Remove display item scope

Remove display item scope

We used display item scope to distinguish display items of the same object and
display item type painted in multiple contexts (e.g. multiple layer fragments).
A scope is assigned to each context and is included in DisplayItem::Id to make
the ids of the display items unique. However, because scope ids are unstable
(that is we can't ensure the same scope ids for the same paintings) among
paintings so we had to disable display item caching for scoped display items,
so basically creating a scope has the same effect of skipping cache.

This CL removes display item scopes and replaces ScopeRecorder with
DisplayItemCacheSkipper, to avoid the complexity and cost of scopes.

Measured performance of the patch (https://ct.skia.org/chromium_perf_runs/
run 920 to 929, 6 on Linux, 2 on Android):
					AVG     MEDIAN
record_time 				-0.15%	-0.10%
record_time_caching_disabled		-0.46%	-0.60%
record_time_construction_disabled	-0.26%	-0.50%
record_time_painting_disabled		-0.31%	-0.52%
record_time_sk_null_canvas		 0.00%	 0.00%
record_time_subsequence_caching_dis..	-0.12%	-0.11%
viewport_picture_size (bytes)		-0.30%	-0.06%

Committed: https://crrev.com/84921bde050ed6d47525df7846065d80e59f3bf0
Cr-Commit-Position: refs/heads/master@{#396984}

[Xianzhu, 2016-05-31 16:35:45]

Description was changed from

==========
Remove display item scope
==========

to

==========
Remove display item scope

We used display item scope to distinguish display items of the same object and
display item type painted in multiple contexts (e.g. multiple layer fragments).
A scope is assigned to each context and is included in DisplayItem::Id to make
the ids of the display items unique. However, because scope ids are unstable
(that is we can't ensure the same scope ids for the same paintings) among
paintings so we had to disable display item caching for scoped display items,
so basically creating a scope has the same effect of skipping cache.

This CL removes display item scopes and replaces ScopeRecorder with
DisplayItemCacheSkipper, to avoid the complexity and cost of scopes.

Measured performance of the patch (https://ct.skia.org/chromium_perf_runs/
run 920 to 929, 6 on Linux, 2 on Android):
					AVG     MEDIAN
record_time 				-0.15%	-0.10%
record_time_caching_disabled		-0.46%	-0.60%
record_time_construction_disabled	-0.26%	-0.50%
record_time_painting_disabled		-0.31%	-0.52%
record_time_sk_null_canvas		 0.00%	 0.00%
record_time_subsequence_caching_dis..	-0.12%	-0.11%
viewport_picture_size (bytes)		-0.30%	-0.06%
==========

[Xianzhu, 2016-05-31 16:36:08]

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

[pdr., 2016-05-31 17:47:23]

On 2016/05/31 at 16:36:09, wangxianzhu wrote:
> 

Woohoo!

LGTM

[chrishtr, 2016-05-31 17:55:33]

https://codereview.chromium.org/2022563002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/platform/graphics/paint/README.md (right):

https://codereview.chromium.org/2022563002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/platform/graphics/paint/README.md:127: It is illegal
for there to be two drawings with the same ID in a display item
This CL makes it legal, so long as it is not cached.

[Xianzhu, 2016-05-31 20:35:12]

https://codereview.chromium.org/2022563002/diff/20001/third_party/WebKit/Sour...
File third_party/WebKit/Source/platform/graphics/paint/README.md (right):

https://codereview.chromium.org/2022563002/diff/20001/third_party/WebKit/Sour...
third_party/WebKit/Source/platform/graphics/paint/README.md:127: It is illegal
for there to be two drawings with the same ID in a display item
On 2016/05/31 17:55:33, chrishtr wrote:
> This CL makes it legal, so long as it is not cached.

Done.

[chrishtr, 2016-05-31 21:07:07]

lgtm

https://codereview.chromium.org/2022563002/diff/40001/third_party/WebKit/Sour...
File third_party/WebKit/Source/platform/graphics/paint/README.md (right):

https://codereview.chromium.org/2022563002/diff/40001/third_party/WebKit/Sour...
third_party/WebKit/Source/platform/graphics/paint/README.md:128: list, except
that the drawings are marked uncacheable
Nit: "except for drawings that are marked uncacheable"

[Xianzhu, 2016-05-31 21:42:30]

The CQ bit was checked by wangxianzhu@chromium.org

[Xianzhu, 2016-05-31 21:42:31]

The patchset sent to the CQ was uploaded after l-g-t-m from pdr@chromium.org,
chrishtr@chromium.org
Link to the patchset: https://codereview.chromium.org/2022563002/#ps60001
(title: " ")

[Xianzhu, 2016-05-31 21:42:39]

https://codereview.chromium.org/2022563002/diff/40001/third_party/WebKit/Sour...
File third_party/WebKit/Source/platform/graphics/paint/README.md (right):

https://codereview.chromium.org/2022563002/diff/40001/third_party/WebKit/Sour...
third_party/WebKit/Source/platform/graphics/paint/README.md:128: list, except
that the drawings are marked uncacheable
On 2016/05/31 21:07:07, chrishtr wrote:
> Nit: "except for drawings that are marked uncacheable"

Done.

[commit-bot: I haz the power, 2016-05-31 21:43:30]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2022563002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/2022563002/60001

[commit-bot: I haz the power, 2016-06-01 00:01:34]

Description was changed from

==========
Remove display item scope

We used display item scope to distinguish display items of the same object and
display item type painted in multiple contexts (e.g. multiple layer fragments).
A scope is assigned to each context and is included in DisplayItem::Id to make
the ids of the display items unique. However, because scope ids are unstable
(that is we can't ensure the same scope ids for the same paintings) among
paintings so we had to disable display item caching for scoped display items,
so basically creating a scope has the same effect of skipping cache.

This CL removes display item scopes and replaces ScopeRecorder with
DisplayItemCacheSkipper, to avoid the complexity and cost of scopes.

Measured performance of the patch (https://ct.skia.org/chromium_perf_runs/
run 920 to 929, 6 on Linux, 2 on Android):
					AVG     MEDIAN
record_time 				-0.15%	-0.10%
record_time_caching_disabled		-0.46%	-0.60%
record_time_construction_disabled	-0.26%	-0.50%
record_time_painting_disabled		-0.31%	-0.52%
record_time_sk_null_canvas		 0.00%	 0.00%
record_time_subsequence_caching_dis..	-0.12%	-0.11%
viewport_picture_size (bytes)		-0.30%	-0.06%
==========

to

==========
Remove display item scope

We used display item scope to distinguish display items of the same object and
display item type painted in multiple contexts (e.g. multiple layer fragments).
A scope is assigned to each context and is included in DisplayItem::Id to make
the ids of the display items unique. However, because scope ids are unstable
(that is we can't ensure the same scope ids for the same paintings) among
paintings so we had to disable display item caching for scoped display items,
so basically creating a scope has the same effect of skipping cache.

This CL removes display item scopes and replaces ScopeRecorder with
DisplayItemCacheSkipper, to avoid the complexity and cost of scopes.

Measured performance of the patch (https://ct.skia.org/chromium_perf_runs/
run 920 to 929, 6 on Linux, 2 on Android):
					AVG     MEDIAN
record_time 				-0.15%	-0.10%
record_time_caching_disabled		-0.46%	-0.60%
record_time_construction_disabled	-0.26%	-0.50%
record_time_painting_disabled		-0.31%	-0.52%
record_time_sk_null_canvas		 0.00%	 0.00%
record_time_subsequence_caching_dis..	-0.12%	-0.11%
viewport_picture_size (bytes)		-0.30%	-0.06%
==========

[commit-bot: I haz the power, 2016-06-01 00:01:35]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2016-06-01 00:03:32]

Description was changed from

==========
Remove display item scope

We used display item scope to distinguish display items of the same object and
display item type painted in multiple contexts (e.g. multiple layer fragments).
A scope is assigned to each context and is included in DisplayItem::Id to make
the ids of the display items unique. However, because scope ids are unstable
(that is we can't ensure the same scope ids for the same paintings) among
paintings so we had to disable display item caching for scoped display items,
so basically creating a scope has the same effect of skipping cache.

This CL removes display item scopes and replaces ScopeRecorder with
DisplayItemCacheSkipper, to avoid the complexity and cost of scopes.

Measured performance of the patch (https://ct.skia.org/chromium_perf_runs/
run 920 to 929, 6 on Linux, 2 on Android):
					AVG     MEDIAN
record_time 				-0.15%	-0.10%
record_time_caching_disabled		-0.46%	-0.60%
record_time_construction_disabled	-0.26%	-0.50%
record_time_painting_disabled		-0.31%	-0.52%
record_time_sk_null_canvas		 0.00%	 0.00%
record_time_subsequence_caching_dis..	-0.12%	-0.11%
viewport_picture_size (bytes)		-0.30%	-0.06%
==========

to

==========
Remove display item scope

We used display item scope to distinguish display items of the same object and
display item type painted in multiple contexts (e.g. multiple layer fragments).
A scope is assigned to each context and is included in DisplayItem::Id to make
the ids of the display items unique. However, because scope ids are unstable
(that is we can't ensure the same scope ids for the same paintings) among
paintings so we had to disable display item caching for scoped display items,
so basically creating a scope has the same effect of skipping cache.

This CL removes display item scopes and replaces ScopeRecorder with
DisplayItemCacheSkipper, to avoid the complexity and cost of scopes.

Measured performance of the patch (https://ct.skia.org/chromium_perf_runs/
run 920 to 929, 6 on Linux, 2 on Android):
					AVG     MEDIAN
record_time 				-0.15%	-0.10%
record_time_caching_disabled		-0.46%	-0.60%
record_time_construction_disabled	-0.26%	-0.50%
record_time_painting_disabled		-0.31%	-0.52%
record_time_sk_null_canvas		 0.00%	 0.00%
record_time_subsequence_caching_dis..	-0.12%	-0.11%
viewport_picture_size (bytes)		-0.30%	-0.06%

Committed: https://crrev.com/84921bde050ed6d47525df7846065d80e59f3bf0
Cr-Commit-Position: refs/heads/master@{#396984}
==========

[commit-bot: I haz the power, 2016-06-01 00:03:33]

Patchset 4 (id:??) landed as
https://crrev.com/84921bde050ed6d47525df7846065d80e59f3bf0
Cr-Commit-Position: refs/heads/master@{#396984}