More precise type lub for numbers

More precise type lub for numbers

R=jkummerow@chromium.org
BUG=v8:2910
Committed: http://code.google.com/p/v8/source/detail?r=17139

[Jakob Kummerow, 2013-10-02 13:35:06]

LGTM with comment.

https://codereview.chromium.org/25722002/diff/1/src/types.cc
File src/types.cc (right):

https://codereview.chromium.org/25722002/diff/1/src/types.cc#newcode134
src/types.cc:134: if (value->ToInt32(&i)) return kOtherSigned32;
How about:
        if (value->ToInt32(&i)) {
          if (Smi::IsValid(i)) return kSmi;
          return kOtherSigned32;
        }

[rossberg, 2013-10-02 13:43:03]

https://codereview.chromium.org/25722002/diff/1/src/types.cc
File src/types.cc (right):

https://codereview.chromium.org/25722002/diff/1/src/types.cc#newcode134
src/types.cc:134: if (value->ToInt32(&i)) return kOtherSigned32;
On 2013/10/02 13:35:06, Jakob wrote:
> How about:
>         if (value->ToInt32(&i)) {
>           if (Smi::IsValid(i)) return kSmi;
>           return kOtherSigned32;
>         }

Hm, right, there might be smis denormalised into heap numbers.

[rossberg, 2013-10-10 11:59:14]

https://codereview.chromium.org/25722002/diff/1/src/types.cc
File src/types.cc (right):

https://codereview.chromium.org/25722002/diff/1/src/types.cc#newcode134
src/types.cc:134: if (value->ToInt32(&i)) return kOtherSigned32;
On 2013/10/02 13:43:03, rossberg wrote:
> On 2013/10/02 13:35:06, Jakob wrote:
> > How about:
> >         if (value->ToInt32(&i)) {
> >           if (Smi::IsValid(i)) return kSmi;
> >           return kOtherSigned32;
> >         }
> 
> Hm, right, there might be smis denormalised into heap numbers.

Done.

[rossberg, 2013-10-10 12:30:11]

Committed patchset #2 manually as r17139 (presubmit successful).