Adds support for UseDartApi vm test on MIPS.

Adds support for UseDartApi vm test on MIPS.

Also, adds:
- xori instruction,
- SubuDetectOverflow (like AdduDetectOverflow).
Committed: https://code.google.com/p/dart/source/detail?r=21396

[regis, 2013-04-12 21:18:04]

LGTM

https://codereview.chromium.org/13884017/diff/1/runtime/vm/intermediate_langu...
File runtime/vm/intermediate_language_mips.cc (right):

https://codereview.chromium.org/13884017/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language_mips.cc:478: __ CompareObject(CMPRES,
right.reg(), left.constant());
Do you still have a reason to dedicate a register for the result of a compare?
If this register (CMPRES) is not implicitly referred to anywhere, you could use
TMP1 explicitly instead.
Never mind, I see you refer to CMPRES implicitly in EmitBranchOnCondition, for
example.
We could try to always use TMP2 as CMPRES and save a dedidated temp register (we
have 3 now), but this may be dangerous.

https://codereview.chromium.org/13884017/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language_mips.cc:1116: // Fall through.
I merged the kSUB case with kADD case on ARM. Since you have a
SubuDetectOverflow, you should keep these cases separate (look at the IA32
implementation) and avoid the issue of the TODO.

[zra, 2013-04-12 22:19:59]

Thanks!

https://codereview.chromium.org/13884017/diff/1/runtime/vm/intermediate_langu...
File runtime/vm/intermediate_language_mips.cc (right):

https://codereview.chromium.org/13884017/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language_mips.cc:478: __ CompareObject(CMPRES,
right.reg(), left.constant());
On 2013/04/12 21:18:04, regis wrote:
> Do you still have a reason to dedicate a register for the result of a compare?
> If this register (CMPRES) is not implicitly referred to anywhere, you could
use
> TMP1 explicitly instead.
> Never mind, I see you refer to CMPRES implicitly in EmitBranchOnCondition, for
> example.
> We could try to always use TMP2 as CMPRES and save a dedidated temp register
(we
> have 3 now), but this may be dangerous.

Yah, I think it's probably less error-prone to keep CMPRES around. Removing TMP2
is probably easier since I've only used it in a couple places so far, but I'll
hold off on that   in case I want it more later, or until we find out that
giving more registers to the allocator improves performance.

https://codereview.chromium.org/13884017/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language_mips.cc:1116: // Fall through.
On 2013/04/12 21:18:04, regis wrote:
> I merged the kSUB case with kADD case on ARM. Since you have a
> SubuDetectOverflow, you should keep these cases separate (look at the IA32
> implementation) and avoid the issue of the TODO.

Done.

[zra, 2013-04-12 22:20:40]

Committed patchset #2 manually as r21396 (presubmit successful).