NumberUntagD is faster when untagging in a temp register

src/ia32/lithium-codegen-ia32.cc

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 //     * Redistributions of source code must retain the above copyright
 //       notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
 //       copyright notice, this list of conditions and the following
 //       disclaimer in the documentation and/or other materials provided
@@ skipping 5267 matching lines @@
       // Pop FPU stack before deoptimizing.
       __ fstp(0);
       DeoptimizeIf(not_zero, env);
     }
     __ jmp(&done, Label::kNear);
   } else {
     ASSERT(mode == NUMBER_CANDIDATE_IS_SMI);
   }
 
   __ bind(&load_smi);
-  __ SmiUntag(input_reg);  // Untag smi before converting to float.
-  __ push(input_reg);
+  // Clobbering a temp is faster than re-tagging the
+  // input register since we avoid dependencies.
+  __ mov(temp_reg, input_reg);
+  __ SmiUntag(temp_reg);  // Untag smi before converting to float.
+  __ push(temp_reg);
   __ fild_s(Operand(esp, 0));
-  __ pop(input_reg);
-  __ SmiTag(input_reg);  // Retag smi.
+  __ add(esp, Immediate(kPointerSize));
   __ bind(&done);
   X87CommitWrite(res_reg);
 }
 
 
 void LCodeGen::EmitNumberUntagD(Register input_reg,
                                 Register temp_reg,
                                 XMMRegister result_reg,
                                 bool can_convert_undefined_to_nan,
                                 bool deoptimize_on_minus_zero,
@@ skipping 35 matching lines @@
       __ j(not_zero, &done, Label::kNear);
       __ movmskpd(temp_reg, result_reg);
       __ test_b(temp_reg, 1);
       DeoptimizeIf(not_zero, env);
     }
     __ jmp(&done, Label::kNear);
   } else {
     ASSERT(mode == NUMBER_CANDIDATE_IS_SMI);
   }
 
-  // Smi to XMM conversion
   __ bind(&load_smi);
-  __ SmiUntag(input_reg);  // Untag smi before converting to float.
-  __ cvtsi2sd(result_reg, Operand(input_reg));
-  __ SmiTag(input_reg);  // Retag smi.
+  // Smi to XMM conversion. Clobbering a temp is faster than re-tagging the
+  // input register since we avoid dependencies.
+  __ mov(temp_reg, input_reg);
+  __ SmiUntag(temp_reg);  // Untag smi before converting to float.
+  __ cvtsi2sd(result_reg, Operand(temp_reg));
   __ bind(&done);
 }
 
 
 void LCodeGen::DoDeferredTaggedToI(LTaggedToI* instr, Label* done) {
   Register input_reg = ToRegister(instr->value());
 
 
   if (instr->truncating()) {
     Label heap_number, slow_case;
@@ skipping 53 matching lines @@
   __ JumpIfNotSmi(input_reg, deferred->entry());
   __ SmiUntag(input_reg);
   __ bind(deferred->exit());
 }
 
 
 void LCodeGen::DoNumberUntagD(LNumberUntagD* instr) {
   LOperand* input = instr->value();
   ASSERT(input->IsRegister());
   LOperand* temp = instr->temp();
-  ASSERT(temp == NULL || temp->IsRegister());
+  ASSERT(temp->IsRegister());
   LOperand* result = instr->result();
   ASSERT(result->IsDoubleRegister());
 
   Register input_reg = ToRegister(input);
   bool deoptimize_on_minus_zero =
       instr->hydrogen()->deoptimize_on_minus_zero();
-  Register temp_reg = deoptimize_on_minus_zero ? ToRegister(temp) : no_reg;
+  Register temp_reg = ToRegister(temp);
 
   HValue* value = instr->hydrogen()->value();
   NumberUntagDMode mode = value->representation().IsSmi()
       ? NUMBER_CANDIDATE_IS_SMI : NUMBER_CANDIDATE_IS_ANY_TAGGED;
 
   if (CpuFeatures::IsSupported(SSE2)) {
     CpuFeatureScope scope(masm(), SSE2);
     XMMRegister result_reg = ToDoubleRegister(result);
     EmitNumberUntagD(input_reg,
                      temp_reg,
@@ skipping 906 matching lines @@
                               FixedArray::kHeaderSize - kPointerSize));
   __ bind(&done);
 }
 
 
 #undef __
 
 } }  // namespace v8::internal
 
 #endif  // V8_TARGET_ARCH_IA32