The X64 change seems ok to me (without understanding it completely).
There is no x64-specific code in it. It's all virtual-frame operations.


http://codereview.chromium.org/487017/diff/8006/10006
File src/x64/codegen-x64.cc (right):

http://codereview.chromium.org/487017/diff/8006/10006#newcode1829
src/x64/codegen-x64.cc:1829: // that it doesn't matter whether a value
(eg, ebx pushed above) is
ebx->rbx

http://codereview.chromium.org/487017/diff/8006/10006#newcode2734
src/x64/codegen-x64.cc:2734: frame_->Push(&receiver);
Could this be made faster by doing, effectively:
  pop(tmp)
  push(Operand(sp))   // aka Dup.
  mov(Operand(sp,kPointerSize), tmp)
It seems like one less operation - but maybe not if the frame is
virtual.

http://codereview.chromium.org/487017/diff/8006/10006#newcode2958
src/x64/codegen-x64.cc:2958: frame_->Push(&receiver);
Again a swap of the top two elements of the frame.
Perhaps give the frame a Swap operation.

http://codereview.chromium.org/487017

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

Reply via email to