[Hit Testing] Optimize for the case of infinite clip rects
* Don't intersect with the overall dirty rect except at changes of root layer, because otherwise the location and rect don't change * Early out intersections when a ClipRect is infinite (since it has no effect) * Don't adjust infinite ClipRects (since it has no effect) * Simplify / resolve TODOs in some cases where we previously had to construct an "infinite" layout rect In local testing, this CL yields a 10% improvement on the intersection-observer/deep-layers.html performance test due to faster hit testing. Bug: 831762 Cq-Include-Trybots: luci.chromium.try:linux_layout_tests_slimming_paint_v2;master.tryserver.blink:linux_trusty_blink_rel Change-Id: Ie7a1e2fb5cbeb8f7f9a267489785266ed7b8ca47 Reviewed-on: https://chromium-review.googlesource.com/1104964 Commit-Queue: Chris Harrelson <chrishtr@chromium.org> Reviewed-by:Xianzhu Wang <wangxianzhu@chromium.org> Cr-Commit-Position: refs/heads/master@{#568631}
Showing
This diff is collapsed.
This diff is collapsed.
Please register or sign in to comment