Support register arguments in more cases....

Support register arguments in more cases.
1. MUL and DIV on SMIs.
2. When calling GenericBinaryOpStub from a virtual frame.
3. When generating code for a loop counter.
Overall performance gain is about 0.6%.

Committed: http://code.google.com/p/v8/source/detail?r=3708

[Kevin Millikin (Chromium), 2010-01-26 09:07:22]

LGTM.

http://codereview.chromium.org/555098/diff/1/2
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/555098/diff/1/2#newcode7122
src/ia32/codegen-ia32.cc:7122: __ mov(edi, Operand(eax));  // backup the 1st
operand
No need for Operand: __ mov(edi, eax).

Comment should be formatted like a sentence: Backup the left operand.

http://codereview.chromium.org/555098/diff/1/2#newcode7277
src/ia32/codegen-ia32.cc:7277: FloatingPointHelper::ARGS_IN_REGISTERS);
Does LoadSse2Smis need the ArgLocation argument any more?  It now expects them
in registers at all call sites.

[Vladislav Kaznacheev, 2010-01-26 10:21:18]

http://codereview.chromium.org/555098/diff/1/2
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/555098/diff/1/2#newcode7122
src/ia32/codegen-ia32.cc:7122: __ mov(edi, Operand(eax));  // backup the 1st
operand
On 2010/01/26 09:07:22, Kevin Millikin wrote:
> No need for Operand: __ mov(edi, eax).
> 
> Comment should be formatted like a sentence: Backup the left operand.

Done.

http://codereview.chromium.org/555098/diff/1/2#newcode7277
src/ia32/codegen-ia32.cc:7277: FloatingPointHelper::ARGS_IN_REGISTERS);
Removed (also for LoadFloatSmis).
On 2010/01/26 09:07:22, Kevin Millikin wrote:
> Does LoadSse2Smis need the ArgLocation argument any more?  It now expects them
> in registers at all call sites.