Use template parameter to reduce branching in style resolve apply passes

Use template parameter to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making an improvement of
1-5% on Android, 6-11% on Linux, 0-13% on Windows and 2-7% on Mac.

BUG=450466, 644148
Committed: https://crrev.com/851bdd47fcb66176959020a63bbdb3b4b090d845
Cr-Commit-Position: refs/heads/master@{#431122}

[alancutter (OOO until 2018), 2016-11-09 03:34:43]

Description was changed from

==========
Use template bool to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

BUG=450466,644148
==========

to

==========
Use template bool to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making a ~3% improvement on Android,


BUG=450466,644148
==========

[alancutter (OOO until 2018), 2016-11-09 03:49:18]

Description was changed from

==========
Use template bool to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making a ~3% improvement on Android,


BUG=450466,644148
==========

to

==========
Use template bool to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making a 1-5% improvement on Android,


BUG=450466,644148
==========

[alancutter (OOO until 2018), 2016-11-09 03:49:19]

alancutter@chromium.org changed reviewers:
+ timloh@chromium.org

[alancutter (OOO until 2018), 2016-11-09 03:57:16]

Description was changed from

==========
Use template bool to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making a 1-5% improvement on Android,


BUG=450466,644148
==========

to

==========
Use template bool to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making an improvement of
1-5% on Android, 6-11% on Linux, 0-13% on Windows and 2-7% on Mac.


BUG=450466,644148
==========

[alancutter (OOO until 2018), 2016-11-09 03:58:05]

Description was changed from

==========
Use template bool to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making an improvement of
1-5% on Android, 6-11% on Linux, 0-13% on Windows and 2-7% on Mac.


BUG=450466,644148
==========

to

==========
Use template parameter to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making an improvement of
1-5% on Android, 6-11% on Linux, 0-13% on Windows and 2-7% on Mac.


BUG=450466,644148
==========

[rune, 2016-11-09 08:24:14]

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

[rune, 2016-11-09 08:24:15]

lgtm

[alancutter (OOO until 2018), 2016-11-09 23:26:41]

The CQ bit was checked by alancutter@chromium.org

[commit-bot: I haz the power, 2016-11-09 23:27:13]

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, 2016-11-10 01:05:12]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-11-10 01:22:17]

Description was changed from

==========
Use template parameter to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making an improvement of
1-5% on Android, 6-11% on Linux, 0-13% on Windows and 2-7% on Mac.


BUG=450466,644148
==========

to

==========
Use template parameter to reduce branching in style resolve apply passes

The optimisation in https://crrev.com/890e3f57caf40adab7ea377cb50ef6591b6b882e
regressed performance on the waterfall perf bots. This patch attempts to
reduce the performance regression by moving one of the branches to
be compile time evaluated as a template parameter.

This optimisation is not being reverted as it is still necessary to
prevent a regression when adding an additional CSS property priority
that increases the number of apply passes we have to perform.

The perf try bots show this change making an improvement of
1-5% on Android, 6-11% on Linux, 0-13% on Windows and 2-7% on Mac.

BUG=450466,644148

Committed: https://crrev.com/851bdd47fcb66176959020a63bbdb3b4b090d845
Cr-Commit-Position: refs/heads/master@{#431122}
==========

[commit-bot: I haz the power, 2016-11-10 01:22:18]

Patchset 2 (id:??) landed as
https://crrev.com/851bdd47fcb66176959020a63bbdb3b4b090d845
Cr-Commit-Position: refs/heads/master@{#431122}