Revert shouldInvalidateNodeListCaches() back to template-unrolled version.

Revert shouldInvalidateNodeListCaches() back to template-unrolled version.

Switch Document::shouldInvalidateNodeListCaches() back to the previous
version of unrolling a loop by way of a recursive template expansion.
A mild slowdown was seen in crbug.com/375810 when using the simpler
loop-based version from http://codereview.chromium.org/280123002/

R=haraken@chromium.org
BUG=375810
NOTRY=true

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=174702

[sof, 2014-05-23 13:51:20]

Please take a look.

Addressing, not sure if fully, a mild non-Oilpan regression on some DOM perf
tests that hit this lookup function hard.

[haraken, 2014-05-23 13:53:39]

LGTM.

Shall we add a FIXME in the code why we need the separate code between oilpan
and non-oilpan?

[sof, 2014-05-23 13:55:58]

On 2014/05/23 13:53:39, haraken wrote:
> LGTM.
> 
> Shall we add a FIXME in the code why we need the separate code between oilpan
> and non-oilpan?

We certainly could, but already have a related comment in Document.h next to
m_nodeLists/m_nodeListCounts; sufficient?

[haraken, 2014-05-23 14:45:04]

On 2014/05/23 13:55:58, sof wrote:
> On 2014/05/23 13:53:39, haraken wrote:
> > LGTM.
> > 
> > Shall we add a FIXME in the code why we need the separate code between
oilpan
> > and non-oilpan?
> 
> We certainly could, but already have a related comment in Document.h next to
> m_nodeLists/m_nodeListCounts; sufficient?

Looks good :)

[sof, 2014-05-23 14:50:59]

On 2014/05/23 14:45:04, haraken wrote:
> On 2014/05/23 13:55:58, sof wrote:
> > On 2014/05/23 13:53:39, haraken wrote:
> > > LGTM.
> > > 
> > > Shall we add a FIXME in the code why we need the separate code between
> oilpan
> > > and non-oilpan?
> > 
> > We certainly could, but already have a related comment in Document.h next to
> > m_nodeLists/m_nodeListCounts; sufficient?
> 
> Looks good :)

The duplication ain't pretty though :/

Sending this along now (i.e., will followup separately if esprehn has additional
comments/concerns.)

[sof, 2014-05-23 14:51:03]

The CQ bit was checked by sigbjornf@opera.com

[commit-bot: I haz the power, 2014-05-23 14:51:29]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/sigbjornf@opera.com/296133014/1

[commit-bot: I haz the power, 2014-05-23 16:08:05]

FYI, CQ is re-trying this CL (attempt #1).
Please consider checking whether the failures are real,
and report flakes to chrome-troopers@google.com.
The failing builders are:
  win_blink_rel on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/win_blink_rel/builds/8909)

[sof, 2014-05-23 16:59:11]

The CQ bit was unchecked by sigbjornf@opera.com

[sof, 2014-05-23 17:00:10]

The CQ bit was checked by sigbjornf@opera.com

[commit-bot: I haz the power, 2014-05-23 17:00:32]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/sigbjornf@opera.com/296133014/1

[commit-bot: I haz the power, 2014-05-23 17:01:02]

Change committed as 174702