Fix performance of ValueToInt32 node when predicting double

[Oliver Hunt]

Fix performance of ValueToInt32 node when predicting double

[Oliver Hunt]

Created attachment 110916 [details]
Patch

[Oliver Hunt]

Committed r97410: <http://trac.webkit.org/changeset/97410>

[Yuqiang Xian]

--- a/Source/JavaScriptCore/dfg/DFGSpeculativeJIT32_64.cpp
+++ b/Source/JavaScriptCore/dfg/DFGSpeculativeJIT32_64.cpp
@@ -851,16 +851,7 @@ void SpeculativeJIT::compile(Node& node)
     }
 
     case ValueToInt32: {
-        if (at(node.child1()).shouldNotSpeculateInteger()) {
-            // Do it the safe way.
-            nonSpeculativeValueToInt32(node);
-            break;
-        }
-        
-        SpeculateIntegerOperand op1(this, node.child1());
-        GPRTemporary result(this, op1);
-        m_jit.move(op1.gpr(), result.gpr());
-        integerResult(result.gpr(), m_compileIndex, op1.format());
+        compileValueToInt32(node.child1());
         break;
     }

I think it should be compileValueToInt32(node);

[Yuqiang Xian]

(In reply to comment #3)

I have included the fix in bug #70069

> 
> I think it should be compileValueToInt32(node);

[Yuqiang Xian]

(In reply to comment #4)

Sorry - should be bug #70096

> (In reply to comment #3)
> 
> I have included the fix in bug #70069
> 
> > 
> > I think it should be compileValueToInt32(node);