[builtins] Introduce proper Float64Exp operator.

src/crankshaft/arm/lithium-arm.h

Index: src/crankshaft/arm/lithium-arm.h
diff --git a/src/crankshaft/arm/lithium-arm.h b/src/crankshaft/arm/lithium-arm.h
index 2ec992fd3ee4183a751e4de77d45b0b8549e7a16..1f18392d69b1a4bdcbfc46ed3882b01cc460980c 100644
--- a/src/crankshaft/arm/lithium-arm.h
+++ b/src/crankshaft/arm/lithium-arm.h
@@ -905,24 +905,11 @@ class LMathClz32 final : public LTemplateInstruction<1, 1, 0> {
   DECLARE_CONCRETE_INSTRUCTION(MathClz32, "math-clz32")
 };
 
-
-class LMathExp final : public LTemplateInstruction<1, 1, 3> {
+class LMathExp final : public LTemplateInstruction<1, 1, 0> {
  public:
-  LMathExp(LOperand* value,
-           LOperand* double_temp,
-           LOperand* temp1,
-           LOperand* temp2) {
-    inputs_[0] = value;
-    temps_[0] = temp1;
-    temps_[1] = temp2;
-    temps_[2] = double_temp;
-    ExternalReference::InitializeMathExpData();
-  }
+  explicit LMathExp(LOperand* value) { inputs_[0] = value; }
 
   LOperand* value() { return inputs_[0]; }
-  LOperand* temp1() { return temps_[0]; }
-  LOperand* temp2() { return temps_[1]; }
-  LOperand* double_temp() { return temps_[2]; }
 
   DECLARE_CONCRETE_INSTRUCTION(MathExp, "math-exp")
 };