Box values of LoadCodeUnits that fall outside of smi range on 32bit platforms.

Box values of LoadCodeUnits that fall outside of smi range on 32bit platforms.

R=fschneider@google.com, vegorov@google.com
BUG=

[zerny-google, 2014-11-17 15:07:39]

This is based on top of the main irregexp CL [1] to avoid making that CL any
bigger to review.

After eliminating the instance calls to the dart impl of LoadCodeUnits for the
unoptimizing pass, I missed that the use of unboxed mints in the case of a 32bit
packed value. This CL adds code for boxing such a value in the unoptimized
compiler pass.

PTAL.

[Florian Schneider, 2014-11-18 12:20:24]

lgtm

https://codereview.chromium.org/731943004/diff/1/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language_arm.cc (right):

https://codereview.chromium.org/731943004/diff/1/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language_arm.cc:1870: const bool kMightBox =
(representation() == kTagged) && !can_pack_into_smi();
Since this is not a compile-time constant, I'd rename it:

s/kMightBox/might_box/

[zerny-google, 2014-11-18 12:34:04]

Thanks for reviewing.

https://codereview.chromium.org/731943004/diff/1/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language_arm.cc (right):

https://codereview.chromium.org/731943004/diff/1/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language_arm.cc:1870: const bool kMightBox =
(representation() == kTagged) && !can_pack_into_smi();
On 2014/11/18 12:20:24, Florian Schneider wrote:
> Since this is not a compile-time constant, I'd rename it:
> 
> s/kMightBox/might_box/

Yes. Done here and in _ia32.cc

[zerny-google, 2014-11-24 13:08:26]

Merged into https://codereview.chromium.org/744853003/