X64: Add inline SwapElements to fundamental code generator on x64 platform.

X64: Add inline SwapElements to fundamental code generator on x64 platform.

Committed: http://code.google.com/p/v8/source/detail?r=6995

[Lasse Reichstein, 2011-03-01 14:01:49]

LGTM

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc
File src/x64/full-codegen-x64.cc (right):

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc#new...
src/x64/full-codegen-x64.cc:2971: VisitForStackValue(args->at(2));
Could we put the values directly into registers, and only push them if we need
to go to runtime? Or will we always need to push them anyway, to free up
registers?

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc#new...
src/x64/full-codegen-x64.cc:2997: __ movq(index_2, Operand(rsp, 0));
Mutiply by kPointerSize here too, just for consistency.

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc#new...
src/x64/full-codegen-x64.cc:3001: __ movq(temp, FieldOperand(object,
JSArray::kLengthOffset));
Make a comment that since the array is in fast elements mode, the length must be
a smi.
(Do we enforce that in ToUInt32, or could we cheat it to be a HeapNumber holding
a potential smi value by doing something like:
  arr.length = 87.5 / 2.5;
?)

[William Hesse, 2011-03-01 14:29:52]

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc
File src/x64/full-codegen-x64.cc (right):

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc#new...
src/x64/full-codegen-x64.cc:2971: VisitForStackValue(args->at(2));
On 2011/03/01 14:01:50, Lasse Reichstein wrote:
> Could we put the values directly into registers, and only push them if we need
> to go to runtime? Or will we always need to push them anyway, to free up
> registers?

This is a direct copy of ia32 code, where there is a register shortage.  I think
it is better to keep the code parallel, in this case.

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc#new...
src/x64/full-codegen-x64.cc:2997: __ movq(index_2, Operand(rsp, 0));
On 2011/03/01 14:01:50, Lasse Reichstein wrote:
> Mutiply by kPointerSize here too, just for consistency.

Done.

http://codereview.chromium.org/6594074/diff/1/src/x64/full-codegen-x64.cc#new...
src/x64/full-codegen-x64.cc:3001: __ movq(temp, FieldOperand(object,
JSArray::kLengthOffset));
On 2011/03/01 14:01:50, Lasse Reichstein wrote:
> Make a comment that since the array is in fast elements mode, the length must
be
> a smi.
Done.
Yes, it is always a smi.


> (Do we enforce that in ToUInt32, or could we cheat it to be a HeapNumber
holding
> a potential smi value by doing something like:
>   arr.length = 87.5 / 2.5;
>