Optimize MutableStylePropertySet::findPropertyIndex

Optimize MutableStylePropertySet::findPropertyIndex

The Intel Vtune Amplifier showed 'MutableStylePropertySet::findPropertyIndex' as one of the biggest hotspot by CPU usage viewpoint for CSS performance tests.
This patch makes it more efficient by using std::find_if instead of previously used linear search, the test results improved as following:

1) CSSPropertySetterGetterMethods.html - approx. 11% better
before
avg 8209.627773541746 runs/s
median 8219.159152799137 runs/s
stdev 95.69280121453318 runs/s
min 7950.488333886082 runs/s
max 8334.299157035108 runs/s

after
avg 9095.94802572367 runs/s
median 9098.099366883693 runs/s
stdev 60.291500266466144 runs/s
min 8946.226065032974 runs/s
max 9180.054245827989 runs/s

2) CSSPropertyUpdateValue.html - approx. 4% better
before:
avg 14598.817565565343 runs/s
median 14620.181782154228 runs/s
stdev 75.59180598169823 runs/s
min 14406.699115207599 runs/s
max 14688.987467305702 runs/s

after:
avg 15207.605281115579 runs/s
median 15217.061385777779 runs/s
stdev 190.52047714899481 runs/s
min 14810.510319831952 runs/s
max 15447.612680018468 runs/s

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=196923

[Mikhail, 2015-06-10 11:44:21]

mikhail.pozdnyakov@intel.com changed reviewers:
+ mkwst@chromium.org, sigbjorn@opera.com

[Mikhail, 2015-06-10 11:45:43]

PTAL

[sof, 2015-06-10 15:08:39]

sigbjornf@opera.com changed reviewers:
+ rune@opera.com
- sigbjorn@opera.com

[rune, 2015-06-11 07:26:50]

Do you have more insight into why find_if makes it faster?

[Mike West, 2015-06-11 07:46:17]

On 2015/06/11 at 07:26:50, rune wrote:
> Do you have more insight into why find_if makes it faster?

LGTM, as the performance numbers you provide look compelling enough to justify
the added complexity. I'm also curious about the source of the speedup, as I
understood `find_if` to be little more than a wrapper around a `for` loop. What
is it doing under the hood that's so much more efficient?

[Mikhail, 2015-06-11 07:48:07]

On 2015/06/11 07:26:50, rune wrote:
> Do you have more insight into why find_if makes it faster?

std::find_if is using more efficient algorithm comparing with simple search

  template<typename _RandomAccessIterator, typename _Tp>
    _RandomAccessIterator
    __find(_RandomAccessIterator __first, _RandomAccessIterator __last,
	   const _Tp& __val, random_access_iterator_tag)
    {
      typename iterator_traits<_RandomAccessIterator>::difference_type
	__trip_count = (__last - __first) >> 2;

      for (; __trip_count > 0; --__trip_count)
	{
	  if (*__first == __val)
	    return __first;
	  ++__first;

	  if (*__first == __val)
	    return __first;
	  ++__first;

	  if (*__first == __val)
	    return __first;
	  ++__first;

	  if (*__first == __val)
	    return __first;
	  ++__first;
	}

      switch (__last - __first)
	{
	case 3:
	  if (*__first == __val)
	    return __first;
	  ++__first;
	case 2:
	  if (*__first == __val)
	    return __first;
	  ++__first;
	case 1:
	  if (*__first == __val)
	    return __first;
	  ++__first;
	case 0:
	default:
	  return __last;
	}
    }

[Mikhail, 2015-06-11 07:57:32]

The CQ bit was checked by mikhail.pozdnyakov@intel.com

[commit-bot: I haz the power, 2015-06-11 07:57:38]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1167383004/20001

[commit-bot: I haz the power, 2015-06-11 08:02:56]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=196923

[Timothy Loh, 2015-06-11 08:31:41]

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

[Timothy Loh, 2015-06-11 08:31:42]

Why don't we just use a lambda?

[Mikhail, 2015-06-11 09:32:59]

On 2015/06/11 08:31:42, Timothy Loh wrote:
> Why don't we just use a lambda?

thanks! lamba would be better indeed