In MacroAssembler::JumpIfNotBothSequentialAsciiStrings a custom mask

In MacroAssembler::JumpIfNotBothSequentialAsciiStrings a custom mask
helps us decide if we have two ascii strings. We don't care if they
are internalized or not. A few days ago we flipped the meaning of the
internalized bit in INSTANCE_TYPE, and that broke this custom mask.
This CL effects a repair.

BUG=
R=yangguo@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=15797

[mvstanton, 2013-07-22 08:49:15]

Hi Yang, here is the additional fix needed for the recent internalized string
change. Thanks!
--Michael

[Yang, 2013-07-22 09:09:46]

Got some comments. It would be great if you could audit all the uses of
*_STRING_TYPE instance types.

https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.cc
File src/arm/macro-assembler-arm.cc (right):

https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.c...
src/arm/macro-assembler-arm.cc:3269: const int kFlatAsciiStringTag =
ASCII_INTERNALIZED_STRING_TYPE;
I actually prefer not to use the defined instance type, but instead something
corresponding to the mask:

kStringTag | kOneByteStringTag | kSeqStringTag

https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.c...
src/arm/macro-assembler-arm.cc:3284: int kFlatAsciiStringTag =
ASCII_STRING_TYPE;
I believe this suffers from the same issue!

https://codereview.chromium.org/19514004/diff/1/src/x64/macro-assembler-x64.cc
File src/x64/macro-assembler-x64.cc (right):

https://codereview.chromium.org/19514004/diff/1/src/x64/macro-assembler-x64.c...
src/x64/macro-assembler-x64.cc:2286: cmpl(scratch, Immediate(kStringTag |
kSeqStringTag | kOneByteStringTag));
I like it this way. Less error prone apparently.

[mvstanton, 2013-07-22 09:40:17]

Hi Yang, thanks for the push to look harder. The audit found another issue in
full-codegen-ia32.cc. And it's good policy to build from elemental tags when
building custom masks.
Thanks,
--Michael

https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.cc
File src/arm/macro-assembler-arm.cc (right):

https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.c...
src/arm/macro-assembler-arm.cc:3269: const int kFlatAsciiStringTag =
ASCII_INTERNALIZED_STRING_TYPE;
On 2013/07/22 09:09:46, Yang wrote:
> I actually prefer not to use the defined instance type, but instead something
> corresponding to the mask:
> 
> kStringTag | kOneByteStringTag | kSeqStringTag

Done.

https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.c...
src/arm/macro-assembler-arm.cc:3284: int kFlatAsciiStringTag =
ASCII_STRING_TYPE;
On 2013/07/22 09:09:46, Yang wrote:
> I believe this suffers from the same issue!

Grr. Removed defined instance check here too, thanks.

https://codereview.chromium.org/19514004/diff/1/src/x64/macro-assembler-x64.cc
File src/x64/macro-assembler-x64.cc (right):

https://codereview.chromium.org/19514004/diff/1/src/x64/macro-assembler-x64.c...
src/x64/macro-assembler-x64.cc:2286: cmpl(scratch, Immediate(kStringTag |
kSeqStringTag | kOneByteStringTag));
On 2013/07/22 09:09:46, Yang wrote:
> I like it this way. Less error prone apparently.

Indeed, way better. thx.

[Yang, 2013-07-22 09:46:06]

On 2013/07/22 09:40:17, mvstanton wrote:
> Hi Yang, thanks for the push to look harder. The audit found another issue in
> full-codegen-ia32.cc. And it's good policy to build from elemental tags when
> building custom masks.
> Thanks,
> --Michael
> 
> https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.cc
> File src/arm/macro-assembler-arm.cc (right):
> 
>
https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.c...
> src/arm/macro-assembler-arm.cc:3269: const int kFlatAsciiStringTag =
> ASCII_INTERNALIZED_STRING_TYPE;
> On 2013/07/22 09:09:46, Yang wrote:
> > I actually prefer not to use the defined instance type, but instead
something
> > corresponding to the mask:
> > 
> > kStringTag | kOneByteStringTag | kSeqStringTag
> 
> Done.
> 
>
https://codereview.chromium.org/19514004/diff/1/src/arm/macro-assembler-arm.c...
> src/arm/macro-assembler-arm.cc:3284: int kFlatAsciiStringTag =
> ASCII_STRING_TYPE;
> On 2013/07/22 09:09:46, Yang wrote:
> > I believe this suffers from the same issue!
> 
> Grr. Removed defined instance check here too, thanks.
> 
> https://codereview.chromium.org/19514004/diff/1/src/x64/macro-assembler-x64.cc
> File src/x64/macro-assembler-x64.cc (right):
> 
>
https://codereview.chromium.org/19514004/diff/1/src/x64/macro-assembler-x64.c...
> src/x64/macro-assembler-x64.cc:2286: cmpl(scratch, Immediate(kStringTag |
> kSeqStringTag | kOneByteStringTag));
> On 2013/07/22 09:09:46, Yang wrote:
> > I like it this way. Less error prone apparently.
> 
> Indeed, way better. thx.

lgtm.

[mvstanton, 2013-07-22 09:55:21]

Committed patchset #2 manually as r15797 (presubmit successful).