Use generated code to create sub strings

Use generated code to create sub strings.

Added a stub to allocate and fill a string object with a substring from another string.

Use the rep movs instruction to copy the string data as it turned out to be the fastest way.

While preparing this I experimented with some SSE2 instructions, so the instructions movdqa and movdqu are still in the IA-32 assembler even though they are not used.

Committed: http://code.google.com/p/v8/source/detail?r=3554

[Mads Ager (chromium), 2010-01-07 09:26:06]

Remove the jump to runtime for two-byte strings and retest?

Remember to file a bug report for the missing x64 and ARM version of this
optimization.

http://codereview.chromium.org/525085/diff/1/4
File src/ia32/assembler-ia32.cc (right):

http://codereview.chromium.org/525085/diff/1/4#newcode756
src/ia32/assembler-ia32.cc:756: EMIT(0xF3);
 EnsureSpace ensure_space(this);
 last_pc_ = pc_;

?

http://codereview.chromium.org/525085/diff/1/2
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/525085/diff/1/2#newcode9337
src/ia32/codegen-ia32.cc:9337: __ jmp(&runtime);
Hmm, remove and make sure to test again?

http://codereview.chromium.org/525085/diff/1/2#newcode9358
src/ia32/codegen-ia32.cc:9358: // As from is a smi it is 2 * the value which
matches the size of a two byte
2 * -> two times

http://codereview.chromium.org/525085/diff/1/3
File src/ia32/codegen-ia32.h (right):

http://codereview.chromium.org/525085/diff/1/3#newcode757
src/ia32/codegen-ia32.h:757: // additional setup and checking in
GenerateCopyCharactersREP adds to much
to -> too

[Søren Thygesen Gjesse, 2010-01-07 09:46:23]

http://codereview.chromium.org/525085/diff/1/4
File src/ia32/assembler-ia32.cc (right):

http://codereview.chromium.org/525085/diff/1/4#newcode756
src/ia32/assembler-ia32.cc:756: EMIT(0xF3);
On 2010/01/07 09:26:06, Mads Ager wrote:
>  EnsureSpace ensure_space(this);
>  last_pc_ = pc_;
> 
> ?

Done.

http://codereview.chromium.org/525085/diff/1/2
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/525085/diff/1/2#newcode9337
src/ia32/codegen-ia32.cc:9337: __ jmp(&runtime);
On 2010/01/07 09:26:06, Mads Ager wrote:
> Hmm, remove and make sure to test again?

Thanks for spotting this. Removed this jmp and added

  __ j(not_equal, &runtime);

after the test for flat two byte string. And this code is expertized by some
tests.

http://codereview.chromium.org/525085/diff/1/2#newcode9358
src/ia32/codegen-ia32.cc:9358: // As from is a smi it is 2 * the value which
matches the size of a two byte
On 2010/01/07 09:26:06, Mads Ager wrote:
> 2 * -> two times

Done.

http://codereview.chromium.org/525085/diff/1/3
File src/ia32/codegen-ia32.h (right):

http://codereview.chromium.org/525085/diff/1/3#newcode757
src/ia32/codegen-ia32.h:757: // additional setup and checking in
GenerateCopyCharactersREP adds to much
On 2010/01/07 09:26:06, Mads Ager wrote:
> to -> too

Done.

[Mads Ager (chromium), 2010-01-07 09:50:38]

LGTM