Changeset 72442 in webkit for trunk/JavaScriptCore/interpreter/CallFrame.h

Timestamp:

Nov 19, 2010, 2:40:12 PM (15 years ago)

Author:

[email protected]

Message:

2010-11-19 Oliver Hunt <​[email protected]>

Reviewed by Geoffrey Garen.

Don't check for constant registers when we can guarantee that the register won't be in the constant pool
​https://bugs.webkit.org/show_bug.cgi?id=49814

Add uncheckedR(int) to CallFrame, and replace all the uses of r() with uncheckedR()
when we can guarantee that the register is not referring to a constant.
This makes the interpreter about 0.5% faster, and makes the CallFrame initialisation
logic correct when we're using a faked callframe (as in the case of the globalExec).

• bytecode/CodeBlock.cpp: (JSC::CodeBlock::createActivation):
• debugger/DebuggerCallFrame.cpp: (JSC::DebuggerCallFrame::thisObject):
• interpreter/CallFrame.h: (JSC::ExecState::uncheckedR):
• interpreter/Interpreter.cpp: (JSC::Interpreter::resolve): (JSC::Interpreter::resolveSkip): (JSC::Interpreter::resolveGlobal): (JSC::Interpreter::resolveGlobalDynamic): (JSC::Interpreter::resolveBase): (JSC::Interpreter::resolveBaseAndProperty): (JSC::Interpreter::callEval): (JSC::Interpreter::unwindCallFrame): (JSC::Interpreter::throwException): (JSC::Interpreter::execute): (JSC::Interpreter::executeCall): (JSC::Interpreter::executeConstruct): (JSC::Interpreter::prepareForRepeatCall): (JSC::Interpreter::createExceptionScope): (JSC::Interpreter::privateExecute):
• jit/JITStubs.cpp: (JSC::DEFINE_STUB_FUNCTION):
• runtime/JSActivation.cpp: (JSC::JSActivation::argumentsGetter):

File:

• trunk/JavaScriptCore/interpreter/CallFrame.h (modified) (1 diff)

trunk/JavaScriptCore/interpreter/CallFrame.h

@@ -127 +127 @@
         // Read a register from the codeframe (or constant from the CodeBlock).
         inline Register& r(int);
+        // Read a register for a non-constant 
+        inline Register& uncheckedR(int);
 
         // Access to arguments.