Pass references to BinComparator.

Pass references to BinComparator.

By passing references to scoped_refptr objects to BinComparator(), we
avoid copying the objects and the consequent reference count churn.
Should make sorting 25% faster.

BUG=268441
R=vmpstr@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=215927

[tomhudson, 2013-08-06 14:41:10]

I think I'm seeing ~10% speedup in cc_perftests' manage_tiles on N7 with this
change Vlad suggested.

Before:

*RESULT manage_tiles_100: 10320.12 runs/s
*RESULT manage_tiles_1000: 762.64 runs/s
*RESULT manage_tiles_10000: 44.95 runs/s

After:

*RESULT manage_tiles_100: 12061.69 runs/s
*RESULT manage_tiles_1000: 916.55 runs/s
*RESULT manage_tiles_10000: 51.48 runs/s

[danakj, 2013-08-06 15:21:10]

On Tue, Aug 6, 2013 at 10:41 AM, <tomhudson@google.com> wrote:

> I think I'm seeing ~10% speedup in cc_perftests' manage_tiles on N7 with
> this
> change Vlad suggested.
>

Wow, maybe we need a presubmit rule to enforce passing scoped refptr by
const& instead of by value.


>
> Before:
>
> *RESULT manage_tiles_100: 10320.12 runs/s
> *RESULT manage_tiles_1000: 762.64 runs/s
> *RESULT manage_tiles_10000: 44.95 runs/s
>
> After:
>
> *RESULT manage_tiles_100: 12061.69 runs/s
> *RESULT manage_tiles_1000: 916.55 runs/s
> *RESULT manage_tiles_10000: 51.48 runs/s
>
>
>
>
https://codereview.chromium.**org/22314011/<https://codereview.chromium.org/2...
>

[danakj, 2013-08-06 15:21:21]

LGTM

[vmpstr, 2013-08-06 15:34:17]

On 2013/08/06 15:21:21, danakj wrote:
> LGTM

lgtm2

[commit-bot: I haz the power, 2013-08-06 15:34:56]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/tomhudson@chromium.org/22314011/1

[commit-bot: I haz the power, 2013-08-06 17:51:42]

Change committed as 215927

[tomhudson, 2013-08-08 11:02:56]

On 2013/08/06 15:21:10, danakj wrote:
> On Tue, Aug 6, 2013 at 10:41 AM, <mailto:tomhudson@google.com> wrote:
> 
> Wow, maybe we need a presubmit rule to enforce passing scoped refptr by
> const& instead of by value.

I haven't found an easy way to audit it, but although cc does a fair amount of
passing scoped refptrs by value I think I'm seeing them at a rate of few/page or
few/second instead of many/frame.