Improve generated code for string encoding tests on ia32.

Improve generated code for string encoding tests on ia32.

Committed: http://code.google.com/p/v8/source/detail?r=4812

[Lasse Reichstein, 2010-06-07 09:29:06]

LGTM

http://codereview.chromium.org/2673001/diff/3001/4001
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/2673001/diff/3001/4001#newcode11310
src/ia32/codegen-ia32.cc:11310: // First check for flat two byte string.
"two-byte"?

http://codereview.chromium.org/2673001/diff/3001/4001#newcode11311
src/ia32/codegen-ia32.cc:11311: __ and_(ebx,
Would andb be enough? (Although I don't think it makes a difference at all when
the register isn't rax, I still like to see the size of the value I'm working
on.)

http://codereview.chromium.org/2673001/diff/3001/4001#newcode11332
src/ia32/codegen-ia32.cc:11332: __ cmp(Operand(edx), Factory::empty_string());
Can you do a direct compare to memory? I.e.,
 cmp(FieldOperand(eax, ConstString::kSecondOffset),
     Factory::empty_string());

http://codereview.chromium.org/2673001/diff/3001/4001#newcode12890
src/ia32/codegen-ia32.cc:12890: __ test(edi, Immediate(kStringEncodingMask));
Could you do an xor (or just an and) here to check that the
asciiness/twobyteness is the same for both strings, instead of testing each for
being two-byte? Or does it use too many registers?

[William Hesse, 2010-06-07 12:17:27]

http://codereview.chromium.org/2673001/diff/3001/4001
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/2673001/diff/3001/4001#newcode11310
src/ia32/codegen-ia32.cc:11310: // First check for flat two byte string.
On 2010/06/07 09:29:06, Lasse Reichstein wrote:
> "two-byte"?

Done.

http://codereview.chromium.org/2673001/diff/3001/4001#newcode11311
src/ia32/codegen-ia32.cc:11311: __ and_(ebx,
All of these changed to test_b.

On 2010/06/07 09:29:06, Lasse Reichstein wrote:
> Would andb be enough? (Although I don't think it makes a difference at all
when
> the register isn't rax, I still like to see the size of the value I'm working
> on.)

http://codereview.chromium.org/2673001/diff/3001/4001#newcode11332
src/ia32/codegen-ia32.cc:11332: __ cmp(Operand(edx), Factory::empty_string());
On 2010/06/07 09:29:06, Lasse Reichstein wrote:
> Can you do a direct compare to memory? I.e.,
>  cmp(FieldOperand(eax, ConstString::kSecondOffset),
>      Factory::empty_string());
> 

Done.

http://codereview.chromium.org/2673001/diff/3001/4001#newcode12890
src/ia32/codegen-ia32.cc:12890: __ test(edi, Immediate(kStringEncodingMask));
We can do a test with the other one and the mask, combined, to check that they
are both ascii.

Done.