Bug 54524

Summary: Allow JSObject to fully utilize cell's capacity for inline storage.
Product: WebKit Reporter: Gavin Barraclough <barraclough>
Component: JavaScriptCoreAssignee: Nobody <webkit-unassigned>
Status: RESOLVED FIXED    
Severity: Normal CC: ggaren, oliver, webkit-ews, webkit.review.bot
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: PC   
OS: OS X 10.5   
Attachments:
Description Flags
The patch, this shows a small sunspider regression, but a similar v8 progression - a wash overall.
ggaren: review-
Fixes for review comments / Qt (not for style though - these all look like preexisting issues). ggaren: review+

Gavin Barraclough
Reported 2011-02-15 19:47:20 PST
Currently JSObject is both directly instantiated for regular JS objects, and derived to implement subtypes. A consequence of this is that we need to ensure that sufficient space from the cell is left unused and available for any data members that will be introduced by subclasses of JSObject. By restructuring the internal storage array out of JSObject we can increase the size in the internal storage for regular objects.
Attachments
The patch, this shows a small sunspider regression, but a similar v8 progression - a wash overall. (89.78 KB, patch)
2011-02-15 22:40 PST, Gavin Barraclough
ggaren: review-
Fixes for review comments / Qt (not for style though - these all look like preexisting issues). (92.78 KB, patch)
2011-02-16 13:16 PST, Gavin Barraclough
ggaren: review+
Gavin Barraclough
Comment 1 2011-02-15 22:40:35 PST
Created attachment 82592 [details] The patch, this shows a small sunspider regression, but a similar v8 progression - a wash overall.
WebKit Review Bot
Comment 2 2011-02-15 22:42:32 PST
Attachment 82592 [details] did not pass style-queue: Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/JavaScriptCore/API/JSObjectRef.cpp'..." exit_code: 1 Source/JavaScriptCore/runtime/JSWrapperObject.h:31: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/JSObject.h:307: Tests for true/false, null/non-null, and zero/non-zero should all be done without equality comparisons. [readability/comparison_to_zero] [5] Source/JavaScriptCore/runtime/JSObject.h:337: Tests for true/false, null/non-null, and zero/non-zero should all be done without equality comparisons. [readability/comparison_to_zero] [5] Source/JavaScriptCore/runtime/JSNotAnObject.h:39: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/ObjectPrototype.h:28: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/JSByteArray.h:35: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/ErrorInstance.h:28: Code inside a namespace should not be indented. [whitespace/indent] [4] Total errors found: 7 in 41 files If any of these errors are false positives, please file a bug against check-webkit-style.
Early Warning System Bot
Comment 3 2011-02-15 23:01:20 PST
Geoffrey Garen
Comment 4 2011-02-16 12:05:21 PST
Looks like you need to update some platform-specific files to fix the Qt build: ../../../Source/WebCore/bridge/qt/qt_runtime.cpp: In member function ‘void JSC::Bindings::QtConnectionObject::execute(void**)’: ../../../Source/WebCore/bridge/qt/qt_runtime.cpp:1835: error: ‘createStructure’ is not a member of ‘JSC::JSObject’
Geoffrey Garen
Comment 5 2011-02-16 12:08:50 PST
Comment on attachment 82592 [details] The patch, this shows a small sunspider regression, but a similar v8 progression - a wash overall. View in context: https://bugs.webkit.org/attachment.cgi?id=82592&action=review r=me if you fix the build. > Source/JavaScriptCore/ChangeLog:17 > + and only allows construction through JSFinalObject::create(). I think you should mention the reason for this change -- now all objects have a direct pointer to their storage, and lea shenanigans are not required. > Source/JavaScriptCore/runtime/JSObject.h:298 > +COMPILE_ASSERT((JSFinalObject_inlineStorageCapacity >= JSNonFinalObject_inlineStorageCapacity), vanilla_storage_is_at_least_as_large_as_non_vanilla); s/vanilla/final/ > Source/JavaScriptCore/runtime/JSTypeInfo.h:73 > + unsigned isVanilla() const { return m_flags2 && (IsJSFinalObject >> 8); } s/Vanilla/Final/
Gavin Barraclough
Comment 6 2011-02-16 13:16:23 PST
Created attachment 82682 [details] Fixes for review comments / Qt (not for style though - these all look like preexisting issues).
Geoffrey Garen
Comment 7 2011-02-16 13:18:06 PST
Comment on attachment 82682 [details] Fixes for review comments / Qt (not for style though - these all look like preexisting issues). I believe the EWS bots will keep running even after an r+, so I'll mark this r+ now.
WebKit Review Bot
Comment 8 2011-02-16 13:19:41 PST
Attachment 82682 [details] did not pass style-queue: Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/JavaScriptCore/API/JSObjectRef.cpp'..." exit_code: 1 Source/JavaScriptCore/runtime/JSWrapperObject.h:31: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/JSObject.h:312: Tests for true/false, null/non-null, and zero/non-zero should all be done without equality comparisons. [readability/comparison_to_zero] [5] Source/JavaScriptCore/runtime/JSObject.h:344: Tests for true/false, null/non-null, and zero/non-zero should all be done without equality comparisons. [readability/comparison_to_zero] [5] Source/JavaScriptCore/runtime/JSNotAnObject.h:39: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/ObjectPrototype.h:28: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/JSByteArray.h:35: Code inside a namespace should not be indented. [whitespace/indent] [4] Source/JavaScriptCore/runtime/ErrorInstance.h:28: Code inside a namespace should not be indented. [whitespace/indent] [4] Total errors found: 7 in 44 files If any of these errors are false positives, please file a bug against check-webkit-style.
Gavin Barraclough
Comment 9 2011-02-16 13:35:48 PST
fixed in r78732.
Note You need to log in before you can comment on or make changes to this bug.