308946 2026-03-01 19:48:16 -0800 contain: layout causes ~8x regression in forced synchronous reflow on in-flow elements 2026-03-28 23:08:03 -0700 1 1 1 Unclassified WebKit Layout and Rendering Safari 26 Mac (Apple Silicon) macOS 26 RESOLVED FIXED InRadar P2 Normal --- 1 chenglou92 yulun_wu bfulgham karlcow simon.fraser webkit-bug-importer zalan oldest_to_newest 2186022 0 478537 chenglou92 2026-03-01 19:48:16 -0800 Created attachment 478537 Repro Adding contain: layout (or contain: strict, contain: content, contain: size layout) to in-flow position: relative elements causes a ~7-8x performance regression during interleaved read/write DOM access patterns (forced synchronous reflow), compared to the same elements without containment. overflow: hidden on in-flow elements exhibits the same regression. Repro: 1. Open the attached HTML file 2. Observe the printed timing ratio Expected: contain: layout should be equal or faster than baseline, since it restricts layout scope. Actual: contain: layout is ~7-8x slower than baseline. Test results on Safari baseline (position:relative): ~69ms + contain:layout: ~539ms ratio: ~7.8x Chrome, same hardware & OS: baseline (position:relative): ~43ms + contain:layout: ~32ms ratio: ~0.75x (faster, as expected) Batched access (all writes then all reads) shows no regression — both ~4-6ms. The issue is specific to forced synchronous reflow triggered by interleaved write-then-read patterns. 2186210 1 simon.fraser 2026-03-02 11:22:09 -0800 Surprising! 2186211 2 webkit-bug-importer 2026-03-02 11:22:16 -0800 <rdar://problem/171545381> 2186218 3 simon.fraser 2026-03-02 11:47:29 -0800 (Look at the relative %, not the actual %): Without contain: Sample Count, Samples%, Normalized CPU %, Symbol 24, 26.7%, 1.2%, WebCore::Element::offsetHeight() (in WebCore) 24, 26.7%, 1.2%, WebCore::Document::updateLayoutIfDimensionsOutOfDate(WebCore::Element&, WTF::OptionSet<WebCore::DimensionsCheck, (WTF::ConcurrencyTag)0>, WTF::OptionSet<WebCore::LayoutOptions, (WTF::ConcurrencyTag)0>) (in WebCore) 22, 24.4%, 1.1%, WebCore::Document::updateLayout(WTF::OptionSet<WebCore::LayoutOptions, (WTF::ConcurrencyTag)0>, WebCore::Element const*) (in WebCore) 22, 24.4%, 1.1%, WebCore::LocalFrameViewLayoutContext::layout(bool) (in WebCore) 22, 24.4%, 1.1%, WebCore::LocalFrameViewLayoutContext::performLayout(bool) (in WebCore) 13, 14.4%, 0.7%, WebCore::LocalFrameViewLayoutContext::flushUpdateLayerPositions() (in WebCore) 13, 14.4%, 0.7%, void WebCore::RenderLayer::recursiveUpdateLayerPositions<(WebCore::RenderLayer::UpdateLayerPositionsMode)0>(WTF::OptionSet<WebCore::RenderLayer::UpdateLayerPositionsFlag, (WTF::ConcurrencyTag)0>) (in WebCore) 9, 10.0%, 0.5%, WebCore::RenderView::layout() (in WebCore) 9, 10.0%, 0.5%, WebCore::RenderBlock::layout() (in WebCore) 9, 10.0%, 0.5%, WebCore::RenderBlockFlow::layoutBlock(WebCore::RelayoutChildren, WebCore::LayoutUnit) (in WebCore) 9, 10.0%, 0.5%, WebCore::RenderBlockFlow::layoutBlockChildren(WebCore::RelayoutChildren, WebCore::LayoutUnit&) (in WebCore) 9, 10.0%, 0.5%, WebCore::RenderBlockFlow::layoutBlockChild(WebCore::RenderBox&, WebCore::RenderBlockFlow::MarginInfo&, WebCore::LayoutUnit&, WebCore::LayoutUnit&) (in WebCore) 3, 3.3%, 0.2%, WebCore::RenderBox::avoidsFloats() const (in WebCore) With contain: Sample Count, Samples%, Normalized CPU %, Symbol 304, 48.3%, 2.1%, WebCore::Element::offsetHeight() (in WebCore) 304, 48.3%, 2.1%, WebCore::Document::updateLayoutIfDimensionsOutOfDate(WebCore::Element&, WTF::OptionSet<WebCore::DimensionsCheck, (WTF::ConcurrencyTag)0>, WTF::OptionSet<WebCore::LayoutOptions, (WTF::ConcurrencyTag)0>) (in WebCore) 299, 47.5%, 2.1%, WebCore::Document::updateLayout(WTF::OptionSet<WebCore::LayoutOptions, (WTF::ConcurrencyTag)0>, WebCore::Element const*) (in WebCore) 299, 47.5%, 2.1%, WebCore::LocalFrameViewLayoutContext::layout(bool) (in WebCore) 299, 47.5%, 2.1%, WebCore::LocalFrameViewLayoutContext::performLayout(bool) (in WebCore) 287, 45.6%, 2.0%, WebCore::RenderView::layout() (in WebCore) 287, 45.6%, 2.0%, WebCore::RenderBlock::layout() (in WebCore) 287, 45.6%, 2.0%, WebCore::RenderBlockFlow::layoutBlock(WebCore::RelayoutChildren, WebCore::LayoutUnit) (in WebCore) 287, 45.6%, 2.0%, WebCore::RenderBlockFlow::layoutBlockChildren(WebCore::RelayoutChildren, WebCore::LayoutUnit&) (in WebCore) 287, 45.6%, 2.0%, WebCore::RenderBlockFlow::layoutBlockChild(WebCore::RenderBox&, WebCore::RenderBlockFlow::MarginInfo&, WebCore::LayoutUnit&, WebCore::LayoutUnit&) (in WebCore) 209, 33.2%, 1.5%, WebCore::RenderBlock::createsNewFormattingContext() const (in WebCore) 20, 3.2%, 0.1%, WebCore::RenderBlock::layout() (in WebCore) 20, 3.2%, 0.1%, WebCore::RenderBlockFlow::layoutBlock(WebCore::RelayoutChildren, WebCore::LayoutUnit) (in WebCore) 16, 2.5%, 0.1%, WebCore::RenderBlockFlow::layoutBlockChildren(WebCore::RelayoutChildren, WebCore::LayoutUnit&) (in WebCore) 2192309 4 yulun_wu 2026-03-20 22:59:42 -0700 Pull request: https://github.com/WebKit/WebKit/pull/61069 2194808 5 ews-feeder 2026-03-28 23:08:00 -0700 Committed 310173@main (541a1c2870a0): <https://commits.webkit.org/310173@main> Reviewed commits have been landed. Closing PR #61069 and removing active labels. 478537 2026-03-01 19:48:16 -0800 2026-03-01 19:48:16 -0800 Repro safari-contain-repro.html text/html 1633 chenglou92 PCFET0NUWVBFIGh0bWw+CjxodG1sPgo8aGVhZD4KPG1ldGEgY2hhcnNldD0idXRmLTgiPgo8dGl0 bGU+U2FmYXJpIGNvbnRhaW46IGxheW91dCByZWdyZXNzaW9uIHJlcHJvPC90aXRsZT4KPC9oZWFk Pgo8Ym9keT4KPHByZSBpZD0ib3V0Ij48L3ByZT4KPGRpdiBpZD0iYzEiPjwvZGl2Pgo8ZGl2IGlk PSJjMiI+PC9kaXY+CjxzY3JpcHQ+CmZ1bmN0aW9uIHNldHVwKGNvbnRhaW5lcikgewogIGNvbnRh aW5lci5zdHlsZS5jc3NUZXh0ID0gInBvc2l0aW9uOnJlbGF0aXZlIjsKICBmb3IgKGxldCBpID0g MDsgaSA8IDUwMDsgaSsrKSB7CiAgICBjb25zdCBkID0gZG9jdW1lbnQuY3JlYXRlRWxlbWVudCgi ZGl2Iik7CiAgICBkLnN0eWxlLmNzc1RleHQgPSAicG9zaXRpb246cmVsYXRpdmU7d2lkdGg6MjAw cHg7cGFkZGluZzoycHggNHB4O2JvcmRlcjoxcHggc29saWQgI2NjYyI7CiAgICBkLnRleHRDb250 ZW50ID0gIndvcmQgIi5yZXBlYXQoNSk7CiAgICBjb250YWluZXIuYXBwZW5kQ2hpbGQoZCk7CiAg fQogIHZvaWQgY29udGFpbmVyLm9mZnNldEhlaWdodDsKICByZXR1cm4gQXJyYXkuZnJvbShjb250 YWluZXIuY2hpbGRyZW4pOwp9CgpmdW5jdGlvbiB0aHJhc2gobm9kZXMpIHsKICBjb25zdCB3aWR0 aHMgPSBbMTgwLCAyMDAsIDE5MCwgMjEwLCAxOTUsIDIwNSwgMTg1LCAyMTVdOwogIGxldCBzaW5r ID0gMDsKICBjb25zdCB0MCA9IHBlcmZvcm1hbmNlLm5vdygpOwogIGZvciAobGV0IGkgPSAwOyBp IDwgMjAwOyBpKyspIHsKICAgIGNvbnN0IG4gPSBub2Rlc1tpICUgNTAwXTsKICAgIG4uc3R5bGUu d2lkdGggPSB3aWR0aHNbaSAmIDddICsgInB4IjsKICAgIHNpbmsgKz0gbi5vZmZzZXRIZWlnaHQ7 CiAgICBuLnN0eWxlLnBhZGRpbmcgPSAoMiArIChpICYgMykpICsgInB4IDRweCI7CiAgICBzaW5r ICs9IG4ub2Zmc2V0V2lkdGg7CiAgfQogIHJldHVybiBwZXJmb3JtYW5jZS5ub3coKSAtIHQwOwp9 CgovLyBUd28gc2VwYXJhdGUgRE9NcyBzbyBuZWl0aGVyIHJ1biBzZWVzIHN0YWxlIHZhbHVlcyBm cm9tIHRoZSBvdGhlcgpjb25zdCBub2RlczEgPSBzZXR1cChkb2N1bWVudC5nZXRFbGVtZW50QnlJ ZCgiYzEiKSk7CmNvbnN0IG5vZGVzMiA9IHNldHVwKGRvY3VtZW50LmdldEVsZW1lbnRCeUlkKCJj MiIpKTsKCi8vIEFkZCBjb250YWluOiBsYXlvdXQgdG8gc2Vjb25kIHNldCBvbmx5CmZvciAoY29u c3QgbiBvZiBub2RlczIpIG4uc3R5bGUuY29udGFpbiA9ICJsYXlvdXQiOwp2b2lkIGRvY3VtZW50 LmdldEVsZW1lbnRCeUlkKCJjMiIpLm9mZnNldEhlaWdodDsKCmNvbnN0IGJhc2VsaW5lID0gdGhy YXNoKG5vZGVzMSk7CmNvbnN0IGNvbnRhaW5lZCA9IHRocmFzaChub2RlczIpOwoKZG9jdW1lbnQu Z2V0RWxlbWVudEJ5SWQoIm91dCIpLnRleHRDb250ZW50ID0KICBgYmFzZWxpbmUgKHBvc2l0aW9u OnJlbGF0aXZlKTogICR7YmFzZWxpbmUudG9GaXhlZCgxKX1tc1xuYCArCiAgYCsgY29udGFpbjps YXlvdXQ6ICAgICAgICAgICAgICAke2NvbnRhaW5lZC50b0ZpeGVkKDEpfW1zXG5gICsKICBgcmF0 aW86ICAgICAgICAgICAgICAgICAgICAgICAgICR7KGNvbnRhaW5lZCAvIGJhc2VsaW5lKS50b0Zp eGVkKDEpfXhgOwo8L3NjcmlwdD4KPC9ib2R5Pgo8L2h0bWw+Cg==