MIPS[64]: Fix `MIPS: Improve Float(32|64)(Max|Min)`.

src/crankshaft/mips64/lithium-codegen-mips64.cc

Index: src/crankshaft/mips64/lithium-codegen-mips64.cc
diff --git a/src/crankshaft/mips64/lithium-codegen-mips64.cc b/src/crankshaft/mips64/lithium-codegen-mips64.cc
index c8383d1efdfec9c8b2ed703beef45dc29b5bf02b..eaa9da174ba5f401da3c6f0c0c9335ecc8e7dc66 100644
--- a/src/crankshaft/mips64/lithium-codegen-mips64.cc
+++ b/src/crankshaft/mips64/lithium-codegen-mips64.cc
@@ -1887,7 +1887,9 @@ void LCodeGen::DoMathMinMax(LMathMinMax* instr) {
     FPURegister right_reg = ToDoubleRegister(right);
     FPURegister result_reg = ToDoubleRegister(instr->result());
     Label nan, done;
-    if (operation == HMathMinMax::kMathMax) {
+    if (left_reg.is(right_reg)) {

[ivica.bogosavljevic, 2016/12/07 14:59:15]

can we move this check to Float64Max?

[Ilija.Pavlovic1, 2016/12/08 13:50:06]

Yes. Now, these checks are in Float(32|64)(Max|Min) functions.

Done.

+      __ Move(result_reg, left_reg);
+    } else if (operation == HMathMinMax::kMathMax) {
       __ Float64Max(result_reg, left_reg, right_reg, &nan);
     } else {
       DCHECK(operation == HMathMinMax::kMathMin);