Improve array to string conversion.

Improve array to string conversion.

BUG=v8:2435
Committed: https://code.google.com/p/v8/source/detail?r=13144

[Yang, 2012-12-03 13:08:51]

Please take a look.

I suggest starting with string.js and uri.js.
%NewString and %TruncateString have been newly introduced into runtime.cc
%_OneByteSeqStringSetChar and %_TwoByteSeqStringSetChar have both a full-codegen
and a crankshaft implementation, but the actual code generating parts are in
codegen-<arch>.cc

[Toon Verwaest, 2012-12-05 14:29:12]

lgtm with comments.

https://codereview.chromium.org/11348349/diff/1/src/arm/codegen-arm.cc
File src/arm/codegen-arm.cc (right):

https://codereview.chromium.org/11348349/diff/1/src/arm/codegen-arm.cc#newcod...
src/arm/codegen-arm.cc:561: __ strh(value, MemOperand(ip, index));
Please add a comment here as well to indicate that LSR 1 takes care of
smi-untag; and maybe a STATIC_ASSERT to document reliance on this fact.

https://codereview.chromium.org/11348349/diff/1/src/ia32/codegen-ia32.cc
File src/ia32/codegen-ia32.cc (right):

https://codereview.chromium.org/11348349/diff/1/src/ia32/codegen-ia32.cc#newc...
src/ia32/codegen-ia32.cc:833: __ mov_w(FieldOperand(string, index, times_1,
SeqString::kHeaderSize),
Can we add a comment that this works because of the Smi-tagging? Or at least a
STATIC_ASSERT that the tagbit is 0 and shifts by 1?

https://codereview.chromium.org/11348349/diff/1/src/ia32/full-codegen-ia32.cc
File src/ia32/full-codegen-ia32.cc (right):

https://codereview.chromium.org/11348349/diff/1/src/ia32/full-codegen-ia32.cc...
src/ia32/full-codegen-ia32.cc:3087: SeqStringSetCharGenerator::Generate(masm_,
true, eax, ebx, ecx);
Can we use an enum rather than bool to indicate two vs one?

https://codereview.chromium.org/11348349/diff/1/src/json-stringifier.h
File src/json-stringifier.h (right):

https://codereview.chromium.org/11348349/diff/1/src/json-stringifier.h#newcod...
src/json-stringifier.h:606:
SeqString::cast(*current_part_)->Truncate(current_index_);
Runtime_Truncate returns the resulting string; and does not support length == 0.
This function on the other hand does support length == 0; and relies on the
internal truncate that does not handle == 0. Can we just unify it all and
support length == 0 directly in the Truncate called here, and make it return the
new string? That way we can just support truncation to length == 0 in
Runtime_Truncate as well and don't need external checks. Then we can also remove
this utility function and just call truncate directly where you'd call
ShrinkCurrentPart.

https://codereview.chromium.org/11348349/diff/1/src/string.js
File src/string.js (right):

https://codereview.chromium.org/11348349/diff/1/src/string.js#newcode819
src/string.js:819: var one_byte = %NewString(n, true);
Can we have a javascript macro that indicates that this is "ONE_BYTE" vs
"TWO_BYTE"? That would be much more readable.