Title: [127643] trunk/Source/_javascript_Core
- Revision
- 127643
- Author
- [email protected]
- Date
- 2012-09-05 13:50:22 -0700 (Wed, 05 Sep 2012)
Log Message
Refactored callee access in the DFG to support it in the general case
https://bugs.webkit.org/show_bug.cgi?id=95887
Reviewed by Phil Pizlo and Gavin Barraclough.
To support named function expressions, the DFG needs to understand the
callee register being used in arbitrary expressions, and not just
create_this.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::getDirect):
(JSC::DFG::ByteCodeParser::getCallee): Remap access to the callee register
into a GetCallee node. Otherwise, we get confused and think we have a
negatively indexed argument.
(ByteCodeParser):
(JSC::DFG::ByteCodeParser::InlineStackEntry::remapOperand): Inlining also
needs to remap, but to the callee in the inline frame, and not the caller's
callee.
(JSC::DFG::ByteCodeParser::parseBlock): Since we support the callee in
the general case now, there's no need to handle it in a special way for
create_this.
Modified Paths
Diff
Modified: trunk/Source/_javascript_Core/ChangeLog (127642 => 127643)
--- trunk/Source/_javascript_Core/ChangeLog 2012-09-05 20:48:23 UTC (rev 127642)
+++ trunk/Source/_javascript_Core/ChangeLog 2012-09-05 20:50:22 UTC (rev 127643)
@@ -1,3 +1,29 @@
+2012-09-05 Geoffrey Garen <[email protected]>
+
+ Refactored callee access in the DFG to support it in the general case
+ https://bugs.webkit.org/show_bug.cgi?id=95887
+
+ Reviewed by Phil Pizlo and Gavin Barraclough.
+
+ To support named function expressions, the DFG needs to understand the
+ callee register being used in arbitrary expressions, and not just
+ create_this.
+
+ * dfg/DFGByteCodeParser.cpp:
+ (JSC::DFG::ByteCodeParser::getDirect):
+ (JSC::DFG::ByteCodeParser::getCallee): Remap access to the callee register
+ into a GetCallee node. Otherwise, we get confused and think we have a
+ negatively indexed argument.
+
+ (ByteCodeParser):
+ (JSC::DFG::ByteCodeParser::InlineStackEntry::remapOperand): Inlining also
+ needs to remap, but to the callee in the inline frame, and not the caller's
+ callee.
+
+ (JSC::DFG::ByteCodeParser::parseBlock): Since we support the callee in
+ the general case now, there's no need to handle it in a special way for
+ create_this.
+
2012-09-05 Mark Hahnenberg <[email protected]>
Remove use of JSCell::classInfoOffset() from virtualForThunkGenerator
Modified: trunk/Source/_javascript_Core/dfg/DFGByteCodeParser.cpp (127642 => 127643)
--- trunk/Source/_javascript_Core/dfg/DFGByteCodeParser.cpp 2012-09-05 20:48:23 UTC (rev 127642)
+++ trunk/Source/_javascript_Core/dfg/DFGByteCodeParser.cpp 2012-09-05 20:50:22 UTC (rev 127643)
@@ -142,6 +142,9 @@
return getJSConstant(constant);
}
+ if (operand == RegisterFile::Callee)
+ return getCallee();
+
// Is this an argument?
if (operandIsArgument(operand))
return getArgument(operand);
@@ -521,6 +524,11 @@
return resultIndex;
}
+ NodeIndex getCallee()
+ {
+ return addToGraph(GetCallee);
+ }
+
// Helper functions to get/set the this value.
NodeIndex getThis()
{
@@ -1126,7 +1134,10 @@
ASSERT(result >= FirstConstantRegisterIndex);
return result;
}
-
+
+ if (operand == RegisterFile::Callee)
+ return m_calleeVR;
+
return operand + m_inlineCallFrame->stackOffset;
}
};
@@ -1835,10 +1846,7 @@
}
case op_create_this: {
- if (m_inlineStackTop->m_inlineCallFrame)
- set(currentInstruction[1].u.operand, addToGraph(CreateThis, getDirect(m_inlineStackTop->m_calleeVR)));
- else
- set(currentInstruction[1].u.operand, addToGraph(CreateThis, addToGraph(GetCallee)));
+ set(currentInstruction[1].u.operand, addToGraph(CreateThis, get(RegisterFile::Callee)));
NEXT_OPCODE(op_create_this);
}
_______________________________________________
webkit-changes mailing list
[email protected]
http://lists.webkit.org/mailman/listinfo/webkit-changes
