy it at the correct type which saves 1
ptr
> (the vtable) for every rare data.

Done.

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBox.h:911: m_rareData =
adoptPtr(createRareData());
On 2016/01/06 19:37:25, esprehn wrote:
> adoptPtr should be at the creation site, not around a virtual call.

Done.

https://codereview.chromium.org/1544423002/diff/160001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBox.h:914: virtual
LayoutBoxRareData* createRareData() { return new LayoutBoxRareData; }
On 2016/01/06 19:37:25, esprehn wrote:
> PassOwnPtr, you should never leak like this.

Done.

[Xianzhu, 2016-01-10 07:30:30]

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

[commit-bot: I haz the power, 2016-01-10 07:30:33]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1544423002/220001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1544423002/220001

[commit-bot: I haz the power, 2016-01-10 07:58:04]

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

[commit-bot: I haz the power, 2016-01-10 07:58:05]

Dry run: Try jobs failed on following builders:
  win8_chromium_ng on tryserver.chromium.win (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.win/builders/win8_chromium_ng/...)

[Xianzhu, 2016-01-10 19:02:27]

Description was changed from

==========
Track positoned/percent-height descendant/container with raredata instead of
maps

This will improve layout performance. Blink_perf.layout tests show 1%~13% gain:

http://storage.googleapis.com/chromium-telemetry/html-results/results-2015-12...

http://storage.googleapis.com/chromium-telemetry/html-results/results-2015-12...

http://storage.googleapis.com/chromium-telemetry/html-results/results-2015-12...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

to

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52%)
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a box has only one positioned container at a time, simplify the positioned
  container map to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>, and modified code to handle
  previously multiple containers (unnecessarily) occurred.

This will improve blink_perf.layout tests performance by 0%~11%:

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

[Xianzhu, 2016-01-10 19:08:43]

Description was changed from

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52%)
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a box has only one positioned container at a time, simplify the positioned
  container map to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>, and modified code to handle
  previously multiple containers (unnecessarily) occurred.

This will improve blink_perf.layout tests performance by 0%~11%:

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

to

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52%)
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~11%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

[Xianzhu, 2016-01-10 20:53:03]

Description was changed from

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52%)
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~11%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

to

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52%)
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

[Xianzhu, 2016-01-10 20:54:22]

Modified according to the rare data stat result
(https://docs.google.com/spreadsheets/d/1i_kqo71EAiqyphgDFKwRThH7Oq69MWHncH-lS...).
See CL description for details.

Ptal.

[Xianzhu, 2016-01-19 20:26:03]

On 2016/01/10 20:54:22, Xianzhu wrote:
> Modified according to the rare data stat result
>
(https://docs.google.com/spreadsheets/d/1i_kqo71EAiqyphgDFKwRThH7Oq69MWHncH-lS...).
> See CL description for details.
> 
> Ptal.

Bump...

[eae, 2016-01-19 22:41:06]

LGTM
The current approach addresses many of the concerns of the original proposal
while maintaining much of the performance benefit. I like it! Thanks for doing
this.

Please allow Elliot another chance to comment before landing though as I'm not
sure if all his concerns have been addressed.

[esprehn, 2016-01-20 00:01:49]

You say:   (rare rate: 1.52%)

but over what data in your description?

I'd also prefer if you added separate methods instead of passing the maps as
arguments. Yeah it's a little bit of copy paste, but it makes understanding the
code a lot easier, and shows up better in the profiler.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBlock.cpp (right):

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:147: for (LayoutBox*
descendant : *descendants) {
you can ASSERT(!descendants->isEmpty()) before the loop?

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:154: for (LayoutBox*
descendant : *descendants) {
ditto?

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:1398: void
LayoutBlock::insertIntoTrackedDescendantsMap(LayoutBox* descendant,
TrackedDescendantsMap*& descendantsMap)
I'd rather you split this into two methods, insertFooObject() methods, and
didn't take the global map as an argument like this.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:1435: if
(!gPositionedContainerMap)
how did we get here if the map was null? Seems like something must be out of
order if that's possible.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBlock.h (right):

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.h:171: bool
hasPercentHeightDescendant(LayoutBox* o) const { return
hasPercentHeightDescendants() && percentHeightDescendants()->contains(o); }
this checks the bit twice, you want to call the Internal() one here.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.h:176:
ASSERT(static_cast<bool>(m_hasPercentHeightDescendants) ==
(percentHeightDescendantsInternal() &&
!percentHeightDescendantsInternal()->isEmpty()));
I would do this as:

(!m_hasPercentHeightDescendants || (percentHeightDescendantsInternal() &&
percentHeightDescendantsInternal()->isEmpty()))

[Xianzhu, 2016-01-20 00:10:09]

Description was changed from

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52%)
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

to

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52% based on top 10k sites -- Run 402
   https://ct.skia.org/chromium_perf_runs/ )
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

[Xianzhu, 2016-01-20 02:33:26]

On 2016/01/20 00:01:49, esprehn wrote:
> You say:   (rare rate: 1.52%)
> 
> but over what data in your description?

It was based on top 10k sites. Updated description.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBlock.cpp (right):

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:147: for (LayoutBox*
descendant : *descendants) {
On 2016/01/20 00:01:48, esprehn wrote:
> you can ASSERT(!descendants->isEmpty()) before the loop?

Done.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:154: for (LayoutBox*
descendant : *descendants) {
On 2016/01/20 00:01:48, esprehn wrote:
> ditto?

Done.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:1398: void
LayoutBlock::insertIntoTrackedDescendantsMap(LayoutBox* descendant,
TrackedDescendantsMap*& descendantsMap)
On 2016/01/20 00:01:48, esprehn wrote:
> I'd rather you split this into two methods, insertFooObject() methods, and
> didn't take the global map as an argument like this.

Inlined the function at call sites.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:1435: if
(!gPositionedContainerMap)
On 2016/01/20 00:01:48, esprehn wrote:
> how did we get here if the map was null? Seems like something must be out of
> order if that's possible.

This is called when a out-of-flow positioned LayoutBox is being destroyed.
gPositionedContainerMap may be null during layout tree construction when layout
state is dirty, e.g. when an image fails to load immediately and falls back to
alttext, or the layout tree is restructured to split inline flows.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBlock.h (right):

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.h:171: bool
hasPercentHeightDescendant(LayoutBox* o) const { return
hasPercentHeightDescendants() && percentHeightDescendants()->contains(o); }
On 2016/01/20 00:01:49, esprehn wrote:
> this checks the bit twice, you want to call the Internal() one here.

Done.

https://codereview.chromium.org/1544423002/diff/260001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.h:176:
ASSERT(static_cast<bool>(m_hasPercentHeightDescendants) ==
(percentHeightDescendantsInternal() &&
!percentHeightDescendantsInternal()->isEmpty()));
On 2016/01/20 00:01:48, esprehn wrote:
> I would do this as:
> 
> (!m_hasPercentHeightDescendants || (percentHeightDescendantsInternal() &&
> percentHeightDescendantsInternal()->isEmpty()))

This will miss check of another side: when m_hasPercentHeightDescendants is true
then percentHeightDescendantsInternal() should be null.

Changed to
ASSERT(m_hasPercentHeightDescendants ? percentHeightDescendantsInternal() &&
!percentHeightDescendantsInternal()->isEmpty() :
!percentHeightDescendantsInternal());

[Xianzhu, 2016-01-21 17:42:57]

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

[commit-bot: I haz the power, 2016-01-21 17:43:48]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1544423002/280001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1544423002/280001

[commit-bot: I haz the power, 2016-01-21 19:12:24]

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

[commit-bot: I haz the power, 2016-01-21 19:12:24]

Dry run: This issue passed the CQ dry run.

[Xianzhu, 2016-01-25 18:16:57]

esprehn@ do you have more comments?

[esprehn, 2016-01-25 21:45:04]

lgtm

https://codereview.chromium.org/1544423002/diff/280001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBlock.cpp (right):

https://codereview.chromium.org/1544423002/diff/280001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:1485: for (auto o :
deadObjects) {
object?

[Xianzhu, 2016-01-25 22:09:38]

The CQ bit was checked by wangxianzhu@chromium.org

[Xianzhu, 2016-01-25 22:09:38]

The patchset sent to the CQ was uploaded after l-g-t-m from eae@chromium.org,
esprehn@chromium.org
Link to the patchset: https://codereview.chromium.org/1544423002/#ps300001
(title: " ")

[Xianzhu, 2016-01-25 22:09:40]

https://codereview.chromium.org/1544423002/diff/280001/third_party/WebKit/Sou...
File third_party/WebKit/Source/core/layout/LayoutBlock.cpp (right):

https://codereview.chromium.org/1544423002/diff/280001/third_party/WebKit/Sou...
third_party/WebKit/Source/core/layout/LayoutBlock.cpp:1485: for (auto o :
deadObjects) {
On 2016/01/25 21:45:04, esprehn wrote:
> object?

Done.

[commit-bot: I haz the power, 2016-01-25 22:11:58]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1544423002/300001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1544423002/300001

[commit-bot: I haz the power, 2016-01-26 00:54:25]

Description was changed from

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52% based on top 10k sites -- Run 402
   https://ct.skia.org/chromium_perf_runs/ )
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

to

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52% based on top 10k sites -- Run 402
   https://ct.skia.org/chromium_perf_runs/ )
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

[commit-bot: I haz the power, 2016-01-26 00:54:26]

Committed patchset #16 (id:300001)

[commit-bot: I haz the power, 2016-01-26 00:57:39]

Description was changed from

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52% based on top 10k sites -- Run 402
   https://ct.skia.org/chromium_perf_runs/ )
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A
==========

to

==========
Improve positioned/percentHeight descendant/container tracking

- Add LayoutBoxRareData::m_percentHeightContainer to track percent height
  containers.
  (rare rate: 1.52% based on top 10k sites -- Run 402
   https://ct.skia.org/chromium_perf_runs/ )
- Still use global maps to track percent height descendants, positioned
  descendants and positioned containers, but add bit field to indicate if
  descendants set exists for a LayoutBlock to avoid hash lookups if the
  block doesn't have percent-height or positioned descendants.
- As a positioned box has only one positioned container at a time, positioned
  container map is simplified to HashMap<LayoutBox*, LayoutBlock*> instead of
  HashMap<LayoutBox*, HashSet<LayoutBlock*>>. Also modify code to handle
  previously (unnecessarily) multiple containers situation.

This will improve blink_perf.layout tests performance by 0%~20%:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-01...

More discussions:
https://groups.google.com/a/chromium.org/forum/#!topic/blink-dev/xWMmQfD_X0A

Committed: https://crrev.com/0ba5554bd01f1f3dd94f38375d02f234e683c9ca
Cr-Commit-Position: refs/heads/master@{#371388}
==========

[commit-bot: I haz the power, 2016-01-26 00:57:40]

Patchset 16 (id:??) landed as
https://crrev.com/0ba5554bd01f1f3dd94f38375d02f234e683c9ca
Cr-Commit-Position: refs/heads/master@{#371388}