Introduce andps for IA32/X64

src/ia32/lithium-codegen-ia32.cc

Index: src/ia32/lithium-codegen-ia32.cc
diff --git a/src/ia32/lithium-codegen-ia32.cc b/src/ia32/lithium-codegen-ia32.cc
index ebeaaa8216a0db4747f6c9aa1820a06a92e7f2a0..042a47080895f340c98feabba695b346f1f548a5 100644
--- a/src/ia32/lithium-codegen-ia32.cc
+++ b/src/ia32/lithium-codegen-ia32.cc
@@ -3883,7 +3883,7 @@ void LCodeGen::DoMathAbs(LMathAbs* instr) {
     XMMRegister input_reg = ToDoubleRegister(instr->value());
     __ xorps(scratch, scratch);
     __ subsd(scratch, input_reg);
-    __ pand(input_reg, scratch);
+    __ andps(input_reg, scratch);

[Benedikt Meurer, 2013/10/28 11:23:18]

Why andps and not andpd?

[Weiliang, 2013/10/28 12:41:10]

For bit-wise operation, I think there is no much difference between andps and
andpd. But andps is one byte less. 

   } else if (r.IsSmiOrInteger32()) {
     EmitIntegerMathAbs(instr);
   } else {  // Tagged case.