Improve keyed loads on strings by using a new stub

Improve keyed loads on strings by using a new stub.

Instead of going through a runtime function for keyed loads
on strings we invoke a separate specialized stub that
assumes string as receiver type and the key to be a number.

The stub calls a JS builtin function to return the corresponding one-character string.



Committed: http://code.google.com/p/v8/source/detail?r=3556

[Kasper Lund, 2010-01-07 09:58:58]

LGTM.

http://codereview.chromium.org/521041/diff/3005/4018
File src/ia32/ic-ia32.cc (right):

http://codereview.chromium.org/521041/diff/3005/4018#newcode402
src/ia32/ic-ia32.cc:402: __ pop(eax);  // return address
Maybe add a comment explaining why popping the return address of the stack is a
good idea in the common case.

http://codereview.chromium.org/521041/diff/3005/4018#newcode419
src/ia32/ic-ia32.cc:419: __ bind(&index_ok);
Maybe add a comment explaining why you need to duplicate the receiver and key on
the stack.

http://codereview.chromium.org/521041/diff/3005/4018#newcode421
src/ia32/ic-ia32.cc:421: __ push(Operand(esp, 1 * kPointerSize));  // key
__ push(edx) ?

[fschneider, 2010-01-07 10:23:55]

http://codereview.chromium.org/521041/diff/3005/4018
File src/ia32/ic-ia32.cc (right):

http://codereview.chromium.org/521041/diff/3005/4018#newcode402
src/ia32/ic-ia32.cc:402: __ pop(eax);  // return address
On 2010/01/07 09:58:58, Kasper Lund wrote:
> Maybe add a comment explaining why popping the return address of the stack is
a
> good idea in the common case.

Done.
I;m not sure if it make a big difference. But I guess performing the load early
is better in the fast case.

http://codereview.chromium.org/521041/diff/3005/4018#newcode419
src/ia32/ic-ia32.cc:419: __ bind(&index_ok);
On 2010/01/07 09:58:58, Kasper Lund wrote:
> Maybe add a comment explaining why you need to duplicate the receiver and key
on
> the stack.

Done.

http://codereview.chromium.org/521041/diff/3005/4018#newcode421
src/ia32/ic-ia32.cc:421: __ push(Operand(esp, 1 * kPointerSize));  // key
On 2010/01/07 09:58:58, Kasper Lund wrote:
> __ push(edx) ?

Done.

[Kevin Millikin (Chromium), 2010-01-07 10:26:33]

LGTM.

http://codereview.chromium.org/521041/diff/3005/4018
File src/ia32/ic-ia32.cc (right):

http://codereview.chromium.org/521041/diff/3005/4018#newcode407
src/ia32/ic-ia32.cc:407: __ mov(ecx, FieldOperand(ebx, HeapObject::kMapOffset));
The next three instructions are abstracted as
MacroAssembler::IsObjectStringType:

Condition is_string = masm->IsObjectStringType(ebx, ecx, ecx);
__ j(NegateCondition(is_string), &miss);