[builtins] Introduce proper Float64Exp operator.

src/crankshaft/mips/lithium-mips.cc

Index: src/crankshaft/mips/lithium-mips.cc
diff --git a/src/crankshaft/mips/lithium-mips.cc b/src/crankshaft/mips/lithium-mips.cc
index 7b481db188042e26cbb38efdd6eacb5fa1659b2d..ea14026d9a46ace973cf79d74fe157b90c90ca75 100644
--- a/src/crankshaft/mips/lithium-mips.cc
+++ b/src/crankshaft/mips/lithium-mips.cc
@@ -1107,12 +1107,8 @@ LInstruction* LChunkBuilder::DoMathClz32(HUnaryMathOperation* instr) {
 LInstruction* LChunkBuilder::DoMathExp(HUnaryMathOperation* instr) {
   DCHECK(instr->representation().IsDouble());
   DCHECK(instr->value()->representation().IsDouble());
-  LOperand* input = UseRegister(instr->value());
-  LOperand* temp1 = TempRegister();
-  LOperand* temp2 = TempRegister();
-  LOperand* double_temp = TempDoubleRegister();
-  LMathExp* result = new(zone()) LMathExp(input, double_temp, temp1, temp2);
-  return DefineAsRegister(result);
+  LOperand* input = UseFixedDouble(instr->value(), f4);
+  return MarkAsCall(DefineFixedDouble(new (zone()) LMathExp(input), f4), instr);
 }