Make Vector<WebCore::NodeEventContext> faster

Make Vector<WebCore::NodeEventContext> faster

The Intel VTune Amplifier profile detects 'NodeEventContext' destructor
(and its subsequent calls) to be the biggest hotspot when running
'EventsDispatchingInDeeplyNestedShadowTrees' performance test.

The main reason for it is inefficient Vector<WebCore::NodeEventContext>
(and the applied VectorMover in particular). This patch makes the
Vector<WebCore::NodeEventContext> more efficient via using
SimpleClassVectorTraits for NodeEventContext.

Results are the folowing (linux desktop x64, GCC 4.6.3,
release content_shell build with enabled 'linux_dump_symbols' flag)
before:
avg 6359.329553799463 runs/s
median 6412.644942400308 runs/s
stdev 220.99651799417947 runs/s
min 5429.032965593617 runs/s
max 6449.434918587015 runs/s

after:
avg 8059.328340764995 runs/s
median 8181.982047553853 runs/s
stdev 465.58159603795553 runs/s
min 6800.9409740083 runs/s
max 8483.844241958124 runs/s

The avg performance gain is around 28%

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=171814

[Mikhail, 2014-04-16 14:02:30]

PTAL

[eseidel, 2014-04-16 17:06:10]

Could you add a comment to explain the code?  I worry the next person seeing
this will not understand what it's for.

Maybe we should have a nicely named macro for this instead?

WTF_CAN_BE_MOVED_WITH_MEMCPY(EventNodeContext);

Is my understanding correct?

[Mikhail, 2014-04-16 20:40:46]

On 2014/04/16 17:06:10, eseidel wrote:
> Could you add a comment to explain the code?  I worry the next person seeing
> this will not understand what it's for.
> 
> Maybe we should have a nicely named macro for this instead?
> 
> WTF_CAN_BE_MOVED_WITH_MEMCPY(EventNodeContext);
> 
> Is my understanding correct?

I like the idea with macro, that could indeed improve code readability not only
in this concrete place but also in all the other places where
'SimpleClassVectorTraits' is used (that is why I think introducing macro is
better than just adding a comment). I would however introduce this macro with a
separate CL (since this one is solving a different issue), wdyt?

[eseidel, 2014-04-16 22:43:26]

lgtm

[commit-bot: I haz the power, 2014-04-16 22:43:44]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/mikhail.pozdnyakov@intel.com/239703005/1

[commit-bot: I haz the power, 2014-04-16 22:54:26]

Change committed as 171814