[Webkit-unassigned] [Bug 103128] LLInt C Loop backend should work on 32 bit platforms, both big and little endian

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=103128


Landry Breuil <landry at openbsd.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |landry at openbsd.org




--- Comment #11 from Landry Breuil <landry at openbsd.org>  2013-03-31 11:06:21 PST ---
Fwiw i'm interested in this too. The  Double2Ints part has been commited in bug #112141 / r145551.

webkitgtk 2.0.0 blows its stack on ppc at startup now, see bug #113638.

I'm currently testing the following diff on OpenBSD/ppc (on top of the other things). Should the PayloadOffset in LowLevelInterpreter32_64.asm use the same CellOffset construct for 64 bits ?

$OpenBSD$
https://bugs.webkit.org/show_bug.cgi?id=103128
--- Source/JavaScriptCore/llint/LowLevelInterpreter32_64.asm.orig       Sun Mar 31 11:51:25 2013
+++ Source/JavaScriptCore/llint/LowLevelInterpreter32_64.asm    Sun Mar 31 11:52:29 2013
@@ -1729,7 +1729,7 @@ _llint_op_next_pname:
     loadi 20[PC], t2
     loadi PayloadOffset[cfr, t2, 8], t2
     loadp JSPropertyNameIterator::m_jsStrings[t2], t3
-    loadi [t3, t0, 8], t3
+    loadi PayloadOffset[t3, t0, 8], t3
     addi 1, t0
     storei t0, PayloadOffset[cfr, t1, 8]
     loadi 4[PC], t1
$OpenBSD$
https://bugs.webkit.org/show_bug.cgi?id=103128
--- Source/JavaScriptCore/llint/LowLevelInterpreter.asm.orig    Sun Mar 31 11:48:40 2013
+++ Source/JavaScriptCore/llint/LowLevelInterpreter.asm Sun Mar 31 19:00:09 2013
@@ -87,6 +87,12 @@ else
     const PayloadOffset = 0
 end

+if JSVALUE64
+    const CellOffset = 0
+else
+    const CellOffset = PayloadOffset
+end
+
 # Constant for reasoning about butterflies.
 const IsArray                  = 1
 const IndexingShapeMask        = 30
@@ -263,13 +269,13 @@ macro assertNotConstant(index)
 end

 macro functionForCallCodeBlockGetter(targetRegister)
-    loadp Callee[cfr], targetRegister
+    loadp Callee + CellOffset[cfr], targetRegister
     loadp JSFunction::m_executable[targetRegister], targetRegister
     loadp FunctionExecutable::m_codeBlockForCall[targetRegister], targetRegister
 end

 macro functionForConstructCodeBlockGetter(targetRegister)
-    loadp Callee[cfr], targetRegister
+    loadp Callee + CellOffset[cfr], targetRegister
     loadp JSFunction::m_executable[targetRegister], targetRegister
     loadp FunctionExecutable::m_codeBlockForConstruct[targetRegister], targetRegister
 end
@@ -824,7 +830,7 @@ macro interpretResolveWithBase(opcodeLength, slowPath)
     getResolveOperation(4, t0)
     btpz t0, .slowPath

-    loadp ScopeChain[cfr], t3
+    loadp ScopeChain + CellOffset[cfr], t3
     # Get the base
     loadis ResolveOperation::m_operation[t0], t2

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.