Fixed perf regression by removing tree traversal for text sibling.

Fixed perf regression by removing tree traversal for text sibling.

Changed contents of HeapHashMap on Document to store a new struct
StyleRecalcData containing both the original ComputedStyle and
the next Text sibling.

Deleted Node::nextTextSibling method that traverses the tree to find
the next text sibling, replaced usage of this method with access to
the Text node from the StyleRecalcData.

Patch yields a 79.21% performance improvement to
blink_perf.dom:select-single-remove benchmark on mac-10-10 bot:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-11-06_17-33-01

BUG=659647
Committed: https://crrev.com/d1fc66d0e6f64e38bcff450b54f908b354eff2aa
Cr-Commit-Position: refs/heads/master@{#432324}

[Bugs Nash, 2016-11-06 22:55:08]

bugsnash@chromium.org changed reviewers:
+ nainar@chromium.org

[Bugs Nash, 2016-11-06 22:55:17]

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

[commit-bot: I haz the power, 2016-11-06 22:55:26]

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

[nainar, 2016-11-06 23:00:38]

LGTM! \o/

Thanks for getting rid of the regression!

[Bugs Nash, 2016-11-06 23:10:19]

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

[commit-bot: I haz the power, 2016-11-07 01:21:19]

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

[commit-bot: I haz the power, 2016-11-07 01:21:20]

Dry run: Try jobs failed on following builders:
  mac_chromium_rel_ng on master.tryserver.chromium.mac (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...)

[Bugs Nash, 2016-11-07 03:33:25]

Description was changed from

==========
Removed tree traversal to find next Text sibling, stored it instead.

Changed contents of HeapHashMap on Document to store a new struct
StyleRecalcData containing both the original ComputedStyle and
the next Text sibling.

Deleted Node::nextTextSibling method that traverses the tree to find
the next text sibling, replaced usage of this method with access to
the Text node from the StyleRecalcData.

BUG=659647
==========

to

==========
Fixed perf regression by removing tree traversal for text sibling.

Changed contents of HeapHashMap on Document to store a new struct
StyleRecalcData containing both the original ComputedStyle and
the next Text sibling.

Deleted Node::nextTextSibling method that traverses the tree to find
the next text sibling, replaced usage of this method with access to
the Text node from the StyleRecalcData.

BUG=659647
==========

[Bugs Nash, 2016-11-07 03:39:30]

Description was changed from

==========
Fixed perf regression by removing tree traversal for text sibling.

Changed contents of HeapHashMap on Document to store a new struct
StyleRecalcData containing both the original ComputedStyle and
the next Text sibling.

Deleted Node::nextTextSibling method that traverses the tree to find
the next text sibling, replaced usage of this method with access to
the Text node from the StyleRecalcData.

BUG=659647
==========

to

==========
Fixed perf regression by removing tree traversal for text sibling.

Changed contents of HeapHashMap on Document to store a new struct
StyleRecalcData containing both the original ComputedStyle and
the next Text sibling.

Deleted Node::nextTextSibling method that traverses the tree to find
the next text sibling, replaced usage of this method with access to
the Text node from the StyleRecalcData.

Patch yields a 79.21% performance improvement to
blink_perf.dom:select-single-remove benchmark on mac-10-10 bot:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-11...

BUG=659647
==========

[rune, 2016-11-07 09:43:19]

rune@opera.com changed reviewers:
+ rune@opera.com

[rune, 2016-11-07 09:43:19]

I realize the documentation I wrote in core/dom/WhitespaceLayoutObjects.md is
already getting out-of-date, but I don't think it makes sense to edit it now
since this code is not yet in its final state.

https://codereview.chromium.org/2476163002/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/dom/Document.h (right):

https://codereview.chromium.org/2476163002/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/dom/Document.h:255: struct StyleRecalcData :
public GarbageCollectedFinalized<StyleRecalcData> {
This is only used for reattachment, right? How about calling it
StyleReattachData or smth. I was worried we store these objects for all elements
being recalc'ed when I first saw it.

Could we perhaps put this class in a separate include file and just forward
declare it in Document.h?

[esprehn, 2016-11-07 22:26:26]

https://codereview.chromium.org/2476163002/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/dom/Document.h (right):

https://codereview.chromium.org/2476163002/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/dom/Document.h:1465: HeapHashMap<Member<Element>,
Member<StyleRecalcData>> m_styleRecalcDataMap;
HeapHashMap<Member<Element>, StyleRecalcData>

that way we avoid tons of heap allocations. That produces some waste inside the
map, but it's okay since this is just temporary inside updateStyle.

[Bugs Nash, 2016-11-14 23:25:53]

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

[commit-bot: I haz the power, 2016-11-14 23:27:19]

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

[Bugs Nash, 2016-11-14 23:31:06]

Have addressed reviews. Rune note that I needed to include the new
StyleRecalcData file instead of forward declaring it because the complete type
is required for using it in the HashMap.

[commit-bot: I haz the power, 2016-11-15 04:14:52]

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

[commit-bot: I haz the power, 2016-11-15 04:14:53]

Dry run: Try jobs failed on following builders:
  linux_chromium_rel_ng on master.tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[rune, 2016-11-15 09:48:35]

lgtm

[Bugs Nash, 2016-11-15 22:19:24]

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

[commit-bot: I haz the power, 2016-11-15 22:19:48]

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, 2016-11-15 22:46:53]

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

[commit-bot: I haz the power, 2016-11-15 22:46:54]

Dry run: Try jobs failed on following builders:
  cast_shell_linux on master.tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/cast_shell_linu...)

[Bugs Nash, 2016-11-15 23:28:37]

The CQ bit was checked by bugsnash@chromium.org

[Bugs Nash, 2016-11-15 23:28:37]

The patchset sent to the CQ was uploaded after l-g-t-m from nainar@chromium.org,
rune@opera.com
Link to the patchset: https://codereview.chromium.org/2476163002/#ps40001
(title: "Rebased")

[commit-bot: I haz the power, 2016-11-15 23:29:20]

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, 2016-11-16 00:45:52]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-11-16 00:48:41]

Description was changed from

==========
Fixed perf regression by removing tree traversal for text sibling.

Changed contents of HeapHashMap on Document to store a new struct
StyleRecalcData containing both the original ComputedStyle and
the next Text sibling.

Deleted Node::nextTextSibling method that traverses the tree to find
the next text sibling, replaced usage of this method with access to
the Text node from the StyleRecalcData.

Patch yields a 79.21% performance improvement to
blink_perf.dom:select-single-remove benchmark on mac-10-10 bot:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-11...

BUG=659647
==========

to

==========
Fixed perf regression by removing tree traversal for text sibling.

Changed contents of HeapHashMap on Document to store a new struct
StyleRecalcData containing both the original ComputedStyle and
the next Text sibling.

Deleted Node::nextTextSibling method that traverses the tree to find
the next text sibling, replaced usage of this method with access to
the Text node from the StyleRecalcData.

Patch yields a 79.21% performance improvement to
blink_perf.dom:select-single-remove benchmark on mac-10-10 bot:
http://storage.googleapis.com/chromium-telemetry/html-results/results-2016-11...

BUG=659647

Committed: https://crrev.com/d1fc66d0e6f64e38bcff450b54f908b354eff2aa
Cr-Commit-Position: refs/heads/master@{#432324}
==========

[commit-bot: I haz the power, 2016-11-16 00:48:42]

Patchset 3 (id:??) landed as
https://crrev.com/d1fc66d0e6f64e38bcff450b54f908b354eff2aa
Cr-Commit-Position: refs/heads/master@{#432324}