Avoid using double temp register explicitly in lithium codegen

Avoid using double temp register explicitly in lithium codegen

BUG=
R=mvstanton@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=16989

[mvstanton, 2013-09-25 07:00:30]

Hi Weiliang, just one question below, thx!
--Michael

https://codereview.chromium.org/23531084/diff/1/src/x64/macro-assembler-x64.cc
File src/x64/macro-assembler-x64.cc (right):

https://codereview.chromium.org/23531084/diff/1/src/x64/macro-assembler-x64.c...
src/x64/macro-assembler-x64.cc:3731: int space =
XMMRegister::kMaxNumAllocatableRegisters * kDoubleSize +
Could you do the same thing on ia32, using
Register::kMaxNumAllocatableRegisters?

[weiliang.lin2, 2013-09-25 07:11:38]

On 2013/09/25 07:00:30, mvstanton wrote:
> Hi Weiliang, just one question below, thx!
> --Michael
> 
> https://codereview.chromium.org/23531084/diff/1/src/x64/macro-assembler-x64.cc
> File src/x64/macro-assembler-x64.cc (right):
> 
>
https://codereview.chromium.org/23531084/diff/1/src/x64/macro-assembler-x64.c...
> src/x64/macro-assembler-x64.cc:3731: int space =
> XMMRegister::kMaxNumAllocatableRegisters * kDoubleSize +
> Could you do the same thing on ia32, using
> Register::kMaxNumAllocatableRegisters?

Thanks Michael, Why I don't do the same thing on ia32 is that in my memory, ia32
will keep some value in xmm0 when calling into runtime. So ia32 may
intentionally save all its 8 XMMRegsiters. An example is like the xmm_scratch in
DoDeferredNumberTagI.

[mvstanton, 2013-09-27 11:00:16]

Okay, it sounds good.
lgtm.

[mvstanton, 2013-09-27 12:43:47]

Committed patchset #1 manually as r16989 (presubmit successful).