Unify canonicalization of HAdd/HSub/HMul a bit.

Unify canonicalization of HAdd/HSub/HMul a bit.

HDiv/HMul are a slightly different story and will be handled in a separate CL.

Committed: http://code.google.com/p/v8/source/detail?r=14322

[Jakob Kummerow, 2013-04-18 09:11:06]

LGTM with a comment.

https://codereview.chromium.org/14296013/diff/1/src/hydrogen-instructions.cc
File src/hydrogen-instructions.cc (right):

https://codereview.chromium.org/14296013/diff/1/src/hydrogen-instructions.cc#...
src/hydrogen-instructions.cc:1444: return this;
Instead of:
    HArithmeticBinaryOperation::Canonicalize();
    return this;
I would prefer:
    return HArithmeticBinaryOperation::Canonicalize();
Otherwise that could lead to surprising behavior when someone puts more logic
into the superclass function and doesn't notice that callers ignore the return
value.

[Sven Panne, 2013-04-18 09:19:18]

Feedback addressed, landing...

https://codereview.chromium.org/14296013/diff/1/src/hydrogen-instructions.cc
File src/hydrogen-instructions.cc (right):

https://codereview.chromium.org/14296013/diff/1/src/hydrogen-instructions.cc#...
src/hydrogen-instructions.cc:1444: return this;
On 2013/04/18 09:11:06, Jakob wrote:
> Instead of:
>     HArithmeticBinaryOperation::Canonicalize();
>     return this;
> I would prefer:
>     return HArithmeticBinaryOperation::Canonicalize();
> Otherwise that could lead to surprising behavior when someone puts more logic
> into the superclass function and doesn't notice that callers ignore the return
> value.

Good point, I even think I had something similar in previous versions. :-/
Changed here and in HSub/HMul.

[Sven Panne, 2013-04-18 09:24:35]

Committed patchset #2 manually as r14322 (presubmit successful).