[Webkit-unassigned] [Bug 235114] REGRESSION (iOS 15): Massive variation in microbenchmark results when tab returns to foreground

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=235114

--- Comment #8 from Chris Dumez <cdumez at apple.com> ---
In the slow case, top samples are:
Sample Count, Samples %, Normalized CPU %, Symbol
10167, 74.2%, 2.8%, vmEntryToJavaScript (in JavaScriptCore)
5205, 38.0%, 1.4%,     ??? (in (null)) [0x53a6858040a4]
5107, 37.3%, 1.4%,         operationGetFromScope (in JavaScriptCore)
2580, 18.8%, 0.7%,             WebCore::JSDOMWindow::getOwnPropertySlot(JSC::JSObject*, JSC::JSGlobalObject*, JSC::PropertyName, JSC::PropertySlot&) (in WebCore)
1467, 10.7%, 0.4%,                 JSC::JSGlobalObject::getOwnPropertySlot(JSC::JSObject*, JSC::JSGlobalObject*, JSC::PropertyName, JSC::PropertySlot&) (in JavaScriptCore)
1136, 8.3%, 0.3%,                     JSC::JSObject::getOwnStaticPropertySlot(JSC::VM&, JSC::PropertyName, JSC::PropertySlot&) (in JavaScriptCore)
873, 6.4%, 0.2%,                         JSC::getStaticPropertySlotFromTable(JSC::VM&, JSC::ClassInfo const*, JSC::HashTable const&, JSC::JSObject*, JSC::PropertyName, JSC::PropertySlot&) (in JavaScriptCore)
510, 3.7%, 0.1%,                             WTF::equal(WTF::StringImpl const*, unsigned char const*) (in JavaScriptCore)


In the fast case, top samples are:
Sample Count, Samples %, Normalized CPU %, Symbol
10164, 72.1%, 2.8%, vmEntryToJavaScript (in JavaScriptCore)
3501, 24.9%, 1.0%,     ??? (in (null)) [0x20370c40ae38]
3363, 23.9%, 0.9%,         ??? (in (null)) [0x20370c4011d8]
3317, 23.5%, 0.9%,             WebCore::jsPerformancePrototypeFunction_now(JSC::JSGlobalObject*, JSC::CallFrame*) (in WebCore)
2072, 14.7%, 0.6%,                 WTF::MonotonicTime::now() (in JavaScriptCore)
1195, 8.5%, 0.3%,                     mach_absolute_time (in libsystem_kernel.dylib)


They do look very different and I suspect this is about JIT'ing vs not JIT'ing. Somebody from JSC should probably take a look (I attached the profiles to the radar).

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20220112/4dc17ba3/attachment.htm>