[webkit-changes] [WebKit/WebKit] 0fa56e: Scoped Arguements needs to alias between named and...
Michael Saboff
noreply at github.com
Mon Dec 18 22:44:23 PST 2023
Branch: refs/heads/main
Home: https://github.com/WebKit/WebKit
Commit: 0fa56ec28b9251f4d6b73e7e963f27838c828bf6
https://github.com/WebKit/WebKit/commit/0fa56ec28b9251f4d6b73e7e963f27838c828bf6
Author: Michael Saboff <msaboff at apple.com>
Date: 2023-12-18 (Mon, 18 Dec 2023)
Changed paths:
A JSTests/stress/arrow-function-captured-arguments-aliased.js
M Source/JavaScriptCore/bytecode/CodeBlock.cpp
M Source/JavaScriptCore/bytecompiler/BytecodeGenerator.cpp
M Source/JavaScriptCore/dfg/DFGByteCodeParser.cpp
M Source/JavaScriptCore/llint/LowLevelInterpreter32_64.asm
M Source/JavaScriptCore/llint/LowLevelInterpreter64.asm
M Source/JavaScriptCore/runtime/GetPutInfo.h
M Source/JavaScriptCore/runtime/ScopedArguments.h
M Source/JavaScriptCore/runtime/ScopedArgumentsTable.cpp
M Source/JavaScriptCore/runtime/ScopedArgumentsTable.h
M Source/JavaScriptCore/runtime/SymbolTable.cpp
M Source/JavaScriptCore/runtime/SymbolTable.h
Log Message:
-----------
Scoped Arguements needs to alias between named and unnamed accesses and across nested scopes
https://bugs.webkit.org/show_bug.cgi?id=261934
rdar://114925088
Reviewed by Yusuke Suzuki.
Fixed issue where an access to a named argument and a seperate access via its argument[i] counterpart weren't recognized throughout
all JIT tiers as accesses to the same scoped value. The DFG bytecode parser can unknowingly constant fold the read access.
Added aliasing via the SymbolTable and its ScopedArgumentsTable for both types of accesses of such values.
related objects
Added watchpoints for scoped arguments, and shared the watchpoint from the SymbolTableEntry for the named parameter with the
ScopedArgument entry for the matching index. Tagged op_put_to_scope bytecodes with a new ScopedArgumentInitialization
initialization type in GetPutInfo to signify this shared watchpoint case. Since currently all tiers write to scoped arguments
via ScopedArguments::setIndexQuickly(), that is where we fire its watchpoint.
Added a new test.
* JSTests/stress/arrow-function-captured-arguments-aliased.js: New test.
(createOptAll):
(createOpt500):
(createOpt2000):
(createOpt5000):
(main):
* Source/JavaScriptCore/bytecode/CodeBlock.cpp:
(JSC::CodeBlock::finishCreation):
* Source/JavaScriptCore/bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::BytecodeGenerator):
* Source/JavaScriptCore/dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
* Source/JavaScriptCore/llint/LowLevelInterpreter32_64.asm:
* Source/JavaScriptCore/llint/LowLevelInterpreter64.asm:
* Source/JavaScriptCore/runtime/GetPutInfo.h:
(JSC::initializationModeName):
(JSC::isInitialization):
* Source/JavaScriptCore/runtime/ScopedArguments.h:
* Source/JavaScriptCore/runtime/ScopedArgumentsTable.cpp:
(JSC::ScopedArgumentsTable::tryCreate):
(JSC::ScopedArgumentsTable::tryClone):
(JSC::ScopedArgumentsTable::trySetLength):
(JSC::ScopedArgumentsTable::trySetWatchpointSet):
* Source/JavaScriptCore/runtime/ScopedArgumentsTable.h:
* Source/JavaScriptCore/runtime/SymbolTable.h:
Originally-landed-as: 267815.345 at safari-7617-branch (99b8814b73d1). rdar://119594814
Canonical link: https://commits.webkit.org/272253@main
More information about the webkit-changes
mailing list