Reviewers: Weiliang, Benedikt Meurer,

Message:
WL.
PTAL.
thanks

Description:
X87: Turn ToNumberStub into a PlatformCodeStub again.

port b6737c56eaf51d63ab83b6529f657c15f834623c (r25187)

original commit message:

  Turn ToNumberStub into a PlatformCodeStub again.

  The HydrogenCodeStub is too expensive and there's no easy way to reduce
  this cost, so turning it into a PlatformCodeStub solves that problem until
  we can use TurboFan for code stubs.

BUG=

Please review this at https://codereview.chromium.org/714573003/

Base URL: https://chromium.googlesource.com/external/v8.git@bleeding_edge

Affected files (+19, -0 lines):
  M src/x87/code-stubs-x87.cc


Index: src/x87/code-stubs-x87.cc
diff --git a/src/x87/code-stubs-x87.cc b/src/x87/code-stubs-x87.cc
index 202dec63bb276351ea1e09fce43142d6cf7eb577..5d38d75455b1994b4c8e59d6d206f95a53d17685 100644
--- a/src/x87/code-stubs-x87.cc
+++ b/src/x87/code-stubs-x87.cc
@@ -2862,6 +2862,25 @@ void SubStringStub::Generate(MacroAssembler* masm) {
 }


+void ToNumberStub::Generate(MacroAssembler* masm) {
+  // The ToNumber stub takes one argument in eax.
+  Label check_heap_number, call_builtin;
+  __ JumpIfNotSmi(eax, &check_heap_number, Label::kNear);
+  __ Ret();
+
+  __ bind(&check_heap_number);
+  __ CompareMap(eax, masm->isolate()->factory()->heap_number_map());
+  __ j(not_equal, &call_builtin, Label::kNear);
+  __ Ret();
+
+  __ bind(&call_builtin);
+  __ pop(ecx);  // Pop return address.
+  __ push(eax);
+  __ push(ecx);  // Push return address.
+  __ InvokeBuiltin(Builtins::TO_NUMBER, JUMP_FUNCTION);
+}
+
+
 void StringHelper::GenerateFlatOneByteStringEquals(MacroAssembler* masm,
                                                    Register left,
                                                    Register right,


--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
--- You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to