Oilpan: Introduce a dedicated heap for CSSValue and RenderObject

Oilpan: Introduce a dedicated heap for CSSValue and RenderObject

According to our object profiling in micro-benchmarks and real-world benchmarks (https://docs.google.com/a/google.com/spreadsheets/d/1EBYDfkvJ1aleIj0iSNmk0zEsq0AqVOsjEyI6XfOoLIU/edit#gid=562560234), the following objects are allocated very frequently.

- Node
- CSSValue
- RenderObject
- HTMLStackItem
- ElementRecord

Given the results, this CL introduces a dedicated heap for CSSValue and RenderObject. (Note: PartitionAlloc also has a dedicated heap for RenderObject.)

This improves various benchmarks a lot:

css/CSSPropertySetterGetter: +35%
css/ClassInvalidation: +32%
css/StyleSheetInsert: +21%
layout/floats-*: ~+20%
layout/large-table-with-*: ~+10%

BUG=420515
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=184362

[haraken, 2014-10-22 11:56:04]

haraken@chromium.org changed reviewers:
+ keishi@chromium.org, oilpan-reviews@chromium.org

[haraken, 2014-10-22 11:56:44]

PTAL

I want to introduce a dedicated heap for HTMLStackItem and ElementRecord as
well, but will do that after more performance study. CSSValue and RenderObject
look promising.

[sof, 2014-10-22 12:21:57]

sigbjornf@opera.com changed reviewers:
+ sigbjornf@opera.com

[sof, 2014-10-22 12:21:57]

(non-owner) lgtm

Given that these are common objects in general, it makes sense to properly
evaluate them being in their own heaps. (iow, not just chasing perf numbers on
artificial benchmarks here.)

[Mike West, 2014-10-22 12:25:37]

On 2014/10/22 12:21:57, sof wrote:
> (non-owner) lgtm
> 
> Given that these are common objects in general, it makes sense to properly
> evaluate them being in their own heaps. (iow, not just chasing perf numbers on
> artificial benchmarks here.)

rs platform/ LGTM.

It might make sense to add a Source/platform/heap/OWNERS file with oilpan folks
in it. I'd suggest that Sigbjorn, Mads, etc. should be able to stamp changes
like this.

[haraken, 2014-10-22 12:27:27]

The CQ bit was checked by haraken@chromium.org

[haraken, 2014-10-22 12:27:56]

On 2014/10/22 12:21:57, sof wrote:
> (non-owner) lgtm

You can upload a CL to add yourself to platform/heap/OWNERS, if you like :)

[commit-bot: I haz the power, 2014-10-22 12:28:14]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/667053009/1

[sof, 2014-10-22 13:42:14]

On 2014/10/22 12:27:56, haraken wrote:
> On 2014/10/22 12:21:57, sof wrote:
> > (non-owner) lgtm
> 
> You can upload a CL to add yourself to platform/heap/OWNERS, if you like :)

Thanks much, but I don't consider myself worthy yet :)

Will continue doing due dilligence on Oilpan "infra" reviews where I can, of
course.

[commit-bot: I haz the power, 2014-10-22 14:53:18]

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

[commit-bot: I haz the power, 2014-10-22 14:53:19]

Try jobs failed on following builders:
  win_blink_rel on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/win_blink_rel/builds/33078)

[haraken, 2014-10-23 04:29:37]

The CQ bit was checked by haraken@chromium.org

[commit-bot: I haz the power, 2014-10-23 04:30:16]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/667053009/1

[commit-bot: I haz the power, 2014-10-23 06:42:29]

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

[commit-bot: I haz the power, 2014-10-23 06:42:30]

Try jobs failed on following builders:
  win_blink_rel on tryserver.blink
(http://build.chromium.org/p/tryserver.blink/builders/win_blink_rel/builds/33203)

[sof, 2014-10-24 14:18:48]

The CQ bit was checked by sigbjornf@opera.com

[sof, 2014-10-24 14:18:49]

lgtm

Flakiness got in the way of this landing it seems.

[commit-bot: I haz the power, 2014-10-24 14:19:22]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/667053009/1

[commit-bot: I haz the power, 2014-10-24 15:05:15]

Committed patchset #1 (id:1) as 184362

[sof, 2014-11-13 08:23:42]

Hmm, I wonder if this is somehow related to the perf drop-off seen here,


https://chromeperf.appspot.com/report?masters=ChromiumPerf&bots=linux-oilpan-...

Investigating.