[x64] Use xorl to materialize smi zero.

Before we always loaded smi zero via a movabs with a 64-bit immediate,
which is pretty expensive compared to the xorl.

R=jarin@chromium.org

Committed: https://crrev.com/f236777bfe6e080ff1ead6baf847cc9b6bb4f9cb
Cr-Commit-Position: refs/heads/master@{#27829}

Review URL: https://codereview.chromium.org/1085153002

Cr-Commit-Position: refs/heads/master@{#32784}

src/x64/macro-assembler-x64.cc

@@ -1100,6 +1100,7 @@ void MacroAssembler::SafePush(Smi* src) {
 
 
 Register MacroAssembler::GetSmiConstant(Smi* source) {
+  STATIC_ASSERT(kSmiTag == 0);
   int value = source->value();
   if (value == 0) {
     xorl(kScratchRegister, kScratchRegister);
@@ -1111,9 +1112,13 @@ Register MacroAssembler::GetSmiConstant(Smi* source) {
 
 
 void MacroAssembler::LoadSmiConstant(Register dst, Smi* source) {
-  // Special-casing 0 here to use xorl seems to make things slower, so we don't
-  // do it.
-  Move(dst, source, Assembler::RelocInfoNone());
+  STATIC_ASSERT(kSmiTag == 0);
+  int value = source->value();
+  if (value == 0) {
+    xorl(dst, dst);
+  } else {
+    Move(dst, source, Assembler::RelocInfoNone());
+  }
 }