Add a fast path to CSSParserFastPaths for CSS transform rotate.

Add a fast path to CSSParserFastPaths for CSS transform rotate.

This cuts the script execution time of the Animometer > Leaves benchmark
by ~40%. Previously we were spending 10% of the time in the fast path
for transforms only to throw the result away entirely and then reparse
the value with the regular CSS parser which is much slower, and also
wasted the work done in the fast path. Adding this fast path makes sure
we never use the regular CSSParser making the style.tranform call much
faster.

This highlights that the fast path is actually a performance regression
whenever you fall out of it, and also that the CSSParser itself has room
for improvement since invoking it is 3x slower than using the fast path.

[esprehn, 2016-04-20 03:49:12]

Description was changed from

==========
Add a fast path to CSSParserFastPaths for CSS transform rotate.

This cuts the script execution time of the Animometer > Leaves benchmark
by ~40%. Previously we were spending 10% of the time in the fast path
for transforms only to throw the result away entirely and then reparse
the value with the regular CSS parser which is much slower, and also
wasted the work done in the fast path. Adding this fast path makes sure
we never use the regular CSSParser making the style.tranform call much
faster.

This highlights that the fast path is actually a performance regression
whenever you fall out of it, and also that the CSSParser itself has room
for improvement since invoking it is 3x slower than using the fast path.
==========

to

==========
Add a fast path to CSSParserFastPaths for CSS transform rotate.

This cuts the script execution time of the Animometer > Leaves benchmark
by ~40%. Previously we were spending 10% of the time in the fast path
for transforms only to throw the result away entirely and then reparse
the value with the regular CSS parser which is much slower, and also
wasted the work done in the fast path. Adding this fast path makes sure
we never use the regular CSSParser making the style.tranform call much
faster.

This highlights that the fast path is actually a performance regression
whenever you fall out of it, and also that the CSSParser itself has room
for improvement since invoking it is 3x slower than using the fast path.

Comparing to Safari this appears to make calls to 
AbstractPropertySetCSSStyleDeclaration::setPropertyInternal nearly 2x
faster for transforms including rotations.
==========

[esprehn, 2016-04-20 03:53:00]

Description was changed from

==========
Add a fast path to CSSParserFastPaths for CSS transform rotate.

This cuts the script execution time of the Animometer > Leaves benchmark
by ~40%. Previously we were spending 10% of the time in the fast path
for transforms only to throw the result away entirely and then reparse
the value with the regular CSS parser which is much slower, and also
wasted the work done in the fast path. Adding this fast path makes sure
we never use the regular CSSParser making the style.tranform call much
faster.

This highlights that the fast path is actually a performance regression
whenever you fall out of it, and also that the CSSParser itself has room
for improvement since invoking it is 3x slower than using the fast path.

Comparing to Safari this appears to make calls to 
AbstractPropertySetCSSStyleDeclaration::setPropertyInternal nearly 2x
faster for transforms including rotations.
==========

to

==========
Add a fast path to CSSParserFastPaths for CSS transform rotate.

This cuts the script execution time of the Animometer > Leaves benchmark
by ~40%. Previously we were spending 10% of the time in the fast path
for transforms only to throw the result away entirely and then reparse
the value with the regular CSS parser which is much slower, and also
wasted the work done in the fast path. Adding this fast path makes sure
we never use the regular CSSParser making the style.tranform call much
faster.

This highlights that the fast path is actually a performance regression
whenever you fall out of it, and also that the CSSParser itself has room
for improvement since invoking it is 3x slower than using the fast path.
==========

[esprehn, 2016-04-20 03:53:05]

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

[commit-bot: I haz the power, 2016-04-20 03:53:13]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1904633002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1904633002/1

[commit-bot: I haz the power, 2016-04-20 05:11:57]

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

[commit-bot: I haz the power, 2016-04-20 05:11:57]

Dry run: This issue passed the CQ dry run.

[Rick Byers, 2016-05-26 04:33:40]

Elliott, any plan to try to land this?  Or do you think we should just be
focusing on http://crbug.com/605792?  I'm also seeing significant CPU time
parsing transforms in the Animometer benchmark I'm looking at
(http://crbug.com/606680).  I'll try applying this patch and see how much win I
get.

[esprehn, 2016-05-26 04:39:21]

On 2016/05/26 at 04:33:40, rbyers wrote:
> Elliott, any plan to try to land this?  Or do you think we should just be
focusing on http://crbug.com/605792?  I'm also seeing significant CPU time
parsing transforms in the Animometer benchmark I'm looking at
(http://crbug.com/606680).  I'll try applying this patch and see how much win I
get.

If we can figure out what I messed up in
https://codereview.chromium.org/1996343003 and reland it and then timloh@ fixes
the inline capacity on the Vectors I'm curious how far off we are from landing
this hack.