Add implementation of control flow and label binding to x64 assembler.

Add implementation of control flow and label binding to x64 assembler.

Committed: http://code.google.com/p/v8/source/detail?r=2057

[William Hesse, 2009-05-26 08:36:45]

Not yet unit tested.  Just out for preliminary review.

[Lasse Reichstein, 2009-05-26 09:01:05]

LGTM

http://codereview.chromium.org/113832/diff/1/2
File src/x64/assembler-x64.cc (right):

http://codereview.chromium.org/113832/diff/1/2#newcode327
Line 327: emit(offset);
This should be safe (code blocks can't be more than 2 gig in size).

http://codereview.chromium.org/113832/diff/1/2#newcode336
Line 336: }
This code might be used identically in other places where a label is emitted.
How about making an emit(Label*) function?

[William Hesse, 2009-05-26 09:12:15]

Unit tests added.

[William Hesse, 2009-05-26 11:49:03]

Ready for comments.  It all lints, compiles, and works.  The unit tests pass. 
The memory operand unit test will be enabled in the next changelist, where
operands will be implemented.

[Lasse Reichstein, 2009-05-26 12:14:40]

LGTM, with initial comments still standing.

http://codereview.chromium.org/113832/diff/9/1013
File src/x64/assembler-x64.cc (right):

http://codereview.chromium.org/113832/diff/9/1013#newcode318
Line 318: EMIT(0xC0 | (dst.code() & 0x7) << 3 | (src.code() & 0x7));
Was this incorrect before?

[William Hesse, 2009-05-26 15:15:31]

http://codereview.chromium.org/113832/diff/1/2
File src/x64/assembler-x64.cc (right):

http://codereview.chromium.org/113832/diff/1/2#newcode327
Line 327: emit(offset);
On 2009/05/26 09:01:05, Lasse Reichstein wrote:
> This should be safe (code blocks can't be more than 2 gig in size).

I'm not sure if this is a comment or a suggestion.  What do you mean?  Do we
need to check something we are not checking?

http://codereview.chromium.org/113832/diff/1/2#newcode336
Line 336: }
On 2009/05/26 09:01:05, Lasse Reichstein wrote:
> This code might be used identically in other places where a label is emitted.
> How about making an emit(Label*) function?
> 

I think there are no further cases to be implemented.  But this is a good idea.

http://codereview.chromium.org/113832/diff/9/1013#newcode318
Line 318: 
On 2009/05/26 12:14:41, Lasse Reichstein wrote:
> Was this incorrect before?

Yes it was.  Cases with two register operands should all be implemented as
src=r/m, dest=reg (an arbitrary choice).

[Lasse Reichstein, 2009-05-26 15:19:59]

LGTM still applies.

http://codereview.chromium.org/113832/diff/1/2
File src/x64/assembler-x64.cc (right):

http://codereview.chromium.org/113832/diff/1/2#newcode327
Line 327: emit(offset);
On 2009/05/26 15:15:31, William Hesse wrote:
> On 2009/05/26 09:01:05, Lasse Reichstein wrote:
> > This should be safe (code blocks can't be more than 2 gig in size).
> 
> I'm not sure if this is a comment or a suggestion.  What do you mean?  Do we
> need to check something we are not checking?

Sorry, just my vague way of saying that I'm only 98% sure this is fine, so do
double-check and correct me if I'm wrong.