Add test, neg, and not instructions to x64 assembler

Add test, neg, and not instructions to x64 assembler

Committed: http://code.google.com/p/v8/source/detail?r=2076

[Lasse Reichstein, 2009-05-29 07:41:55]

LGTM

http://codereview.chromium.org/112066/diff/1/4
File src/x64/assembler-x64-inl.h (right):

http://codereview.chromium.org/112066/diff/1/4#newcode74
Line 74: // REX.W is set.
For completeness, you might write that REX.X is clear.

http://codereview.chromium.org/112066/diff/1/4#newcode84
Line 84: emit(0x40 | (reg.code() & 0x8) >> 1 | op.rex_);
Shouldn't this always be optional (i.e., emit nothing if no extended registers
are used)?
I.e., is there any case when 0x40 should be written?
(ditto for reg,reg-version).

http://codereview.chromium.org/112066/diff/1/2
File src/x64/assembler-x64.cc (right):

http://codereview.chromium.org/112066/diff/1/2#newcode522
Line 522: emit_rex_64(rax, dst);
"rax" here is just a placeholder for a zero REX.R bit.
Maybe create an emit_rex_64(Register) and emit_rex_64(Operand) that does that
explicitly.

http://codereview.chromium.org/112066/diff/1/2#newcode533
Line 533: emit_operand(rbx, dst);  // Operation code 3
Ditto for rbx here, it's just a way to specify the value 3.
It would be better with an emit_operand(int opcode_extension, Operand) to be
explicit.

http://codereview.chromium.org/112066/diff/1/2#newcode623
Line 623: }
Use emit_optional_rex_32 here?

[William Hesse, 2009-05-29 08:43:03]

http://codereview.chromium.org/112066/diff/1/4
File src/x64/assembler-x64-inl.h (right):

http://codereview.chromium.org/112066/diff/1/4#newcode74
Line 74: // REX.W is set.
On 2009/05/29 07:41:55, Lasse Reichstein wrote:
> For completeness, you might write that REX.X is clear.

Done.

http://codereview.chromium.org/112066/diff/1/4#newcode84
Line 84: emit(0x40 | (reg.code() & 0x8) >> 1 | op.rex_);
On 2009/05/29 07:41:55, Lasse Reichstein wrote:
> Shouldn't this always be optional (i.e., emit nothing if no extended registers
> are used)?
> I.e., is there any case when 0x40 should be written?
> (ditto for reg,reg-version).

No, there are cases where a REX prefix is still needed (to switch byte-register
codes from AH, BH, etc. to SIL, DIL, etc.)

http://codereview.chromium.org/112066/diff/1/2
File src/x64/assembler-x64.cc (right):

http://codereview.chromium.org/112066/diff/1/2#newcode623
Line 623: }
On 2009/05/29 07:41:55, Lasse Reichstein wrote:
> Use emit_optional_rex_32 here?

No, because then we use the old ia32 set of byte registers.