LGTM.

http://codereview.chromium.org/146029/diff/2001/2006
File src/x64/builtins-x64.cc (right):

http://codereview.chromium.org/146029/diff/2001/2006#newcode205
Line 205: __ EnterConstructFrame();
What is the calling convention of this function on entry, and also after
EnterConstructFrame?

http://codereview.chromium.org/146029/diff/2001/2006#newcode227
Line 227: __ movq(rbx, rax);  // store result in ebx
rbx, not ebx

http://codereview.chromium.org/146029/diff/2001/2006#newcode230
Line 230: // ebx: newly allocated object
rbx

http://codereview.chromium.org/146029/diff/2001/2006#newcode253
Line 253: __ push(Operand(rbx, rcx, times_pointer_size, 0));
Isn't there a three-argument version of Operand(Register, Register,
ScaleFactor)?  If not, we should create one, I think.

http://codereview.chromium.org/146029/diff/2001/2006#newcode276
Line 276: __ CmpInstanceType(rcx, FIRST_JS_OBJECT_TYPE);
CmpInstanceType can now be written inline as a single instruction,
cmpb(Operand(rcx, typeOffset), Immediate(FIRST_JS_OBJECT_TYPE))
Update CmpInstanceType, or else write instructions directly.
I have just put a bunch of these in - maybe they should all be changed
to CmpInstanceType().

http://codereview.chromium.org/146029/diff/2001/2008
File src/x64/codegen-x64.cc (right):

http://codereview.chromium.org/146029/diff/2001/2008#newcode1403
Line 1403: Result temp = allocator()->Allocate();
Use kScratchRegister here.

http://codereview.chromium.org/146029/diff/2001/2008#newcode1647
Line 1647: __ testl(value.reg(), Immediate(kSmiTagMask));
We have many of these throughout the code, and have been using testq for
them.  Since the high bits of the mask are zero, it doesn't matter which
we use, but testl can lead to a shorter encoding, I suppose.  We should
be consistent, for now.  If you plan to replacing these throughout all
the code, that is OK with me.

http://codereview.chromium.org/146029

--~--~---------~--~----~------------~-------~--~----~
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
-~----------~----~----~----~------~----~------~--~---

Reply via email to