Skip geometry mapping when source and destination spaces are equal

Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms

BUG=678597
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2

Review-Url: https://codereview.chromium.org/2621483002
Cr-Commit-Position: refs/heads/master@{#442186}
Committed: https://chromium.googlesource.com/chromium/src/+/4a496bb8248a82782841ecee92df37d3db00bb29

[pdr., 2017-01-08 04:29:46]

Description was changed from

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms
==========

to

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
==========

[pdr., 2017-01-08 04:29:58]

Description was changed from

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
==========

to

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms

CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
==========

[pdr., 2017-01-08 04:30:25]

The CQ bit was checked by pdr@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-01-08 04:30:36]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[pdr., 2017-01-08 05:15:54]

pdr@chromium.org changed reviewers:
+ chrishtr@chromium.org, xianzhanzhu@chromium.org

[commit-bot: I haz the power, 2017-01-08 05:51:48]

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

[commit-bot: I haz the power, 2017-01-08 05:51:48]

Dry run: This issue passed the CQ dry run.

[Xianzhu, 2017-01-08 23:54:09]

This CL isn't in my incoming review list. Found it when searching your another
CL. lgtm.

I also tried a similar change locally (just uploaded as
https://codereview.chromium.org/2619163002) for crbug.com/678597, but the result
was not as good as yours. On Linux the result was not stable. On Android, the
result was stabler, but not as good as expected (paint-offset-changes.html):
non-spinvalidation: 310ms, original spinvalidation: 410ms, spinvalidation with
the CL: 390ms.

Let's land this to see what will happen on bots.

[Xianzhu, 2017-01-09 00:00:13]

The CQ bit was checked by wangxianzhu@chromium.org

[commit-bot: I haz the power, 2017-01-09 00:00:28]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[Xianzhu, 2017-01-09 00:01:03]

Description was changed from

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms

CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
==========

to

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms

BUG=678597
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
==========

[commit-bot: I haz the power, 2017-01-09 00:03:38]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1483920013997360, "parent_rev":
"ac83e63737f7e2cfc0cc382a3c9f03bafeabf526", "commit_rev":
"4a496bb8248a82782841ecee92df37d3db00bb29"}

[commit-bot: I haz the power, 2017-01-09 00:04:02]

Description was changed from

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms

BUG=678597
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
==========

to

==========
Skip geometry mapping when source and destination spaces are equal

This patch skips geometry mapping when it would be a no-op which
avoids an expensive mapping through the identity matrix.

Performance test results:
appending-text: 7.5% improvement
paint-offset-changes: 8.2% improvement

Speeds up prePaint by ~14% on https://www.w3.org/TR/html5/single-page.html:
before: Last FrameView::prePaint is ~210ms
after: Last FrameView::prePaint is ~180ms

BUG=678597
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2

Review-Url: https://codereview.chromium.org/2621483002
Cr-Commit-Position: refs/heads/master@{#442186}
Committed:
https://chromium.googlesource.com/chromium/src/+/4a496bb8248a82782841ecee92df...
==========

[commit-bot: I haz the power, 2017-01-09 00:04:03]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/chromium/src/+/4a496bb8248a82782841ecee92df...

[pdr., 2017-01-09 17:29:12]

On 2017/01/08 at 23:54:09, wangxianzhu wrote:
> This CL isn't in my incoming review list. Found it when searching your another
CL. lgtm.
> 
> I also tried a similar change locally (just uploaded as
https://codereview.chromium.org/2619163002) for crbug.com/678597, but the result
was not as good as yours. On Linux the result was not stable. On Android, the
result was stabler, but not as good as expected (paint-offset-changes.html):
non-spinvalidation: 310ms, original spinvalidation: 410ms, spinvalidation with
the CL: 390ms.
> 
> Let's land this to see what will happen on bots.

I found this looking at https://www.w3.org/TR/html5/single-page.html and only
later found it helped address 678597, sorry about stepping on your bug. There's
still a lot of room to go though, because paint invalidation is still
significantly slower on https://www.w3.org/TR/html5/single-page.html with
spinvalidation. I'll file a bug so we can track this.

[pdr., 2017-01-09 17:29:58]

On 2017/01/09 at 17:29:12, pdr. wrote:
> On 2017/01/08 at 23:54:09, wangxianzhu wrote:
> > This CL isn't in my incoming review list. Found it when searching your
another CL. lgtm.
> > 
> > I also tried a similar change locally (just uploaded as
https://codereview.chromium.org/2619163002) for crbug.com/678597, but the result
was not as good as yours. On Linux the result was not stable. On Android, the
result was stabler, but not as good as expected (paint-offset-changes.html):
non-spinvalidation: 310ms, original spinvalidation: 410ms, spinvalidation with
the CL: 390ms.
> > 
> > Let's land this to see what will happen on bots.
> 
> I found this looking at https://www.w3.org/TR/html5/single-page.html and only
later found it helped address 678597, sorry about stepping on your bug. There's
still a lot of room to go though, because paint invalidation is still
significantly slower on https://www.w3.org/TR/html5/single-page.html with
spinvalidation. I'll file a bug so we can track this.

Oh, also, there are still some no-op maps in GeometryMapper that we could
remove. I just got the ones that I could prove helped paint invalidation on the
html5 spec.

[pdr., 2017-01-09 23:10:39]

As expected, this ended up as an 11.8% win on paint-offset-changes with
spinvalidation:
https://chromeperf.appspot.com/report?sid=1582db6af4d16ebc1512dc9fe06ccde60cc...