X64 Implementation: Make codegen load literals and assign to local variables.

X64 Implementation: Make codegen load literals and assign to local variables.

Committed: http://code.google.com/p/v8/source/detail?r=2204

[William Hesse, 2009-06-16 12:54:26]

Does not yet compile and run.  But please comment.

[William Hesse, 2009-06-16 14:17:39]

Compiles, runs, and returns the wrong answer from
{
  39;
  _result = 42;
  return _result;
}
(That is the AST implicitly created from 39;42;)

I would like to commit it, though.

[William Hesse, 2009-06-17 11:23:33]

This now works, and returns the right answer.  Please review.

[Lasse Reichstein, 2009-06-17 11:35:05]

LGTM

http://codereview.chromium.org/126198/diff/1010/1012
File src/x64/codegen-x64.cc (right):

http://codereview.chromium.org/126198/diff/1010/1012#newcode133
Line 133: // Function compiles and runs, but returns a JSFunction object.
Comment no longer correct?

http://codereview.chromium.org/126198/diff/1010/1012#newcode336
Line 336: */
What is the commented part for?

http://codereview.chromium.org/126198/diff/1010/1011
File src/x64/virtual-frame-x64.cc (right):

http://codereview.chromium.org/126198/diff/1010/1011#newcode520
Line 520: 
Much of this code looks like it's identical to the ia32 version, except for a
few lines of assembler. 
Would it be possible to abstract the identical parts into a shared template
function so the architecture specific bits are only assembler operations?

[William Hesse, 2009-06-17 12:02:32]

http://codereview.chromium.org/126198/diff/1010/1012
File src/x64/codegen-x64.cc (right):

http://codereview.chromium.org/126198/diff/1010/1012#newcode133
Line 133: // Function compiles and runs, but returns a JSFunction object.
On 2009/06/17 11:35:05, Lasse Reichstein wrote:
> Comment no longer correct?

Done.

http://codereview.chromium.org/126198/diff/1010/1012#newcode336
Line 336: */
On 2009/06/17 11:35:05, Lasse Reichstein wrote:
> What is the commented part for?

Handling multiple return statements in the same function.  Added as a TODO and
commented.

http://codereview.chromium.org/126198/diff/1010/1011
File src/x64/virtual-frame-x64.cc (right):

http://codereview.chromium.org/126198/diff/1010/1011#newcode520
Line 520: 
On 2009/06/17 11:35:05, Lasse Reichstein wrote:
> Much of this code looks like it's identical to the ia32 version, except for a
> few lines of assembler. 
> Would it be possible to abstract the identical parts into a shared template
> function so the architecture specific bits are only assembler operations?

Actually, the use of kScratchRegister makes a number of things much simpler.  I
forgot to remove lines 473 and 474 - that is an example.