I talked to Chris and we can try a targeted 2d optimization in GeometryMapper::SourceToDestinationProjection. I'll look into this.

This benchmark has 5000 layers and does 1000 hit tests per frame, so the code is largely a microbenchmark of PaintLayerClipper::CalculateRectsWithGeometryMapper (via PaintLayer::HitTest). There's a !rls fast-path that avoids GeometryMapper::SourceToDestinationProjection and removing that drops the regression from 19% to 15%. If we make an RLS-compatible fastpath, the regression only drops to 9.5% because we still need to do a little extra work for RLS (a few comparisons and a few LayoutRect moves). I don't think it makes sense to microoptimize this microbenchmark to avoid GeometryMapper::SourceToDestinationProjection unless we can see it on real (or more representative) content.

I'm going to look at the remaining 4% regression that isn't explained by calling GeometryMapper::SourceToDestinationProjection and see if there's anything there that we might be able to fix.

I found that the linux-rel bot is not updating. Lets use the following graphs for this bug:
https://chromeperf.appspot.com/report?sid=69cbbe7d2e2ade46ee2f3e8ad999728923259e32282d2d4f90769c297a12e7f5&start_rev=524947&end_rev=541841

"Check LayoutBox hit test clip optimization before hit testing any phase" / https://chromium-review.googlesource.com/c/chromium/src/+/953378 / git position #541659 just landed which should improve this graph.

This is looking pretty good on the graphs. I'd like to let it bake for another few days to ensure the perf results are good, then merge to M66.

I have another patch ready to go (https://chromium-review.googlesource.com/c/chromium/src/+/957363) with similar results. With these two simple patches together, the hit testing regression should be completely gone (and more).

The upcoming patch turned out to be too complex to merge. I'd like to merge just this performance patch: https://crbug.com/dd5e972c015c14842c39eaae7b76a74e59b9416a

Please add affected OSs.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb

commit b3e06524f7eb66c456c7b63fdf6df5d98fe59deb
Author: Philip Rogers <pdr@chromium.org>
Date: Mon Mar 12 17:32:02 2018

HitTestLocation should store a LayoutRect instead of an IntRect

A large amount of time hit testing is spent converting LayoutRects
to IntRects for the purposes of intersecting. This patch updates
HitTestLocation's bounding_box_ rect to be a LayoutRect which is
consistent with the LayoutPoint point and reduces conversions.

With this patch, blink_perf.events.hit-test-lots-of-layers improves
by between 2% and 5%.

Bug:  818772 
Cq-Include-Trybots: master.tryserver.blink:linux_trusty_blink_rel;master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
Change-Id: Ic3bc3b846b89e392b7b9f29a6affb42a5be81152
Reviewed-on: https://chromium-review.googlesource.com/957363
Reviewed-by: Chris Harrelson <chrishtr@chromium.org>
Commit-Queue: Philip Rogers <pdr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#542529}
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/exported/WebPluginContainerImpl.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/exported/WebPluginContainerTest.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/input/EventHandler.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/input/EventHandler.h
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/input/PointerEventManager.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/inspector/InspectorCSSAgent.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/layout/HitTestLocation.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/layout/HitTestLocation.h
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/layout/HitTestResult.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/layout/HitTestResult.h
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/layout/LayoutInlineTest.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/layout/line/EllipsisBox.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/layout/line/LineBoxList.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/page/TouchDisambiguation.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/paint/PaintLayer.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/testing/Internals.cpp
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/core/testing/data/plugin_container.html
[modify] https://crrev.com/b3e06524f7eb66c456c7b63fdf6df5d98fe59deb/third_party/WebKit/Source/platform/geometry/LayoutRectOutsets.h

Note that this merge request is only for https://crbug.com/dd5e972c015c14842c39eaae7b76a74e59b9416a. I don't want to merge the patch that just landed in comment #8.

Your change meets the bar and is auto-approved for M66. Please go ahead and merge the CL to branch 3359 manually. Please contact milestone owner if you have questions.
Owners: cmasso@(Android), cmasso@(iOS), josafat@(ChromeOS), abdulsyed@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Please merge this change as soon as possible

https://crrev.com/dd5e972c015c14842c39eaae7b76a74e59b9416a has been merged to M66:
https://chromium-review.googlesource.com/c/chromium/src/+/961108

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/debbb8554587f0e15b88484d94917ff76025c706

commit debbb8554587f0e15b88484d94917ff76025c706
Author: Philip Rogers <pdr@chromium.org>
Date: Wed Mar 14 00:12:49 2018

[CI] Add LayoutRect::IntersectsInclusively for edge-inclusive intersection

This patch adds a bool check version of LayoutRect::InclusiveIntersect.

Bug:  818772 
Change-Id: I0c844628c6b79359e5b88c2bf54b4a6600479d46
Reviewed-on: https://chromium-review.googlesource.com/958411
Reviewed-by: Stefan Zager <szager@chromium.org>
Commit-Queue: Philip Rogers <pdr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#542958}
[modify] https://crrev.com/debbb8554587f0e15b88484d94917ff76025c706/third_party/WebKit/Source/platform/geometry/LayoutRect.cpp
[modify] https://crrev.com/debbb8554587f0e15b88484d94917ff76025c706/third_party/WebKit/Source/platform/geometry/LayoutRect.h
[modify] https://crrev.com/debbb8554587f0e15b88484d94917ff76025c706/third_party/WebKit/Source/platform/geometry/LayoutRectTest.cpp

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/0954e0672f34d88424ead8763c12ddfd66ce4a26

commit 0954e0672f34d88424ead8763c12ddfd66ce4a26
Author: Philip Rogers <pdr@chromium.org>
Date: Wed Mar 14 00:21:25 2018

[CI] Document FloatQuad and FloatRoundedRect inclusive intersections

The following intersection functions are edge-inclusive:
FloatQuad::IntersectsRect
FloatQuad::IntersectsCircle
FloatQuad::IntersectsEllipse
FloatRoundedRect::IntersectsQuad

This is different from FloatRect::Intersects and LayoutRect::Intersects
which are edge-exclusive. This patch adds a comment above each of these
functions specifying that they are edge-inclusive, and adds tests.

Bug:  818772 
Change-Id: I7c1212a08e5e8874fb9679c5a676f91a97d3fce7
Reviewed-on: https://chromium-review.googlesource.com/960841
Reviewed-by: Stefan Zager <szager@chromium.org>
Commit-Queue: Philip Rogers <pdr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#542961}
[modify] https://crrev.com/0954e0672f34d88424ead8763c12ddfd66ce4a26/third_party/WebKit/Source/platform/geometry/FloatQuad.cpp
[modify] https://crrev.com/0954e0672f34d88424ead8763c12ddfd66ce4a26/third_party/WebKit/Source/platform/geometry/FloatQuad.h
[modify] https://crrev.com/0954e0672f34d88424ead8763c12ddfd66ce4a26/third_party/WebKit/Source/platform/geometry/FloatQuadTest.cpp
[modify] https://crrev.com/0954e0672f34d88424ead8763c12ddfd66ce4a26/third_party/WebKit/Source/platform/geometry/FloatRoundedRect.h
[modify] https://crrev.com/0954e0672f34d88424ead8763c12ddfd66ce4a26/third_party/WebKit/Source/platform/geometry/FloatRoundedRectTest.cpp

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/870c262728c7c2c5423d553a83de9a239634ee86

commit 870c262728c7c2c5423d553a83de9a239634ee86
Author: Philip Rogers <pdr@chromium.org>
Date: Fri Mar 16 20:17:04 2018

[CI] Inline FloatSize::IsZero and FloatRoundedRect::IsZero

blink_perf.events.hit-test-lots-of-layers is a microbenchmark of
hit PaintLayer::HitTestLayer which uses GeometryMapper. One of the
hottest functions is GeometryMapper::LocalToClipRectInternal which
creates a lot of FloatClipRects that check FloatRoundedRect::IsZero
in their constructor. By inlining the IsZero functions, we get a
3%-4% improvement on the entire benchmark.

Bug:  818772 
Change-Id: I0fb4c02152b799399aa19d521ba91ed46e326c3c
Reviewed-on: https://chromium-review.googlesource.com/966797
Reviewed-by: Chris Harrelson <chrishtr@chromium.org>
Commit-Queue: Philip Rogers <pdr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#543802}
[modify] https://crrev.com/870c262728c7c2c5423d553a83de9a239634ee86/third_party/WebKit/Source/platform/geometry/FloatRoundedRect.cpp
[modify] https://crrev.com/870c262728c7c2c5423d553a83de9a239634ee86/third_party/WebKit/Source/platform/geometry/FloatRoundedRect.h
[modify] https://crrev.com/870c262728c7c2c5423d553a83de9a239634ee86/third_party/WebKit/Source/platform/geometry/FloatSize.cpp
[modify] https://crrev.com/870c262728c7c2c5423d553a83de9a239634ee86/third_party/WebKit/Source/platform/geometry/FloatSize.h

This issue has been approved for a merge. Please merge the fix to any appropriate branches as soon as possible!

If all merges have been completed, please remove any remaining Merge-Approved labels from this issue.

Thanks for your time! To disable nags, add the Disable-Nags label.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot